Home

Software Visualization in Prolog

1. Java method public String java math BigDecimal toStringO public String Javesnath BleDectnal toStrinat public String Javasnath BicDecinal toStrinsO e E Vmax Java method public String java math DigDecimal toString aX Eile Edit Udu Options Cantent Forn Action Debuo Help mag EAE Siber bo Sing intStrng mv al abet ost int insert onPomnt infSinng lengh scal fetur Senan 220705 nso i gt buf new StrincBufierlmEting out insenfinsenionPein 7 Neigra ele buf emnend niStino Tatlin bu toString Java method public String java math BigDecimal toString 6 E Vmax Java method public String java math BigDecimal toString ax File Edit lUiredow Options Content Form ction Debug Help Sip return inal ro Sino Stine ur buf String misting mival abs Mesta fin sionumn soun T i eS Signo PHOTO PS els i gt ew String Boffer Siina but incengnsertionPoirt bub nseri io iSong length Dut spnendint tind Tetum buitoSiinad Java method public String java math BigDecimal toString0 1F lt seale DP 1F lt seale 0 i 2 h 108 5 6 JAVA VISUALIZATION Java method public Str EigDecimaldoString ax Java method public String java math Big Decimal to String IX File Edit Undu Options Content File Edit Window O
2. 0 0 4 7 1 Disambiguating Actions 4 7 2 Menus 4 7 3 Navigation 4 7 4 Selecting Graphical Form 4 7 5 Modifying the Action Relation The Legends bli hee aes 4 8 1 Displaying the Legend 4 8 2 Modifying the Visualization Relation 4 9 The Bookmarks 4 10 The Preview Window 4 11 Chapter Summary Introduction o Architecture 2 o o Text Processing Directory Visualization 5 4 1 MIME Types Analyzing Java Source Code 5 5 1 Constructing the Program Database 5 5 2 Analyzing the Program Database Java Visualization 5 6 1 Visualizing Files Classes and Packages 5 6 2 Visualizing Cross references 5 6 3 Visualizing Methods CONTENTS CONTENTS 7 5 7 Analyzing Run time Data 2 2 2 2 0000 00000 113 5 7 1 Adding Trace Points sise s aoa es a a a 2 0 00 0000 113 5 7 2 Generating the Trace File ms e emoi an e Da ee 113 5 7 3 Reading the Trace File aoaaa a 115 5 8 Run time Data Visualization a 116 5 8 1 Value Visualization ee ee 120 5 9 Prolog Visualization e esger mee e AP Oe AAA RE A 127 O92 gt Editing Prolog 20h Pad hm Oy a ea eR da 129 5 10 The Help System coue Tane 224 604 h eae bob de 131 5 14 Chapter Summary cece epee oe eS Goa ee ee ees 133 6 Discussion 135 6 1 Classifying Vmax 0c BA ee ee ba ee 135 61L SCOPE as rel eR eel a Beek Oe wee a Fe Boe 135 621 2 Content a Beh do eee ta aed eee oe A 137 6 1 32 FORM 2 2 4606 wii
3. 4 vmax Examine value array sil EJ ae File Edit Window Options Content Form File Edit Window Options Content Form Examine value array Examine value array a Frame 1 b Frame 2 HE vmax Examine value array DX A vmax Examine value array E File Edit Window Options Content Form File Edit Window Options Content Form Examine value array Examine value array c Frame 8 d Frame 9 Figure 5 44 Animating an event line 120 5 8 RUN TIME DATA VISUALIZATION An individual input can be visualized by clicking on it in a view or by clicking on its marker on the event line Pressing the n key or navigating to the Next input view of an input will navigate to the chronologically next input of the same event Pressing the p key or navigating to the Previous input view of an input will navigate to the previous input of the same event The Any next input view displays the next input from any event and Any previous input displays the previous input from any event The Next 5 inputs view of an input will display the next five inputs from the same event and the Any next 5 inputs will display the next five inputs from all events shown in Figure 5 45 The statement at which the trace occurred can be navigated to by selecting the Go to statement view of the event line 3 Vinax Trace next 5 inputs EONS File Edit Window Options Content Form Action Debug Help 175 from W 180 array 204 u0 Al 209 10 Aa Tr
4. Figure 2 4 Microsoft Visual Basic 5 0 showing dialog box design and the project navigation tree Genitor Genitor from Genitor Corp goes further than ordinary development environments because it offers a suite of tools for automatic navigation project management and documentation Its class editor shown in Figure 2 5 provides a navigation tree that visualizes the method structure and can be manipulated via menus and drag and drop to change the program objects While Genitor has access to a great deal of data about the program its graphical output is very limited A navigation tree shows only the structure of the file and other interrelationships are not illustrated but accessed through navigation Its views are hard coded using just lists and a nested list view Source Code Analyzers Source Navigator from Cygnus Software is a development tool for C Java and COBOL that provides tree and graph views of the system It shows the class hierarchy cross references between classes symbols used in the source code in a tree and a graph of header file inclusion It can manage projects in excess of 100 000 lines of source code CC Rider from Western Wares uses a source code analyzer to view the structure of C programs It can show class hierarchies class ancestries function call trees class nesting header file inclusion symbols and program statistics It can be used to navigate to parts of the source code and provide
5. Output is Drunk the tea Drunk the coffee Drunk the coffee C 6 A Scientific Calculator This example gives the implementation of the scientific calculator in Figure 6 2 Calculator application implemented in SVT Written by Calum Grant o 6 o module svt multifile view_content 2 view 3 action 3 reaction 2 object_component 3 visual_content 3 visual_component 2 visual_context 2 visual 5 visual_context_description 2 popup calculator popup frame calculator 0 0 1 true 0 calculator svt svt o Management of calculator state calculator evaluate A B A B calculator evaluate A B A B calculator evaluate A B A B calculator evaluate A B A B calculator evaluate A B x y exp A B calculator evaluate _ N N A 22 A A A sqrt B A gt 0 gt B is sqrt A B is nan calculator evaluate A sin sin A A cos cos A A tan tan A calculator evaluate calculator evaluate calculator evaluate calculator evaluate 190 C 6 A SCIENTIFIC CALCULATOR calculator evaluate A 1 x 1 A calculator evaluate A ln B A gt 0 gt B is log A B is nan calculator evaluate A exp exp A calculator evaluate _ C 0 calculator evaluate A x B A gt 170 gt B is inf A2 is truncate A A gt 0 gt factorial A2 B B is nan calculator memory _ R MC 0 R calculator memory
6. long identifier Dir Name CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 85 view_content directory view Dir text Text text_node Dir Text view content directory view Dir file File Name member Dir File file File identifier File Name view content directory view Dir directory File Name member Dir File directory File identifier File Name view_content directory view Dir file size File Size member Dir File file size File Size view content directory view Dir file age File Age member Dir File file age File Age view content directory view Dir executable File member Dir File executable File view content directory view Dir source code File member Dir File source _code File Three different visualizations of this view are shown in Figure 5 4 vmax Directory Momesicamgl00 0cal 8 TPX File Edit Window Options Content Form File Edit Window Options Content Form y t gt a a UE E bookmark htm dynfonts java lib libjsd so bookmark htm dynfonts java lib libjsd so a m plugins lib dynfonts registry spell nethelp A Me movemaiksre java vreg netscape dynMotif m ye libnullplugin dynMotifso LICENSE movemail src libnullplugin dynMatifso LICENSE movemail sre pe a all movernail nethelp netscape dynMotif netscape R a E movemail nethelp netscape dynMotif netscape Net
7. Zeus uses Modula 3 s class hierarchy to define views of an algorithm Each running algorithm generates animation events which call the animation library to provide output to the user The effort of setting up new animations is high because the algorithm must be implemented within the Zeus framework and hence does not correspond to real world programs The programmer must manually insert calls to Balsa and Zeus to perform the animation actions requiring a priori detailed knowledge of the algorithm Pavane Pavane 84 86 is a visualization system that is capable of visualizing the run time behaviour of parallel and massively parallel architectures Parallel distributed and concurrent execution is a very common subject of visualization because concurrent execution is so difficult to understand CHAPTER 2 PREVIOUS WORK 29 The implementation consists of five modules that can be distributed across a network pro viding a parser to an intermediate language a run time execution monitor visualization utilities a renderer and a viewer This architecture allows remote and distributed visualization Visual izations in Pavane are specified in a declarative style 85 where a mapping is defined between the data space and objects in the visual space This gives a powerful degree of abstraction of the visualization from the data Smooth animation can also be defined Its output is shown in Figure 2 11 Figure 2 11 An animation of sorting by P
8. wl public static long com icetar TarHeader parse Octal byte header int offset int length throws Invalid HeaderException wile publi static StringBut fer com ice tar TarHeader parseName byte header int offset int length throws InvalidHeaderException a pubic com ice tar Tar Entry TarEntry Tar rchive archive byte headerBuf throws InvalidHeaderException alo pubic TarHeader com ice tar TarEntry parseTarHeadertbyte header throws InvalidHeaderException a pubic void com ice tar Tar Archive JistContentsQthrows InvalidHeaderException IO Exception wi public void com ice tar TarArchive extract Contents File dest Dir throws InvalidHeaderException IOException wie pubio Tarfntry comice ta Tarfrohive readEntryQthrows InvalidHeaderException IOException Places where com ice tar InvalidHeaderException is thrown public static long com ice tar TarHeader parselctal byte header int offset int length throws ease Figure 5 24 Methods where a class is thrown CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 101 E Methods reachable from public void JavaToolbar actionPerfonned ActionEvent evt Hui File Edit Window Options Content Form Action Debug public void JavaToolbar actionPerformed Action Event evt private woid pr n ny private void JavaToolbar show Output public eee filters String description public JavaFileFitter JavaFileFitter String extension String description Public void Java File F
9. EnMsgAdapter J DatabasePanel j CommandPanel AboutPanel Java files under homes camg100 Work SVT bin samples Java com ice sqlclient homes camg100 Work SYT bin samples Java com ice sqlclient UpdatePanel java j Figure 5 6 View showing just Java files in a directory The other views of a directory are selected from its Content menu shown in Figure 5 5 The list of Java files in a directory is shown in Figure 5 6 This view is declared view Dir java_dir Dir Java files in directory directory Dir view_content java_dir Dir title Java files under Name long identifier Dir Name view_content java_dir Dir class File Name member Dir File java_source code File short_identifier File Name view content java_dir Dir directory File Name member Dir File directory File identifier File Name Directories are structured in trees which can be displayed in a number of ways The basic directory tree is specified view Dir dir tree Dir Directory tree directory Dir view content dir _tree Root title Directory tree under Name long _identifier Root Name view content dir _tree Root directory Dir self or contains Root Dir view content dir _tree Root contains Dir1 Dir2 self or contains Root Dirl directory Dirl1 member Dir1 Dir2 directory Dir2 self_or contains Self Self self o
10. bolVisual for an identifier Symbolld The type of the visual is Type The graphical symbol Symbol Visual that depicts the view content Content is retrieved by visual_content Content VisualContext Visual visual Visual _ _ Symbolld SymbolVisual Once the view content and its accompanying symbol have been declared using the content 3 and visual 5 predicates they are automatically displayed in the legend whenever they are used Example 25 The content type file 2 is declared content file File Text Id str Text A file named Text The visual labeled_icon 3 is declared visual labeled_icon Id Text Icon Id str Text A labeled icon of a Icon Symbol labeled icon Symbol Text Icon icon Icon 4 8 2 Modifying the Visualization Relation The second function of the legend allows the user to edit the visualization relation to dynami cally modify how information is graphically represented The legend has an action context that displays a pop up menu when an entry in the menu is clicked on by the mouse This menu dis plays a list of alternative graphical forms for the data and the current graphical representation is indicated in the menu as shown in Figure 4 25 Selecting an option from this menu changes the visualization relation to use the selected graphical form Both the visualization relation and the legend are updated to reflect the changes Figure 4 26 shows an example of changing a vis
11. hen ntal nSirng Siringuifor buf String int St nt nseries reote oi else HmeatenPot 0 buf row Stings a i etna uteeort 9 a buf now SinngeuterG neettiorBont insire length B ppoe O 0 ou font Ps 1 ieerlon oint H ul apando uf appending tut but tasting Java method public String java math BigDecimal toString CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION CI Sx Symbol Meaning StringBuffer buf buf new StringBuffer intString fort i 0 i lt insertionPoint i jody A declaration statement An expression statement A for loop cone aen ifthen statement econ true ae iethen else statement If false Stang oStingd Method declarator retum intVal toString Exit statement Sequential statements Legend for Text b The initial legend E Vmax Java method public String java math BigDecimal toString TA File Edit Widow Options Content Form fetion Debug Help Stang toting a return int Val toShinal Sigan tut Shing ining nivel abs gaara lat signum siarum int inserionPoint Bling lengik Seale insertan cim O t 1 else irsertianPoint gt 0 i etum signum gt E O e Sng buf new StringBuferfint String nacio o iffsignum lt 0 bufinset buf new StineBulfer insertionPoint int Sting length al oe 5 eo forint 0 1 lt ingertionPoirt i kuf apponi
12. marks that store views for later retrieval Such an interface would provide an easy means to access data 9 Provide a text editing window that is synchronized with graphical views of the source code This would provide code browsing and a complete development environment 1 2 Work Done These aims have been met by redesigning the visualization engine that sits at the core of the SV system to generate the views A specification method for information visualization has been developed by decomposing the visualization process into a series of relations that can be specified in Prolog This provides a formal foundation to information visualization and therefore to SV The formal model is called a semantic model because the information content of a view or its semantics is always defined explicitly The model also incorporates a formal model of interaction with visualizations This method has been shown to be practical by implementing a general purpose information visualization tool Semantic Visualization Tool SVT using this approach SVT can in princi ple be used for for any type of visualization and it is specified entirely in Prolog It provides a method for interfacing any knowledge based environment in Prolog to any graphical output algorithm SVT abstracts the information content of a view from its graphical representation The map ping from the information content to the graphical representation is dynamic so that different mappin
13. v2 tree edge A B brown visual content wire A B neutral v2 tree edge A B blue visual content wire A B earth v2 tree edge A B green The alternative forms are selected by changing the visualization context Example 12 What is wrong with the following visualization relations visual _content man X v box X visual content woman X v box X visual content ship S Name w ship icon S Visualization relation v is not injective so two different types of view content can map to the same visual Both man X and woman X are represented by the same visual Thus a user seeing a box X would not be able to tell whether it represented a man X or a woman X Visualization relation w loses information by another means It does not convey the name of the ship Name to ship_icon 1 A visual type system described in Section 3 5 can prevent such information loss Example 13 Given 46 3 5 A VISUAL TYPE SYSTEM person A man A person A woman A view_content c person X alive X person A view_content c married X Y married X Y view_content c children X Y married X Y child X C child Y C visual content person X v circle X red visual content married X Y v line X Y visual content children X Y v line colour X Y green in view context c and visualization relation v what can a user deduce from a green line interconnecting
14. virtual int Window width 0 virtual int Window height 0 virtual void GL_redraw 0 virtual void Mouse cursor SVT_cursor 0 virtual void Pulldown_menu char title int id char 0 virtual void Inactivate pulldown int id 0 virtual void Popup menu char 0 virtual void Status text const char 0 virtual void Set_text_window const char text int size bool ro 0 virtual char Get_text window int amp length 0 virtual void Set_text_position int position 0 virtual void Lock _editor 0 virtual void Set_title const char 0 virtual SVT type Get_type 0 virtual void Set_frame size int int 0 virtual void Destroy 0 virtual void Start background process 0 virtual void Stop background process 0 hg class MainBase class Main data public MainBase const char path 0 virtual MainBase void Boot int add _data char pred int data int remove data char pred int data int load prolog char filename char Boot_error public These functions are called by SVT virtual bool Break 0 virtual ViewBase Create popup 0 virtual ViewBase Create viewer 0 virtual char Get_string const char title const char prompt const char def 0 virtual int Get_option const char title const char prompt const char buttonlist 0 virtual char Get_user_arguments int amp 0 endif Appendix C Further Examples This appendix contains
15. CHAPTER 4 SEMANTIC VISUALIZATION TOOL 63 a A hierarchy java lang Throwable H_java lang Class Format Error java lang Exception ninitializerError java lang Incompatible Class Change Eror java tang Exception java lang Class NotFoundException _java lang Clone Not Supported Exception java lang legalAccess Exception java lang Instantiation Exception J java lang InterruptedException java lang No Such Field Exception T java lang AbstractMethod Eror java lang No SuchMethodException H_javatang No Class Def FoundEror_ H_java lang Class CastException H java lang Unsatisfied Link Eror J java lang llegalArqument Exception H java lang verity Eror java lang lllegalThread State Exception java lang ThreadDeath java lang NumberFormat Exception java lang MrtualMachine Error java lang legalMonitorState Exception java tang intemal Eror java lang legal State Exception java Jang Out OfMemory Error java lang Index Out Of Bounds Exception java Jang Stack Overflow Error java lang Array Index Out Of Bounds Exception java lang Unknown Eror java lang Stringlndex Out Of Bounds Exception H _java lang Negative Aray Size Exception java lang Null PointerException java lang Security Exception b A nested list c A hierarchical list Figure 4 13 Tree layou
16. Input is an input event in a trace file File with associated event number EventNo that occurred at time Time and whose value is Value trace next_input Inputl Input2 Not extensible Finds the next input event Input2 after Input1 that has the same event number trace prev_input Inputl Input2 Not extensible Finds the input event Input2 preceeding In putl on the same event line Jjava trace_point EventNo String Statement Dynamic EventNo is an event number that ex amines value String at java statement Statement trace value Value String Term Dynamic Value is a value in a trace file that examines the variable String and whose value is encoded in term Term Term is array Values An array with list of values Values float Float A float or double with value Float integer Int An int with value Int e null A null object object Id Class Fields An object with object id Id class name Class and Fields is a list of members of the form field FieldName Value object Id An object with object id Id 176 B 4 VISUALS DEFINED BY SVT B 3 4 Prolog Source Code exec_clause Clause Not extensible Declares that the clause Clause is executable i e not in the scratch area user_clause Clause Term Dynamic Stores a clause Clause with the structure of the clause stored by the term Term user_term Functor Arity Description Extensible Defines a new term to go in the Change functor menu with n
17. ProxyConnection com ruleofeig corn ruleofeight sf Server Subtypes of java lang Thread Ready Figure 5 16 The subtypes of a class A MA A EE ALUMS cee X File Edit Window Options Content Form Action Debug com ruleofeight mail EmailMessage com ruleofeight mail JavaMailMessage com ruleofeight mail OROEmailMessage com ruleofeight mail SocketMessage Implementations of com ruleofeight mail EmailMessage com ruleofeight mail EmailMessage A Figure 5 17 The implementations of an interface CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 97 Sea Vmax All Java classes EA File Edit Window Options Content Form Action Debug Help om nueofeight util Quote Delegate bomruleofeight sy stem_StdOutConsumer com ruleofeight mail Gmailessage java lang Runnable java lang Cloneable j com ruleofeight util StringWrapper Eom rueoteigie util Quote Tool com_ruleofeight atil HmiDoc bom ruleofeight util File Util com ruleofeight system GnuLauncher hom ruleofeight system Gnulnfol com Juleofeight lt f Simple Connection Bom nueoteigi st Server bom neo feigit s Prox y Connection bom ruleofeight st Connection com ruleofeight objserv _ServerTester com ruleofeight objsery ServerConnection bom ndeofeigt objserv Server com tukot eight mal Test Applet com ruleofeight mail Sockethessage com tuleofeight mail_OROEmailess
18. and pre attentive vision Taking advantage of perceptually rich computer environments In SPIE Proceedings on Human Vision Visual Processing and Digital Display HI volume 1666 pages 504 513 1992 83 B E Rogowitz and L A Treinish Data structures and perceptual structures In SPIE Proceedings on Human Vision Visual Processing and Digital Display HI volume 1913 pages 600 612 1993 84 G C Roman and K C Cox A taxonomy of program visualization systems IEEE Com puter 26 12 11 24 December 1993 85 G C Roman and K C Cox A declarative approach to visualizing concurrent computa tions IEEE Computer 22 10 25 36 1989 86 G C Roman K C Cox C D Wilcox and J Y Plun Pavane a system for declarative visualization of concurrent computations Journal of Visual Languages and Computing 3 161 193 1992 87 M Sarker and M H Brown Graphical fisheye views for graphs In Human Factors in Computing Systems Proceedings of CHI 92 pages 81 92 1992 88 M Sarker S S Snibe O J Tversky and S P Reiss Stretching the rubber sheet A metaphor for viewing large layouts on small screens In Proceedings of the ACM Sym posium on User Interface Software and Technology UIST 93 pages 81 92 1993 rc aS J J Shilling and J T Stasko Using animation to design object oriented systems Object Oriented Systems 1 1 5 19 September 1994 90 B Shneiderman Direct manipulation A step beyond programming lan
19. int syncDemo Object p synchronized p 4 hashtable ad dfx p i hashtable ad d x py return compute p x a Black and white b Colour Figure 5 33 A synchronized block CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 109 void tryDemo void tryDemo loadFiles status INTERNAL status FAILED closeFiles exitCleanupl closeFiles exitCleanup a Black and white b Colour Figure 5 34 A try catch finally block 110 5 6 JAVA VISUALIZATION Y Vmax Java method public int RobotCommand execute Time s xX File Edit Window Options Content Form Action Debug Help Java method public int RobotCommand executeTime Ready a Traditional horizontal switch statement layout Y vmax Java method public int RobotCommand executeTime xX File Edit Window Options Content Form Action Debug Help nt executeTimeQ int p1 paaneterl 0 paraneteri parmeteri R dl y p EESE EASE E A E Java method public int RobotCommand executeTime Ready b A vertical layout of a switch statement Figure 5 35 Alternative graphical forms of a switch statement CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 111 E Method block structure and refs Method block structure Method control flow Visual block structure References of method Reachable definitions cached Reachable definitions Dverriding methods Overridden met
20. mapping is the visualization relation and navigators are the action relation D 2 CHAPTER 6 DISCUSSION 139 A user of Vmax needs to be familiar with Java in order to understand concepts such as in heritance overloading variable usage or function calls The user also needs to understand the source code in order to know where trace points should be inserted Prior understanding of the algorithm is necessary in many systems to create effective output but systems such as Eliot 61 and Leonardo 20 do not require the selection of trace points D 2 1 The visualization system is almost completely decoupled from the target system There is only a small amount of code that is dynamically linked to the target system via Java s class loader to output run time data structures to a trace file 5 7 2 D 2 2 6 1 5 Interaction The interaction style is more advanced than in other systems because it uses a browser interface 84 7 3 Interaction using the mouse and keyboard can be completely specified in Prolog 3 8 or dynamically modified 4 7 5 and can specify menus 4 7 2 and direct manipulation 5 9 E 1 Navigation is easy with a browser interface 84 7 3 which is very important because there are 70 different types of view to choose from E 2 Other SV systems do not provide specification for interactive behaviour Elision control is high as selecting different views of the same object from menus 84 7 3 can
21. right left null BinNode value M left null right null right right Examine value tree left right Ready Ready a e Run time data in a graph f Run time data in a table Ann o lt 4 aculator File Edit Window Options Content Form Action Debug Help m X M court 4 N count ais Mi Hl core SERIA coher SR Copa con is a parent of an ancestor i an cesos o e X is father to Y E is an ancestor or MJ Ris a parent of i Bis a parent or W is mother to Y X is father to Y X is a parent of Y All Clauses g Prolog visualization h GUI specification Figure 1 1 The scope and output of Vmax 16 1 5 NOTATION AND TYPOGRAPHY e New visualizations for program objects including coloured Nassi Shneiderman Diagrams 72 have been developed 1 5 Notation and Typography Text that is written in a slanted font refers to Prolog or C source code e g virtual void On_reset_size Text inset in a Courier font is a larger example of Prolog or C source code e g This is C source code Prolog atoms and functors begin with a lowercase letter e g charles or parent charles harry Prolog compound terms are written as a functor followed by a comma delimited list of
22. visual_content if term Id A B query query Id if A B visual_content term Id Text List query pl_term Id Text List visual_content variable term Id Text query colour _ text Id Text yellow o Block structured visual_context_description block Block structured visual_content tpm predicate Id H T block tpm predicate Id H T visual_content or_term Id A B block pl_or Id A B visual_content and_term Id A B block pl_and Id A B visual content if_term Id A B block infix text ld lt A B visual_content if_term Id A B block block clause Id A B visual_content term Id Text List block pl_term Id Text List visual_content variable term Id Text block colour_text Id Text yellow t El o English visual_context_description prolog_english English visual content tpm _ predicate Id H T prolog english tpm predicate Id H T visual_content or_term Id A B prolog english infix _text Id or A B visual_content and_term Id A B prolog english infix _text Id and A B visual_content if _term Id A B prolog english infix _text Id if A B visual_content term Id Text List prolog_english pl_english Id Text List visual content variable term Id Text prolog english text Id Text o Tree visual_content tpm predicate Id H T tree tpm predicate Id H T visual_content or_term Id A B tree tree n
23. would return which variables are concurrently live with variable v 3 3 Specifying View Content When the data has been gathered it is ready to be viewed A query is issued to the knowledge database to extract the information to display This is a set of terms called the view content which is shown in Figure 3 1 In this way every datum in the view is represented by one term If T is the set of terms defined in Appendix A the view content C is Definition 1 View content Let To C T be the set of terms that represent view content Let C C To be the the view content Example 4 The view content C for a small family tree could be the set person charles person diana person elizabeth person harry mother elizabeth charles father charles harry mother diana harry In a system with multiple views each view is identified by a view context which is represented by a term Definition 2 View context Let Tx C T be the set of terms that are view contexts Let x Tx be a view context Example 5 The view of the entire family tree could have the view context family tree The view showing just the ancestors of harry could have view context ancestors harry The view content C is specified by a relation between the view context x and C CHAPTER 3 A MODEL OF INFORMATION VISUALIZATION 43 Definition 3 View relation A view relation 2 is a relation between Tx and To The view relation defines which terms are in
24. 3 Visualizing Methods Navigating to a method will give a graphical representation of the method body as a Nassi Shneiderman Diagram 72 which is an established graphical representation for block structured code Colour has been introduced to highlight the different types of statement The basic method view is shown in Figure 5 30 See Vmax Java method public String java math BigDecimal toString SELES File Edit Window Options Content Form Action Debug Help public String toString a E a O if scale 0 No decimal point public int java math BigDecimal signum private StringBuffer buf private String intString private int insertionPoint private int signum return intVal toString Insert decimal point E E StringBuffer buf private int javamath BigDecimalscak private BigInteger java math BigDecimal int val String intString intVal abs toString int signum signum StringtoStirg int insertionPoint intString length scales if insertionPoint 0 Point goes right befo return signum lt 0 0 0 intStrings 3 else if insertionPoint gt 0 Point goes insid buf new StringBuffertintString buf insert insertionPoint if signum lt 0 buf insert 0 3 else We must insert zeros between point and i buf new Buffertine Sting but new Buffer nt int String Jeng ae ee AST Eas ee ee intStr futinsentnsenonPom i fb 7 U orpona OEO ee fo
25. 5 11 CHAPTER SUMMARY Chapter 6 Discussion This chapter discusses where Vmax fits into the taxonomies of SV which enable a comparative evaluation of Vmax with other systems In many criteria Vmax outper forms existing systems Vmax s main contribution is in its tailorability flexibility and automation and it is shown that the underlying approach is sufficiently broad to cover all aspects of SV SVT and Vmax are not commercial products and have not been evaluated as such They are intended to demonstrate how Prolog can specify visualization An empir ical evaluation would be needed to demonstrate the effectiveness of Vmax but that is beyond the scope of the work Because of the different approach of Vmax and SVT there are many research di rections that emerge from this work The modular design allows components to be replaced without affecting the rest of the system Further work is discussed includ ing the possibilities of voice visual or interactive specification techniques improved graphical constraint layout and interfacing SVT to commercial visualization and de velopment systems 6 1 Classifying Vmax As well as outlining future directions Price et al 79 provide a framework for comparison and classification This indicates whether Vmax has been successful in addressing the issues raised in the paper The classification of Vmax in their taxonomy is given in Table 6 1 A classification of xx xxx means that Vmax is p
26. A SV system that is based upon this model and specification method has been imple mented e Some new visualizations and visual languages including the use of colour in Nassi Shneid erman diagrams 72 have been developed New methods of interaction such as a browser style interface and interactive modification of the legend have been demonstrated e The new techniques integrate SV with visual programming and text based development environments 7 7 Future Directions There seems little doubt that the recent advances in software tools will continue Graphics visualization and interactive techniques look certain to play a role as will greater reusability and automation of software creation Programming languages are continually evolving to meet the 156 7 7 FUTURE DIRECTIONS needs of the day and the hardware on which they run As methods of programming change then so must the methods of visualization There are also many uncertainties for example what will be the next major programming paradigm shift the use of parallelism and formal methods or even whether textual programming can be replaced entirely by natural language visual immersive interactive or artificial intelli gence systems Maybe a combination of all of these technologies will play a role In software visualization the goal must still be to serve the user through power and usability More information should be presented in more ways But power and flexibility sh
27. Addison Wesley 1990 32 G Franck and C Ware Representing nodes and arcs in 3D networks In Proceedings 1994 IEEE Symposium on Visual Languages October 4 7 1994 St Louis Missouri pages 189 190 1994 33 M R Garey and D S Johnson Crossing number is NP complete SIAM Journal of Alge braic and Discrete Methods 4 3 312 316 1983 34 G Golovchinsky T Kamps and K Reichenberger Subverting structure Data driven diagram generation In Proceedings of IEEE Visualization 1995 pages 217 223 1995 35 J Gosling B Joy and G Steele The Java Language Specification Addison Wesley 1996 BIBLIOGRAPHY 159 36 C A M Grant Visual language editing using a grammar based visual structure editor Journal of Visual Languages and Computing 9 35 1 374 1998 37 T R G Green Cognitive dimensions of notations In Proc British Computer Society HCI 89 pages 443 460 1989 38 T R G Green and D Benyon The skull beneath the skin entity relationship models of information artifacts Int J Human Computer Studies 44 6 801 828 1996 39 T R G Green and M Petre Usability analysis of visual programming environments a cognitive dimensions framework Journal of Visual Languages and Computing 7 131 174 1996 40 C Hand A survey of 3D interaction techniques Computer Graphics Forum 16 5 269 281 December 1997 41 D Heller and P M Ferguson Motif Programming Manual O Reilly and Associ
28. An integer with value Int e null A null object object Id Class Fields An object with object id Id class name Class and fields Fields Fields is a list of members of the form field FieldName Value object Id An object with object id Id If changes are made to the source code or trace points have been added or removed then the program can be recompiled and executed to generate a new trace file The option Reload trace on the trace file s Action menu purges the run time data from the database and loads in the new trace file Because the trace file is stored off line it can be re examined at any time 5 8 Run time Data Visualization Clicking on a trace file navigates to it which reads it in and analyzes it if necessary The default view is shown in Figure 5 41 which gives event lines for each value traced An event line is a line representing time on which marks have been indicated where the event occurs Time is along the horizontal axis and different events are on the vertical axis Each event line is labeled with the expression or variable that was traced Every value in the trace file can be visualized by selecting the All values view of the trace file This view is shown in Figure 5 42 The values in a single event line can be displayed by clicking on an event line shown in Figure 5 43 with the time in the left column input expression in the middle column and the value in the right hand column This event line can
29. Class and interface declarations form the body of the Java file Each declaration Declaration in the file is stored as member File Declaration and java declaration Declaration ClassType Modifiers Extends Interfaces ClassHeader ClassType is the type of the class either a class or an interface Modifiers is a term representing the modifiers specified for the declaration of the form modifiers Abstract Final Access Static Transient Volatile Native Synchronized Extends is the name of the class this declaration extends Interfaces is a list of the names of the declarations this class implements ClassHeader is the text node of the class header Additionally text_node Declaration Text stores the text node of the entire declaration Each declaration s members Member are stored in the member Declaration Member pred icate The type of the member is indicated by the java variable Member and the java method Member predicates to determine whether the member is a field or a method Variables are stored as java variable Variable Type Modifiers where Variable is the iden tifier of the variable Type is a term denoting its type and Modifiers is a term describing the modifiers applied to the variable If a variable is a parameter to a method then this fact is stored java parameter Variable If a variable has an initializer then this value is stored Jjava initializer Variable Expression where Expression is a term denoting the e
30. DATA VISUALIZATION IIX File Edit Window Options Content Form Action EE kegen OOOO O O O O O O O U BinNode value 3 left BinNode Symbol Meaning value 1 A en rin 3 Nested integer value right BinNode null Null object ClassName value 2 left null Field1 Value1 Java run time object right null Field2 Value2 right null Examine value tree left Legend for Text a The initial view b The initial legend Vmax Examine value tree left EEX Vmax Examine value tree left eS File Edit Window Options Content Form Action File Edit Window Options Content Form Action 2 null null Examine value tree left Examine value tree left c d Vmax Examine value tree left MERX AMEN Ex Symbol Meaning Nested integer value Null object Java run time object Examine value tree left Legend for Text e The final view f The final legend Figure 5 50 Transforming a textual display of a value into a graphical display using the Legend CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 127 5 9 Prolog Visualization It is natural to ask what other languages Prolog and SVT are capable of analyzing and displaying The answer is all languages because Prolog is Turing powerful and Prolog is no more tailored to Java than to any other language However different programming languages require different structures and rules Vmax can display and edit Prolog as a visual language
31. Delphi is a trademark of Inprise Corp Genitor is a registered trademark of Genitor Corp IBM Visualization Data Explorer is a trademark of International Business Machines Inc Java is a trademark of Sun Microsystems Inc LabVIEW is a trademark of National Instruments Corp Linux is a registered trademark of Linus Torvalds Microsoft is a registered trademark of Microsoft Corp Motif is a trademark of Open Software Foundation Inc OpenGL is a registered trademark of Silicon Graphics Inc Open Inventor is a trademark of Silicon Graphics Inc Pentium is a trademark of Intel Corp Prograph is a trademark of Pictorius International UNIX is a trademark of X Open Company Ltd Visual Basic is a trademark of Microsoft Corp Visual FoxPro is a trademark of Microsoft Corp Visual Studio is a trademark of Microsoft Corp Windows is a trademark of Microsoft Corp X Window System is a trademark of Massachusetts Institute of Technology All other trademarks are the properties of their respective owners Software Visualization in Prolog Calum A McK Grant Abstract Software visualization SV uses computer graphics to communicate the structure and behaviour of complex software and algorithms One of the important issues in this field is how to specify SV because existing systems are very cumbersome to specify and implement which limits their effectiveness and hinders SV from being integrated into professional software developmen
32. K svt functionpress K current_view context calculator _ o Define the key bindings calculator key calculator key E calculator key c C calculator key x calculator key oe reaction functionpress Fun svt Fn functionpress Fun current _view_context calculator R F O M 1I Mem calculator input Fn R F 0 M 1 Mem R2 F2 02 M2 12 Mem2 set _view context calculator R2 F2 02 M2 12 Mem2 R2 R refresh object result result R2 restructure view reaction digitpress Fun svt Fn digitpress Fun current_view_ context calculator R F O M 1 Mem calculator input Fn R F 0O M 1 Mem R2 F2 02 M2 12 Mem2 set_view_ context calculator R2 F2 02 M2 12 Mem2 R2 R refresh object result result R2 restructure view o Make the calculator appear in the action menu action menu action _ Calculator svt calculator popup APPENDIX C FURTHER EXAMPLES 193 reaction calculator popup svt popup calculator o Create a calculator popup_calculator
33. Methods and variables used by a method E X Sea Vmax References within class com ice sqiclient CommandPanel File Edit Window Options Content Form Action Debug 0 ol void actionPerformed Action Event event JSplit Pane split Pane JTextArea resultText JText Area commandText void establishContentsQ JButton command Button JPanel commandPanel JPanel result Panel a a b JTextField databaseText CommandPanelO References within class com ice sqlclient CommandPanel private JSplitPane com ice sqlclient CommandPanel splitPane i Figure 5 28 Variable usage within a class CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 103 AE Vmax Definitions reachable from public JavaFileFilter JavaFileFilter String filters String description File Edit Window Options Content Form Action Debug fer String filters String description ee aii Mr O O public woid ee public void Sacer e ion String description private String filters private String description private Hashtable JavaFileFilter fitters private String extension private String JavaFileFitter fullDescription private String JavaFileFilter description private String description Definitions reachable from public JavaFileFilter JavaFileFilter String filters String description public JavaFileFilter JavaFileFilter String 1 filters String description Figure 5 29 Variables and methods reachable from a method 104 5 6 JAVA VISUALIZATION 5 6
34. StringBuffer intString buf insert insertionPaint buf a return buftoString Java method public String java math BigDecimal toString public String java math Biglecimal toString ti c Coloured Nassi Shneiderman Diagram A vmax Java method public String java math BigDecimal toString ES File Edit Window Options Content Form Action Debug Help public String java math BigDecimal toString dj d 3 D block structure Figure 5 31 Different representations of block structured code E Vmax Java method public String java math BigDecimal toString ax File Edit lsndow Options Consent Forn Action ebro Help Sting toString Meca 0 return intValtoSimngd stinger bul Sinng inisitrng itVal sbe toStiing int signum cir int n amp ertianport In Strng length scale rsertienPeint 0 return ignurneD 2 2 0 iniSinng elsa insertionpoint gt 0 1 buf new Simi BA asaran E RECE utreen 2 buf new StingFuffar insettionPoint intString length buf append signum lt t 0 far rt ES ieinsenionE oint i bal apre buf appendiniSieno y PROD Java method public String java math BigDecimal toString public String Javesnath BiaDectnal toStrinat El a The initial view EE Vmax Java method pubiic String java malh EigDecimal oS tin EE File Edit Udu Options Content Form Action Tebuo Help Sting teStrngg i iecae D
35. Text visual_component right text Id Text string Id Text alignment Id right Ye visual component calculator title Calculator border _line border calculator 5 table keypad 4 0 calculator calculator result keypad 1 visual calculator _button Id Str Pos Id str Str num Pos Calculator button Sym button Sym Str visual component calculator button Id Text Pos button Id Text alignment Id fill contains keypad Id order Id Pos Te object component calculator Id Result Keypad Id 192 C 6 A SCIENTIFIC CALCULATOR contain Result pl1 4 4 p2 4 4 p3 4 4 p4 4 4 contain Keypad p3 0 5 p4 0 5 b3 b4 l line p1 p2 p4 p3 pll 1 thickness 2 1 object component border _line Id Sub S Id contain Sub a S S b S S c S S d S S line a b d c al 1 Interaction with the calculator Define the mouse cursor changes action move calculator digit _ _ svt indicate action action move calculator function _ _ svt indicate action action click calculator digit D _ _ svt digitpress D action click calculator function F _ svt functionpress F action key _ K svt digitpress D digit K D current_view context calculator _ action key _ K svt functionpress F current view_context calculator _ calculator key K F action key _
36. Window System Addison Wesley 1996 51 H Koike The role of another spatial dimension in software visualization ACM Transac tions on Information Systems 11 3 266 286 1993 52 H Koike A fractal based method for controlling information display ACM Transactions on Information Systems 13 3 1995 53 H Koike and M Aida A bottom up approach for visualizing program behavior In Pro ceedings 1995 IEEE Symposium on Visual Languages September 5 9 1995 Darmstadt Germany pages 91 98 1995 54 H Koike and H Chu VRCS Integrating version control and module management using interactive three dimensional graphics In Proc 1997 IEEE Symposium on Visual Lan guages pages 170 175 1997 160 BIBLIOGRAPHY 55 H Koike and T Takada A framework and a system for visualizing parallel Linda pro grams Technical Report UEC IS 1997 8 University of Electro Communications 1997 56 H Koike and H Yoshihara Fractal approaches for visualizing huge hierarchies In Proc of the 1993 IEEE Symposium on Visual Languages pages 55 60 1993 57 E Kraemer and J T Stasko The visualization of parallel systems An overview Journal of Parallel and Distributed Computing 18 105 117 1993 58 G Kramer Solving Geometric Constraint Systems MIT Press 1992 59 Intelligent Systems Laboratory SICStus Prolog User s Manual Swedish Institute of Computer Science PO Box 1263 SE 164 26 Kista Sweden 1999 60 D Ladret and M Ru
37. Y Z A sample point on a graph with identifier Id at position X Y Z hier_edge Id A B An edge between two visual objects A and B that is laid out horizontally as a hierarchy 178 B 7 ACTIONS DEFINED BY SVT icon Id Type An icon Id with type Type insert Object Position Child Passes a child object Child to a visual object Object in position Position insertable_object Object Name Declares an insertable object Object with name Name lens_edge Id A B An edge Id between two visual objects A and B laid out as a fish eye list_edge Id A B An edge Id between two visual objects A and B laid out as a vertical list multi_container Object Creates a visual object Object which can contain any number of objects which are laid out in a 2 D palette using the contains 2 constraint order Object Order Assigns an order Order to a visual object Object which is used to present visual objects in a certain sequence scale Id Declares a scale Id in the view for comparing numerical values size Object Scale Value Sets the size of the visual object Scale according to the value Value in scale Scale table Id Cols Rows Creates a table Id with Cols columns and Rows rows If Cols is zero then the table is filled from top to bottom and the number of columns is variable If Rows is zero then the table is filled from left to right and the number of rows is variable Otherwise the table is filled left to right text_window Text
38. _ reaction zoom _ context ViewContext svt visual_context ViewContext VisualContext new_zoom_view ViewContext VisualContext Naturally this behaviour can be altered to display something completely different in the pre view window Another means of identifying the object beneath the mouse cursor is to look in the status bar When the mouse cursor is moved over a visual object X if it has a textual description Text provided by the predicate long identifier X Text then that text is shown in the status bar by calling the predicate set_status_text Text action move X svt display _text Text long_identifier X Text reaction display text Text svt set_status_text Text 4 11 Chapter Summary SVT is an implementation of the formal model of visualization described in Chapter 3 It is a general purpose visualization workbench that is specified using compiled Prolog It provides multiple windows for 3 D output and incorporates a 3 D model viewer and a window for editing text The executable is built around Prolog Motif and OpenGL and runs under UNIX Each view issues a query to Prolog to return the graphical constraints that compose the scene These constraints are cross referenced to create a scene graph Graphical layout algorithms tra verse the scene graph to assign 3 D coordinates to the objects in the scene These algorithms are held in virtual methods of the nodes of the scene graph and are invoked
39. _ graph function F I java method F long _identifier F I view content call graph calls A B java method A java cached_call A B Method calls within a class shown in Figure 5 26 view C cached_calls C Method calls within class cached java class C visual context cached_calls _ vs view content cached_calls C title Method calls within I long _identifier C I view content cached_calls C function F I java class method C F identifier F I view_content cached_calls C calls A B java class method C A java cached_call A B java class method C B References to a variable view V cached variable refs V References to variable cached java variable V visual context cached_variable refs _ vs view_content cached variable refs V title References to variable I long _identifier V 1 view content cached variable refs V function M I java cached reference M V long _identifier M 1 C 2 Visualization Relations The following example shows the visualization relations for block structured Prolog shown in Figure 5 53 o Query structured visual_context_description query Query visual_content tpm predicate Id H T query tpm predicate Id H T visual_content or_term Id A B query query Id or A B visual_content and_term Id A B query query Id and A B APPENDIX C FURTHER EXAMPLES 185
40. a general purpose logic based framework for visualization 23 24 Components of a scene in Alpha are pulled from the predicates by querying them The predicates specify screen coordinates directly The alternative push style of imperative languages is void draw Window _W for int RecID 0 RecID lt num RecID Rectangle _W RecID 10 RecID 20 10 15 a RecID RectangleColor _W RecID a RecID lt 50 DARKGREY LIGHTGREY Line W _L 5 10 10 num 20 10 The same functionality in Alpha is specified in 23 as Window Out _W Rectangle _W Out RecID Out X Out Y Out L Out H For RecID InRange RecID 0 num 1 Assign X 10 RecID 20 Y 10 L 15 H a RecID RectangleColor _W RecID Out DARKGREY If a RecID lt 50 Line _W Out _L Out 5 Out 10 Out 10 num 20 Out 10 As can be seen the Alpha program is actually longer and more obscure than the imperative language In the above example the draw function could be called automatically every time the data state changes Nevertheless as a graphics language Alpha is very expressive and provides abstraction between the data and the view Alpha is still low level because there is a low degree CHAPTER 2 PREVIOUS WORK 31 of abstraction between the output of Alpha and the graphics on the screen The Alpha script must be reprogrammed to change the output Eliot Eliot 61 extends Polka 96 to make it automatic to use Algorithms written
41. a good layout It is not possible to determine exactly which methods get called where For example some sections of code are semantically dead and will never get executed Itis generally not possible to determine which method gets executed in a method invocation because that method may be overridden by a subtype and only methods that are final can be guaranteed not to be overrid den Further program analysis could restrict which methods might be called but Vmax does not attempt this The Overriding methods view shows the methods that override a given method and the Overridden methods view shows all of the methods that have been overridden by a given method The References of method view in Figure 5 27 shows all of the methods and variables used by the method The References within class view in Figure 5 28 show which class methods access which member variables The Reachable references view of a method in Figure 5 29 combines the reachable methods with the method references to give a tree of all the variables that are reachable from the current method Clicking on any class member will navigate to it Navigating to a variable will go to the type of the variable if available or to a view showing all of the methods where the variable is referenced Because method calls and variable references can be time consuming to compute especially for a view that has to scan the entire database there is the option of precomputing cross referenc
42. a higher and more abstract level In declaring a view an implementor does not need to be concerned with the graphical representation and vice versa Also visual representations and view declarations can be reused rather than reimplemented If the mapping between the view content and the visual content is a bijection then a dynamic legend can be implemented It is also useful to specify interaction because this adds more flexibility to the system and can be used for many different things such as bookmarks menus changes to the mouse cursor data manipulation and navigation The user interface can be completely tailored by the data Declarative specification methods are arguably more suitable than imperative languages be cause visualization is data orientated and not sequential A knowledge engineering environment may also help to represent heterogeneous data 7 4 The Role of Prolog in Software Visualization It has been demonstrated that Prolog can play a central role in information visualization and SV systems and can specify every visual and interactive aspect of a SV system But that does not necessarily make it superior to other approaches Prolog has certain characteristics that make it particularly suitable for SV The other declara tive SV systems base their implementations on logic languages 20 86 It provides an environ ment for knowledge engineering that manages the data and analysis part of SV and it provides a query langu
43. a relation be tween the view context x and the set of graphical constraints G that represent the image EREA is the relation between the view context x and the set of visuals V that describes the image 5 is the relation between the view content C and the graphical constraints G Figure 3 2 illustrates how the relation ss derives the graphical constraints for the view context family_tree The resulting relation is shown in Figure 3 3 y person harry labeled_icon harry harry person text text harry harry 1 v_stack harry icon harry text harry 7 icon icon charles person e person charles labeled_icon charles charles person text text charles charles person elizabeth labeled_icon elizabeth elizabeth person lt N person diana labeled_icon diana diana person father charles harry line charles harry blue edge edge elizabeth charles elizabeth charles mother elizabeth charles line elizabeth charles green colour edge elizabeth charles green edge edge diana harry diana harry mother diana harry line diana harry green View View View Visualization Visual Graphical Context Relation Content Relation Content Constraints x 2 c v ale G Figure 3 2 A graphical representation of the relations shaded showing how graphical con straints are derived from the view context The graphical constraints Constraint for a view context ViewContext and visualization con te
44. analysis such as method calls variable uses and cross refer ences 2 Provide a high level specification method This should specify the following aspects of SV e View content what a view is showing This must provide a generic query mecha nism e Graphical abstraction how the contents of the view are rendered e Interaction how to interact with items in the view The specification methods of existing systems are very awkward to use and do not specify either view content or interaction 3 Provide dozens of different views of the program and run time data Previous systems provide only a small number of views because of the difficulties in setting up new views CHAPTER 1 INTRODUCTION 13 4 Base the implementation on a theoretical foundation Existing SV systems are not based on formal models 5 Provide a rich and varied graphical output This offers a more powerful and interesting means of expression 6 Provide flexibility in graphical representation Users should be able to interactively change the way data is represented 7 Provide a legend that describes the contents of the view This is necessary if the graphical representation is so flexible and useful for users unfamiliar with the system No other systems provide an automatic legend 8 Provide navigation by browsing to program objects in the scene Provide a preview win dow that graphically displays the object beneath the mouse cursor and graphical book
45. arguments in round brackets e g parent charles harry Prolog lists are written in square brackets e g 1 2 3 Prolog variables begin with an uppercase letter e g X Id and Object4 The variable 7 is unbound Prolog predicates or compound terms may written as Functor Arity e g parent Parent Child would be written parent 2 parent 2 is understood to be distinct from parent 3 Prolog modules are written as Module Predicate e g java method Method Often the module prefix is omitted 1 6 Dissertation Overview Chapter 2 describes the previous work in the field from SV systems and visual programming to the state of the art in commercial systems The taxonomies of SV 79 84 and visual program ming 71 provide important reference points for future work and have motivated much of this work Chapter 3 describes a formal model of information visualization and how visualization can be specified in Prolog This is the theory that underlies the rest of the work Chapter 4 describes how the model and specification techniques described in Chapter 3 can be implemented in practice A generic visualization tool Semantic Visualization Tool is developed that can be used to visualize any kind of data including the structured and relational data in computer software Chapter 5 describes Vmax a tool for software visualization that is the main product of this work Chapter 6 presents the results and evaluation of this work and discuss
46. be placed above the text Each graphical constraint has identifiers that reference other constraints for structuring the image Example 18 A coloured line visual could be declared visual_primitive line From To Colour Primitive Primitive edge edge From To From To Primitive colour edge From To Colour The line between two objects with identifiers From and To is given the identifier edge From To CHAPTER 3 A MODEL OF INFORMATION VISUALIZATION 49 Section 4 6 describes even more succinct specifications for the visuals in Examples 17 and 18 Section 4 6 1 describes how more complex visual objects may be specified using a set of geometric constraints The set of graphical constraints G is next solved out and rendered as an image I Definition 11 Image Let I be the output image I is a function of G I G I could be regarded as a set of coloured pixels a colour function over R 2 ora subset of R or R3 where R denotes the set of real numbers I G is implemented by a constraint solver and renderer that assembles the graphical con straints into an image The implementation could use any layout or rendering algorithms which will not be the subject of formal analysis The set Tg of graphical constraints is implementation defined The graphical constraints provided by SVT are given in Section B 6 3 7 Generating Views sond R A V G N The whole visualization process is the combined relation E which is
47. class hierarchy x File Edit Window Options Content Form Action Debug Help a ca Array Florales DSC Object FDBigint EZ Throwable Math gt FareadGroul Number E E DS Thread Ca Cy System String a Sea Lae java lang ClassLoader dj Figure 5 20 The whole class hierarchy CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 99 5 6 2 Visualizing Cross references The information from program analysis in Section 5 5 2 is used to construct a variety of views to show how program objects relate to each other The class references view in Figure 5 21 shows which declaration types are present in the given class The references between all classes is shown in Figure 5 22 This view can be restricted to a particular package The Declarations of class view in Figure 5 23 gives all the places in methods and classes where the given class has been declared The Places class thrown view in Figure 5 24 gives all the methods where the class is thrown as an exception The reachable methods of a method in Figure 5 25 show all of the methods that are reachable from a given method The Calls to method view shows all of the methods that may call the given method The Method calls within class view of a class shows which methods call other methods within a class in Figure 5 26 The complete call graph can be shown but this is less useful because it is so large and the graphical layout algorithm has difficulty finding
48. complete and fully functioning examples of the specification techniques For reasons of conciseness they were not be presented in full in the main body of this dissertation C 1 Views All views Shown in Figure 4 3 view views All views visual_context views views view_content views title Welcome to Vmax view_content views view navigate_to C D view_context C D D separator All Java interfaces shown in Figure 5 17 view java_interfaces Java interfaces visual_context java_interfaces vs view_content java_interfaces title Java Interfaces view_content java_interfaces interface I T java interface I long _identifier I T The Java class hierarchy shown in Figure 5 20 view java_classhier Java class hierarchy visual_context java_classhier vs visual_context java_classhier fisheye visual context java_classhier tree view_content java_classhier title Java class hierarchy view_content java_classhier class Class Name java class Class short_identifier Class Name view_content java_classhier contains Super Sub java class Sub 183 184 C 2 VISUALIZATION RELATIONS java direct supertype Sub Super Complete call graph view call_graph Complete call graph cached visual context call graph vs view content call _ graph title Complete call graph view content call
49. container top left top right bottom left and bottom right respectively and also give some positional constraints An anchor expression is a term that specifies an anchor and also constraints on that anchor It is a notation that is intended to give a direct and concise method of specifying the position of graphical components in a visual object Anchor expressions are converted into graphical constraints in the same way as other object components An anchor expression may be of the form e Anchor The identifier of an anchor This expression creates the graphical constraint anchor anchor Object Anchor CHAPTER 4 SEMANTIC VISUALIZATION TOOL 67 e Anchor Offset An offset from a named anchor This expression creates the graphical constraints an chor NewAnchor offset NewAnchor Anchor Offset Offset is a 2 D or 3 D vector of numbers Anchor2 Anchor Offset An offset between two named anchors creating the graphical constraints offset anchor Ob ject Anchor2 anchor Object Anchor 1 midpoint Anchorl Anchor2 A midpoint between two named anchors creating the graphical constraints anchor New Anchor midpoint NewAnchor anchor Object Anchor1 midpoint NewAnchor anchor Object Anchor2 e Anchor3 midpoint Anchorl Anchor2 A named midpoint between two named anchors creating the graphical constraints mid point anchor Object Anchor3 anchor Object Anchor1 midpoint anchor Object An chor3 anchor Obj
50. content of the view has changed and the view needs to be refreshed the option Refresh view will update the view The drawing speed options affect the quality of the rendering because it may be necessary to decrease the rendering quality to increase the rendering speed for smooth viewpoint control on slow hardware The Legend option creates a window to show the legend explaining the symbols present in the current visualization and is described fully in Section 4 8 The Zoom option creates a window that shows a graphical view of the object beneath the mouse cursor as described in Section 4 10 The Create new viewer option creates a new viewing window containing the initial view while Duplicate creates a new viewing window with the same content as the existing one SVT man ages multiple viewing windows Toggling the View window option hides or reveals the graphical view while toggling the Text window option hides or reveals the text window 4 5 Generating Visualizations Each view is generated by issuing a query to Prolog that returns the graphical constraints for a particular view context and visualization context These constraints are structured into a scene graph laid out and rendered CHAPTER 4 SEMANTIC VISUALIZATION TOOL 57 EET Vmax Java method public void com ice yallery Gallerylmage Panel old ContactDirectory NA SES Vmax Event values array 0X File Edit Window Options Content Form Action Debug Help File Edit Window Options
51. detailed information about program objects URL http www genitor com 2URL http www cygnus com 3URL http www westernwares com 24 2 4 SOFTWARE VISUALIZATION SYSTEMS Genitor Class Editor olx File View Edit Search Buffer Key Text Options Tools Window Help QD String Class QD CLogPage Class General Information D General Information Cul Include Files Description fe Member Data Items Source File e tmp ney Constructors d Owner Andy i Ent String Default Constructor Keywords sample test brochure packaging if String const _ThisClass amp rhs Copy C Sample Code Stringlunsigned short us Conversion Base Classes String short s Conversion Construct Al CPropertyPage Stringlunsigned long ul Conversion E Member Data Items Stringllong I Conversion Constructor Eng Constructors Stringlunsigned int ui Conversion Cc Destructors Stringlint i Conversion Constructor y CLogPage String char c int nCount 1 Construc Member Functions String const char psz UINT uiMaxLer CPub EVIND DoD ata xchange CD ataE xchange pDX String const char psz Constructor GPU GetFilename Const String TStdStringRep pRep Constr DninitDialoa DnCmdBrowse Retum Type void Function Body operator Jlint nPos INT uiLen Co Messages Description 5 ON_COMMAND ID_Browse Retum Type _ThisClass Operators Parameters ey EE Custom Code Function Body E Global Functio
52. ew od a BSG Sees ee Nw BS 137 6 1 4 Method estaca ra a bed ea Se ea ea MS amp bat 138 6 15 Interaction 2 65 2 a Ga ee Re a be 139 6 1 6 gt Effectiveness cocino eae is foe Fae o 139 6 2 Evaluating Vmax isre at ee es eh gk Re eek Ge Se ed ee Bes 139 621 Benefits e can ee Ave A de Shave oe Bate y 140 6 2 2 Performance 265555 55 6 dG sek a wee ERE e 140 6 2 3 Extensibility ii RA ep eas Be 141 63 Evaluating SVT erer E Ae we ee ea we Pace oe eek 141 6 3 1 The Specification Language 141 6 3 2 Modularity ste 22 066 ee eee RS OW A ee we be 142 6 3 3 Application Areas vera ee oe a A RA 142 6 4 Commercial Relevance o e 143 6 5 Further Work ses is e bt eee e a EA 145 6 5 1 Measure Usability of Vmax o o a 145 6 5 2 Measure Programmers Needs o e 145 6 5 3 Improve the Graphical Constraint Solver 146 65 4 Animation z RA A Re Ee ots Bee 146 6 5 5 Sound e i040 oh awe aa ioe bh Gade ee EAE eA 146 6 5 6 Automatic Choice of Graphical Layout 146 6 5 7 Improve the Graphical User Interface 146 6 5 8 Investigate other Specification Languages 146 6 5 9 Investigate Visual Specification Languages 147 6 5 10 Natural Language Interface o o 147 6 5 11 Formal Theory of Visualization o 147 6 5
53. f ti tn ET For example f 1 parent node 123 456 and harry are all terms Definition 18 Relations f A and B are sets then R is a relation between A and B if and only ifR C Ax B aRb gt a b R Definition 19 Composition of relations If Ry is a relation between A and B and Ra is a relation between B and C where A B and C are sets then R Ra is the relation between A and C defined by a c E RyRy gt Th B a b R and b c Ra Note that relation composition is associative because for relations R1 Ra and R3 Ri R2 R3 Ri R2R3 163 164 Appendix B Documentation These notes provide a brief summary of the predicates and functions available to developers who wish to create visualizations using Semantic Visualization Tool and to extend its capabilities Only the publicly available predicates have been listed B 1 Running SVT SVT is invoked svt options file argl arg2 argn where file contains the specification for the visualization system in fcompiled SICStus Prolog Optional arguments arg argn are passed to Prolog and may be read by the user arguments l predicate The optional options may include e a argument Argument argument is passed to the Prolog program e file Loads the fcompiled Prolog file Files are loaded in the order they appear on the command line e p path Specifies the SVT directory Otherwise this value is obtained from the environ
54. file File imports all the declarations from package Package e g java io would have package io java Id is the identifier of the import statement import_declaration File Id Declaration Dynamic Java file File imports the declaration Dec laration e g java io OutputStream would have declaration OutputStream io java Id is the identifier of the import statement initializer Variable Expression Dynamic The variable Variable is initialized by the expres sion Expression interface Interface Not extensible Interface is a Java interface declaration is_abstract Method Not extensible Method is an abstract method All method declarations in interfaces are considered to be abstract is_native Method Not extensible Method is a native method is_private Member Not extensible Class member method or variable Member is private has private access is_protected Member Not extensible Class member Member is protected is_public Member Not extensible Class member Member is public svt java_source_file File Not extensible File is a file containing Java source code This is determined solely by its file extension labelled_statement Statement Label Dynamic Statement is a label statement with label La bel make_qualified_name List Text Not extensible Returns text Text for a qualfied name List e g Object lang java returns java lang Object method Method Not ex
55. functor has the same name as the class of the object with a prepended V Thus the constraint string 1234 Peter instantiates a class Vstring with identifier 1234 and string Peter The functor s arguments are passed directly to the object s constructor for initialization The complete list of graphical constraints provided by SVT is given in Section B 6 This class hierarchy is designed to be completely extensible to incorporate any method of graphical output A C file can extend one of the classes of Figure 4 9 and declare a handler to instantiate the new graphical constraint for inclusion in the scene graph Once the graphical constraints have been instantiated they are structured into a graph The objects must be cross referenced which means creating C pointers between the nodes in the scene graph Each Vobject contains a term called its identifier each Vattribute contains a term identifying the object to which it applies and each Vassociation contains two terms identifying the objects that it associates Cross referencing is implemented using a hash table Each Vobject inserts its identifier and address into the hash table Each Vattribute looks up the address of its object from the hash table Each Vassociation looks up the addresses of the objects it associates from the hash table It follows that the cost of cross referencing is linear with the number of graphical constraints in the scene 4 5 3 Structuring The
56. in C are compiled in C to overload the C operators to invoke visualization events For example each type of program object such as an array or integer is created as an Eliot class and any attempt to make assignments to the data calls the overloaded operator which updates the display The user can choose the graphical representation of the run time data from a menu and display different outputs in different windows Eliot is much simpler to use than any of the other systems in this section because it is com pletely automatic and the user s intervention with the source code is minimal It also has much greater potential for intelligence and can be used in conjunction with the standard Polka calls Unfortunately Eliot still requires some modification of the source code and its output is not as flexible as animation systems that program their animation explicitly Eliot can only visualize integers reals characters arrays and binary trees Visualizing other structures would require special implementation Eliot can only show the current data state 2 4 5 Run time Visualization Systems Run time visualization displays what is happening in a program as it is running This can monitor code including program statements threads or object usage and data Data visualization differs from algorithm animation in its purpose and degree of abstraction Data visualization reveals real data in running systems for analysis while algorithm animation is m
57. in context Context expression_reference Expression Context Variable Not extensible The expression Expres sion references the variable Variable in context Context expression_statement Statement Expression Dynamic Statement is an expression statement with expression Expression Expression is a term storing the structure of the expression expression_type Expression Context Type Not extensible The expression Expression has type Type in context Context find_declaration Type Declaration Not extensible Finds the declaration class or interface Declaration that is referred to in a Java type Type An array of a class will return the class find_variable Name Context Variable Not extensible Returns the variable Variable that the name Name refers to when used in the context Context Context contains the local decla rations for_statement Statement Init IText Condition CText Incr IncText Body Dynamic State ment is a for statement with initializer expression Init and text node IText condition ex pression Condition and text node CText incrementation expression Incr and text node IncText and statement body Body APPENDIX B DOCUMENTATION 173 has_name Object QualifiedName Dynamic Variable declaration or method Object has the fully qualified name QualifiedName e g java lang Object would have qualified name Object lang java import_all_declaration File Id Package Dynamic Java
58. is gathered from a Java program by inserting trace points into the source code and recompiling it When the Java program executes these trace calls output the run time data to a trace file Vmax reads this file and adds the data to Prolog s knowledge database for visualization 5 7 1 Adding Trace Points Trace calls can be inserted into the source code manually simply by typing them or automat ically The Action menu of a statement obtained by clicking with the middle mouse button on the statement has the option Trace expression which pops up a dialog box prompting for the expression to trace at that statement This is shown in Figure 5 39 Open directory Go to statement Move back Move forwards Execute command Fj ee BC SS X Show log Empty log Enter expression to trace Add view DK Cancel Trace expression Recalculate cross references Calculator 2i a A statement s Action menu b The trace expression dialog box Figure 5 39 Tracing an expression by clicking on a statement The easiest way to trace a variable is to drag the variable to a statement which will trace the variable at that statement This has exactly the same effect as entering the name of the variable in the dialog box in Figure 5 39 b The fully qualified name of the variable is used as the trace expression to disambiguate it from other variables with the same name Figure 5 40 illustrates th
59. it should be laid out The On_reset_size method of the Vwindow object is called to initialize the entire scene On_set_size is called for each object to change its size particularly to shrink it to fit the view in the viewing window The On_set_size method of the Vwindow object is called to ensure that the whole view fits in the viewing window and composite objects call On_set_size for each of their child objects On_set_position is called for each object to position it correctly within the view Composite objects set the position of their children Finally On_draw renders the object on the screen using OpenGL Each object stores its own coordinates 4 5 5 Graphical Layout Algorithms The architecture allows any kind of layout algorithm to be applied to the scene graph Some of the layout algorithms used in SVT are described here None of these algorithms are particularly sophisticated but they show how different algorithms can be applied to the scene graph Graph Layout There are many types of graph layout algorithm 6 and the one chosen here is based upon the shortest path from a root vertex as first described by Sugiyama et al 98 mainly because of its speed and simplicity A root vertex is selected and preference is given to the vertex with the least number of edges leading towards it Then the undirected unweighted distance to each other vertex in the graph is computed using a breadth first search Each vertex is assign
60. j a A Vex Definition 5 Visualization relation The visualization relation is a relation between Tc and Ty The visualization relation maps members of C to members of V If L isa bijection then there is an unambiguous one to one correspondence between data and its graphical representa tion The visualization would then be complete every datum is mapped to something visible and unambiguous no two data are mapped to the same representation If a user can see all of the visuals V in the image i e they are not too small and indistinguishable then he she can invert 2 to recover its meaning C If the user knows the query Bae he she can deduce from his her knowledge of C which propositions must be true or false The approach works is because the structure of the image always matches the structure of the view content CHAPTER 3 A MODEL OF INFORMATION VISUALIZATION 45 Multiple visualization relations can be provided for the same view This gives flexibility in changing the graphical representation for a particular view This would be useful if for example different users had particular preferences for colours or the way things were presented if different graphical forms better suited a data set or if different visualization relations emphasized or de emphasized different parts of the view content Because different visualization relations can coexist each has a visualization context to identify it oye E p V Defini
61. natural language interface Many natural language systems use Prolog so it would be quite straightforward to embed a natural language interface into SVT Queries such as show the marital status of the salesmen who work in Ipswich who sold no speakers last month could be converted into a clause for the view content A command such as show salesmen as circles could change the visualization relation 148 6 5 FURTHER WORK 6 5 11 Formal Theory of Visualization The semantic model in Chapter 3 is a fully formalized theory of visualization based upon set theory By formulating the relations in logic the model is essentially an inverse of description logic 28 that is used to describe rather than generate the graphical properties of an image The constraints form the basis of a constraint multiset grammar 65 and the scene graph relates to graph grammars 81 The theoretical work in Chapter 3 could be reconciled with these other visual language formalisms 6 5 12 Improve Program Analysis There is much information about the source code that has not been analyzed For example data flow in methods or which methods get called in which circumstances abstract interpretations of the data or which values can be stored in which variables Adding these analyses to Vmax would enhance the information presented to the programmer 6 5 13 Visualize other Languages such as C Vmax focuses on Java but other programming languages could be visu
62. node_structure TextNode Type Children Not extensible Returns the type Type for a text node TextNode and its list of child text nodes Children object_size Object Size Not extensible Returns the size in bytes Size of an object Object s text node parse_buffer Buffer TextNode Not extensible Applies the buffer Buffer s lexical analyser and parser and returns the text node TextNode that is the root of the parse tree parse_tree Buffer Tree Not extensible Applies the buffer Buffer s associated lexical analyser and parser and returns a term Tree describing the entire parse tree The term Tree is of the form node Type TextNode C1 Ca Cn where Type is an integer TextNode is the address of its text node and C1 Cn are subterms of the same format reload_buffer Buffer TextNode Not extensible Re reads the contents of the text buffer Buffer from the external source typically the filing system and returns the new text TextNode save_buffers Not extensible Ensures that all of the text buffers that have changed are written to disk Use before running programs that could examine these files text_node Object TextNode Dynamic Assigns the text node TextNode to an entity Object token TextNode Type Not extensible Returns the type Type of a text node TextNode B 2 6 Miscellaneous Predicates colour Colour R G B Extensible Defines colour names Colour for use in object compo nents and visuals
63. notifyAll void wait long timeout void wait long timeout int nanos void wait ol oll oh void finalize void main String argv void instanceMain String args void printUsage a a a String PROP_PREFIX String PROP_PACKAGE String LOCAL_PROPERTIES a a a a a String DEFAULTS_RSRC String VERSION_STR Gallery app int slideDelay boolean slideShow Inherited members in com ice gallery Gallery public boolean java lang Dbject equals Object obj A Figure 5 12 A class view including all inherited members and methods EE Vmax Public members in com ice gallery Gallery El File Edit Window Options Content Form Action Debug Wo i oi Class getClass int hashCode boolean equals Object obj String toString void notify Db T T void notifyAllQ void wait long timeout void wait long timeout int nanos void wait i s 0 a void main String argv void instanceMain String args void printUsageQ String VERSION_STR a a a Gallery app int slideDelay boolean slideShow Public members in com ice gallery Gallery public boolean java lang Dbject equals Object obj A Figure 5 13 A class view of all public inherited members and methods CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 95 ES Vmax Supertypes of com ruleofeight objserv ServerConnection Mus File Edit Window Options Content Form Action Debug java lang Object java lang Thread com ruleofeight sf Connection com ruleofeight objsery Ser
64. number are commercially viable VPLs that have had particular commercial importance include LabVIEW 104 ProGraph 19 77 VisualAge from IBM Visual AppBuilder from Novell AVS from Advanced Visual Systems and AppWare from Novell Visual languages are more successful as end user languages for specific applications than as general purpose pro gramming tools li Y 2 2 Quicksort EAEE Figure 2 6 The quicksort algorithm written in ProGraph ProGraph originally described by Pietryowski 77 is one of the most successful and widely used VPLs Figure 2 6 shows an example of a ProGraph program It is plausible to suggest that visual languages might one day supersede the textual program ming paradigm At present practical issues such as scalability 17 and tools 36 65 71 need to be solved to make visual programming a viable alternative to textual programming 2 4 3 Code Visualization Systems Code visualization provides graphical information about the source code of a program It in cludes high level overviews program structure and source code itself It is different to visual programming in that code visualization systems do not necessarily modify graphical representa tions of code SEE SEE 4 is a source code pretty printer that goes much further than syntax colouring Syntax colouring is often used in text editors which identify the lexemes present in the source text an
65. represent lexemes This parse tree is analyzed by pattern matching the parse tree with the production rules present in the Java grammar For example this code excerpt program file node 118 P 1 T File package_statement P File Package import _statements I File type declarations T locale Package File 1 1 matches one possible structure of the root node There is a clause for each production rule in the Java grammar The predicates that traverse the parse tree also store information about the Java file in Prolog s database this part of of Prolog s database could be called the program database The predicate java program _file Tree File adds the Java file File to the program database given its parse tree Tree The file s buffer Buffer is stored java buffer File Buffer Note that the predicates relating to Java have been put in the separate module java The first part of a Java file is its package declaration and its import declarations The predicate java package File Package stored the file s package Package as a list of names The predicate java import_declaration File Id Import stores that the file imports the package Import so for example java util Hashtable would be represented Hashtable util java Id is the unique identifier of the import statement The predicate java import all _declaration File Id Import stores that the file imports all classes from the package Import
66. represent the state of the art commercially and offer rapid application development compared with text only environments These tools form a basis for more advanced SV and the visual representations navigation techniques and human computer interaction issues in these systems are directly relevant to SV Better SV techniques are likely to be incorporated into development systems as the sophistication of these tools grows A survey in Personal Computer World 76 reviews the leading visual development tools for the desktop market where the leading platform is Microsoft Windows 95 NT The products un der review are Delphi 2 0 Optima 1 5 Power Objects PowerBuilder 5 0 VisualAge Basic Visual Basic 4 0 Visual C 4 2 Visual FoxPro 5 0 Java Workshop Visual J Visual Inter Dev Visual Basic 5 0 C Builder and Visual Caf Most offer database connectivity object orientation application builders visual design visual navigation and embedded help systems The programming languages Java C Pascal FoxPro and Basic are used Visual aspects of these systems include e A graphical user interface e A windowed environment e Dialog boxes tool bars and pull down menus e What you see is what you get WYSIWYG form designers e WYSIWYG graphical user interface design menus dialog boxes and graphical resources e Visual browsing of program resources including graphical resources and program li braries e Code navigation us
67. the view content for a particular view context Relations are used instead of functions because they are more general can be one to many are bi directional and can be expressed in first order logic The relation 2 could be regarded as the query that determines the data in the view content All Prolog predicates can be regarded as relations between terms The Prolog predicate is used as an indicator function which finds all members of the relation that match its arguments This means that Prolog can be used to define the view relation and a binary predicate view content ViewContext ViewContent specifies the binary relation between the view context and the view content The first pa rameter specifies the view context and the second view content for the view context This is a one to many relation Clauses are added to the view content 2 predicate to add data to the view content Because Prolog is Turing powerful and because terms can represent arbitrary data any com putable mapping from data source to the view content can be implemented This method of specifying view content is therefore completely generic The view contexts Ty and view content Tc are implementation defined Example 6 The family tree could be declared by the view relation view content family tree mother X Y parent_of X Y woman X view content family tree father X Y parent_of X Y man X view content family tree pers
68. two red circles From seeing the view it is deduced that the visual content V must be circle A red circle B red line A B line colour A B green for some A and B to account for the image on the screen Inverting the visualization relation v the view content must be person A person B married A B children A B From the rules and the view relation it can be deduced that for some A B and C alive A person A man A woman A alive B person B man B woman B married A B child A C child B C 3 5 A Visual Type System Restrictions can be placed on 2 to ensure that the visualization relation is valid Nonsensical mappings could be a relationship being mapped to a size attribute or a string mapped to a colour The resulting image would be invalid due to these errors called visualization errors To help prevent such errors a type system can ensure that members of the view content are only represented by suitable visuals The visualization relation should only map content to visu als of the same type as the content The type system is optional but it has a secondary function of allowing users to modify the legend of a diagram by selecting type compatible visuals as described in Section 4 8 2 By itself the type system cannot guarantee the integrity of the output because the structure of the data may be unsuitable for the specified visualization method For example a relationship in the view conten
69. 07 poses the question Given the range of information required in programming can a VPL highlight enough of the important information to be of practical benefit The only way the trade off between visual complexity and information completeness can be resolved is through elision control to selectively choose what to display by providing many different views The user interface uses graphical browsing techniques which have been shown to be a very easy successful and effective means of information retrieval and utilizes direct manipulation which is argued to be a natural interface to data 90 Usability testing is beyond the scope of this work There are no objective metrics for compar ing the effectiveness of different SV systems 79 and there has been little empirical evaluation of the other systems described in this dissertation So while there is indirect evidence that the facilities offered by Vmax are useful and that it is easy to use this has not been confirmed by user studies 6 2 2 Performance The performance figures shown in Table 6 2 were obtained on a 400MHz Pentium II PC with 128MB of RAM running Linux 2 2 They were measured by enabling a logging function in SVT that times these tasks Code input rate 6 320 lines per second Cross referencing 5 144 lines per second View generation 900 constraints per second Trace file input rate 19 503 lines per second Table 6 2 The performance of Vmax The figure given for
70. 100 user 16214 May 28 00310 READM r r r 1 camgi00 user 4674 Oct 18 1994 Keys r r r 1 camgl00 user 11481 May 28 00310 bookm libnullplugin dynMotif so LICENSE movernail src movemail nethelp cern E nO yet TA be 5 ET ne drwxr xr x 2 camgl00 user 1024 Aug 22 13 24 lib r xr xr x 1 camgl00 user 38275 May 28 00 10 libjs a a 3 r xr xr x 1 camg100 user 14177 May 28 00 10 libnu E r xr xr x 1 camgl00 user 5360 May 28 00310 movem netscape dynMotif netscape Netscape ad plugins README registry E z ETE eee a bo a a pees r xr xr x 1 camgl00 user 12109808 May 28 00310 netsc ss A r xr xr x 1 camgl00 user 9199900 May 28 00 10 netsc EM 2 drwxr xr x 2 camgl00 user 1024 Aug 22 13 24 plugi Il rw r r 1 camgl00 user 2569 Aug 22 13 24 regis spell vied NceysymDS drwxr xr x 2 camgl00 user 1024 May 28 00 10 spell rwxr xr x 1 camgl00 user 22812 May 28 20301 vreg Directory homes camg100 local netscape 7 i Figure 5 3 A view showing the contents of a directory The information in this view includes the directory name the files their names their sizes their ages and the type of the file If the text window is open the directory is listed there This is a lot of information to present so the visualization relation can select what to display and present it in different ways The view is declared view Dir directory view Dir Directory contents directory Dir view_content directory view Dir title Directory Name
71. 12 Improve Program Analysis o o 148 6 5 13 Visualize other Languages such as C o 148 6 5 14 Data Persistence e 148 6 5 15 Provide Visual Programming 148 6 5 16 Interface to Compilers o o e 148 6 5 17 Interface to the Java Virtual Machine Debug Interface 148 CONTENTS 6 5 18 Improve Efficiency e 0 0 0 0 00000 148 6 6 Chapter Summa 2 0 0 eee eee ee ee ee eee 149 Conclusions 151 DA Vmak so do oe A a E OE OR oe A oe Re eee a 151 711A Strengths of Vmax ssoi es ee ea ee a 151 7 1 2 Weaknesses of Vmax 2 2 0 00 e 152 7 2 Implementing SV Systems 0 0 02 0 00000 152 13 Specifying SV Systems voi eh ee ee eee ee ae ae 153 7 4 The Role of Prolog in Software Visualization 0 0 153 7 4 1 Advantages of using Prolog as a Specification Language 154 7 4 2 Disadvantages of Prolog o e 154 7 5 Information Visualization in SVT o 155 7 5 1 Formalizing Information Visualization 155 TO Contributions taria a a de a 155 TI Future Directions sera ba a A Ba Bae A 155 Definitions 163 Documentation 165 B 1 Runing SVD cds a A a A A ae 165 B 2 SVT Predicate Summary 165 B 2 1 View Predicates 0200000000004 166 B 2 2 Visualization Predicates o
72. 166 B 2 3 Legend Predicates ia cb le dt Geka ee ae 167 B 2 4 Interaction Predicates o e 168 B 2 5 TextPredicates 20 00 a ss ds esa 168 B 2 6 Miscellaneous Predicates o o e 169 B 3 Vmax PredicateSummary iiss c os 0000000022 eee 170 B 3 1 Filing Systems se eb ae EEN ee ee eee oy 170 B 3 2 Java Source Code ai ge eke Po a ee Ae ee ee E 171 B 3 3 Java RUM 0M6 2 Pak he oe She ee Beek ha ad Bea 175 B 3 4 Prolog Source Code 0 00000000 176 B 4 Visuals Defined by SVT o segara gorto Sk oe ee 176 B 5 Visual Object Components Defined by SVT o o ooo oo 176 B 6 Graphical Constraints Defined by SVT ooo 177 B Actions Defined by SVE mear te lA A ta RA 178 B 8 Views Defined by Vmax e 178 Bo The C Interface e AA A A Bet 179 B 9 1 Adding Graphical Constraints o 179 B 9 2 The Graphical User Interface o o o 181 Further Examples 183 CI VIEWS mr Ais A A eS A ee at e A Sree Ss 183 C 2 Visualization Relations o o e 0 000000004 184 ES Visia es e padre cBe dete EE a ns ls ole Fs 185 CA Visual Objects 52 0 eae e poe a EA e tie wpe 186 CONTENTS C 5 Interaction EL y MEUS Y edhe ce de e Roe ben Gh deck de ee SA te WA Ho ne ee G C 5 2 An Implementation of CCS o oo C 6 AScientific Calculator e 10 CONTENTS Chapte
73. 4 EIDE id Figure 6 2 A fully functional scientific calculator specified in SVT H A oO x eaj eje OA aj sj EA Fase a Tcl Tk b Prolog Figure 6 3 Similar calculators CHAPTER 6 DISCUSSION 145 6 4 Commercial Relevance Programming is difficult and time consuming and therefore expensive Tools which make pro gramming easier more reliable and faster reduce the cost of writing and maintaining programs and are therefore important Visualization is merely an extension of the existing trend of provid ing graphical environments in which to program which provide graphical support for using the compiler editing navigating modifying code creating skeleton applications and creating user interfaces Visualization can reduce the cognitive burden on the programmer thereby improving productivity These benefits are already being exploited in commercial programming products The approach used by SVT and Vmax could be used in commercial products A stand alone programming environment like Vmax could provide a working alternative to existing environ ments and provide a much greater range of visualization Otherwise an environment such as Mi crosoft Visual Studio could have an embedded Prolog interpreter linked to its project databases Stand alone visualization tools could also be based upon or augmented by SVT s approach to visualization Existing architectures such as IBM s Visualization DataExp
74. 5 55 The menu to change the functor of a term a Actions on terms b Actions on clauses Figure 5 56 Menus of operations on terms CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 131 terms to the window The dynamic clause scratch term Id indicates those terms that belong in the scratch view and these terms are not executable Vmax Edit clause is an ancestor of EEX Vmax Edit clause is an ancestor of D Xx File Edit Window Options Content Form Action Debug File Edit Window Options Content Form Action Debug is an ancestor of is an ancestor of MM za isan ancestor of Edit clause is an ancestor of Ready 3 Reeds a b a Vinax Edit clause is an ancestor of EAS File Edit Window Options Content Form Action Debug Help la Edit clause is an ancestor of c Figure 5 57 Direct manipulation of clauses Since Prolog can be edited graphically it should be possible to specify SVT entirely through this visual language The proviso is that these dynamic changes to Prolog are not persistent because they do not write back to the source code A better implementation would load and save textual representations of Prolog The Add new view option on the action menu prompts the user for the name of the new view creates skeleton clauses for the new view and navigates to Edit view relation for the new view This is shown in Figure 5 58 5 10 The Help Sys
75. 544164 a SeeSoft b SeeSys Figure 2 7 Line oriented software statistics The display is interactive allowing users to drag small windows over each source file to dis play actual source code and change the colour scale This visualization technique is sometimes referred to as a mural 49 where a long display is compressed in one dimension so that it may be contained within the screen and a superimposed slider then magnifies the region of interest Murals obey Shneiderman s visual information seeking mantra overview first zoom and filter then details on demand 91 Baker and Eick 5 attempt to solve the problem of screen size and scalability by develop ing a system called SeeSys to visualize software systems using line oriented software statistics Instead of using fixed width bars with lengths proportional to the number of lines of code the screen is divided into rectangular regions as shown in Figure 2 7 b with rectangle areas propor tional to the lines of code The system can visualize files directories and subsystems and there is a zoom facility to zoom in on any subsystem Additional information is presented to the user as the mouse moves over the visualization The output of SeeSoft and SeeSys is inflexible so that the output data and format it fixed There is also only one type of information available the modification history of each line of source code and the size and modification history of each source f
76. A TOOL FOR SOFTWARE VISUALIZATION 123 A MESHES CECA o E OX 1 max Examine value tree left 2 xX File Edit Window Options Content Form File Edit Window Options Content Form BinNode value 3 left BinNode value 1 iet mal value left nigh ight BinNod ng io value left right value 2 a null left null null null right null right null Examine value tree left Examine value tree left a Text b Horizontal SS Vmax Examine value tree left I E x Vmax Examine value tree left oo E x File Edit Window Options Content Form File Edit Window Options Content Form value MMMM 2 left null right null Examine value tree left Examine value tree left c Vertical d Plain graph 124 5 8 RUN TIME DATA VISUALIZATION Soe vmax Examine value variable lt x File Edit Window Options Content Form Action Ss Vmax Examine value tree left MA X File Edit Window Options Content Form Examine value tree left Examine value variable Ready mi Ready 3 e Graph with coloured edges A doubly linked list TR File Edit Window Options Content Form Action Debug Help g A ring Figure 5 48 Different visualizations of structures CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 125 a Depth 1 b Depth 3 c Depth 5 d The whole structure Figure 5 49 Visualizing a structure to a given depth 126 5 8 RUN TIME
77. Content Form Action Debug Help SS AS Color bg this gstBackgrounde ageName tringlthis files elementAt fidx agePath __ this dir getPath Fils separator imageName this feedback setStatusMessage Adding imagePath j img this getToolkit getimage imagePath pre geting y me Trace tracker new MediaTracker this tracker addimage img 0 ita getGraphics i A this removeAll this add this imageScroller BorderLayout CENTER this invalidate this validate ICETracer traceblithStack NULL Image buffer gt a Pan and zoom of Figure 4 5 b Rotation Figure 4 6 Examples of viewpoint manipulation Reset viewpoint Refresh view Fastest rendering Fast rendering F Medium rendering Good rendering Best rendering New viewer Ctrl N Bookmarks Duplicate Ctrl D Preview Legend Viewer a Text Ctrl T Figure 4 7 The Window menu 4 5 1 Acquiring the Visual Primitives Each view stores a record of its current view context visualization context action context and reaction context To generate a view the Prolog query get_constraint ViewContext VisualizationContext Constraint is issued to return the graphical constraints Constraint for its current view content View Context and visualization context VisualizationContext All queries are issued through the C interface and the textual representation is shown only for illustration The g
78. M R M M R R calculator memory M R M M R R calculator memory M MR M M calculator input digitpress D R2 F2 02 M2 12 Mem2 calculator R2 F2 02 M2 12 Mem2 calculator input functionpress 0 F 0 0 1 true Mem calculator input digitpress D R2 F 0 M true Mem M 1 M 1 R2 calculator input digitpress D R2 F 0 M2 true Mem M2 is M 0 1 calculator input functionpress R F 0 M2 true Mem M 1 M 1 M2 is calculator input functionpress R2 F 0 M2 1 Mem R2 is R M2 is M calculator input functionpress R2 Fn R2 M false Mem calculator evaluate 1 1 R2 is R3 calculator input functionpress R2 F 0 M false Mem calculator evaluate R Fn R3 calculator input functionpress R2 F 0 M false M2 input digitpress D M 0 _ F O false Meml 0 F O 1 true Mem 1 2 _ F O false Mem R F O M true Mem is R 10 M D R F O M true Mem R2 is R D M R F O M true Men Ls R F O M 1I Mem Fn R F 0O M _ Mem Fn _ calculator evaluate O R F R3 Fn R F 0O M _ Mem R2 is R3 Fn R F O M _ Men calculator memory Mem R Fn M3 R3 M2 is M3 R2 is R3 digit Atom Digit name Atom N name 0 Zero Digit is N Zero Digit lt 10 Digit gt 0 factorial M 1 M lt 1 factorial A B A gt 1 N is A 1 factorial N X B is A X Define
79. MAX NODES Graph Out 1 Node Out N 1 For InRange N 0 gNumNodes 1 Arc X Y 1 If g X Y 1 Alpha declarations are simply logical predicates It is similar to Prolog but has a much more awkward syntax and has to define parameters as either inputs or outputs and lacks unification and compound terms Variables can only be integer types Expressions from the target language can be embedded into Alpha directly Whenever an Alpha predicate is called expressions in the 30 2 4 SOFTWARE VISUALIZATION SYSTEMS File Edit Search Run Windows Programs Help 2246 97 Kamada Kawai Graph 1 Kamada Kawai View 1 ADU AWN O Kamada Kawai exe E Py kmi dy Imi Iteration 1 E_Pxx kmi 1 Imi Iteration 2 E_Pxy kmi Imi dx Iteration 3 EPyy kmi 1 Imi nodearrl nodearr Lindex Px E_Px nodearr Lindex Py E Py nodearr Lindex Pxx ElPxx B nodearrlindex Pxy EPxy gt nodearrl index1 Pyy E Pyy mr ff end of calDeltaE Program Trace dan EE IS Th Figure 2 12 The Leonardo animation system target language are evaluated according the the current state of the the virtual machine Thus the view is always synchronized with the virtual machine Graphical components in the scene are declared directly as predicates so for example the Arc 3 predicate declares the presence of arcs in the scene This approach has been proposed as
80. N TOOL 65 object component Visual Identifier Component Figure 4 15 shows the anchors and graphical components of a box with a shadow a a 5 3 _ _ _ b 5 3 b 0 4 i b 4 4 Child p2 c 5 3 _ d 5 3 Cc c 4 0 d pi 4 4 Figure 4 15 The anchors and graphical components of a box with a shadow The complete specification for this visual object is object component shadow_box Box Child Box contain Child a 5 3 b 5 3 c 5 3 d 5 3 l line a b d c a pl polygon c 4 0 d d 4 4 c 4 4 p2 polygon b 0 4 b 4 4 d 4 4 dl 1 The name of the visual is shadow_box Box Child where Box is the identifier of the visual object and Child is the identifier of the object that is inserted into it The second argument to object_component 3 specifies the identifier of the visual object in this case Box The third argument specifies the components of the object Object components such as I line a b d c a are translated into graphical constraints by the component primitive 3 predicate that defines how object components map to graphical con straints component_primitive 3 is called from visual_primitive Visual Primitive object component Visual Id Component component primitive Component Id Primitive The complete list of object components provided by SVT is given in Section B 5 The visual shadow_box 2 can be used like any other vis
81. Name is a list of atoms Literals are represented by literal String where String is the characters in the literal Method calls are represented method call Method Parameters where Method is an expression specifying the method and Parameters is a possibly empty list of expressions A member access is represented field_access Expr Field where Field is the name of the member A class instantiation is represented new class Type Parameters where Type is the class to be instantiated and Parameters is the list of expressions An array instantiation is represented new_array Type DimExprs Dims where Type is the type for instantiation with a list of dimension expressions DimExprs of Dims dimensions Types in expressions or variables are also represented by terms A primitive type is rep resented primitive P where P is one of the primitive types such as boolean or double A named type is of the form named Name File where Name is a list of identifiers for exam ple java util Hashtable would be Hashtable util java and File is the file that the type appears in The file is important because different names may refer to different classes depend ing upon the file they are in An array of a type is of the form array Type where Type is a type term Array types can be nested The Java source code can be rescanned using the java rescan file File predicate to rescan the file File This traverses the program database and cal
82. Node Places the contents of the text node TextNode in the editing area of the viewer title Title Sets the title of the view to be the string Title v_stack Id A B Creates a new visual object Id that puts visual object A above visual object B B 7 Actions Defined by SVT click Object Buttons Times The mouse has been clicked on object Object with the mouse buttons Buttons Times times Buttons is the term Left Middle Right Shift where Left is the left mouse button Middle is the middle mouse button Right is the right mouse button and Shift is the shift key which are either on or off drag From To Buttons A mouse drag from object From to object To with the mouse buttons Buttons key Object Key A key Key has been pressed with the mouse cursor over object Object linger Object The mouse cursor is resting over an object Object menu MenuContext MenuText An item from a menu MenuContext with menu text MenuText has been selected move Object Buttons The mouse cursor has moved over an object Object with mouse buttons Buttons B 8 Views Defined by Vmax This information is displayed by selecting List views from the Debug menu APPENDIX B DOCUMENTATION 179 B 9 The C Interface B 9 1 Adding Graphical Constraints The base classes Vnode Vobject Vattribute Vcompound and Vassociation are declared in the file svtvisual h listed below An instance of the class SVT_declare_visual should examine the Prolog ter
83. Prolog files can be loaded by the Load Prolog option on the File pulldown menu shown in Figure 4 4 The Close option closes the current window and the Quit option exits the program 4 4 View Manipulation One of the main issues in visualization is to present exactly what the user wants but it is often the case that too much information is presented to the user called information overload 7 Information filtering is necessary to remove the unwanted information while leaving the required information While the view relations filter data semantically viewpoint control can filter data spatially and such spatial filtering has common applications in information murals 49 lensing techniques 87 and rubber sheets 88 The screen has a finite size and hence has a theoretical upper limit on the amount of infor mation it can convey In the cases where the data is simply too large to be fitted in a window the feature size of the output is automatically reduced such that the whole view still fits in the window An example is shown in Figure 4 5 This approach provides an initial overview of the whole data even though detailed information may be lost and this approach follows Shnei derman s visual information seeking mantra overview first zoom and filter then details on demand 91 A user may have an area of interest or in the 3 D case a volume of interest There are many different techniques for scrolling and zooming to select such
84. Prolog to return the graphical constraints that compose a scene The graphical constraints returned from the query form the nodes of a scene graph A scene graph is a data structure of nodes and links that stores the graphical structure of the scene 106 Much of SVT is supporting architecture that provides a graphical user interface between the user and the visualization In addition SVT provides a number of algorithms for graphical layout that assign 3 D coordinates to the objects in the scene graph When the graphical components have been assigned coordinates the scene graph is rendered 53 54 4 2 ARCHITECTURE 4 2 Architecture SVT is implemented in C and integrates OpenGL 50 to provide 3 D graphical output SICS tus Prolog 3 7 59 and Motif 1 2 41 for the graphical user interface Each of these components provide C interfaces for compilation into a single executable file The architecture of SVT is shown in Figure 4 1 User Application Figure 4 1 The architecture of SVT The structure of SVT in Figure 4 2 implements the semantic model shown in Figure 3 1 The colours match Figure 4 1 Data Input ry a E gs Interaction User Figure 4 2 The structure of SVT 4 3 Running SVT SVT is an executable file svt that runs the Prolog file specified on its command line Additional command line arguments are passed to the Prolog program For example running svt myvis argl arg2 from a UNIX sh
85. Scene Graph The cross referenced nodes form a scene graph A Vwindow object is created at the root node of the scene graph and every object in this graph except the root node will have pointer to its parent Any objects that do not have parents are automatically added to the Vwindow to ensure that the scene graph is connected Before each node is cross referenced the node is initialized by calling its On initialize method After each node has been cross referenced further initialization is done by calling the CHAPTER 4 SEMANTIC VISUALIZATION TOOL 59 Valignment Vcolour _Vcompound_ Vanchor Vline_colour Vbitmap Vbox Vfill_colour Vgraph_datum Vanchor_point Vgraph Vdialog_frame Vgraph_point Vaxis_align Vellipse Vinsert Vscale Vx_align object_containen Vorder Vstring Vy_align Vpolygon Vthickness Vtitle Vz_align Vpolyline Vtransparency Vunknown Vbeyond Vmu
86. Section 4 5 new_view_context l or new _view 2 are typically called from reaction 2 as a response to a user input 72 4 7 INTERACTION SVT provides an action context svt that navigates to an object by clicking on it just as in World Wide Web browsing A click on the background of the view navigates backwards which is can also be done from the Action menu The mouse cursor changes to indicate the different types of navigation Each object declares which view contexts are applicable to it using the view 3 predicate view Object ViewContext Description declares that the view context ViewContext can be used to visualize object Object and De scription is a textual description of the view For example view Directory file view Directory Files in directory directory Directory view Directory directory tree Directory Sub directory tree directory Directory view Class inherited members Class Class members inherited in Name java class Class identifier Class Name view all_picture files Picture files declares two view contexts for a directory one for a Java class and one that is not associated with an object When there are multiple views of an object the first declaration is used forming the default view context for that object Where possible the current view context is reused so that the target view is of the same type as the original Navigation is implemented by acti
87. Software Visualization in Prolog Calum A McK Grant Queens College Cambridge Dissertation submitted for the degree of Doctor of Philosphy December 1999 Copyright Calum Grant 1999 Vmax and SVT are copyright Calum Grant 1999 All rights reserved Permission is hereby granted to distribute this document without modifica tion in any format or electronic storage system Declaration This dissertation is the result of my own work and includes nothing which is the outcome of work done in collaboration I hereby declare that my dissertation thesis entitled Software Visualization in Prolog is not substantially the same as any that I have submitted for a degree or diploma or other qualification at any other University I further state that no part of my dissertation thesis has already been or is being concurrently submitted for any such degree diploma or other qualification This dissertation is approximately 44 000 words long excluding appendices Acknowledgments Thanks to everyone who has helped me complete this work and in particular Peter Robinson my supervisor the EPSRC for funding this work the proof readers and all members of the Rainbow Group for their help and ideas Thanks also to the departmental secretaries and to my parents Trademarks AVS Express is a registered trademark of Advanced Visual Systems Inc Borland is a registered trademark of Inprise Corp C Builder is a trademark of Inprise Corp
88. This demonstrates SVT s potential for visual programming and offers a possible solution to end user specification of visualization Prolog clauses are stored user _clause Clauseld Clause where Clauseld is a unique identifier for the clause and Clause is the structure of the clause There is no external textual representation of Prolog at present The clauses that the user has defined can be viewed by the Clauses or Predicates view shown in Figure 5 51 All clauses can be visualized in the All Clauses view shown in Figure 5 52 Vmax List of Predicates ES File Edit Window Options Content Form Action Debug HM count Mis fatherto Mis motherto W is an ancestor of E is a parent of List of Predicates Ready Figure 5 51 The user defined predicates Y RU TS E X File Edit Window Options Content Form Action Debug Help Wcoure 4 M count Nis W count elizabeth is mother to charles eg is father to harry 7 EEE I san ancestor of MU e parr of Ml Bis a proof MI sterol CENA ee M morero Bl iad ARIES Xis father to IX is father to Y All Clauses Ready Aj Figure 5 52 Visualizing all clauses simultaneously Clicking on one of the clauses navigates to it The structure of the clause can be visualized in a number of ways as shown in Figure 5 53 As before the Legend can be edited to transform any one of these views into any other 128 5 9 PROLOG VISUALIZATION is a parent of X Y is mo
89. USSION 147 a similar specification technique The viewing window does not have scroll bars but these are also an effective method for viewpoint control Balloon help could describe objects beneath the mouse replacing the legend 6 5 8 Investigate other Specification Languages There are many variations of Prolog and other query based languages that would be able to implement the semantic model of visualization Prolog was chosen because of its established use its formal basis and its simplicity If the core language is sufficient to express a visualization system then any language which builds on Prolog is also sufficient The differences in syntax between the dialects of Prolog would change the appearance of the specifications but not change their meaning significantly Prolog itself provides a means of modifying its syntax using the op 3 predicate This example shows how new operators could replace view_content 2 and visual content 3 family tree Q gt person X man X woman X person X V gt labeled_icon X X person using graph A possible use for SVT is in visualizing databases Prolog can be interfaced to external databases but it may make sense to implement the semantic model in the database language itself The relationships between the the view context the view content the visual content and the graphical constraints might be expressible in a relational database language such as SQL 21 Database languages are o
90. Vmax can be extended by loading users Prolog Vmax also visu 14 1 3 RESULTS alizes Prolog and can edit Prolog as a visual language so that users can write their own queries visually to provide new views Prolog can specify other types of visualization such as directory browsing and graphical user interfaces SVT and Vmax were implemented to investigate whether using Prolog is a realistic approach to SV and how such a system would be designed 1 3 Results SVT and Vmax provide a comprehensive SV environment Because the approach is generic the system has a very broad scope and a very wide range of graphical output form Figure 1 1 shows a range of sample output On a modern desktop PC the performance of Vmax is quite reasonable making it suitable for visualizing projects of around 100 000 lines of code The code input rate is about 6 300 lines of Java per second The time to generate views is typically under half a second The final system has 151 different types of view or 70 if alternative visualizations are discounted These include 44 views for Java source code 7 views for Java run time 8 views for Prolog 6 views for directories and one view for help A taxonometric classification of Vmax has been made using the taxonomy of Price et al 79 given in Table 6 1 It shows the new approach to compare favourably to previous systems in many criteria including automation scope information content graphical output form specific
91. Work is needed to manage large data sets and control elision more effectively 2 8 Conclusions Software visualization is an emerging field and is becoming increasingly important in under standing documenting debugging and constructing software The great success of visual tech 38 2 8 CONCLUSIONS niques in professional development tools has greatly cut application development time and im proved application quality and is set to continue as more techniques move from experimental systems into commercial ones As software complexity grows more powerful techniques must be devised to maintain large software systems The trend is that software development is becoming more visual although visual program ming languages are still inadequate to replace text entirely in most circumstances Like visual programming languages software visualization systems must demonstrate their benefits before they will gain widespread acceptance Work must be done to find applicable formalisms for both visual languages and software visualization to allow progress Software visualization is currently impeded by lack of high level specification which could provide solutions for many problems including view specification quantity of views graphical mapping flexibility granularity elision control navigation scalability and automation Chapter 3 A Model of Information Visualization Software visualization presents unique challenges because the data presen
92. a view 84 7 3 and automatically tailors menus to the data 4 7 2 Potentially Prolog could analyze the data structure and decide itself how to display it However users still need to choose the view and visualization context themselves D 1 1 Almost every aspect of visualization 3 3 3 4 and interaction 3 8 4 7 2 can be specified in Prolog making the system highly tailorable Tailorability can be done by the user by changing the visualization relation 4 7 4 modifying it using the Legend 84 8 2 or editing the action relation 4 7 5 D 1 2 A high degree of customization is possible just by selecting different views 4 7 3 or graph ical forms 84 7 4 but the legend itself can be edited to customize output completely 4 8 2 Graphical queries can be added that customize the information in a view 5 9 Additional Pro log can be dynamically loaded that contains more view specifications 4 4 Other declarative systems can only be customized by textual programming 20 86 D 1 2 1 Vmax is connected to the program via loading its source code 5 5 loading its run time data 5 7 3 and automatic annotation by inserting trace points into the source code via drag and drop 85 7 1 The connection technique is similar to Viz 27 where players are event lines in Vmax states are input events the history is the trace file view is the view context filters are the view relations
93. ace next 5 inputs Ready y Figure 5 45 A sequence of five inputs 5 8 1 Value Visualization An individual value can be visualized by clicking on its input mark on an event line or from a view Clicking on a member of a compound structure navigates to the sub structure indicated by the mouse The correct name for each sub structure such as array 5 x is automatically derived for the value and displayed in the title of the view This name is automatically displayed in the status bar as the mouse moves over the view The value view is specified view Value trace value Value Value trace value Value _ _ view Input trace value Value Value of input trace input Input _ _ _ Value visual context trace value Value trace view content trace value Value title Value Name identifier Value Name view content trace value Value Content trace value content Value Content trace value content Value Content trace value Value _ Term Term array Values CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 121 in_list Value2 Values trace value content Value2 Term Term array Values Content array Value Values Term integer I Content integer Value I Term float F Content float Value F Term null Content null Value Term object Value Class Fields Content object Value Class Fields Term object Val
94. ackages and classes it also visualizes small scale structures such as methods and expressions in visual languages Run time data is also visualized and due to the ability to edit the visualization relation can be displayed in a virtually unlimited number of ways The views are highly interconnected and browser style navigation makes it simple to select a desired view Vmax augments SVT with parsers for directories Java source code and trace files and pro vides predicates for reading in analyzing and visualizing them A structure of text nodes stores the parse tree and each text node represents a block of text that can be edited in the text window This structure adapts as editing proceeds Run time data is acquired by inserting trace calls into Java source code recompiling and executing it Java s Reflection package is used to examine the structure of each run time value it receives This generates a file which is read and analyzed and the results can be browsed displayed in many different forms and animated Prolog can be edited as a visual language in several different graphical forms This can be used to add new views as an alternative to textual specification Vmax demonstrates the potential SVT has to define and implement visualization systems The implementation shows that Prolog is capable of specifying all aspects of SV and that the browser environment SVT provides can be used effectively as an interface to information 134
95. age com ruleofeight mail MailTest com ruleofeight mail Mail Pack kom meofe igt mail Javal hilMessage bom leo fag htm ail EmailAddress Exception com ruleofeight mail EmailAddress J icom ruleofeight find Regex Filename Filter Eom rueofeigtr find NegatedRe gexFilenameF iter com ruleofeight find GnuNativeFind com_ruleofeight find Find com_muleofeight find AlFilenameFiter java lang void java lang Mrtualivachine Error lava lang Verify ror lava lang Lhsatisfied Link Bror java lang Unknown Error java lang Throwable java lang Thread Group java lang ThreadDeath java lang Thread lava lang System fava Jang Stringing jexOut O fBoun dsExc eption na Jang StringBuffer java Jang String java lang Stack Overflow Error java lang Short java lang_Null SecurityManager java lang SeourityManager fava tang Security Exception J java lan g Runtime Exception java lang Runtime java lang Process lava lang OutOfivemory Eror java lang NumberFormat Exception java lang Number java lang Nul Point erExce ption ava lang No Such Method Exception java lang No SuchtethodError J java lang NoSuchField Exception java lang NoSuchFieldError java lang NoClassDef Found Etor ava lang Negstive Aray SizeExc ep on ava lang Mah java lang Linkage Eror J java lang InterruptedException J java lang IntemalError java lang instantiation Exception fava lang Instanti
96. age for the contents of the views This work has shown how these two functions can be combined For specialized systems with limited scope and fixed visualization behaviour there is little advantage to be gained from the flexibility that Prolog provides However it may be easier to implement such a system using a Prolog based visualization tool like SVT Systems that are knowledge orientated require user queries and offer multiple views and flexible visualizations could use the specification environment that SVT provides Systems that focus on numerical and volumetric data would derive only limited benefit from Prolog Existing techniques for numerical data are already very powerful Although the speed of Prolog is not as high as C and some algorithms may be better imple mented in C it is not a choice between Prolog and C Most Prologs have a C interface so C and Prolog can work together to achieve the best of both worlds Some Prologs such as BinProlog and GNU Prolog will actually compile to native code which is linked into the executable The use of Prolog does not have to be restricted to SV It provides a suitable meta language for many other types of knowledge URL http www binnetcorp com 2URL http pauillac inria fr diaz gnu prolog 154 7 4 1 7 4 THE ROLE OF PROLOG IN SOFTWARE VISUALIZATION Advantages of using Prolog as a Specification Language Prolog is a query language Data queries can be formulated d
97. ainly used for teaching Algorithms and algorithm animation are more abstracted from the underlying program The University of Washington Program Illustrator The University of Washington Program Illustrator UWPI 42 offers a source level run time debugger of Pascal programs allowing single instruction steps in the source code that provides an automatic visualization of the data in the program A layout strategist written in Lisp applies AI techniques to determine the best way to render the program data and displays the current data state graphically in a window According to Price et al 79 UWPI is the best example to date of automatic SV with any kind of intelligence But UWPI can only illustrate small programs involving sorting and graphs and only possesses a shallow understanding of numerical program data GROOVE GROOVE 89 from the Georgia Institute of Technology illustrates the run time behaviour of object oriented systems using animation so that programmers can better understand the class hierarchy and relationships Figure 2 13 shows GROOVE GROOVE animations are created by manually creating visual objects in the scene from func tion calls Therefore while GROOVE is flexible in its output there is a degree of effort to create visualizations and the system is only suitable for small programs Look Look from Objective Software Technology gt is a commercial product that allows visual display of the run time b
98. alability 17 that visual representations are not as compact as textual ones 71 awkwardness to edit in visual notation when text and keyboards can be faster 36 or that purely visual programming systems are no better than traditional ones for large projects The emphasis of current research is to tackle these problems In fact Brooks 13 is entirely in favour of using diagrams to assist software development and has enthusiasm for using diagrams as thought and design aids Brooks also concedes that multiple diagrams are required for different aspects of software and that some aspects are more suitable for visualization than others It is the awkwardness of visual representation at the lowest level code that seems to be impeding visual programming Removing text entirely may indeed not be desirable or practical but visual support for software development has proved to be very useful 2 3 Taxonomies of Software Visualization The most comprehensive taxonomy of SV is provided by Price et al 79 Myers s taxonomy 71 precedes Price s and offers taxonomies for VP as well as PV Roman and Cox 84 offer a simpler and less comprehensive taxonomy of PV Taxonomies serve to classify quantify and describe types of SV and visualization systems assist in finding weaknesses and areas of research and offer ways of comparing and evaluating SV systems The taxonomies also discuss and classify various SV systems Price et al classify and
99. alized There should be no difficulties in providing visualization for other languages and to visualize many different languages simultaneously 6 5 14 Data Persistence SICStus Prolog provides predicates for persistent data storage in an external database Vmax does not make use of it and uses the non persistent internal database This means that large software systems have to be stored in main memory which can be exhausted and the data must be reloaded each time Vmax is run which could be time consuming It could be better to store the program database externally to avoid this A distributed client server architecture could centralize this database to have an entire work group cooperating in the same system Views could be served as VRML over the internet and navigation within the VRML viewer could connect to the server to request the new view 6 5 15 Provide Visual Programming Vmax provides visual programming for Prolog The graphical notations for Java could also be manipulated but that is beyond the scope of this work Drag and drop techniques could rearrange source code at the statement level and tool bars could automatically construct code for programming constructs or members of classes 6 5 16 Interface to Compilers Program analysis is performed by Prolog which is a slower language than C or C and reim plements much of the work of existing compilers Instead of performing program analysis itself Prolog could interface
100. alized and at each discrete sample point lies the sea depth and its temperature The predicate sea_depth X Y Depth would convey the depth of the sea at the specified location while sea_temperature X Y Temperature would convey the temperature at each location Data would be added to these predicates from an external source A query such as sea_depth X Y D D gt 10 D lt 12 would return all locations where the depth was between 10 and 12 metres Example 2 Suppose wiring of a house is visualized The predicate socket Socket Pure Prolog implements first order logic ISO Prolog 1 contains features such as cuts that can break the logic but are a practical necessity to prevent unwanted backtracking 42 3 3 SPECIFYING VIEW CONTENT would identify the sockets while the predicate location Socket Location would identify the location of each socket perhaps by a description or a set of co ordinates Wires could be indicated by wire Socketl1 Socket2 Type to indicate the existence of the wires between sockets with Type being live neutral or earth Example 3 Suppose the liveness of variables was visualized over a function Variables could be represented as variable Variable and statements by statement Statement and liveness could be indicated by live Variable Statement The query live V s would return which variables were live for a particular statement s while live v S live V S
101. als with all levels of granularity The output of other systems is textually programmed whereas Vmax can be modified by the end user Interaction and navigation is much more advanced than in other systems The legend describes each view so the views should be clear but this has not been demonstrated in an empirical evaluation The specification strategy differs from other declarative systems such as Pavane 85 and Leonardo 20 because it uses three stages of visualization and is based upon an established language with efficient implementations Specifications in Vmax are reusable and provide en capsulation of graphical objects this is not possible in the other approaches As well as improvements to the implementation there are many avenues for further academic research The modular design allows any component of the system to be replaced 150 6 6 CHAPTER SUMMARY Chapter 7 Conclusions This dissertation has described a new paradigm for specifying and implementing SV A knowledge engineering environment stores and reasons about the data and views are generated by querying the graphical constraints for a particular view and visualization method By subdividing this query every aspect of the visualization system can be specified The work is aimed at addressing problems with automation flexibility granularity elision and specification that have been very difficult to solve in existing systems Its wide ranging scope makes it a str
102. an area of interest 40 but the method used here is the one used Open Inventor 106 The controls are given in Table 4 1 Such controls are a standard method for manipulating 3 D models This method also works for 2 D models which provides single mouse movements for continuous movement of the image to any magnification and position over the scene While rotation is necessary for viewing 3 1A loose upper bound is the amount of video RAM 56 4 5 GENERATING VISUALIZATIONS Ses Vmax Java method public void com ice gallery Gallerylmage Panel oldContactDirectory Mus File Edit Window Options Content Form Action Debug Help Java method public void com ice gallery GallerylmagePanel oldContactDirectory private InterruptedException ex A Figure 4 5 A complex view is shrunk to fit in the viewing area Viewpoint movement lt ctrl gt mouse button 1 pan in direction of mouse move lt ctrl gt mouse button 2 rotate the view with the mouse movement a fast movement spins it lt ctrl gt mouse button 3 zoom in or out with vertical mouse movement Table 4 1 The controls for viewpoint manipulation D models this feature is less useful for 2 D and can be ignored by the user to maintain an orthogonal viewing direction to the scene Examples of viewpoint manipulation are shown in Figure 4 6 The Window menu is shown in Figure 4 7 The option Reset viewpoint moves the viewpoint back to the overview of the scene If the
103. ang Object is output CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 115 ClassName class ObjectNo id Fields fields members where class is the name of the object s class id is an integer assigned to the object fields is the number of fields in the Object and members are the members of the class If the object has been output previously in the same event then instead it is output ObjectNo id The field members are output FieldName name value where name is the name of the field and value is its value Fields and array members may themselves be objects or arrays 5 7 3 Reading the Trace File A trace file is read by navigating to it The ensure trace loaded File predicate is called which reads the file File The trace file is read into a buffer and a lexical analysis is performed on it using Lex The lexemes are parsed using the DCG 75 trace file gt uN trace file H T gt input H trace file T input input E T V gt event E time T value V input input E N T V gt event E name N time T value V name N gt Name string N event E gt Event integer E time T gt TimeStamp integer T value integer 1 gt intvalue I value float F gt floatvalue F value array L gt array L value 0 gt object 0 intvalue 1 gt IntegerValue integer I floatvalue F gt FloatValue float F array L g
104. antages of Prolog Prolog is not very suited to applications programming because of its execution model and data handling In fact Prolog is even used to implement parts of Microsoft Windows NT 46 Its execution speed is inferior to low level languages like C There is overhead in providing a Prolog virtual machine and execution environment Many sources of data are provided in C which must be converted into Prolog However Prolog can access such data through a C interface Large numerical data sets are more efficiently manipulated in C But Prolog can call C functions for intensive computations The output must be fed to a graphical constraint solver The graphical constraint solver is slower and much trickier to implement than directly mapped graphics CHAPTER 7 CONCLUSIONS 155 7 5 Information Visualization in SVT SVT provides a very general framework for information visualization By implementing a knowledge database in Prolog and declaring the visualization environment at a high level any type of information visualization can be implemented SVT implements a complete graphical browsing environment for the data This approach provides an alternative to the traditional low level imperative programming method of implementing visualization By using a more high level language to declare the visu alization system directly development times could be greatly reduced 7 5 1 Formalizing Information Visualization The underlying sem
105. antic model was demonstrated to be an effective method of expressing vi sualization A very wide range of views have been expressed in the model By ensuring that each visualization is decomposed into a bijection between the semantics and its representation or syntax the task of specification is simpified and the mapping between the semantics and the graphical representation can be made flexible The specification and implementation of SVT uses this formal model A formal theory of visual types was proposed and was also shown to be an effective basis for implementation It is due to this work that an interactive legend could be implemented The model of interaction is also formal and can even be used to express other formal models of interaction such as CCS 69 given in Section C 5 2 There has been almost no previous work on the formal theory of SV 7 6 Contribution A theoretical model of visualization has been devised that expresses the high level relation ship between computer graphics and its underlying semantics While other such models exist 44 they are not applicable to SV because they are restricted to numerical data e A general technique for specifying visualization visual languages interaction and SV in Prolog has been developed e A generic visualization tool has been implemented that uses the theoretical model and specification method e A link has been established between visualization and knowledge engineering e
106. arsing directories parsing run time data analyzing Java analyzing run time data and visualizing Prolog Many views and visuals have been defined Its architecture is shown in Figure 5 1 Rules for Java Rules for file system Rules for Prolog ee a Ms Views Rules for run time data Visuals Operating System Figure 5 1 The architecture of Vmax 5 3 Text Processing Text processing is a significant part of Vmax because it needs to parse directories source code run time trace files and allow the user to edit parts of files To enable this a data structure of text nodes and text buffers has been implemented as part of SVT Text buffers store text files in a single array of text and text nodes are nodes in the parse tree of a text file CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 83 A text node stores pointers to a segment of text as offsets into a text buffer so each text node represents a block of text The address of this text node can be passed to Prolog which can query its contents or display it in the editing window Figure 5 2 shows a data structure of text nodes and text buffers A Root Text Node A AA AA Text Node Text Node Text Node A AoA A A AhAAA A Text Node Text Node Text Node AAA Text Node Parse Tree Figure 5 2 A data structure of text nodes in a text buffer If a text node s buffer is tagged as rea
107. articularly strong in that area while x means that Vmax does not support the feature at all As there are no objective metrics for measuring these criteria they should be regarded as no more than a rough guide The taxonomy of Price et al is used in favour of the taxonomy of Roman and Cox 84 because it is more comprehensive and more suited to comparative evaluation 6 1 1 Scope The scope of Vmax is very broad because it is comprehensive in program and run time visual ization No other systems combine both program and run time visualization to the same degree Systems using a declarative specification 20 85 lack the power of data manipulation for man aging and displaying program databases and are restricted to numerical data Where Vmax is deficient it should be possible to extend Vmax s scope by adding more views 4 4 A 1 Vmax s support for Java is much more extensive than it is for Prolog A 1 3 At present there is no specific support or view for concurrency although the architecture time stamps input events which would record concurrent behaviour 5 7 2 The specification mechanism could cope with concurrency A 1 3 1 Because Prolog stores the complete program database 85 5 135 136 6 1 CLASSIFYING VMAX A Scope A 1 Generality A 1 1 Hardware A 1 2 Operating System A 1 3 Language A 1 3 1 Concurrency A 1 4 Applications A 1 4 1 Speciality A 2 Scalability A 2 1 Program A 2 2 Data Set B Content B 1 Pr
108. ata on what information to provide 6 5 3 Improve the Graphical Constraint Solver There are limitations in the types of graphical output possible by SVT because of limitations in the graphical constraint solver It lacks many layout algorithms and its graph layout is quite primitive It does not have strong support for numerical visualization including producing many kinds of graph or volumetric data Even so the graphical output it does produce is sufficiently powerful to implement various visual languages and notations The modular architecture allows any kind of graphical constraint solver to be used The graphical constraints returned by Prolog can drive any constraint layout mechanism The existing constraint solver can be extended through its C interface to augment the layout algorithms already present The geometric constraint solver used to lay out visual objects can only manage linear con straints and is inadequate to solve more complicated geometric constraints or find optimal solutions 6 5 4 Animation SVT was not originally designed to be an animation tool Animation constraints could be pro vided that control animation in the scene and the viewer could be reimplemented to provide animation Paths of smooth motion such as those in Tango 95 could also be specified via con straints Inter frame animation could be implemented by automatically providing smooth path transitions between the same object in sequential vie
109. atement is a syn chronized statement with lock Expression text node Text and statement body Body throw_statement Statement Expression Dynamic Statement is a throw statement that throws the expression Expression try_statement Statement Block Statements Catches FBlock FStatements Dynamic State ment is a try catches finally block with main body Block of statements Statements a list Catches of catch blocks of the form catch Text Variable CBlock CStatements and a finally block FBlock of statements FStatements variable Variable Not extensible Variable is a variable variable Variable Type Modifiers Dynamic Variable is a variable of type Type and modifiers Modifiers while_statement Statement Condition CText Body Dynamic Statement is a while state ment with condition expression Condition condition text node CTerm and statement body Body B 3 3 Java Run time java add_tracepoint Statement Expression TracePoint Not extensible Adds a trace point Ex pression before a statement Statement and returns the identifier of the trace point Trace Point trace any_next_input Input1 Input2 Not extensible Finds the next input event Input2 after Inputl trace any_previnput Input1 Input2 Not extensible Finds the input event Input2 immediately preceeding Input ensure_trace_loaded File Not extensible Loads the contents of trace file File trace event Input File EventNo Time Value Dynamic
110. ates 1994 42 R R Henry K M Whaley and B Forstall The University of Washington Illustrating Compiler In Proceedings of the ACM SIGPLAN 90 Conference on Programming Lan guage Design and Implementation White Plains New York 20 22 June pages 223 233 1990 43 W L Hibbard C R Dyer and B E Paul Display of scientific data structures for algorithm visualization In Proc IEEE Visualization 1992 pages 139 146 1992 44 W L Hibbard C R Dyer and B E Paul A lattice model for data display In Proc IEEE Visualization 1994 pages 310 317 1994 45 J M Hoc T R G Green and R Samurcayn Psychology of Programming Academic Press 1990 46 D Hovel Using Prolog in Windows NT network configuration In Proceedings of Practi cal Applications of Prolog 3rd 6th April 1995 Paris France pages 317 339 April 1995 47 W Hower and W H Graf A bibliographical survey of constraint based approaches to CAD graphics layout visualization and related topics Knowledge Based Systems 9 7 449 464 1996 48 J Jaffar and M J Maher Constraint logic programming A survey Journal of Logic Programming 19 20 503 581 1994 49 D F Jerding and J T Stasko The information mural A technique for displaying and nav igating large information spaces In Proceedings of the IEEE Symposium on Information Visualization Atlanta GA October 1995 pages 43 50 1995 50 M J Kilgard OpenGL programming for the X
111. ates to the semantics of visual languages From a theoretical standpoint a visual language is a set of diagrams which are valid sentences in that language where a diagram is a collection of symbols in a two or three dimensional space 65 One of the main issues in visual language theory is defining membership of visual languages The main meth ods for specifying visual languages are graph grammars attributed multiset grammars logical formalisms and algebraic formalisms The main drawback with visual language theory is that it deals with interpretation not view generation This is useless for visualization where the task is to generate the diagram in the first place One grammar that is generative is given by Grant 36 where the grammar defines the geometric transformations of objects that are contained in one another In its broadest sense parsing images is a vision problem In visual language theory diagrams are used where the symbolic constituents of the diagram are known Such a pre processing stage is analogous to lexical analysis Parsing diagrams is more costly than parsing text because of the unordered nature of the constituents of a diagram and because geometric conditions must be tested Visual Grammars The two main types of grammar are graph grammars and attributed multiset grammars Graph grammars 81 regard the view as a set of interconnected symbols Graphical rewrite rules spec ified in the grammar tran
112. ation tailorability navigation granularity and elision control An empirical study of usability and usefulness of the system lies beyond the scope of this work It is intended only to demonstrate new techniques and there are no standard methods for empirical evaluation of SV systems The system is easy to use due to 1ts browser interface and degree of automation but this has not been confirmed by usability studies The reader is invited to download Vmax 1 4 Contribution The work demonstrates a number of new techniques that can be applied to SV e It has been demonstrated that the programming language Prolog can be used to specify all aspects of SV e A SV system has been constructed that compares favourably to existing systems in many taxonometric criteria including scope information content graphical output form speci fication tailorability navigation granularity and elision control e A semantic model of information visualization provides a link between visualization first order logic and knowledge engineering Visualization systems and knowledge engineer ing systems can be integrated e A general purpose information visualization tool has been implemented that can be speci fied entirely in Prolog e A dynamic mapping between the information content of a view and the graphical form of a view has been implemented This can be queried to provide an automatic legend and modified by the user Different mappings can be
113. ation Eror E walang Index Out Of Bounds Exception java lang Incompatible Class Change Eror J java lang_lllegalThread State Exception java lang Illegal State Exception java Iang legli nitorSt ateExo eption jav a lang llleg alAcgument Exception java lang egalAccess Exception J java lang MegalAccess Error J java lang F DBigint J java lang FloatingDecimal J java lang Float J lava Jang ExceptionhinitialerError J java Jang Exception java lang Error java lang Double iava lang Compiler java lang Clone Not Supported Exception jaya lang Class Not Found Exception iava lang ClassLoader lava lang ClassFormat Emor igya lang Class Circularty Error java lang Class Cast Exception java lang Class llaya lang Byte aya lang Boolean iava lang Array Store Exception Loxa tng ray nd exOutO fBoundsException Ba Jang Arithmetic Exception java lang Abstracthtethod Emor iava lang Object com ice gallery About Dialog All Java classes com ruleofeight mail EmailMessage J Figure 5 18 All classes and interfaces File Edit Window Options Content Form Action Debug com ruleofeight util QuoteDelegate j com ruleofeight system StdOutConsumer j com ruleofeight mail EmailMessage j java lang Runnable j ava lang Cloneable Java Interfaces com ruleofeight util QuoteDelegate Figure 5 19 All interfaces 98 5 6 JAVA VISUALIZATION Y vmax Java
114. automatically cus tomized depending on what object is conveyed in the menu context Menu contexts do not have to be parameterized The order items appear in the menu matches the order in which they are declared The special description seperator is not displayed verbatim in the menu but inserts a horizontal divider for aesthetic purposes More complex menu structures such as sub menus could in principle be declared in a similar way Another example of a menu specification is given in Section C 5 1 Menus can be created as either pop up or pull down The former is displayed at the mouse cursor position while the latter is invoked from the menu bar A pop up menu is created by calling popup menu MenuContext which pops up the given menu typically from the reaction 2 predicate Pull down menus are declared pulldown menu MenuContext Description which declares that the menu MenuContext with label Description should be put on the menu bar When an action is selected from a menu its corresponding reaction is executed 4 7 3 Navigation In SVT navigation is achieved by changing the view context This is done using the new view context l predicate or the new_view 2 predicate which also changes the visualization context new_view_ context NewViewContext new _view NewViewContext NewVisualContext The new contexts are passed to the current viewer which deletes the old scene and constructs a new scene as described in
115. avane Pavane visualizations are very complicated to set up The target program in C is manually annotated by inserting a number of calls that initialize Pavane and declare the data to watch via the VisualMonitor function Event points are inserted into the source code by calling VisualUpdate which tells Pavane to update the visualization with the new state A file containing visualization rules declares how to produce the visualization Only numer ical values and arrays can be passed to these rules Each rule declares objects in the scene and their screen coordinates depend on the values obtained from the C state Expressions map the input data to coordinates in the scene The syntax of these visualization rules is very complex Once the system has been set up quite powerful visualizations are possible The main problem with Pavane is the manual effort and complexity in defining new visualiza tions Its specification language is very cumbersome and low level Leonardo Leonardo 20 is an algorithm animation system that provides a logic programming language called Alpha to define graphics corresponding to data of a program executing in a virtual machine Leonardo claims to be much easier to use than Pavane 85 and its virtual machine can be stepped forwards as well as backwards An output from Leonardo is shown in Figure 2 12 Declarations in Alpha are embedded in the object program within and for example int g MAX NODES
116. be animated as shown in Figure 5 44 The animation is implemented by trace animate file File repeat trace input _ File _ _ Value frame_rate 1 0 navigate _ to trace value Value user_interrupt where backtracking over the predicate will navigate to each value Value in the trace file File A variation of this action animates groups of inputs and another variation animates backwards The repeat predicate loops the query frame rate 1 0 pauses to ensure an interval of one second between calls and user_interrupt throws an exception when the user presses the Escape key terminating the animation CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 117 Ses Vmax Trace file homes camg100 Work SVTibin samples Traces Quicksort trac diam iD File Edit Window Options Content Form Action Debug numbers numbers array from Trace file homes camg100 WVork SVT bin samples Traces Quicksort trace svt Figure 5 41 All the events in a trace file S s Vmax All input event values a File Edit Window Options Content Form Action Debug Figure 5 42 All values in a trace file 118 5 8 RUN TIME DATA VISUALIZATION Sl Vmax Event values aray ES File Edit Window Options Content Form Action Debug Help 224 array 274 array 317 array 361 array Event values array Ready Ji Figure 5 43 All the values in an event line CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 119 HE vmax Examine value array e a x
117. be stored lang java package_declaration Id Dynamic Id is a package declaration parameter Variable Dynamic Indicates that the given variable is a parameter to a method program_file ParseTree File Not extensible Performs the analysis of the parse tree ParseTree of the file File reachable Method1 Method2 Prev Visited Not extensible Method2 is reachable from Me thodl and Prev is the method that calls Method2 Visited is a list of methods already visited to prevent looping which is initially return_statement Statement Dynamic Statement is a return statement with no return value return_statement Statement Expression Dynamic Statement is a return statement that re turns the expression Expression scan_directory Directory Not extensible Reads in all of the Java source files in a given direc tory Directory statement Statement Dynamic Statement is a Java statement statement_entry_point Statement EntryPoint Not extensible Gives the statement entry point EntryPoint the first statement executed within a compound statement Statement statement_exit_point Statement ExitPoint Not extensible Gives the statement exit point Ex itPoint the last statement executed within a compound statement Statement May give multiple exit points upon backtracking statement_follows Method Stmtl Stmt2 Type Not extensible In method Method statement Stmtl is followed by statement Stmt2 with transition ty
118. ble Declares that in action context Action Context the user s input Action has response Reaction mouse_cursor Cursor Not extensible Changes the mouse cursor to Cursor Valid cursors are hand normal watch zoomin and zoomout popup menu Menu Not extensible Posts a pop up menu at the mouse cursor Menu is the menu context identifying the menu to display popup modal Title Message Buttons Result Not extensible Pops up a modal dialog box one that must be dismissed before this predicate returns with title Title message string Message a list of strings Buttons giving a list of buttons Result returns the button that was pressed or O for cancel popup text Title Prompt Default Result Not extensible Pops up a modal dialog box prompt ing the user to enter text which is returned in Result Title is the title of the dialog box Prompt is the explanatory text and Default is the initial text If the user hits the Cancel button the unit atom is returned in Result pulldown_menu MenuContext Text Extensible Declares a pull down menu with menu bar text Text and menu context MenuContext reaction ReactionContext Reaction Extensible This predicate implements responses to user input and is called whenever a reaction Reaction needs to be executed in reaction context ReactionContext set_status_text Text Not extensible Changes the text in the status bar to Text B 2 5 Text Predicates buffer Object Buffer Dyna
119. bles stored dates and even which functions or statements apply date calculations Using SV a program analysis would determine this information and display it in graphical form Although development tools have provided enormous productivity gains their degree of vi sualization is still very limited and visual programming has not proved as successful as was initially hoped 107 The recent advances in software development tools have come from im proved user interfaces and graphical support and SV could improve that productivity further In spite of the obvious benefits there are considerable technical difficulties to overcome The main problem preventing SV from being used in professional development tools is its inflexi bility and high set up costs A system for automatic SV is needed to make SV accessible to programmers At the same time it should be flexible enough to provide the desired information is the desired format In general SV systems are either automatic or flexible they are not both 64 The aim of a SV system is to provide useful information about a program to a user Program mers need many different types of information so complete systems should visualize all aspects of software including the program structure low level details high level overviews and run time data Most SV systems are very specific and have a very limited scope In addition they have a very poor range of granularity and do not view the program on man
120. ce supporting visual programming Brooks 14 states that A favorite subject for Ph D dissertations in software engineering is graphical or visual programming the application of computer graphics to software design Nothing even convincing much less exciting has yet emerged from such efforts I am persuaded that nothing will This argument should not be taken too seriously because it was written before windowing systems he only con siders flow chart techniques and confuses superimposed views with multiple views Brooks also claims that limited size of display renders visual programs unscalable although this argument collapses when one considers multiple views Scaling is indeed a problem in visual languages 17 but the problem has been overcome in textual languages O Brien 73 states beware the claims of visual programming Drawing lines between objects becomes bafflingly web like Purely visual programming is not yet and may never be viable This argument is also flawed because structured networks such as those used in LabVIEW 104 or ProGraph 19 77 is just one visual representation of code The empirical evidence supporting visual programming languages is discussed at length by Whitley 107 These arguments are directed at the practicalities of VP where the text in the system has been replaced entirely by diagrams rather than at SV Visual programming has many weaknesses that need to be overcome including lack of sc
121. com pare in detail twelve systems Sorting out Sorting 3 Balsa 16 Zeus 15 TANGO 95 ANIM 8 Pascal Genie from Chariot Software Group UWPI 42 SEE 4 TPM 12 Pavane 86 LogoMedia 25 and CenterLine ObjectCenter 93 Price et al then construct a taxonomy based upon a tree structure where each leaf in the taxonomy offers a different and orthogonal classifi cation criterion The complete taxonomy is shown in Figure 2 2 Price et al justify their first level classification by a generalized model of the software vi 20 2 3 TAXONOMIES OF SOFTWARE VISUALIZATION Figure 2 2 Price et al s taxonomy of software visualization CHAPTER 2 PREVIOUS WORK 21 sualization process Scope relates to the source program and the input domain of the software visualizer Content describes the particular aspect of the software that is visualized Form de scribes the output of the system Method describes how the implementation is specified and how the SV system works Interaction describes how the user can interact with the view or the SV system to control navigate or modify the visualization Effectiveness offers evaluation crite ria for SV systems By making their taxonomy tree structured they allow their taxonomy to be extended as the field evolves Roman and Cox offer a related taxonomy shown in Figure 2 3 Data State Scope Control State Behaviour Direct Rep
122. contains Root Descendent Not extensible This is the transitive closure of member 2 contains Root Descendent Depth Not extensible The same as contains 2 but only searches to a maximum depth of Depth current_time Y Mo D H Mi S Not extensible Returns the current time as year Y month Mo day of month D hour H minute Mi and second S darken Colour1 Colour2 Not extensible Takes a colour Colour and returns a colour Colour2 that is darker ensure_fresh Object Extensible Makes sure that the given entity Object is loaded and is up to date This predicate is called every time the view context is changed identifier Object Identifier Extensible Declares that the entity Object has a textual identifier Identifier lighten Colour1 Colour2 Not extensible Takes a colour Colour and returns a colour Colour2 that is lighter long_identifier Object Identifier Extensible Declares that entity Object has a long textual identifier Identifier member Parent Child Extensible Dynamic Declares a relationship between Parent and Child short_identifier Object Identifier Not extensible Declares that entity Object has a short tex tual identifier Identifier system_call Command Not extensible Executes the shell command Command and waits for the process to terminate before returning 170 B 3 VMAX PREDICATE SUMMARY system_process Command Not extensible Executes the shell command Command and does not wait f
123. creen as both the hierarchy and the list can make poor use of space The algorithm works from the leaf vertices up by summing the widths and the heights of a vertex s subtrees If the total width exceeds the total height then the list layout is used otherwise the hierarchy layout is used More sophisticated versions of this algorithm could do an even better job of packing trees into a small space 4 6 Expanding the Graphical Vocabulary The set of visuals Ty forms a graphical vocabulary with which to communicate information A large graphical vocabulary means a more varied graphical output which is visually more interesting and can be used to mimic other notations This can be implemented in several ways in SVT The first is through the C interface which can define new graphical constraints by using OpenGL directly to output graphics The second is to add clauses to the visual primitive 2 predicate that composes graphical constraints into visuals thereby augmenting Ty This is still a low level approach but is much more convenient than programming in C A more high level approach is to specify the composition of visuals directly by visual_component Visual Component where Visual is the new visual being declared and Component is a visual that is already declared Visuals can be defined in terms of other visuals and it is straightforward to augment and combine existing visuals visual _component 2 is called from visual primitive 2 by
124. ctions as a graph e Class hierarchy browser Shows the class hierarchy and a window lists class members e Routine and data access profile display Displays run time usage of data e Speedup and parallel execution extrapolation display Shows graphs of the performance of parallel C Tau s visualization is too limited to classify it as a SV system Tau is implemented manually there is no easy way of extending the views or the system and its output form is limited to numerical graphs and a class hierarchy 4URL http www takefive com CHAPTER 2 PREVIOUS WORK 25 2 4 2 Visual Programming Systems Visual programming systems consist of a visual editor to construct visual programs in a visual programming language and a means of executing the visual programs In Price et al s definition 79 SV encompasses VP because VP is a specialization of SV that allows interactive mod ification of the program VP systems use many visual programming languages VPLs visual representations of software and techniques to interact with visualizations to modify the program Visual programming languages have existed since 1966 99 and there are now hundreds of VPLs in existence VPLs are becoming as prolific as textual languages although there is not the same tool support such as Lex and Yacc 2 for visual programming 36 making it consider ably more difficult to implement visual programming languages than a textual ones 71 and only a small
125. current view which is useful if the information content of the view is known to have changed retrieve_thumbnail_view Not extensible Sets the view and visualization context of the main viewer to the view and visualization context of the current bookmark view save_thumbnail_view Not extensible Sets the view and visualization context of the current bookmark view to the view and visualization context of the main viewer swap_thumbnail_view Not extensible Swaps the view and visualization contexts of the current bookmark view and the main viewer viewport_object Object Not extensible Returns the object Object being visualized B 2 2 Visualization Predicates add_component Visual Id Component Extensible This predicate can be used in place of object_component 3 but is used to add object components to an existing object anchor_expression_id Expression Object Id Not extensible Returns an identifier Id for an anchor expression Expression in a visual object Object anchor _expression_primitive Expression Object Id Primitive Not extensible Returns the graphical constraints Primitive of an anchor expression Expression in a visual object Ob ject with anchor identifier Id APPENDIX B DOCUMENTATION 167 object_component Visual Id Component Extensible Declares the object components for a visual object Visual with identifier Id Component is typically a list of components object_primitive Component Id Primiti
126. d colour them according to their type for example reserved words or comments Syntax colouring gives demonstrable benefits to programmers resulting in fewer syntax and typing errors and faster comprehension of the source code 4 SEE converts source code into the style of a technical manual for automatic documentation Type faces bold and italic emphasis font size and shading are used to display the source text 26 2 4 SOFTWARE VISUALIZATION SYSTEMS with one procedure per page and procedure parameters clearly laid out in a table Cross refer ences to other procedures are generated automatically There are other source formatting utilities available including vgrind in BSD Unix and WEB part of the TEX typesetting system SEE is very limited in its output its output format is fixed and it does not make extensive use of graphics SeeSoft and SeeSys SeeSoft 29 by AT amp T Bell Labs is a SV system that produces line oriented software statistics of the source code in a software project Each line of source code is represented by a single hor izontal bar contained in a vertical bar representing each source file Modification statistics are compiled for each line of source code which are then displayed using a coloured scale continu ous or discrete showing update frequency or time of last update Files can be grouped according to function and Figure 2 7 a shows the screen layout SUBSYSTEM STATS 136041 272082 408123
127. d only then so is its text in the editing window When text is edited text moves in the buffer and so the text nodes must move their pointers to new positions This is done by traversing the parse tree and can be implemented more efficiently if text nodes store text offsets relative to their parents offsets A text buffer can pipe the input or output of a shell command into the text buffer or load and save files File text buffers have an associated filename Each text buffer can also have an associated lexical analyzer and parser An instruction to execute the lexical analyzer in Lex or parser in Yacc on that text buffer will invoke the correct one since there are multiple lexical analyzers and parsers compiled into Vmax It is even feasible to parse a single text node allowing local updates to the parse tree A complete list of operations Prolog can perform on text is given in Section B 2 5 A text file is loaded by calling create file buffer Filename Buffer TextNode Type which creates a file buffer for file Filename and type Type and returns the address of the buffer Buffer and text node of the whole file TextNode The characters CharList in a text node TextNode can be retrieved by calling lexeme_string_list TextNode CharList A lexical analysis can be performed by calling lex_buffer Buffer Lexemes which returns the list of lexemes Lexemes for the buffer Buffer using its lexical analyzer The buffer can be parsed by ca
128. d visualization setup are written in C and inserted into the source program There are user controls to pause control speed and pan the display and the visualizations are 2 and 24 dimensional animated in real time Figure 2 9 a shows XTango Tango for X Windows that visualizes animation scripts generated from a running program gt a E Ea a Bin packing in b Minimum spanning subtree in Polka c Quicksort in XTango Polka 3D Figure 2 9 Visualizations produced by Tango and its derivatives Tango has now been superseded by Polka 96 which is well suited for colourful smooth animations of parallel programs Animation calls are made from a running algorithm that call a 28 2 4 SOFTWARE VISUALIZATION SYSTEMS C library to effect animation events in a display Objects are created that can be moved around the scene Samba is a front end to Polka that accepts animation directives from a text file Polka supports 2 D and 23 D visualizations shown in Figure 2 9 b but Polka 3D produces full 3 D visualizations shown in c Several visualization systems have been built upon Polka including PARADE 97 to visual ize parallel and distributed systems ConchViz to visualize the execution of cluster and message passing environments PVaniM visualizes and animates data from a Parallel Virtual Machine which sends communication and user event statistics back to a monitor for visualization T
129. description Description This predicate can be used instead of content type 2 and con tent_description 2 content_description Content Description Extensible Declares that the information content Content has textual description Description This text appears in the legend alongside the symbol content_type Content Type Extensible Declares the type relation that assigns types Type to information content Content visual Visual Type Description Symbolld Symbol Extensible The visual term Visual has type Type description Description and has a symbol Symbol in the legend with object id Symbolld Symbol is the visual that appears in the legend This predicate combines visual_description 2 visual_type 2 and legend symbol 2 visual_context_description VisualContext Description Extensible Declares that the visual ization context VisualContext has textual description Description This text apperas in the Form pull down menu that selects alternative visualization contexts for the data visual_description Visual Description Extensible Declares that the visual Visual has tex tual description Description This text appears in the pop up menu that selects alternative graphical forms for the content visual_type Visual Type Extensible Defines the type relation y that assigns types Type for the visual Visual 168 B 2 SVT PREDICATE SUMMARY B 2 4 Interaction Predicates action ActionContext Action Reaction Extensi
130. e Any part of the visualization process can be modified by changing either the view or the visualization relation Navigation is achieved by changing the view context and there is no limit to the number of views that can be defined The graphical representation of data can be changed simply by changing visualization context ee A Interaction is modeled as a relation gt between user actions and system reactions The interactive behaviour is specified directly in Prolog The next chapter describes an implementation of this formal model in a visualization system Chapter 4 Semantic Visualization Tool Semantic Visualization Tool SVT is a visualization tool that implements the se mantic model of visualization described in Chapter 3 It is intended to be gen eral purpose allowing any kind of data to be presented in any format although its strengths lie in presenting semantic networks The most unusual aspect of this tool is the Prolog engine used for querying the scene data SVT is in essence a graphical constraint solver sitting on top of Prolog Queries issued to Prolog return a large set of graphical constraints that are constructed into a scene graph and rendered SVT provides a graphical user interface a 3 D model viewer and an integrated text editor Its architecture allows its presentation capa bilities to be extended through extensive C and Prolog application programmer interfaces The entire system is specified usi
131. e declared visual_type highlight X X content _type is_tall X X 3 6 Specifying Graphical Constraints The visual content V is a high level graphical description of the view that is mapped to a set of low level graphical constraints G that drive a graphical constraint solver The graphical con straints are shown in Figure 3 1 The set of graphical constraints completely defines the output image Definition 9 Graphical constraints Let Tg C T be the set of terms that represent graphical constraints Let G C Tg be the graphical constraints of the view Each visual in V is mapped to a set of graphical constraints that compose the visual Definition 10 Visuals Let E be a relation between Ty and Tg ig implemented in Prolog by the predicate visual primitive 2 visual_primitive Visual Constraint implements the relation gt and defines the graphical constraints Constraint for a visual Vi sual Clauses can be added to this predicate to define new visuals in the visualization system Example 17 A labeled icon could be declared visual_primitive labeled_icon Id Text Icon Primitive Primitive icon icon Id Icon Primitive string text Id Text Primitive v_stack Id icon Id text Id The object that is defined has an identifier Id and has an icon of type Icon with identifier icon Id and a string of Text with identifier text Id v_stack Id icon Id text Id specifies that the icon should
132. e mapping from view content to visual content without redefining the views It also makes it possible to provide multiple graphical representations for the same data Previous systems have only provided a fixed mapping which must be edited in text to change the view e These mappings can be fully formalized and embedded in first order logic It is possi ble to reason about heterogeneous data visualization Previous systems do not base their specification on a formal model of visualization e These mappings can be specified straightforwardly in Prolog This means that there is no need to devise a new language and is therefore much easier to implement and will run much faster e Because Prolog is introspective it can query its own program and dynamic it can modify itself these mappings can be queried and modified by the user at run time Other systems offer fixed visualization behaviour and cannot display an automatic legend describing the view The fact that the mapping between view content and visual content is a bijection permits this e It is possible to pose arbitrary queries to a knowledge database Previous systems restrict this to run time data e Visuals are specified independently of the view data and can be reused This is a big problem for previous systems which effectively have to start from scratch each time a new view is created e Multiple views can be defined independently and used within the same system By label in
133. e task F 3 Vmax is new software and has not had a chance to be used in real projects It is not quite of production quality and lacks user manuals The research aims were not to produce commercial software F 4 6 2 Evaluating Vmax Evaluation poses a particular problem for SV systems There has been little empirical evaluation of the other systems described here 71 79 An empirical evaluation is beyond the scope of this work so conclusions about Vmax s usability or the merits of adding graphics to a programming environment will not be made One approach is to evaluate the usability of the system using a cognitive dimensions frame work 37 39 This is a framework for discussion about programming environments providing many different evaluation aspects and criteria Often changes in one dimension will affect the 140 6 2 EVALUATING VMAX performance in others However it is more focused on tasks of data manipulation and visual programming while Vmax does not provide these functions Cognitive dimensions can describe particular notations and environments but Vmax uses a general approach that does not fix the notation or environment From a cognitive dimensions perspective the programming task is still textual Cognitive dimensions only examines the user s perspective and does not discuss the implementational aspects The approach we shall use is to compare Vmax s capabilities with existing systems and Price et al
134. ect Anchor2 Additional constraints may be imposed on anchors depending on the context in which anchor expressions are used For example if an anchor is specified as the corner of a container then it is aligned with the other corners of the container Anchor expressions appear within specifications for polygons text and containers but can also appear as object components Some constraints can only be specified as object components such as e xalign Anchorl Anchor2 yalign Anchorl Anchor2 zalign Anchorl Anchor2 Constrains the given axis to be aligned in both anchors Generates the graphical constraint axis_align anchor Object Anchor1 anchor Object Anchor2 N where N is 0 1 or 2 Laying out Visual Objects The constraints on the anchors must be solved to give screen coordinates for the anchors and hence for the graphical primitives When a visual object receives an On reset_size call it must assign relative coordinates to all of its anchors to find their extrema and the size of the visual object The visual object can traverse the scene graph to determine the set of constraints that it needs to solve Such constraints form a set of linear inequalities Solving such linear systems is well established an example of such an algorithm can be found in 11 Even some non linear geometric constraints can be solved and algorithms for solving these types of constraint can be found in 58 The point to emphasize is that any system of c
135. ection between the semantics and the visuals needs to be defined By editing the legend this bijection does not need to be specified at all Alpha and Pavane require textual programming to change the graphical display but this can be done by the end user through a GUI in SVT Both Alpha and Pavane can specify animation in their scenes Animation in SVT is quite primitive and is limited by the viewer and not by the specification language 6 3 2 Modularity SVT and Vmax are modular in their design The window manager is accessed through a well defined interface so that SVT can be used with any windowing system The layout constraint solver could be completely replaced by any layout mechanism that inter faces Prolog to OpenGL The existing hierarchy could be extended with better layout algorithms The 3 D graphics toolkit could be replaced or the declarative language that is used to drive the constraint solver Even C can drive the constraint solver directly thus bypassing the need for Prolog to generate visualizations Other aspects of the data gathering such as the program database do not need to be managed by Prolog either but could be stored using existing tools and accessed by Prolog 6 3 3 Application Areas Information Visualization SV is a particularly challenging example of information visualization because of the size and complexity of the data structures the wide range of possible output and the highly heterogeneous nature o
136. ection with this line The geometry for such transformations can be found in 31 In general a bounding box intersection test with the line is all that is needed By calculating the distance from the object to the viewer precedence of actions can be given to objects that are nearest the viewer Not every visual object in the scene needs to be tested Because objects are often nested child objects need only be tested if the line intersects their parent object Once the correct action is found its reaction is executed by the reaction 2 predicate CHAPTER 4 SEMANTIC VISUALIZATION TOOL 71 4 7 2 Menus The predicate action 3 can also specify menus Each menu has a term to identify it called a menu context The action term is of the form menu MenuContext Description where MenuContext is the name of the menu and Description is the text that appears in the menu For example a menu indicating actions on an object could be specified action menu actions File Compress file svt compress File file File action menu actions File Delete Name svt delete file File file File directory File identifier File Name action menu actions Class Make all members public svt to public Class java class Class action menu actions _ seperator svt action menu actions _ Exit SVT svt exit specifies the menu actions Object in an action context svt This menu is
137. ed in a number of ways The Add new clause option on the action menu of the Predicates view adds a new user clause Clicking on the clause navigates to it and its Clause view shows just that clause while its Predicate view shows all the clauses in its predicate The Delete clause option on a clause s action menu removes the clause A right mouse click on a term will pop up a menu giving alternative names for its functor or atom Selecting an alternative from this list will change the functor and redisplay the view The Set functor option lets the user enter the name of the functor This menu is shown in Figure 5 55 and the menu is defined by the user_term 3 predicate This list could get quite long so maybe a system of sub menus could be implemented to manage this A middle mouse click on a term will pop up a menu giving operations on the term as shown in Figure 5 56 The Increase arity option increases the arity of the compound term by one and Decrease arity decreases the arity by one A compound term with an arity of zero becomes an atom Cut Copy Paste Delete and Duplicate perform the described action on the term Dragging a term onto another will replace the destination with the source term of the drag and this also works between windows The effect of such a drag is shown in Figure 5 57 There is also a separate Scratch terms view which can be used to deposit terms temporarily by dragging 130 5 9 PROLOG VISUALIZATION Figure
138. ed to a layer equal to its distance from the root vertex so that the root vertex is on layer zero This ensures that adjacent vertices appear on adjacent layers or on the same layer Finally an ordering of vertices on each layer minimizes the number of edge crossings This algorithm works reasonably for sparse graphs and is shown in Figure 4 10 Multi container Layout A multi container is a composite visual object that lays out its children in rows as shown in Figure 4 11 This algorithm resets the sizes of its children and finds the optimum width of the container such that the aspect ratio is close to an ideal value This width is found by interval bisection The ordering of the children is not changed because their order may convey information CHAPTER 4 SEMANTIC VISUALIZATION TOOL 61 void list Recursively Std Out Consumer consumer int depth String f i list WectordescendList new Vector retum a ja retum Enumeration subFinds descendList elements find ls Figure 4 10 Graph layout bin cgi bin freetype 1 1 gcc 2 95 gccinst glt23 glut33 gprolog 10 0 img include kdeold kde build lesstif lib netcache netscape qt Share tmp vprol xmed Figure 4 11 A multi container Fish eye Lens A fish eye lens is a very wide angle lens that shows places nearby in detail while also showing remote regions in successively less detail 87 A simple fish eye lens has been implemented that s
139. ed_line A B coloured_line A B red the visual red_line 2 is defined in terms the visual coloured Jine 3 which is itself composite No additional Prolog is required to use the visuals labeled icon 3 coloured Jine 3 and red line 2 4 6 1 Specifying Visual Objects Visual objects are graphical entities on the screen They are effectively pictograms that com municate entities and data in the view content Visual objects can also communicate textual and numerical information and combine other visual objects It is the way visual objects com bine that is particularly difficult 36 as illustrated in Figure 4 14 but this complex behaviour is necessary to implement visual languages com ice sqlclient SQLClientHandler T Nos Yes Figure 4 14 Visual objects adapt their layout in response to other visual objects Such behaviour could of course be programmed directly through the C interface but SVT provides a means of specifying this expansion behaviour in Prolog Each visual object consists of a set of anchors that are points in space and graphical components such as lines and polygons attached to these anchors These object components are specified with the object component 3 predicate which specifies the components Component of an object with visual Visual and iden tifier Identifier 2 Anchors could be points in space time for animation but this possibility has not been explored CHAPTER 4 SEMANTIC VISUALIZATIO
140. ehaviour of C systems It is claimed that this allows programmers to under stand the system much more quickly find bugs much faster reduce redundant object behaviour and improve the overall code quality Programmers can view references between classes the SURL http www objectivesoft com 32 2 4 SOFTWARE VISUALIZATION SYSTEMS public virtual void draw int int protected COLOR mycolor countfct Figure 2 13 Screen shot of GROOVE showing an object oriented design class hierarchy object creation and deletion events and concurrent threads Each view shows the current method acting on the objects Look views are hard coded and actually provide just four different visualizations class references class hierarchy object creation and code clustering While the system is automatic it is dedicated to just these views and cannot be extended by the user The views themselves show only object names and method invocations 2 4 6 Architectures Viz Viz 26 27 is a framework developed at Open University in which run time SV systems may be built Viz uses players which can be any program object such as a variable a function or a code block to generate history events shown in Figure 2 14 a l F x gall ally tingly keyboard mouse instructions from user a Acquiring and filtering history data b The architecture of Viz Figure 2 14 Viz T
141. eher VPL A visual logic programming language Journal of Visual Languages and Computing 2 163 188 1991 61 S P Lahtinen E Sutinen and J Tarhio Automated animation of algorithms with Eliot Journal of Visual Languages and Computing 9 337 349 1998 62 T V Le Techniques of Prolog Programming with implementation of logical negation and quantified goals John Wiley and Sons Inc 1993 63 G L Lohse K Biolsi N Walker and H H Rueter A classification of visual representa tions Communications of the ACM 37 12 36 49 December 1994 64 D Manuel The impossible dream Towards general and automatic visualizations Tech nical Report 322 University of Exeter Computer Science Department 1995 65 K Marriott and B Meyer Visual language theory Springer Verlag 1998 66 B H McCormick T A DeFanti and M D Brown Visualization in scientific computing a synopsis IEEE Computing Applications and Graphics 7 4 61 70 1987 67 D W McIntyre comp lang visual Frequently Asked Questions comp lang visual 1999 68 Sun Microsystems Java Virtual Machine Debug Interface Reference 1999 69 A J R G Milner A calculus for communicating systems Lecture Notes in Computer Science 92 1980 70 B Mohr A Malony and J Cuny Tau In G Wilson editor Parallel Programming using C M I T Press 1996 71 B A Myers Taxonomies of visual programming and program visualization Journal of Visual Lang
142. ell would run the pre compiled Prolog bytecode myvis q in SVT The argu ments arg and arg2 are passed as input to myvis The loaded Prolog files contain data interfaces rules for reasoning about the data and define the view visualization action and reaction relations to completely define the behaviour of SVT A main viewing window is created by calling create viewer ViewContext VisualContext ActionContext ReactionContext with view context ViewContext visualization context VisualContext action context Action Context and reaction context ReactionContext This window consists of a main view area a text area pull down menus and a status bar shown in Figure 4 3 CHAPTER 4 SEMANTIC VISUALIZATION TOOL 55 A MASAS E X File Edit Window Options Content Form Action Debug Calculator Home directory Pressing Help on the menu bar navigates to the main help system A hand cursor indicates that navigation is possible by Root directory tree a single left mouse click A spider cursor navigates backwards by a single left mouse click Selecting the Legend option from the Window menu explains all of the graphical notation Selecting the Samples view provides sample data to view Files are only loaded when they are navigated to Welcome to Vmax Figure 4 3 The main viewing window Load Prolog Ctrl L Close Ctrl W Quit Ctrl Q Figure 4 4 The File pulldown menu Additional
143. ents a specification language that defines data structures of numbers and mappings between elements in the data structures to graphical display coordinates For example map earth location to xz plane map temperature to y axis In later work by Hibbard et al 44 a lattice model describes the relationship between data and display They concede that their technique is inadequate for complex information processing such as recursive data structures and they foresee considerable technical difficulties to manage such data 2 5 3 Specification Languages for SV SV systems fall into three categories those that are not extensible at all those which are extended in the native programming language and those that provide their own language for specification Only Pavane Leonardo VIS AD and Samba fall into the last category Samba is merely a sequence of visualization and animation directives Both Pavane and Leonardo use a declarative specification to map live run time data to graph ical objects The drawback with both systems is that they only provide a direct mapping to graphical objects on the screen Thus they are both very low level approaches Another prob lem they have is that they lack encapsulation for example to define libraries of objects and the mapping from the data to the graphics is fixed A change in the mapping can require significant reprogramming The implementations of Pavane and Leonardo are very different howeve
144. ere is little theory that underpins SV Current approaches do not use any theoretical basis and are implemented ad hoc Work should be done on the formalization of information display which might lead to more principled methods for designing SV systems The specification techniques for SV systems are very awkward SV systems are specified at a very low level and they are therefore cumbersome to use Work on specification languages is needed to provide high level specification and scripting of SV which would make SV much easier to implement and provide much more powerful systems In spite of these technical difficulties SV systems provide valuable insights into the workings of complex software for many different kinds of computer language and task The challenge SV presents is to provide a system which is very broad in scope very flexible in its output has dozens of different types of view is completely automatic can be interactively customized and can be specified at a high level 1 1 Aims Price et al 79 identify many of these difficulties and outline directions for future work The aim of this work is to produce a SV system that addresses some of these problems by achieving the following aims 1 Have a broad scope visualizing all aspects of computer software Previous systems are too limited in their scope Aspects to include are e Program structure e Object orientation e Methods e Run time data e Results from program
145. ers It would seem clear that a feature or view should be deemed useful if it is used regularly by a user If a wide range of views was used or views and navigation paths were taken that are not available in traditional programming systems then the worth of Vmax is demonstrated Questionnaires could also give supporting evidence 6 5 2 Measure Programmers Needs No attempt has been made to optimize the graphical views and notations used in Vmax which are only a first approximation of what might be useful Discovering which notations programmers find most useful what information programmers require and how they wish to navigate would lead to a configuration of Vmax that was more tailored to programmers needs Choice of graphical notation can have a significant impact on user performance 37 Com parative performance of various graphical notations can be measured but care must be taken to 2URL http www ibm com dx 3URL http www avs com 146 6 5 FURTHER WORK ensure that the experiments are not swamped by noise and become inconclusive 107 The ar chitecture of Vmax allows very different notations and views to be constructed relatively easily so would be an appropriate test bed for different notations A study of how users navigated through Vmax would indicate important navigation paths and perhaps yield the kind of information programmers were interested in Questionnaires asking programmers needs could provide d
146. ert Container Object Inserts an object Object into an empty container P1 further P2 Constrains the anchor P1 to be further than anchor P2 fillcontainer Object Sets the alignment of the visual object Object to fill its container Pl leftof P2 Constrains the anchor P1 to be left of anchor P2 line Points Declares a sequence of line segments along the list of anchors Points Id line Points Declares a sequence of line segments Id along the list of anchors Points P1 nearer P2 Constrains the anchor P1 to be nearer than anchor P2 APPENDIX B DOCUMENTATION 177 oscale Scale A B Value Vector Offsets the anchor B by r Vector from A where r is the weighting of the value Value in the scale Scale taken from the origin polygon Points Creates a polygon whose vertices are given by the list of anchors Points The polygon should not be self intersecting and the vertices should go clockwise or anti clockwise Id polygon Points Creates a polygon with identifier Id with list of vertices Points Id multicontainer A B C D Creates an empty container with comers A B C and D into which multiple objects can be inserted P1 rightof P2 Constrains the anchor P1 to be right of anchor P2 scale Scale A B Value Vector Offsets the anchor B by r Vector from A where r is the weighting of the value Value in the scale Scale text Text A B C D Declares a text component with string Text with corner anchors A B C and D Id text Te
147. es from the Action menu This constructs the dynamic predicates cached calls M M2 and cached_references M V which store which methods M use other methods M2 or variables V These behave exactly the same as method calls 2 and method refs 2 but are much faster EE Vmax Members declarations in com ice sqiclient MySQLCient x File Edit Window Options Content Form Action Debug com ice sqlclient MySQLClient com ice sqlclient MySQLAdapter com ice sglclient SQLClientHandler Members declarations in corm ice sqlclient MySQLClient com ice sqlclient SQLClientHandler A Figure 5 21 Classes contained within a class 100 5 6 JAVA VISUALIZATION A vmax Cass cross references EIX File Edit Window Options Content Form Action Debug Help farEntr ice tar Tat Diss com ice tar TarButter JavaBorder J Component Base hom ice sqlolient SOLClient som ice saiclient My SOLClient om ice sgl dient Tab ke Defs Panel Icom ice sglclient Tables Panel lcom ice saiclient Main Panel com ce sgloient Host Info Panel com ice sglclient Command Panel com ice sglclient UpdatePanel Icom ice sglclient Query Panel fom ice sglolient Min Frame Y com ice gallery Gallery Image Panel com ice gallery ContactThread omice gallery FikTree fom ice gallery FileTreeData Bombo galery Galbo Geom ice gallery Fle TreeFiter Romice gallery Image Canas Remice gales GalleryUserreedback c
148. es where Vmax fits in to the taxonomies of SV The conclusions are in Chapter 7 The appendices contain documentation for SVT and Vmax and further examples to illustrate the specification techniques Chapter 2 Previous Work Although computer programming is traditionally a textual discipline the use of graphical illustrations of software has always played a useful role in software de velopment and writing computer programs Illustrations are used to plan and doc ument software although the idea of using illustrations to create programs visual programming has been around for a long time 99 as well as automated utilities to document and browse finished software Visual techniques are now commonplace in software development particularly in the design and generation of user interfaces the construction of skeleton applications and the navigation of source code This survey describes the issues and areas of research within software visualization a number of systems both commercial and research that use software visualization techniques and a number of taxonomies for describing and comparing software visualization systems The survey concludes with research problems that need to be solved for software visualization to deliver real demonstrable gains to software developers 2 1 Basic Definitions There is no universal consensus in the literature for the definitions of broad terms such as visual visualization visual programming so
149. esentation There are advantages in using the third dimension including added realism and tangibility of program objects increased information density and an extra quantitative axis Shneiderman 90 argues that programmers think of their code as structural and tangible and that direct manipulation of abstract objects in the computer using a visual paradigm would be the easiest and most natural way to interact with a computer The 3 D approach adds reality and tangibility to the program objects thus concretizing the soft ware more effectively for the programmer However there is evidence that making objects more realistic makes it more difficult for users to extract the information or abstract the semantics that the symbol conveys 63 There is also evidence that 3 D visualizations can convey between 1 6 and 3 0 times more information than an equivalent 2 D view and users are able to understand complex relationships better when presented in 3 D 105 Lohse et al 63 derive a classification of various basic types of visual representation by ask ing subjects to sort a set of sixty images of a full range of visual representations By performing statistical tests they compiled a taxonomy to grade each of the images according to the high level 36 2 7 RESEARCH DIRECTIONS properties of spatiality temporality comprehensibility level of abstraction continuity attrac tiveness subdivision numerical level dynamic vs static and informa
150. et_constraint 3 predicate is implemented as get_constraint ViewContext VisContext Constraint view_content ViewContext Content visual_content Content VisContext Visual visual_primitive Visual Constraint 58 4 5 GENERATING VISUALIZATIONS ee Figure 4 8 The base classes of graphical constraints Q V G and implements the relation gt described in Section 3 7 4 5 2 Generating the Scene Graph Each graphical constraint Constraint is a term that represents one node of a scene graph The nodes in the scene graph are instances of the base classes shown in Figure 4 8 Thus each type of node is either a visual object or an attribute and visual objects may be composite or binary associations This base class hierarchy implements the basic graphical types in Section 3 5 which are objects attributes and associations Vobject refers to all constraints that can be rendered on the screen Vcompound refers to all constraints that compose sub objects Vassociation is an association between two visual ob jects and Vattribute is an attribute applied to visual objects This base class hierarchy covers all graphical entities attributes and associations that can make up an image The C header file svtvisual h implementing this class hierarchy is included in Section B 9 1 The complete class hierarchy of visuals provided by SVT is shown in Figure 4 9 AS a naming convention the graphical constraint s
151. etric constraints to augment these basic ones 4 7 Interaction While viewpoint manipulation is implemented in C interaction with the data is specified entirely in Prolog User input events such as mouse moves and key presses are passed through the GUI to SVT SVT receives the screen coordinates of mouse inputs and interrogates the scene graph to determine the objects beneath the mouse cursor that the user was intending to interact with The result is an action term that encodes the user s input This encodes a high level descrip tion of the action such as a mouse move over a particular object rather than low level screen coordinates The action terms that SVT generates are e click Object Buttons Times The mouse has been clicked over object Object with the mouse buttons Buttons Times times Buttons is the term Left Middle Right Shift 70 4 7 INTERACTION where Left is the left mouse button Middle is the middle mouse button Right is the right mouse button and Shift is the shift key which are either on or off e drag From To Buttons A mouse drag from object From to object To with the mouse buttons Buttons e key Object Key A key Key has been pressed with the mouse cursor over object Object e linger Object The mouse cursor is resting over an object Object e menu MenuContext MenuText An item from a menu MenuContext with menu text MenuText has been selected e move Object Buttons The mouse curso
152. ew functor Functor new arity Arity and description Description user_view_content ViewContext Content Dynamic A dynamic version of view content 2 that can be modified by the user at run time user_view_context Object ViewContext ViewDescription Dynamic A dynamic version of view 3 that can be modified by the user at run time user_visual_context ViewContext VisualContext Dynamic A dynamic version of visual con text 2 that can be modified by the user at run time scratch_term Clause Dynamic Declares that the clause Clause is in the scratch area B 4 Visuals Defined by SVT This information is displayed by selecting List visuals from the Debug menu B 5 Visual Object Components Defined by SVT H T A list of object components P1 above P2 Constrains the anchor P1 to be above anchor P2 alignment Object Align Sets the alignment of the visual object Object to be Align within its container anchor E An anchor expression E P1 below P2 Constrains the anchor P1 to be below the anchor P2 Id colour Colour Attributes the colour Colour as defined by the colour 2 predicate or as R G B to the object component Id contain Object A B C D Contains a sub object Object with corners A B C and D Id container A B C D Creates an empty container Id with corners A B C and D Id container A B C D E E G H Creates an empty container Id with the corners of its bounding cuboid A B C D E F G and H ins
153. expression A conditional statement is stored java conditional_statement Statement Condition CText Then Else where Condition is a term representing the condition expression CText is the text node of the condition Then is the identifier of the then statement and Else is the identifier of the else statement A while statement is stored java while_statement Statement Condition CText Body where Body is the statement of the body of the loop The other types of statement are stored java break statement l Jjava break_statement 2 java case_statement 3 java conditional statement 4 java continue state ment l java continue_statement 2 java declaration statement 2 java default statement 1 java do_while_statement 4 java empty_statement l java for_statement 8 java labelled statement 2 Jjava return_statement 1 java return_statement 2 java synchronized statement 4 java throw stat ement 2 and java try_statement 6 these predicates are documented in Section B 3 2 Expressions in statements or declarations are represented by terms Prefix postfix and infix operators Operator are represented by prefix Operator E postfix Operator E and infix Oper ator A B respectively on expressions A B and E A conditional is represented by condi tional Cond A B for a condition expression Cond Cast expressions are represented cast Type Expr where Type is the type of the cast Qualified names are represented as qualified name Na me where
154. f the data Other applications are possible provided that the data is interfaced to Prolog Visual Programming Visualization is an intrinsic part of visual programming because visual language systems need to generate diagrams to communicate the visual program The constraint based output of SVT can specify visual languages which relates the diagram to the language structure and semantics Vmax uses several graphical notations to represent programs in Prolog or Java There is no facility to manipulate Java graphically which distinguishes program visualization from visual programming The combination of visualization and interaction provides a complete specification of visual programming SVT can emulate block structured visual languages but is more limited with node and link based programming Extending SVT s layout capabilities would extend the possible visual language notations Graphical User Interface Specification A graphical user interface GUI is essentially an interactive graphical view and can therefore be specified in SVT Visuals define the widgets in the GUI which are generated automatically from the data in the view The dialog box in Figure 6 1 is specified view_content dialog font_option Font Name Status font_data Font Name widget _data Font Status CHAPTER 6 DISCUSSION 143 Font select Figure 6 1 A dialog box specified in SVT view_content dialog style_option Style Name Status s
155. filter out what is not of interest Different visualization relations 4 7 4 or editing the legend 84 8 2 can also hide various types of information Elision is manually controlled through menu selections E 2 1 Temporal control is also quite good because the program data can be stepped forwards or backwards and animated forwards or backwards at any frame rate 5 8 E 2 2 There is no facility to record and play back user inputs Textual scripting is possible by executing sequences of Prolog commands 84 3 E 3 6 1 6 Effectiveness The purpose of Vmax is different to many SV packages because it is designed to be used on real programs and not small examples for teaching F 1 This change of emphasis makes it imperative that views can be generated quickly and automatically The automatic legend describes each view so therefore the views should be clear and self explanatory 4 8 No other SV systems provide an automatic legend for each view The views themselves have not been optimized to programmers needs but the system can be completely tailored to provide the most appropriate views 3 3 3 4 and interaction 83 8 F 2 An empirical evaluation is beyond the scope of this work Vmax is aimed at software de velopers and measuring the performance of the whole development process is difficult It is therefore not possible to say whether Vmax is effective in real situations or whether Vmax is at all appropriate for th
156. ften imperative or functional which lack backtracking capability and the syntax of such queries may not be so user friendly However Datalog 102 offers a language similar to Prolog that implements a relational model to access an underlying database 6 5 9 Investigate Visual Specification Languages The main problem with Prolog is that it is slightly too complicated to be used as an end user lan guage Its execution model is quite difficult to understand and most users are used to imperative programming styles Thus only expert users would be able to add new views to the visualization system The visual languages Vmax provides to edit Prolog clauses are fully operational albeit changes are not persistent The benefit of this notation and means of editing over textual Prolog has not been demonstrated and it is unclear whether users would find this approach or any visual lan guage a more accessible means of specifying visualization A more comprehensive study of graphical notations and editing for specifying visualization queries may yield results on the merits of graphical specification and yield notations that are particularly effective Work in visual query languages 18 and end user queries could be used to specify the information content in views Visuals and visual objects could be designed by graphical tools instead of textual specification 6 5 10 Natural Language Interface Another approach to end user specification is to use a
157. ftware Visualization Algorithm Visualization Program Visualization Data Animation Static Static Algorithm Visualization Static Code Data V via gramming A Visualization Visualization Algorithm Animation Figure 2 1 Venn Diagram showing relationships between the various fields of software visual ization reproduced from 79 Myers excludes VP from PV while Price et al include it Price et al classify the SV fields and nomenclature according to Figure 2 1 although they concede that their classification is an oversimplification because these fields are strongly interrelated Sometimes products such as Delphi or Visual Basic are regarded as VP because visual tech niques are being used to create software But these are not VP systems because textual program ming is still necessary navigation and the design of the user interfaces is visual but that is not considered to be programming The above fields all use visual languages which Myers defines as all systems that use graphics including VP and PV systems and visual programming languages VPLs are visual languages used for programming The definitions given by Myers 71 are probably the de facto standard but are by no means definitive 2 2 Benefits of Software Visualization Visual representations seek if not always successfully to assist software development By pro viding representations of software that are in some wa
158. ftware visualization program visualization and algorithm animation Although the concepts themselves are fairly clear different sources place emphasis on different concepts in forming their definitions The Concise Oxford Dictionary 100 defines visual as of concerned with or used in see ing and visualize as make visible esp to the mind thing not visible to the eye make visible to the eye Price et al 79 prefer the definition the power or process of forming a mental picture or vision of something not actually present in sight from the Oxford English Dictionary 92 with the emphasis that visualization is a mental process rather than just the task of vision and can therefore include all of the senses This contrasts with the definition given by McCormick et al 66 as the study of mechanisms in computers and in humans which allow them in convert to perceive use and communicate visual information with the emphasis clearly on the process of communication These definitions are not mutually exclusive because visual communication precedes the cognitive model In this text visual is taken to mean visible and visualization to mean conversion to visible form even though this excludes the other senses In the context of computers visual means visible on a computer display and visualization is communicating data with graphics This specifically includes all graphical forms as well as text whic
159. g JTextArea messageText Java class com ice gallery AboutDialog Figure 5 9 The main class of a file ee vin lava file AboutDialog java g com ice gallery com sun java swing border com sun java swing java lang java awt event java awt com ice util AWTUtilities com ice gallery AboutDialog Java file AboutDialog java Figure 5 10 The overview of a Java file CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 93 visual_context java_decl Decl class view_content java_decl Decl title Java Type Name java declaration Decl Type _ _ _ _ long _identifier Decl Name view content java_decl Decl text T text _node Decl T view content java_decl Decl variable Var Name member Decl Var java variable Var identifier Var Name view_content java_decl Decl method Method Name member Decl Method java method Method identifier Method Name view_content java_decl Decl native Method member Decl Method java is native Method view_content java_decl Decl abstract Method member Decl Method java is abstract Method view content java_decl Decl public Member member Decl Member java is public Member The alternative views of a class are selected from its Content menu shown in Figure 5 11 Java class Variables Methods Inherited members Inherited structure a Inherited variables str
160. g each view and visualization mapping directives such as visualize function Fn using coloured nassi_shneiderman diagram can be issued Support for multiple view defini tions in other systems is much more limited CHAPTER 3 A MODEL OF INFORMATION VISUALIZATION 41 e Arbitrary data structures can be passed through the visualization pipeline Previous sys tems deal only with numbers which is too limited e It is possible to interact with objects in the view because they are assigned identifiers of objects in the database Interaction is also specified formally It isn t possible to specify interaction in other declarative systems 3 2 Representing Data Data gathering is an essential first step in any visualization system Therefore it is necessary to formalize the notion of the data from which to visualize Previous approaches 23 44 86 restrict this domain to numbers this domain must be extended to encompass all types of data The approach we shall use is to regard this data as a set of propositions that can be computed to be true Such a notion has little meaning in an imperative language but in logic programming 1t is the set of propositions that can be proven from the inference rules of the logic program The canonical form of logic programming is the programming language Prolog although there are many other logic programming languages We shall restrict ourselves to the core language Using Prolog has several advantages f
161. gs can be selected from a menu to render the information differently The mapping can be queried so that a legend can automatically describe what is in the view The mapping can be completely tailored by the user by interacting with the legend to change the way individual items are graphically represented SVT provides a browser interface so that any item in a view is navigated to by clicking on it Alternative queries of the current object being viewed can be selected from a menu and alternative methods of visualizing the current information can also be selected from a menu A preview window automatically provides a graphical display of the object beneath the mouse cursor and a bookmark window provides five thumbnail views that can be used to store and retrieve views SVT provides a framework for SV A system called Vmax has been implemented to browse Java source code and Java run time data Vmax is specified in Prolog and uses Prolog to analyze source code and run time data and store the program database Views defined in Vmax can query the program database to provide any information about the program SVT provides Vmax with a specification language that allows all aspects of SV to be spec ified at a high level SVT also automatically manages navigation and the legend and provides graphical algorithms to display the output Vmax has a text editor to edit source code and provides a complete development environment with integrated end user SV
162. guages IEEE Computer 16 8 57 69 1983 91 B Shneiderman The eyes have it A task by data type taxonomy for information visu alizations In Proceedings 1996 IEEE Symposium on Visual Languages September 3 6 Boulder Colorado pages 336 343 IEEE Computer Society Press 1996 92 J A Simpson and E S C Weiner eds The Oxford English Dictionary Oxford Univer sity Press 1989 93 CenterLine Software ObjectCenter Reference CenterLine Software Inc Cambridge Massachusetts 1991 94 I Somerville Software Engineering Addison Wesley 1992 95 J T Stasko Tango A framework and system for algorithm animation IEEE Computer 23 9 27 39 1990 96 J T Stasko Polka animation designer s package Technical Report 30332 0280 College of Computing Georgia Institute of Technology Atlanta GA 1992 97 J T Stasko The PARADE environment for visualizing parallel program executions A progress report Technical Report GIT GVU 95 03 Graphics Visualization and Usability Center Georgia Institute of Technology Atlanta GA January 1995 162 BIBLIOGRAPHY 98 K Sugiyama S Tagawa and M Toda Methods for visual understanding of hierarchical systems EEE Trans on Systems Man and Cybernetics SMC 11 2 109 125 1981 99 W R Sutherland On Line Graphical Specification of Computer Procedures PhD thesis School of Computer Science Cambridge MA 1966 100 J B Sykes The Concise Oxford Dictiona
163. gure 5 35 b which also reflects Java s semantics that unless there is a break statement case rules are executed sequentially Different method displays are obtained from the method s Content menu shown in Figure 5 36 Another means of representing the method code is through a flow graph This shows control flow between statements in a method Different types of transition such as the branch of a condition are labeled and indicated by different colours as shown in Figure 5 37 Exceptions present problems because potentially they can be thrown at any point during a try block it is difficult to analyze which exceptions could be thrown at which points which handler would handle it and could obfuscate the flow chart For these reasons they are not shown in the flow graphs Synchronized blocks are also difficult to represent in a flow graph and only the entry point of a synchronized block is indicated A final representation for a method extends Nassi Shneiderman diagrams to visualize expres sions within statements An example is shown in Figure 5 38 2Thanks to Alan Blackwell for suggesting these CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION Steal max Java method public String java math BigDecimal toString SENS File Edit Window Options Content Form Action Debug Help Song toString if scale 0 return intVal toString StringBuffer buf String intString intVal abs toString int signum signumd int insert
164. gure 5 7 Some different views of a directory tree 88 5 5 ANALYZING JAVA SOURCE CODE 5 4 1 MIME Types A MIME type is a multipurpose internet mail extension used for declaring the content of a file MIME types can be implemented straightforwardly in Prolog by mime_extension Type Extension to declare that a particular MIME type Type has the filename extension Extension Subtyping of MIME types is declared mime subtype SuperType SubType for MIME types SuperType and SubType Applications to handle MIME types are declared mime application Type Description Application where Description describes the application and Application is the location of the executable MIME types provide relevant application programs for a file which appear automatically on the Action menu of a file as shown in Figure 5 8 Selecting an application executes it with the given file Edit plaintext View hex contents Compress Open directory Rename file Delete file Open in Kwrite Open in Kedit Open in Emacs Move up Move back Move forwards Execute command Show log Empty log New clause Add view Recalculate cross references Calculator Figure 5 8 MIME types add applications to the Action menu of a file 5 5 Analyzing Java Source Code 5 5 1 Constructing the Program Database Before Java can be visualized it must be read in to Prolog s knowledge database This is achieved by parsing Java
165. h is of course visible These definitions are more straightforward than those offered by Price et al or McCormick et al Myers 71 defines visual programming VP as any system that allows the user to specify a program in a two or more dimensional fashion and specifically excludes textual languages which he considers to be one dimensional picture definition languages and drawing packages 17 18 2 2 BENEFITS OF SOFTWARE VISUALIZATION Price et al define VP as the use of visual techniques to specify a program in the first place Price s definition is better because textual languages are visual and have a two dimensional ren dering Myers confuses the display with the internal representation of the textual language Myers distinguishes program visualization PV from VP as the program being specified in a conventional textual manner and the graphics is used to illustrate some aspect of the program or its run time execution Price et al prefer the term software visualization SV to encompass algorithm animation animating the operations of an algorithm code visualization a visual form of the program code and data visualization a visual form of the program data and defines it to be the use of the crafts of typography graphic design animation and cinematography with modern human computer interaction technology to facilitate both the human understanding and effective use of computer software So
166. he effort required to get there SVT is very flexible in specifying its behaviour so can be tailored heavily to suit this kind of environment The views and interactive behaviour of Vmax merely illustrate the possibilities of visualiza tion and navigation but no investigation has been done here of programmers needs This would require questionnaires timed completion tasks and detailed measurement of programmers activ ities which is beyond the scope of this work Vmax produces activity logs that would facilitate such a study Vmax should however offer benefits over ordinary textual programming but there is no empirical evidence to support this Because Vmax provides all of the functionality of a textual programming environment it should certainly do no worse than a text editor 81 82 5 2 ARCHITECTURE Vmax stores information as facts in Prolog s internal database This means that the data is not persistent and needs to be reloaded between sessions Directories and source code are loaded on demand when the user attempts to visualize them Lex and Yacc 2 are used to lexically analyze and parse input files and the parse tree is passed to Prolog for analysis This constructs the program database which is stored in dynamic Prolog predicates Each entity in the program database is assigned a unique identifier so for example any file directory method variable class or statement are all represented by unique integers uid Id retur
167. he system architecture is shown in Figure 2 14 b and consists of histories views individual visualization windows mappings translation from view information into graphical output and navigators to manipulate and move between views CHAPTER 2 PREVIOUS WORK 33 Viz itself provides only a framework for implementation and does not provide a specification language Vogue Vogue 51 is a 3 D visualization system that has been applied to parallel concurrent program ming object orientation and software engineering Systems built using Vogue include Visu aLinda 55 Integrating Version Control and Module Management 54 Fractal Views 52 56 and Bottom up Program Visualization 53 Vogue s primary operation is to extrude an array of 2 D networks into a 3 D image The argument is made that more information can be fitted onto the screen and the evolution of a system can be viewed in a static diagram However Vogue is an architecture that provides only nodes and links that must be created manually in the source code Its output capabilities are very limited using just colour shapes and lines to communicate information Vogue is actually just an output library and is not an end user system because a considerable programming effort is required to generate visualizations 2 5 Specification 2 5 1 Visual Language Theory Visual language theory 65 explores formalisms for structuring the visual domain and how graphical structure rel
168. here are other visualizers for parallel processes including XPVM Xab PVMTrace and Pablo 57 The main drawback with Polka is that 1t offers little more than an animation library in C Because the animator must manually write all of the animation calls it is very time consuming to create animations in Polka A degree of expertise is required to learn the library Balsa and Zeus Zeus 15 by DEC SRC is written in Modula 3 and is based upon an algorithm animation sys tem called Balsa 16 Balsa itself was used as a teaching aid for algorithms using algorithm animation and provided a windowed graphical display panning zooming and allowed users to view the execution of several algorithms running simultaneously and can provide synchronized multiple views of the same algorithm The user is able to set the graphical display of data items and a source window with pretty printed text is also available Balsa works by inserting interest ing event points into the source code also known as annotating the source code which generate visualization scripts that can be replayed a b c Figure 2 10 Output from the Zeus algorithm animation system Zeus supports synchronized multiple configurable views Zeus can run in a multi threaded multi processor environment making it suitable for visualizing parallel programs some sound extensions have been incorporated into Zeus and Zeus now supports 3 D Figure 2 10 shows Zeus in operation
169. hods Reachable methods cached Reachable methods Calls to method cached Calls to method Home Figure 5 36 The Content menu for methods ERA String java math BigDecimal toString fl OX A ARA String java math BigDecimal toString File Edit Window Options Content Form Action Debug He File Edit Window Options Content Form Action Debug Java method public String java math BigDecimal toString Java method public String java math BigDecimal toString public String java math BigDecimal toString A return intVal toString a a Black and white b Colour Figure 5 37 Control flow within a method 112 5 6 JAVA VISUALIZATION 8 Vmax Visual method public String java math BigDecimal toString File Edit Window Options Content Form Action Debug String toString Visual method public String java math BigDecimal toString a The view Symbol Meaning a b a b Q a b c Etatement A declaration statement Eh a b lfthen statement Ifthen else statement ath D A literal O Method declarator pl Method call scale Name scale new Type a b Statement Exit statement Sequential statements Expression statement Legend for Nassi Shneiderman vertical b The legend Figure 5 38 Nassi Shneiderman Diagrams with colour and expression visualization CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 113 5 7 Analyzing Run time Data Run time data
170. hows a tree with the base of the tree with the highest detail The root vertex is placed in the centre of the graph and children are placed in concentric rings around the centre The sizes of the children are reduced exponentially with distance from the root so that they fit into the graph without overlapping A fish eye lens is shown in Figure 4 12 The algorithm works by assigning each vertex a sector of the circle and a depth d from the root vertex Each vertex divides its sector equally between its children The magnification of each vertex is given by a 4 1 and its distance from the centre by R 1 8 4 2 62 4 6 EXPANDING THE GRAPHICAL VOCABULARY rs j NN 2 2 Aa gt eo ight com ice Figure 4 12 A fish eye lens where d is the distance from the root node R is the radius of the fish eye lens a is the decay constant for the size and is the radius ratio In this implementation was chosen to be 0 5 and 5 was 0 4 In the case where the size of the vertices exceed their sector size their radii are reduced to prevent overlap Hierarchical List Layout A hierarchical list is a hybrid of a hierarchy shown in Figure 4 13 a and a nested list shown in Figure 4 13 b and is shown in Figure 4 13 c It aims to achieve a better aspect ratio ratio width to height than either the hierarchy or list alone Perhaps a better aspect ratio can make better use of the screen real estate and fit more data onto the s
171. i view editing In Pro ceedings of the IEEE Workshop on Visual Languages Kobe Japan October 1991 pages 4 9 1991 16 M H Brown and R Sedgewick Techniques for algorithm animation EEE Software 2 1 28 39 1985 157 158 BIBLIOGRAPHY 17 M M Burnett M J Baker C Bohus P Carlson S Yang and P Vanzee Scaling up visual programming languages Computer 28 3 45 54 1995 18 T Catarci M F Costabile S Levialdi and C Batini Visual query systems for databases A survey Journal of Visual Languages and Computing 8 215 260 1997 19 R G Cote ProGraph CPX purely visual Byte 20 1 179 1995 20 P Crescenzi C Demetrescu I Finocchi and R Petreschi LEONARDO a software visualization system In Proceedings of the Ist Workshop on Algorithm Engineering pages 146 155 1997 21 C J Date An Introduction to Database Systems Addison Wesley 1994 22 R S Day Alternative representations In The Psychology of Learning and Motivation Vol 22 G H Bower ed pages 261 305 Academic Press New York 1988 23 C Demetrescu and I Finocchi A general purpose logic based visualization frame work In Proceedings of the 7th International Conference in Central Europe on Computer Graphics Visualization amd Interactive Digital Media 99 pages 55 62 1999 24 C Demetrescu and I Finocchi A technique for generating graphical abstractions of pro gram data structures In Proceedings of the 3rd Inter
172. il Bookmark thumbnail save view action menu thumbnail Retrieve thumbnail retrieve view action menu thumbnail Swap thumbnail swap views action menu thumbnail Refresh thumbnail refresh Menu to modify the structure of terms shown in Figure 5 56 action menu term_menu Term Increase arity svt increase arity Term action menu term_menu Term Decrease arity svt decrease arity Term action menu term_menu Term Cut svt cut_term Term action menu term_ menu Term Copy svt copy term Term action menu term_menu Term Paste svt paste term Term action menu term_ menu Clause Delete clause svt delete clause Clause user _ clause Clause action menu term_menu Clause Duplicate clause svt duplicate_clause Clause user _ clause Clause C 5 2 An Implementation of CCS CCS 69 is a theoretical calculus of communicating systems This example shows that formal models of interaction can be expressed within SVT s action reaction model Demo of CCS in action reaction model ccS terms are represented P1 P2 P3 for concurrency and T1 T2 is a process Transitions of X and X are reduced The silent transition tau is automatically reduced Restriction of a name X is marked nu X Renaming is marked New Old accepts no reductions AHP A A oP V V P V action tra
173. ile Nevertheless they provide a useful browsing and overview facility Graph Visualizer 3D Graph Visualizer 3D GV3D 32 105 is a tool to create three dimensional visualizations of networks and in particular of computer software and object oriented code Figure 2 8 shows CHAPTER 2 PREVIOUS WORK 27 visualizations it can produce By itself GV3D is just a graphical output library that can be interfaced to analysis tools s a Calls b Class hierarchy c Modules Figure 2 8 Renderings of almost 6 million lines of code by Graph Visualizer 3D 2 4 4 Algorithm Animation Systems Algorithm animation systems illustrate the principles of an algorithm by illustrating its operations on a data structure Sometimes the animations don t relate to a real program at all but are simply animation scripts Of special interest is how the underlying program is connected to the output graphics which is often no more complicated than embedded calls to a specialized graphics library Tango and Polka Tango 95 is an algorithm animation system by J Stasko at the Georgia Institute of Technology that can create smooth animations using smooth path transitions between execution steps An imations are designed using trajectories and changes in size colour and visibility The source code is annotated to generate abstract data events and under execution the data events map to animation events to generate the animation Animation calls an
174. ile Edit Window Options Content Form Action Debug Help q q S a q a a g privateintto private int u0 private int arrays private int I0 privateint from private int pivot privateintl1 private int t void sortrange int array int from int to q sortrange array from 10 aaa private int ul Java method private static void QuickSort sortrange int array int from int to private intl array Figure 5 40 The variable is drag and dropped onto the statement to automatically insert a trace of the variable in front of the statement then its members are traced recursively but a hash table java util Hashtable stores the Objects that have already been encountered to prevent the trace routine from entering infinite loops in cyclic data structures Each time svt trace trace Event Name Expression is called it outputs Event eventno Name name TimeStamp time to the trace file where eventno is the event number Event name is the name of the event and the time is the time in milliseconds from the start of the program The next output is Expression An int is output IntegerValue value where value is the value of the integer A float or a double is output FloatValue value where value is the value of the float An array is output ArrayLength length contents Where length is the length of the array and contents is length values that make up the array A null object is output Object null and a java l
175. ilter add Extension String extension Public void JawaFile Filter set De serption String description Methods reachable from public void JavaToolbar actionPerformed ActionEvent evt public void JavaToolbar actionPerformed ActionEvent evt A Figure 5 25 Reachable methods Se Vmax Method calls within JavaToolbar iO XxX File Edit Window Options Content Form Action Debug a a void ia args void actionPerformed ActionEvent evt T T T JavaToolbar String a void getFile ActionEvent evt void showOutput a a void windowClosing WYindowEvent e void rung Method calls within JavaToolbar public void JavaToolbar actionPerformed ActionEvent evt A Figure 5 26 Method calls within a class 102 5 6 JAVA VISUALIZATION Ses Vmax References made by public void JavaToolbar action Performed Actio Miamis File Edit Window Options Content Form Action Debug ol ol private void JavaToolbar showOutputQ private void JavaToolbar getFile ActionEvent evt private JButton JavaToolbar oWWin private Objecto private Thread JavaToolbar runner a a private int JavaToolbar what private JButton JavaToolbar viewFile al a private JButton JavaToolbar runFile private JButton JavaToolbar compileFile private ActionEvent evt private JButton JavaToolbar selectFile References made by public void JavaToolbar actionPerformed ActionEvent evt public void JavaToolbar actionPerformed ActionEvent evt Figure 5 27
176. ing visual browsers of procedures and classes e Code navigation between code and GUI events Visual modification of the properties of some types of program object and resources Dialog box based skeleton code creation for components and applications There are only three visual notations used each supporting interaction e Direct representation of graphical resources forms menus dialog boxes and source code CHAPTER 2 PREVIOUS WORK 23 e Nested list representation for hierarchical structures such as procedures classes methods help and graphical resources e Table or dialog box representation for program object properties Here we see a success of visual languages in designing spatial things graphical user inter faces spatially These products do not however provide visual support for the programming itself and merely insert code templates which the programmers need to flesh out themselves While clearly useful there are only a very limited number of application views using only the nested list notation and are used just for code navigation But even these techniques offer greater ease in building large high quality applications Figure 2 4 shows a screen shot of Microsoft Visual Basic aa Microsoft Visual Basic Project ON Eie Edt view Insert Format Debug Run Tools Window Help ja m4 me oc e tale ThisDocument Private Sub Document C Tc rr E a m Document Requester Mi Document Requester
177. interface is done through the classes MainBase and View Base declared in the file svt h These are abtract base classes and the actual interface is imple mented in their derived classes An interface for Motif 1 2 has been implemented The File svi h GUI interface for SVT Written by Calum Grant Copyright C Calum Grant 1998 Hifndef SVT_H Hdefine SVT_H namespace SVT enum SVT_cursor SVT_ARROW SVT_ WAITING SVT_ZOOMIN SVT_ZOOMOUT SVT_POINT enum SVT_type SVT_MAIN SVT_THUMBNAIL SVT_ZOOM SVT_PLAN SVT_LEGEND SVT_POPUP class ViewBase public class View data public ViewBase virtual ViewBase void GL_draw void Drawing speed int void Input_event int input int d0 int dl int d2 void Key press int key void Mouse _move int x int y int shift int ctrl void Mouse pause int x int y void Mouse press int x int y int 1 int m int r int shift int ctrl void Mouse release int x int y int 1 int m int r int shift int ctrl void Main viewer void Reset _view void Editor insert int offset char text void Editor delete int offset int length void Editor update void Resize view void Refresh view 182 B 9 THE C INTERFACE void Menu_select int id int item void Popup select int void Duplicate view ViewBase void Reset viewpoint void Background process void Help These functions are called by SVT
178. ion Work is therefore needed to make SV easier to specify and implement The comparative difficulty of specification extends to layout interaction if possible at all and navigation which are all manually coded It makes it more difficult to change the graphical representation once the view has been implemented Intelligence and flexibility are required because a representation that is appropriate for 10 elements is unlikely to be appropriate for 100 000 AVE 34 and Rogowitz and Treinish 83 suggest ways to correct this Work is needed to improve the tailorability and flexibility of SV systems It can be very time consuming to set up visualizations and add new views to the point where the cost of setting up the view outweighs the insight gained from the visualization particularly if the insight is difficult to gain because the data is represented inappropriately Standard empirical evaluation methods for SV do not exist because the programming task is difficult to quantify Hence evidence of worth is lacking for many SV systems SV systems must be made easier and more automatic to use Scaling up systems to tackle large applications is a problem for many systems caused by lack of proper data management to handle large data sets effectively poor view granularity poor elision control poor automation and poor flexibility in graphical representation Hence most of the visualization systems presented here only display small quantities of data
179. ion const double virtual void On draw const di class Vcompound public Vobject public int p 2 Internal data Vobject First_child int amp const Vobject Next_child int amp const Vcompound int virtual Vcompound virtual void On_reset virtual void On_reset_size virtual void On set_size const double virtual void On set _position const double virtual void On draw const fa class Vassociation public Vcompound public int p 14 Internal data Vobject From const Vobject To const Vassociation int id int from int to virtual Vassociation virtual void On reset virtual void On structure J class Vattribute public Vnode public int p 13 Internal data Vobject Object const Vattribute int virtual Vattribute virtual void On reset class SVT_declare visual public SVT_declare visual virtual Vnode Construct_visual int 0 APPENDIX B DOCUMENTATION 181 SVT_declare visual next E char StringTerm int Convert string terms to char double FloatTerm int Convert float terms to double int IntTerm int Convert int float terms to int void VectorTerm int double Convert X Y Z terms to double endif B 9 2 The Graphical User Interface The functionality of SVT is completely independent of the type of graphical user interface All communication with the graphical user
180. ionPoint intString length scale iffinsertionPoint 0 return signum lt D D O intString else iflinsertionPoint gt 0 buf new StringBuffer intString buf insert insertionPoint if signum lt 0 buf insert else buf new Siringaufert insenionP oint intString length buf append signum lt O 0 0 for int O i lt insertionPoint i but append buf append intString return buftoString Java method public String java math BigDecimal toString public String java math Biglecimal toString i a Indented text Steal max Java method public String java math BigDecimal toString EIE File Edit Window Options Content Form Action Debug Help A if scale A return intvaltoStria OOOO StringBuffer buf String intString intval abs toString lint signum aa OOOO buf new StringBuffer 3 insertionPoint intString K buf append signum lt 0 0 0 Mo if sianum lt OD Affor intizO i lt insertionPoint i return buftoString Java method public String java math BigDecimal toString public String java math Biglecimal toString ai b Nassi Shneiderman Diagram 105 106 5 6 JAVA VISUALIZATION MES File Edit Window Options Content Form Action Debug Help Sess Vmax Java method public String java math BigDecimal toString return int ValtoString StringBuffer bu String intString intVal abs toString A buf new
181. irectly in the language Pre vious declarative approaches to SV do not provide a database to query Prolog is a good meta language for reasoning about source code Programming language rules can be expressed quite naturally in Prolog which allows program structures to be queried Specifications in Prolog are directly executable There is no need to implement a language interpreter because existing ones can be used Existing Prolog implementations are likely to be far more efficient than ad hoc declarative implementations Prolog has good theoretical foundations This allows formal reasoning using existing vi sual language theory Prolog is Turing powerful Prolog is already widely used in other applications Unlike imperative languages Prolog is a high level language good for knowledge engi neering Prolog has a built in mechanism for parsing text Its untyped execution allows heterogeneous data to be passed through the visualization pipeline Prolog is dynamic The code can be modified at run time for example to add new queries and change the visualization and action relations New code can be loaded at run time Prolog is introspective This means that it can query itself for example to show the visu alization relation in the legend This is impossible in imperative languages Any graphical output mechanism can be used As discussed in Section 6 5 8 these advantages are not necessarily peculiar to Prolog 7 4 2 Disadv
182. is method which works with both block structured and flow chart notations of the source code When a trace point is inserted the fact java trace point EventNo Expression Statement is added to Prolog s database where EventNo is a unique identifier Expression is the trace expression text and Statement is the identifier of the statement The text svt trace trace EventNo Expression Expression is inserted into the text buffer before the statement Statement 3 All the trace points in a Java file can be removed by selecting the Remove trace points option from the file s Action menu 5 7 2 Generating the Trace File After all of the trace points have been added the program is recompiled and executed When a java lang Object is passed to svt trace trace it uses the Reflection package java lang reflect to probe the contents of the object This allows run time querying of data types including field names types and values and is used to extract all of the information about the object for output to the trace file Unfortunately java lang reflect can only probe public members which is a necessary security feature If a compound object derived from java lang Object is encountered 3This could change the semantics of the program so ideally a more sophisticated method is required 114 5 7 ANALYZING RUN TIME DATA Sex Java method private static void QuickSort sortrange int array int from int to SE xX F
183. ization can be thought of as a mapping from the object to be visualized the view context to the image on the screen From a generic model of visualization this mapping can be decomposed A V G i ides into a series of relations ay gt and gt that model different stages in the visualization process These are formal mappings in first order logic The advantages of this are twofold Firstly a theoretical foundation has been laid that reveals certain properties of visualization such as the stages of visualization and the decomposition of graphical communication in terms of objects attributes and relationships and their correspon dence to a semantic network It is possible to prove what a user can and cannot deduce from a view and theoretical work on the types of visual entities also follows from this work Most importantly however are the practical implications It shows how knowledge based systems implemented in Prolog can be interfaced to a visualization system and suggests that Prolog s strengths in knowledge engineering make it a suitable front end for information visual ization Knowledge based queries can be formulated directly in Prolog to filter away unwanted data depending on the context or interests of the user Prolog provides a very concise specifica tion language for all stages of visualization All specifications are directly executable in Prolog As well as being completely generic this method is very flexibl
184. l Data 2 4 2 oe na se E AA e he 2 5 3 Specification Languages for SV o oo Information Visualization ee 26 11 Perception ira A tn a 2 6 2 Representations 4 ste a ge hoes RAE Bk ee BER 2 6 3 Architectires 2 4 5 4 94 a 64 eae a ad ES 2 6 4 Graphical Layout s e eien ici Poe Se Se Ee eos Res arch Directions s 2 sea oka eR Oe Oe OR ATE YEE ae Concliisions is 4 4 8 4 a tie tee Bost Rao eb ee Ra SS Soke da A Model of Information Visualization 3 1 3 2 3 3 3 4 3 5 Tntrod ction ss s i y Seater eS Vg eka deal eo eri RE ak Eee G Representing Data garne cia eh eo A Bae ae N Specifying View Content 0200000000008 Specifying Visual Content o 02 0000 00008 A Visual Type System 2 2 ee 3 6 3 7 3 8 3 9 Semantic Visualization Tool 4 1 4 2 4 3 4 4 4 5 4 6 4 7 4 8 A Tool for Software Visualization 5 1 5 2 3 3 5 4 5 3 5 6 Specifying Graphical Constraints Generating Views Specifying Interaction 3 8 1 Actions 3 8 2 Reactions Chapter Summary Introduction Architectures Sika ts Oa hme ee 8 Running SVT View Manipulation Generating Visualizations 4 5 1 Acquiring the Visual Primitives 4 5 2 Generating the Scene Graph 4 5 3 Structuring The Scene Graph 4 5 4 Graphical Layout 4 5 5 Graphical Layout Algorithms Expanding the Graphical Vocabulary 4 6 1 Specifying Visual Objects Interaction
185. larations and the class The predicate java expression reference Ex pr Context Variable returns variables Variable referenced by the expression and java express ion_calls Expr Context Method2 returns the method Method2 called by the given expres sion Even though Method2 may be overridden this predicate does not check this possibility java expression_type Expr Context Type returns the type of an expression java find declara tion Type Class gives the class Class of a type if it exists The predicate java method calls Me thod1 Method2 returns the methods Method2 that may be called by method Methodl java re achable Method1 Method2 Prev Visited returns all methods Method2 that are reachable from Method1 where Prev is the calling method of Method2 and Visited is a list of methods the search algorithm has been before to ensure termination The complete list of predicates for Java source code is given in Section B 3 2 These pred icates merely scratch the surface of algorithms that can provide meaningful information about a program Unfortunately there are no short cuts in their implementation but direct access to the program database and the ability to add new predicates that build on old ones mean that Prolog is a suitable meta language for programs Prolog is suited to analyzing Java source code because its data structures expressions types parse trees are all trees which can be represented straightforwardly as Prolog term
186. le File File is Prolog source code with suffix pl readable File Dynamic File File is readable root_directory Dir Not extensible Directory Dir is the root directory source_code File Not extensible File File contains source code This is determined by the filename extension svt_path PathName Not extensible Returns an atom PathName containing the value of VT PATH writable File Dynamic File File is writable B 3 2 Java Source Code These predicates are in the module java block Block Statements Dynamic Block is a block of statements Statements block_statement Statement Block Statements Dynamic Statement is a block Block of state ments Statements buffer File Buffer Dynamic The java file File has buffer Buffer break_statement Statement Dynamic Statement is a break statement break_statement Statement Label Dynamic Statement is a break statement to label Label cached_calls Method1 Method2 Dynamic This is a precomputed version of method calls 2 that stores that Method calls Method2 cached_reference Method Variable Dynamic This is a precomputed version of method refs 2 that stores that Method references the variable Variable case_statement Statement Expression Text Dynamic Statement is a case statement with case Expression with text node Text class Class Not extensible Class is a Java class class_member Class Member Not extensible Returns members Member which a
187. lete file X file X user action click X on off off 1 svt navigate_to X user action naviagate_to X view _context X _ _ L adds a single middle button click action a delete file reaction and specifies navigation be haviour for a left mouse click In the case where multiple actions are declared for an object all the actions are executed 4 8 The Legend Graphical communication can only be effective 1f the user is able to interpret the image Because the graphical output can be so varied even for the same data there is great potential for confusion with notation rendering the visualizations useless To help prevent this a legend is provided by SVT to explain all of the graphical notation in a view This is a window giving an example of each symbol and a textual description of what it denotes Examples are shown in Figure 4 24 S tem EE Symbol Meaning Symbol Meanin y g An exception catch Statements A directory dynfonts Etatement A declaration statement dynfonts Statement An expression statement lfthen else statement File is executable A T Method declarator A file LICENSE Sequential statements LICENSE hidden The age of a file A try catch block The size of a file Big a A variable private ActionEvent event private ActionEvent event Legend for File sizes a Legend for Nassi Shneiderman vertical J a A legend for a directory b A legend for a method Fig
188. lization contexts visualization context descriptions visuals visual components visual objects menus actions reactions and content types to declare This may seem very complicated but it provides much more flexibility and automation for the end user When specifying one aspect of the system the implementor does not need to be concerned about how the other aspects are specified In particular the view content is specified completely independently to the visual content which is specified independently to the graphical constraints of the visuals By subdividing the problem each component is much simpler to specify and can be reused The specification method SVT uses may be directly compared with Pavane 85 and Leonardo 20 Leonardo uses a logic based language Alpha but the resulting notation is more awkward than Prolog s and can only manage integer values not compound terms so it would be useless for knowledge engineering A strength of Alpha is that it can embed numerical expressions from the object language directly into it Pavane s declarative approach is more awkward than Alpha s The languages are very similar in their usage but Pavane offers an iterative style of programming and generates graphical ob jects as tuples in a data space Pavane uses named parameters to pass attributes to the graphical objects it constructs SVT uses parameter position to pass arguments to its visuals Pavane s im plementation actually tra
189. llcontainer True contain False e h f d fillcontainer False l1 line g a b hl l1 thickness 2 12 line g 10 0 al 12 thickness 2 13 line h 10 0 bl 13 thickness 2 1 A labeled dialog selection button visual _component select Id Text on select Id Text darkgrey lightgrey darkred visual component select Id Text off select Id Text lightgrey darkgrey grey object component select Id Text Top Bottom Face Id t text Text a b p4 10 2 c p3 p4 10 0 p2 p4 0 10 pl p2 10 0 il p1 2 2 12 p2 2 2 13 p3 2 2 Ti p r 2 2 face polygon il i2 i4 i3 face colour Face 11 polygon p1 p2 i2 i1 11 colour Top 12 polygon p1 p3 i3 i1 12 colour Top d1 polygon p3 p4 14 13 dl colour Bottom d2 polygon p2 p4 14 12 a2 colour Bottom 1 C 5 Interaction Interacting with the bookmarks action click _ on off off off 1 thumbnail retrieve view action click _ off off on off 1 thumbnail swap_menu reaction swap _ views thumbnail swap thumbnail view reaction retrieve view thumbnail retrieve thumbnail view reaction save_ view thumbnail save thumbnail view reaction swap menu thumbnail popup menu thumbnail reaction refresh thumbnail refresh view 188 C 5 INTERACTION C 5 1 Menus The bookmarks menu shown in Figure 4 28 action menu thumbna
190. lling parse_buffer Buffer Tree which returns the whole parse tree Tree as a large term conveying all of the text nodes in the parse tree for buffer Buffer A slight drawback of this method is that a data structure has been constructed for the entire parse tree which is slow Perhaps the parser should construct Prolog terms instead of text nodes Prolog can itself parse using definite clause grammars DCGs 75 that can replace the Yacc parser Parsing with a DCG is slower because it executes in Prolog and not C 84 5 4 DIRECTORY VISUALIZATION 5 4 Directory Visualization Vmax can visualize the computer s filing system It reads in directories off disk and stores the directory data in Prolog s knowledge database The predicate ensure directory loaded Path loads the given directory into the knowledge database It executes the command Is Lal and pipes the output into a text buffer This buffer is then parsed using a simple DCG that describes the format of a directory listing The DCG reads in the directory data and puts it into the database Each file and directory in the filing system is assigned a unique identifier Each file or directory also has a path which is represented as a list so that for example usr bin netscape would be represented as netscape bin usr The predicate path id Path Identifier determines the unique identifier Identifier for each path Path A complete listing of the filing system predicates is given i
191. lorer and AVS Express could be interfaced to SVT and provide access to both the data sources and the output display and scene graphs SVT and Vmax are not commercial products but demonstrate the technical viability of im plementing SV in a knowledge engineering environment 6 5 Further Work 6 5 1 Measure Usability of Vmax The usability of Vmax has not been formally measured and therefore the benefits of using it have not been proven Studies that compare the performance of users using a text only system and those using a graphical system such as Vmax would answer this question Unfortunately such broad studies are rarely conclusive and may not relate to real world tasks 45 Vmax would have to be evaluated in real situations to measure the performance of users using the system Vmax generates output messages monitoring its activity which can be saved into a log file for future analysis Changes in view context and visualization context are recorded and time stamped Navigation graphs between view contexts could reveal how users actually used the system rather than what is possible Usage statistics such as mean time between navigations number of view contexts used number of times the visualization relation is changed use of bookmarks use of preview window use of the legend or the way objects were visualized would provide evidence of worth for that feature It would provide a workbench for measuring the psychology of programm
192. ls retract 1 to remove the data from the database Updating the program database is necessary in an editing environment where the source code can change 5 5 2 Analyzing the Program Database Vmax adds a number of predicates for deriving various types of information about the program which augment the program database to provide information for visualization The rules used to implement these follow from the Java Language Specification 35 The predicate java direct_supertype 2 finds the immediate supertype of a class java supertype 2 finds any supertype of a class java direct_subtype 2 finds a class s immediate subtypes and java subtype 2 finds all subtypes of a class java direct_superinterface 2 finds the interfaces that a class implements while java superinterface 2 finds all interfaces that a class implements The predicate java class_member Class Member gives class members Member of a class including those inherited from supertypes java class_method Class Method gives the methods Method belonging to a class including those inherited from supertypes The predicates java is_abstract l java is_native l java is_private l java is_protected 1 and Java is_public l return information about the status of members and methods CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 91 The predicate java contains_statement Statement1 Statement2 returns all statements State ment2 that are contained within the possibly compound statement Sta
193. lti_container Vsize Vabove Vwindow Vstring_attrib Vbelow Vnode_link HH Vx_padding Vfurther Vany_node_link Vy_padding Vleftof Viable Vnearer Vvisual_object Vrightof Vcontains Vcorner Vbottom_left Vbottom_right Vtop_left Vtop_right Vedge Vfill_edge Vhier_edge Vlens_edge Vlist_edge Vexternal_anchor Vinternal_anchor Vlinear_scale Vorigin_scale Vmidpoint Vposition Vv_stack Vrelative Figure 4 9 SVT s complete class hierarchy of graphical constraints 60 4 5 GENERATING VISUALIZATIONS node s On_structure method These functions are declared virtual void On initialize virtual void On structure Additional structuring finds subgraphs connected by the Vedge association 4 5 4 Graphical Layout Each Vobject has the following virtual methods for graphical layout virtual void On reset _size virtual void On set_size double new_size virtual void On set _position double new position virtual void On _draw that are called in that order Each object determines its natural size expressed in pixel units as a 3 D bounding box On _reset_size sets the size of the visual object to its natural size and if it is a composite object it will first call On_reset_size of its child members The Vobject has pointers to all of its attributes and associations so that it can determine how
194. m in 3 D as convex as gridded as directed as undirected as hierarchical etcetera A comprehensive survey of graph theory and methods is given by Battista et al 6 2 7 Research Directions Both Myers 71 and Price et al 79 identify several deficiencies in current systems that should be addressed in future research The reason SV systems are not more successfully applied is because of these difficulties Both Price et al and Myers say that formal specification of SV and VP systems is limited and although theoretical work has been done in visual languages it has so far proved to be impractical for SV Without the foundations for specification and without formal data representation it is difficult to provide specification for scripting Scripting facilities for SV systems are extremely low level CHAPTER 2 PREVIOUS WORK 37 Relational properties Graphical relations strict continuous linear order a Mapping data to graphical structure abstract visualization constraint network generation b Architecture reproduced from 34 geometric layout Figure 2 15 The AVE visualization system often consisting of in line C code and the data representation and management is usually ad hoc This means that it is cumbersome to add views to systems so SV systems tend to have a very limited number of views The lack of views leads to fixed granularity poor elision control and poor or no navigat
195. m passed to the method Declare_visual and return a new instance of the specified class or return 0 if no match is found These handlers form a chain The instantiated visual is automatically added to the scene graph The C code is linked to the SVT object files to produce a new executable with the new constraint The File svt_visual h Defines classes for creating visual constraints in SVT Written by Calum Grant Copyright C Calum Grant 1998 1999 ifndef SVIVISUAL H define SVIVISUAL H namespace SVT class Vnode public int p 5 Internal data Vnode virtual Vnode virtual void On reset virtual void On structure class Vobject public Vnode public double d 14 Internal data int p 251 Internal data Structure query class Vassociation First_assoc int amp const class Vassociation Next_assoc int amp const class Vattribute First_attrib int amp const class Vattribute Next_attrib int amp const Bounding box manipulation void Set_size const double void Set_position const double const double Get _size const const double Get_min pos const void Get_max pos double const Vobject int virtual Vobject virtual void On reset virtual bool On attribute class Vattribute 180 B 9 THE C INTERFACE virtual void On_reset_size virtual void On set_size const double virtual void On set _posit
196. ment variable S VT_PATH e Xt options To customize the appearance of the user interface B 2 SVT Predicate Summary In addition to the built in Prolog predicates 59 SVT provides many predicates for visualization and interaction Only the most important predicates are listed here Vmax adds many more predicates for Java file system browsing and run time data structures Predicates are either e Extensible These predicates are declared as multifile and are intended to be augmented by user code to implement the visualization system e Not extensible These are predicates that provide functionality that would not normally be augmented by the user e Dynamic These extensible predicates are declared as dynamic and may be modified at run time to store data through the assert I and retract 1 predicates SICStus Prolog s module system is used to prevent name clashes Unless otherwise specified all SVT and Vmax predicates are in the module svt 165 166 B 2 SVT PREDICATE SUMMARY B 2 1 View Predicates create_viewer ViewContext VisualContext ActionContext ReactionContext Not extensible Creates a main viewing window with a text pane status bar and pull down menus The initial view shown in the main viewing area has view context ViewContext visualization context VisualContext action context ActionContext and reaction context ReactionCon text current_view_context ViewContext Not extensible Returns the view contex
197. mic Assigns the text buffer Buffer to entity Object create_file_buffer Filename Buffer TextNode Type Not extensible Creates a text buffer Buffer and loads the file Filename into it The text node of this file is TextNode and the buffer is given an associated type Type delete_text_node TextNode Not extensible Deletes the text in the text node TextNode from 1ts buffer insert_before TextNode Text Not extensible Inserts the string Text immediately before the text node TextNode in the text node s buffer lex_buffer Buffer Lexemes Not extensible Applies the buffer Buffer s lexical analyser and returns a list of lexemes Lexemes Each list member is a pointer to a text node lexeme_float TextNode Float Not extensible Takes the address of a text node TextNode and returns the floating point number Float in the text lexeme_integer TextNode Integer Not extensible Takes a text node TextNode and returns the integer Integer in the text lexeme_length TextNode Length Not extensible Returns the length Length of a text node TextNode lexeme_string TextNode String Not extensible Returns the contents of a text node TextNode as an atom lexeme string_list TextNode String Not extensible Returns the contents of a text node TextN ode as a string of chars APPENDIX B DOCUMENTATION 169 move_text_cursor_to TextNode Not extensible Moves the text cursor in the editing window to the given text node TextNode
198. mponent identifiers only have scope within the visual object Note that in the definition of shadow box colour 2 the anchors a b c and d are in the same name space as shadow _box 2 because they refer to the same object Figure 4 16 shows the name expansion for two different instances of a shadow box 2 with identifiers 123 and 312 123 a vc 123 1 anchor 123 b vc 321 1 anchor 321 b ESA E ee 1235 age he ead ea ere a Ad l 1242 i 1251 l I ve 23 p2 i vc 821 p2 A A 1237 e a 1246 nchor 123 d nchor 321 d vc 123 p1 1239 vc 321 p1 1248 Figure 4 16 Object component name expansion to prevent name clashes Anchor Expressions The appearance of a visual object is governed by the way its anchors are positioned Anchor coordinates cannot be absolute because they must move to accommodate child objects such as in the shadow_box 2 example and must move with the position of the visual object SVT implements a very general approach to positioning anchors by declaring anchors and imposing arbitrary constraints on them Such geometric constraints can be very difficult to solve and is a separate topic 58 Completely arbitrary constraints are in general insoluble These constraints are specified in anchor expressions in the object components contain Child a S 3 b 5 3 c 5 3 d 5 3 contains the anchor expressions a 5 3 b 5 3 c 5 3 and d 5 3 These anchor expressions specify the four corners of the
199. n usr file_timestamp File Y Mo D H Mi S Dynamic returns the creation date of the file File as year Y month Mo day of month D hour H minute Mi and second S filename_information Path Status Not extensible Returns the type of file Status of a file Path in the filing system This queries the operating system directly and returns O if the file does not exist 1 if it is a regular file 2 if it s a symbolic link and 3 if it s a directory get_list_from_path Filename List Not extensible Determines the file s path List from its full filename Filename For example usr bin netscape would return netscape bin usr header_file File Not extensible File File is a header file home_directory Dir Not extensible Directory Dir is the user s home directory home_directory path PathName Not extensible Returns an atom PathName giving the user s home directory HOME java_source_file File Not extensible File File is Java source code APPENDIX B DOCUMENTATION 171 make_file_fullname DirName Filename FullName Not extensible Returns the complete file path FullName for a file Filename in directory DirName makefile File Not extensible File File is a makefile as determined by its filename path_id Path Id Not extensible Returns a unique identifier Id for a given path Path picture_file File Not extensible File File is graphics as determined by its filename extension prolog_source File Not extensib
200. n Section B 3 1 Each file is stored as a file Identifier and each directory as directory Identifier If a file is a member of a directory then member Directory File stores that file File is a member of the directory Directory The contains 2 predicate gives the transitive closure of the member 2 predicate The file File s name Name is stored in filename File Name and its full path Path in file path File Path The flags readable File writable File and executable File store information about the status of the file file_fullname File FullName gives the full filename FullName of the file or directory and file_extension File Extension returns the file extension Extension e g txt of the file file_ barename File BareName gives the filename without its extension file_timestamp File TimeStamp gives the file s creation date TimeStamp and file age File Age gives the age Age of the file in seconds The above predicates provide enough information for a complete file browser The default directory view is shown in Figure 5 3 Y Vmax Directory fhomes camg100Aocal netscape r x File Edit Window Options Content Form Action Debug Help total 21591 a drwxr xr x 9 camgl00 user 1024 Aug 22 13 24 x druxr xr x 26 camgl00 daemon 1024 Aug 22 13 23 E 7 r r r 1 camg100 user 16154 May 28 00 09 LICEN bookmark htm dynfonts java lib libjsd so r r r 1 camgi00 user 556930 May 28 00 09 Netsc r r r 1 camg
201. national Conference on Visual Infor mation Systems 99 pages 785 792 Lecture Notes in Computer Science Springer Verlag 1999 25 C J DiGiano and R M Baecker Program auralization sound enhancements to the pro gramming environment In Proceedings of Graphics Interace 92 Vancouver Canada 11 15 May pages 44 52 1992 26 J Domingue B A Price and M Eisenstadt A framework for describing and implement ing software visualization systems In Proceedings of Graphics Interface 92 Vancouver Canada 13 15 May 1992 pages 53 60 Morgan Kaufmann Palo Alto CA 1992 27 J Domingue B A Price and M Eisenstadt Viz A framework for describing and im plementing software visualization systems In D J Gilmore R L Winder and F Dtienne editors User Centred Requirements for Software Engineering Environments pages 197 212 1994 28 J F Sowa ed Principles of Semantic Networks Explorations in the Representation of Knowledge Morgan Kaufmann 1991 29 S G Eick J L Steffen and E E Sumner SeeSoft a tool for visualizing line oriented software statistics IEEE Transactions on Software Engineering 18 11 957 968 1992 30 M Eisenstadt and M Brayshaw The Transparent Prolog Machine TPM an execution model and graphical debugger for logic programming Journal of Logic Programming 5 4 1 66 1988 31 J D Foley A D van Dam S K Feiner and J F Hughes Computer Graphics Principles and Practice
202. nces within class cached References within class Class references Declarations of class Home Figure 4 20 A Content menu used to change the information displayed about an object visual_context ViewContext VisualContext The Form menu on the menu bar changes the visualization context and is shown in Figure 4 21 The description for each menu item is declared by the predicate visual context description 2 and the menu is declared pulldown _menu form ViewContext Form current_view context ViewContext action menu form ViewContext Text svt set_vis context VisualContext visual_context ViewContext VisualContext visual_context_description VisualContext Text reaction set_vis context VisualContext svt new_visual_ context VisualContext Query Block structured Prolog English Hierarchy Figure 4 21 A Form menu used to change the visualization context 4 7 5 Modifying the Action Relation Interactive behaviour can be displayed and modified by selecting the Interaction view which has view context actions ActionContext for an action context ActionContext This allows a more user friendly specification of the action relation and provides information to the user about the interactive behaviour This view is shown in Figure 4 22 Selecting the Add action option on the Action menu adds a new entry to the table Clicking on the action text pops up a menu showing the different type
203. nd hence the number of references are preserved between the view content and the visual content Data that makes no references to other entities is called an entity data with a single reference is an attribute and data with two or more references is called a relationship The correspondence between content types and visual types is given in Table 3 1 Visual type entity visual object attribute visual attribute relationship visual association Table 3 1 The correspondence between information and its display Examples of visual objects include text shapes polygons pictograms and icons Examples of visual attributes include colours transparency font size changes in form and thickness Exam ples of visual associations include containment adjacency relative position connecting lines and layout This list is not exhaustive and is implementation defined These visuals form a graphical vocabulary with which the data in the view content is communicated to the user Any kind of data can be composed into a network of entities attributes and relationships The correspondence between the view content and the visual content is maintained because the type system ensures that entities in the view content must be mapped to visual objects in the visual content attributes map to visual attributes and relationships map to visual associations This cor respondence is informally observed by Green and Benyon 38 who decompose graphic displays in
204. ng Prolog and pre compiled Prolog files are dynamically loaded into the system to provide visualizations for the user Although SVT is completely generic its primary aim is to visualize the kind of data present in program databases and program run time information Because this data and its presentation is so varied a tool with maximum flexibility has been im plemented Vmax is a visualizing text editor built on top of SVT which provides graphical views of the program database and run time information and is described fully in Chapter 5 4 1 Introduction The theoretical model presented in Chapter 3 leaves much scope in its implementation This chapter presents one possible implementation to demonstrate the validity of the model and to provide a platform for software visualization which is the aim of this dissertation SVT provides the capability to script visualization in Prolog The actual visualization applica tion is written in Prolog and optionally also in C The architecture allows C code to be called from Prolog At start up SVT loads the Prolog files that contain the specifications for view re lations visualization relations interactive behaviour and rules for querying and processing the input data This chapter deals with the final stages in the visualization process given in Figure 3 1 It is assumed that the data gathering and the view and visualization relations have been specified so that SVT can simply issue a query to
205. nning program e Lack of smooth data animation 7 2 Implementing SV Systems The work has highlighted several techniques that could be used in SV systems The browser interface worked very well and seems like a good interface for data exploration The 3 D output could be used with good effect particularly for numerical data structures but was slightly less convincing for communicating program structures With modern hardware and graphics libraries there is little extra overhead in producing 3 D rather than 2 D output A wide range of graphical output is useful to tailor the output and enrich the graphical com munication A wide range of graphical layout algorithms is useful to deal with different types and structures of data If users need to specify more advanced layout control geometric constraints offer an approach Multiple views are already common and proved to be necessary to retrieve different types of data without cluttering the view Changing the view and visualization method from pull down menus provided a convenient way of doing that Graphical bookmarks and a preview window function well and are potentially useful A legend to describe each view would reduce possible confusion over notation These functions do not necessarily have to be implemented in Prolog For systems that do not require the flexibility and automation of Vmax or the ability to pose queries Prolog is not necessary In this case standard graphical output me
206. ns Sample Code E String operator Jlint nPos UINT uiLen Const Function Body Mill operator retascringRep AdjustSubstring nPos uilen Lengtb operator return String operator const char HnPos uilen pas paa Y How 13 gz xz s gan Perform default processing _Inherited DoDataExchange pDX perform specialized processing DDX_Text pDX ID Logfile m Logfile DEF Damien TY mm Tamfidod e Figure 2 5 Genitor s class editor SNiFF from Take Five Software has source code analyzers for C C Java Fortran Assembler Python Tcl Perl and Bash It provides project and documentation management views of class hierarchies class browsing and include browsing and source code navigation Projects in excess of a million lines of code can be analyzed In these systems the views are fixed the information they provide is limited their output vocabulary is limited and they have no support for run time information However they offer a great improvement over text only programming Tau Tau 70 is a collection of tools to navigate and profile C source code and incorporates class browsing and displays detailed run time profile information visually Views include e File and class display This provides list boxes to browse methods and classes in source code e Call graph extended display Shows the calls made between fun
207. ns tau S tau ccs S action trans A B T ccs A1 B action trans A T ccs Al action trans A B T ccs A B1 action trans B T ccs B1 action trans X S X ccs S action trans X S X ccs S action trans H1 S T ccs H2 S action trans H1 T ccs H2 action trans nu X S0 T ccs nu X S action trans S0 T ccs S restrict X T action trans B A S0 T ccs B A S action trans S0 T0 ccs S rename A B TO T action trans A0 B0 tau ccs A B action trans A0 X ccs A action trans B0 Y ccs B neg X Y APPENDIX C FURTHER EXAMPLES 189 action trans write X S tau ccs S write X neg X X neg X X restrict X X fail restrict X X fail restrictd oy rename A B A B rename A B A B rename _ _ A A reaction S0O ccs action trans S0 tau ccs S reaction S ccs reaction _ ccs Dispenser TeaDispenser CoffeeDispenser TeaDispenser t p p p tea TeaDispenser CoffeeDispenser c p p coffee CoffeeDispenser Personl c p p coffee drunk_ coffee Person2 t p coin coin coin coin tea drunk_tea M1 drunk coffee write Drunk the coffee 1 M1 M2 drunk_tea write Drunk the tea M2 reaction Dispenser Personl Personl Person2 M1 M2 ccs
208. ns a new identifier Id which is guaranteed not to clash with any other identifier uid 1 is implemented as a counter which returns a different integer each time it is called Representing entities by unique integers gives efficient storage and indexing Run time data is gathered by adding trace points to the Java source code which inserts a call to output a value to a trace file The Reflection package java lang reflect analyzes the structure of the value to output its contents Trace calls can be added to the source code simply by dragging a variable to the point in the source code where the trace should occur The generated trace file is read in and visualized Java source code can be visualized as a visual programming language either as a control flow diagram or as a Nassi Shneiderman Diagram 72 Even expressions can be visualized Prolog can also be visualized as a visual language and even be used to define views in a more user friendly manner Vmax provides 70 view contexts 133 visuals and includes 151 different views if the different visualization relations for each view context are counted These views can be extended by users Prolog as can the analysis routines to provide further useful information for the programmer 5 2 Architecture Vmax extends SVT s predicates with many of its own for software visualization Some of these are implemented in C There are predicates for lexical analysis using Lex parsing using Yacc 2 p
209. nslates Swarm based declarations into Prolog although the implementor does not see this Both Alpha and Pavane construct low level geometric objects by specifying their graphical coordinates directly which is little improvement over an ordinary graphics toolkit Alpha can assign integers to objects in the scene SVT works very differently by generating high level constraints on the output image which are automatically structured Constraints are necessary to combine different notations and to handle arbitrary graphical structures Every object in the scene has a label which can be any Prolog term Because Alpha and Pavane use directly mapped objects to the screen they provide no encap sulation of visual objects Even something as simple as a red arc has to be programmed in two separate predicates in Alpha In SVT constraints can be composited arbitrarily and are com positions of primitive graphical constraints providing encapsulation and specification at a much higher level While SVT provides three stages of visualization Pavane and Alpha provide just one thereby fixing their output 142 6 3 EVALUATING SVT Alpha and Pavane cannot specify interaction The objects in their scenes do not have seman tics so there is no means of interacting with them SVT can specify any interactive behaviour for any object or component in the scene The size of specification is small in SVT because SVT provides encapsulation via visuals so only the bij
210. ode Id A B red visual_content and_term Id A B tree tree _node Id amp A B orange visual content if term Id A B tree tree node ld lt A B red visual content term Id Text List tree tree _node Id Text List palegreen visual content variable term Id Text tree tree _node Id Text yellow o Prolog visual_context_description prolog Prolog visual_content tpm predicate Id H T prolog tpm predicate Id H T visual_content or_term Id A B prolog infix _text Id A B visual_content and_term Id A B prolog seq text Id A B visual_content if _term Id A B prolog pl_if Id A B visual_content term Id Text List prolog prolog text Id Text List visual content variable term Id Text prolog text Id Text C 3 Visuals A labeled edge in a hierarchy visual labeled_hier edge A B Text Colour A B str Text A colour hierarchical edge labeled Text Symbol container Symbol text p Symbol A text c Symbol B contains Symbol p Symbol labeled hier edge p Symbol c Symbol Text Colour 1 Colour red Colour blue Colour green visual component labeled hier edge A B Text Colour 186 C 4 VISUAL OBJECTS hier edge colour A B Colour edge label A B Text e A coloured if then else structure in a Nassi Shneiderman Diagram visual nsd_c
211. ogram B 1 1 Code B 1 1 1 Control Flow B 1 2 Data B 1 2 1 Data Flow B 2 Algorithm B 2 1 Instructions B 2 1 1 Control Flow B 2 2 Data B 2 2 1 Data Flow B 3 Fidelity and Completeness B 3 1 Invasiveness B 4 Data Gathering Time B 4 1 Temporal Control Mapping B 4 2 Visualization Generation Time C Form C 1 Medium C 2 Presentation Style C 2 1 Graphical Vocabulary C 2 1 1 Colour C 2 1 2 Dimensions C 2 2 Animation C 2 3 Sound C 3 Granularity C 3 1 Elision C 4 Multiple Views C 5 Program Synchronization D Method D 1 Visualization Specification Style D 1 1 Intelligence D 1 2 Tailorability D 1 2 1 Customization Language D 2 Connection Technique D 2 1 Code Ignorance Allowed D 2 2 System Code Coupling E Interaction E 1 Style E 2 Navigation E 2 1 Elision Control E 2 2 Temporal Control E 2 2 1 Direction E 2 2 2 Speed E 3 Scripting Facilities F Effectiveness F 1 Purpose F 2 Appropriateness and Clarity F 3 Empirical Evaluation F4 Production Use KKK KK any desktop PC Linux Java KKK any in language none kK kK OK kK 6 320 lines s 115 000 lines 19 500 lines s kk KKK kK KK KKK kk kk K kk kk kk k kk KKK compile and run time dynamic to static and dynamic post mortem colour monitor points lines polygons tables graphs position size enclosure pictograms plain text KKK KK kK KKK automatic KKK KK Kk KKK interactive or Prolog automatically annotated kk kk browser mouse menu
212. oloured_if Id C T F 1d C T F NSD coloured if then else Symbol nsd_coloured_if Symbol cond Symbol true Symbol false Symbol nsd_statement true Symbol if true nsd_statement false Symbol if false string cond Symbol condition visual_component nsd_coloured_if If Cond True False nsd_if If Cond True False nsd_if colour If C 4 Visual Objects A labeled interconnecting arrow visual jvl_assign Id Text L R Id L R str Text left arrow Sym string left Sym a string right Sym b jvl_assign Sym Text left Sym right Sym ip ae object_component jvl_assign Id Text L R Id text Text p q x1 5 0 y1 5 0 x midpoint c d y midpoint e f contain L a c b d contain R e g f h yalign c e yalign d f line x1 x 4 0 yl y 4 0 polygon x1 4 3 x1 x1 4 3 1 An if then else structure in a Nassi Shneiderman Diagram visual nsd_if ld C T F Id C T F NSD if then else Symbol nsd_if Symbol cond Symbol true Symbol false Symbol nsd_statement true Symbol if true nsd_statement false Symbol if false string cond Symbol Condition 1 object component nsd_if Object Condition True False Object APPENDIX C FURTHER EXAMPLES 187 cond contain Condition a 10 2 b 10 2 g 10 2 h 10 2 contain True g e c f fi
213. om ice gallery SlideThread com ice tartar JavaTouba avafie Fite Eomice 13 nvalidHeaderException com ice sqlclient ErvsgAdapter kom ice gallery_InputField j com ice gallery FileTreeNode J gom ice gallery FileTreeModel bom ice gallery FileTree Cell Renderer fom begallery ContactButton com oe gallery foot Diag com ice sqidient Ab outPand Class cross references JavaCanvas A Figure 5 22 Membership of classes within others EE max Declarations of class com ice sqiclient My SQLAdapter OS File Edit Window Options Content Form Action Debug Help com ice sqiclient TablesPanel com ce sglclient TableDefsPanel com ice sqiclient QueryTablePanel com ice sqlclient MySQLClient 0 public JDBCAdapter com ice sqlclient MySQLClient performTableSQLQuery String query throws SQLException T public JDBCAdapter com ice sqlclient MySQLClient getTableColurnninfo String tableName throws SQLException public JDBCAdapter com ice sqlclient MySQLClient getDatabaseTablelnfo throws SQLException com ice sglciient DatabasePanel Declarations of class com ice sqlclient MySQLAdapter public JDBCAdapter com ice sqlclient MySQLClient performTableSQLQuery String query throws SQLException ai Figure 5 23 Places where a class is used ox File Edit Window Options Content Form Action Debug Help Se Vmax Places where com ice tar InvalidHeaderException is thrown
214. on X man X view content family tree person X woman X if the predicates woman 1 man l and parent_of 2 are defined The view context is family tree which has in its view content the term mother X Y to denote that X is mother to Y father X Y to denote that X is a father to Y and person X to denote that X is a person ancestor_of 2 the transitive closure of parent_of 2 can be defined as ancestor of X Y parent_of X Y ancestor of X Y parent_of X Z ancestor of Z Y Then the view context descendents P containing just the descendents of P could be defined view_content descendents P person X P X ancestor of P X view_content descendents P mother X Y ancestor of P Y parent _of X Y woman X view content descendents P father X Y ancestor_of P Y parent_of X Y man X Example 7 The view relation view content wiring socket Socket Location socket Socket location Socket Location view content wiring wire S1 S2 Type wire S1 S2 Type 44 3 4 SPECIFYING VISUAL CONTENT has view context wiring with the view content socket Socket Location giving a socket with location and wire S1 S2 Type denoting a wire between two sockets of the specified type Example 8 The view relation sea_temperatures view content sea_temperatures temperature point X Y Temp sea temperature X Y Temp gives the sea bed
215. on click Object on off off off 1 svt navigate to ViewContext view_context Object ViewContext _ reaction navigate to ViewContext svt new_view_context ViewContext Another way to change the view context is through the Content menu The menu navi gate Object displays all the view contexts that can be used to visualize a particular object Object The user then selects a view context from this menu and the system navigates to it The Content menu is declared action menu navigate Object Description svt navigate to ViewContext view Object ViewContext Description action menu navigate _ seperator svt action menu navigate _ Home svt navigate _to home When the user presses the right hand mouse button on an object in the scene the Content menu is displayed to navigate to any view of that object The Content menu on the menu bar selects a different view of the current object A Content menu for a class is shown in Figure 4 20 The option Home goes to the initial view shown in Figure 4 3 4 7 4 Selecting Graphical Form The visualization contexts VisualContext that can be used with a view context ViewContext are declared 3 Additional checks are present to ensure that the given object has been correctly loaded CHAPTER 4 SEMANTIC VISUALIZATION TOOL 73 F Java class Subtypes of class Supertypes of class Method calls within class cached Method calls within class Refere
216. on relation from a pop up menu A bookmarks view provides thumbnails for storing views for later retrieval and a preview view automatically displays the object beneath the mouse cursor SVT provides the capability to specify a visualization environment at a high level and inter face it to a knowledge database SVT can only provide the scripting and flexibility because it uses Prolog to define the behaviour of the system By itself SVT does not provide any visualizations these must still be supplied by an application written in Prolog and optionally C Chapter 5 describes one such system that is used to visualize software Chapter 5 A Tool for Software Visualization This chapter describes Vmax a programmers text editor built on SVT that provides software visualization Vmax provides a browser window for visualizing the filing system Java source code run time data and can edit Prolog graphically Vmax deals with all levels of granularity from high level program structures to visualizing source code SVT s architecture allows arbitrary queries of a program database for very specialized views providing an information rich environment for the program mer The aim of implementing Vmax is to investigate how software visualization can be specified and scripted within the environment provided by SVT The system demon strates how to specify SV alternative views of the same data browsing a broad scope showing all levels of granulari
217. one of the most important facts about a SV system Myers 71 offers three different taxonomies for programming systems program visualiza tion systems and language specification technique He classifies programming systems using 22 2 4 SOFTWARE VISUALIZATION SYSTEMS three orthogonal criteria giving eight categories The criteria are visual vs textual program ming example based vs not example based and interpreted vs compiled The taxonomy for PV systems uses a 2x3 grid with static vs dynamic on one axis and code visualization data visualization and algorithm visualization on the other A programming system may belong to more than one category The taxonomy by specification technique lists 14 different methods of language specification and examples of visual languages belonging to each category Myers taxonomies are forerunners to Price et al s which is far more comprehensive and more applica ble to SV 2 4 Software Visualization Systems A range of SV systems are presented in this section Unfortunately a comprehensive survey is beyond the scope of this work There are hundreds of programming systems using visual techniques which are roughly divided into development tools utilizing visual techniques visual programming systems and software visualization systems 2 4 1 Development Tools Almost every professional development tool uses visual techniques to assist application develop ment These visual development tools
218. ong alternative to other declarative and imperative implementations However the added complexity has performance implications 7 1 Vmax Vmax is very different to use from other systems because of its user interface The popularity of the browser interface indicates that this might be a good choice for software browsing It is also very different because the number of views it provides is an order of magnitude greater than previous systems It is much more flexible and tailorable in its information content the graphical display changed from the Form menu or the Legend and its interactive behaviour Minimum user input is required to select the appropriate view But are these features useful Without direct empirical measurements it is not possible to be certain There is plenty of indirect evidence which suggests that alternative graphical rep resentations are necessary for different tasks and that the ability to filter out different kinds of information is necessary to avoid cluttering the display Graphical browsers are widely used and effective and without a legend views can be difficult to interpret Price et al 79 also cite prob lems with elision control that Vmax solves because of its query mechanism and granularity which Vmax solves by its generic view generation mechanism The classification in Table 6 1 summarizes the strengths and weaknesses of Vmax 7 1 1 Strengths of Vmax e Broad scope and content The data Vmax presen
219. onstraints can be specified in Prolog and implemented in C A simple local propagation solver is implemented in SVT The anchors form a network of values that are connected by constraints The basic constraint types are T Y C T gt Yy C my nz 4 3 where x y and z are floating point variables and c m and n are fixed values The constraint network is constructed for the three coordinate values in each anchor in the visual object Each anchor expression of the form A B x y z generates the constraints Ay Bi 2 Ay Bot y Az B3 2 where the suffix indicates the axis 1 2 or 3 Each constraint of the form xalign A B yalign A B or zalign A B generates one of the constraints 68 4 6 EXPANDING THE GRAPHICAL VOCABULARY Ay B Ay B Az B3 Each midpoint expression of the form A midpoint B C generates the constraints 1 1 1 1 1 1 A B As B Az B 1 1 36 223 at 362 3 3 3 363 Each containment component of the form contain Child A B C D where the dimensions of Child are x y z generates the constraints A gt Co y By gt Apts Ci Ai Di B By Ags C3 As Dz Cr Bz As D As Each containment component of the form contain Child A B C D E F G H where the dimen sions of Child are x y z generates the constraints A gt Coty Ci A Ei A Gi El Cs Az Fy A Go Cy Es gt Az 2 Gz Ex B gt Ai e2 Di B FA B H F B gt A Da C Fy E H G B As Dz A
220. ontal boxes Vertical boxes F Block clause Infix lt Infix Infix if Infix and Infix or Infix Vertical NSD if then NSD coloured if then horizontal seq Text if then Right stick Indexed right arrow Nothing A red text node amp A red text node A red text node lt c Change the visualization relation Bee vmar a euses O File Edit Window Options Content Form Action Debug Help M count elizabeth is mother to charles names is father to harry X is an ancestor of Y if anal is mother to harry Xis a parent of a if Xis an ancestor of A i Kis a parent of A if Xis mother to A Xis a parent of Y Pis father to Y a is father to y All Clauses d The new view THE PREVIEW WINDOW A ETE FEIE Symbol Meaning AB Conjunction ade Condition al Disjunction a A term x Prolog variable Legend for Block structured i b The initial legend 2 Legend ES Symbol Meaning AB Conjunction a i Condition AI Disjunction mada A term A Prolog variable Legend for Block structured e The new legend Figure 4 26 Changing the visualization relation from the legend CHAPTER 4 SEMANTIC VISUALIZATION TOOL 79 Welcome to Vinax Figure 4 27 A row of bookmarks Bookmark Retrieve Swap Refresh Figure 4 28 The Bookmarks menu action linger Obj svt zoom context ViewContext view Obj ViewContext
221. or data representation First of all it is a very high level language for expressing relationships and inferences between data and can express high level rules very naturally Prolog is a good meta language so it is suitable for representing and reasoning about computer languages which are entirely rule based The work described in Chapter 5 shows that all aspects of computer languages can be modeled in Prolog A Prolog program constitutes a set of logical inferences in first order logic In other words Prolog is itself a formal notation which removes the need to develop further notation in this chap ter Formal treatment of Prolog and first order logic is well established and is treated elsewhere 80 A common application of Prolog is in knowledge engineering and automated reasoning 62 Visualization has potential uses for displaying the results of automated reasoning while auto mated reasoning adds flexibility to derive new knowledge implicit knowledge for presentation in a visualization system Information visualization is the presentation of knowledge so it is natural to integrate knowledge engineering into visualization Prolog is Turing powerful so it can compute everything that a language such as C can although the language has different applications Prolog can perform any perform any kind of data manipulation The following examples illustrate Prolog s approach to data representation Example 1 Suppose sea bed data is visu
222. or the process to terminate before returning timer_start Not extensible Resets the internal timer timer_stop Time Not extensible Returns the elapsed time Time in seconds since timer st art 0 uid UID Not extensible Returns a unique identifier UID This is extremely useful when dy namically declaring new entities with unique names The returned UID is an integer so that it can be indexed efficiently uid 1 is implemented as a counter user_arguments List Not extensible Returns the list List of arguments passed to SVT when it was invoked B 3 Vmax Predicate Summary B 3 1 Filing System compressed File Not extensible File File is compressed directory File Dynamic File is a directory ensure_path_loaded Path Not extensible Reads in the directory specified by Path executable File Dynamic File is an executable file file File Dynamic File is a file file_age File Age Not extensible Returns the age Age in seconds of a file File file_barename File BareName Not extensible Returns the filename BareName without its file extension for a file File file_extension File Extension Not extensible Returns the filename extension Extension for a file File file_fullname File Name Not extensible File File has full name Name for example usr bin netscape file_path File Path Dynamic File File has path Path which is a list of directories For exam ple usr bin netscape would have path netscape bi
223. ould not come at the expense of usability Automation must make sure that the increased power and flexibility is accessible without unduly impeding the user Intelligence is required to anticipate the correct data and graphics to output and to make views adapt to the data and search criteria Interaction with the views and the application could manipulate data and provide direct means of querying and navigation Specification through graphical and interactive techniques would tailor visual ization solutions with minimum effort Finally integration of the different types of SV with other development tools would realize these potential benefits Vmax goes some way to tackling these issues but much further work discussed in Section 6 5 is needed SV based on knowledge engineering in Prolog provides a strong alternative to existing approaches but can only complement rather than supersede existing work Bibliography 1 Information techology Programming languages Prolog Part 1 General core ISO IEC 133211 1 1995 1ISO TEC 1995 2 A V Aho R Sethi and J D Ullman Compilers Principles Techniques and Tools Ad dison Wesley 1986 3 R M Baecker Sorting out sorting In ACM SIGGRAPH 81 and excerpted in ACM SIGGRAPH Video Review No 7 1981 30 minutes 4 R M Baecker and A Marcus Human Factors and Typeography for More Readable Pro grams Addison Wesley Reading Massachusetts 1990 5 M J Baker and S G Eick S
224. out appenednt sting EST Java method public String java math BigDecimal toStringO public String Java BleDectnal toStrinat public String Java nath Fietsrinal toString c EE Vmax Java method public String java math BigDecimal to Strine ax File Edit Udu Options Content Fern Action Tebus Help String toSting epee return iniValioBtina SningBurer buf Sinna inStrng iva aksit Sting int signum siorumi int inser on Powt in Sinng lengik scale inserionPovnt 0 1 ele insertianPcint gt 0 r new Stina iste uf insor inecttionP ait 1 a Puf new StingBuffere neetiorPont imString length buf appendisignumet 0 0 Prue ot Pont 7 buf agpend 0 Tut append intSinn Tetun bo Sina Java method public String java math BigDecimal toString SETIEN tam sonar 1 7 E d ax File Edit Uid Options Content Form Action Debug Help Seeng Tetum inval testinal EME String intSting inval abaia Siral nt sirum signum jnl nsertionP oint intStrina lenat scale 7G oium stand DTD UE else FinsedionPant gt 0 buf npa Snes buf nseri neortionPort fisignum lt 0 buf now soporte Sena length buf spoendsienum D 0 0 9 fered a bafappend 0 Buf aocondintStuna tun bultos ring
225. pace filling software visualization Journal of Visual Lan guages and Computing 6 119 133 1995 6 G D Battista P Eades R Tamassia and I G Tollis Algorithms for drawing graphs an annotated bibliography Computational Geometry Theory and Applications 4 235 282 1994 7 N J Belkin and W B Croft Information filtering and information retrieval Two sides of the same coin Communications of the ACM 35 29 38 1992 8 J L Bentley and B W Kernigham A system for algorithm animation Computing Systems 4 1 5 30 1991 9 J Bertin Graphics and graphic information processing de Gruyter Berlin New York 1981 10 A F Blackwell Metacognitive theories of visual programming What do we think we are doing In Proceedings 1996 IEEE Symposium on Visual Languages September 3 6 1996 Boulder Colorado pages 240 246 IEEE Computer Society Press 1996 11 A Borning K Marriott P Stuckey and Y Xiao Solving linear arithmetic constraints for user interface applications In Proc 1997 ACM Symposium on User Interface Software and Technology pages 87 96 October 1997 12 M Brayshaw and M Eisenstadt A practical graphical tracer for Prolog International Journal of Man Machine Studies 35 597 631 1991 13 FP Brooks The Mythical Man Month Addison Wesley 1975 14 FP Brooks No silver bullet IEEE Computer 20 4 10 19 April 1987 15 M H Brown Zeus a system for algorthim animation with mult
226. pe Type Type may be one of case Label default else for for_exit sequence then while while exit statement_text Statement Text Dynamic Java statement Statement has text Text static_initializer Id Block Statements Dynamic Static initializer Id has a block Block of statements Statements switch_statement Statement Condition CText Block Statements Dynamic Statement is a switch statement with condition expression Condition condition text node CText and a block Block of statements Statements sub_expression Expression1 Expression2 Not extensible Finds a sub expression Expres sion2 of an expression Expression sub_expression_decl Statement Decls Expr NewDecls Not extensible Finds the sub express ion Expr nested within a statement Statement and returns the list NewDecls of declara tions that apply to that sub expression subinterface Interface Class Not extensible Class or interface Class directly or indirectly im plements interface Interface subtype Supertype Subtype Not extensible The transitive closure of direct subtype 2 Java class Subtype is a subtype of Supertype superinterface Class Interface Not extensible Class or interface Class directly or indirectly implements interface Interface supertype Subtype Supertype Not extensible Java class Supertype is a supertype of Subtype APPENDIX B DOCUMENTATION 175 synchronized_statement Statement Expression Text Body Dynamic St
227. ptions Content Form fection Debug Help String toString AEEA SETENTA Siig ma Sure nitro Valsad OSE lint Sionum signum ret umn intval toString ini insertionPoint in Sinng lenglh scale StringBuffer buf fiipsedionP ant String intString int Val abaf toString esa int insertionPoint intString ler um rara 0 insarisoP cin gt 0 buf new Sosa but incor insarionPoint balines al but rew Stingeuterintstangh Jour i Sta neta rum signum lt O 200 3 0 inSting EU rew SA sien ASA Ena TER EMANAN buf append buf enpend ntString Tati bur toString Java method public String java math BigDecimal toString public String Javemath leDectnal toStrinat Java method public String java math BigDecimal toString public String Java ath Fiete gt inal toStrine i j The final view EE Legend ox Symbol Meaning A declaration statement alement An expression statement iftrue N it l Ifthen else statement a Method declarator Bo Statement Exit statement Statement 1 Statement 2 Sequential statements Statement 3 Legend for Text k The final legend Figure 5 32 A textual representation of code is transformed into a visual one by selecting different graphical forms in the legend int sync Demo Object p
228. ptions of the declaration declaration Declaration ClassType Modifiers Extends Interfaces ClassHeader Dynamic Declaration is a declaration of a ClassType which is either class or interface Its modifiers modifiers is the term are modifiers Abstract Final Access Static Transient Volatile Native Synchronized Extends is the name of the class it extends and Interfaces is a list of classes it interfaces ClassHeader is the text node of the header of the class declaration_statement Statement Decls Dynamic Statement is a declaration statement declar ing the list of declarations Decls default_statement Statement Dynamic Statement is a default label direct_subinterface Class Interface Not extensible Class or interface Class directly imple ments Interface direct_subtype Supertype Subtype Not extensible Class Subtype extends class Supertype direct_superinterface Interface Class Not extensible Class or interface Class directly imple ments Interface direct_supertype Subtype Supertype Not extensible Class Subtype extends class Supertype do_while_statement Statement Condition CText Body Dynamic Statement is a do while statement with condition expression Condition condition text node CTerm and statement body Body empty_statement Statement Dynamic Statement is an empty statement expression_calls Expression Context Method Not extensible The expression Expression calls the method Method
229. r VIS AD also provides a declarative specification for mapping data to the screen but is re stricted to simple numerical data structures In spite of their limited success declarative specification languages offer significant advan tages over imperative programming methods 20 For example they offer a higher degree of abstraction between the data and the graphics the languages are specialized to the task and are potentially easier to use 2 6 Information Visualization The wider field of information visualization looks at techniques for expressing all types of in formation in a visual form Visualization has proved an extremely important use of computers where the information they store needs to be output and often visualization communicates in formation far more effectively that textual data Visualization has applications in virtually every field including business science finance management databases information retrieval and software engineering CHAPTER 2 PREVIOUS WORK 35 2 6 1 Perception The study of visualization examines what information to visualize what visual representations to use i e what visual languages to use and how the visualization is created The study of choosing visual representation lies largely in psychology which regards visualization as a mental process and performs experiments on human subjects to find metrics for human responses and the effectiveness of various types of notation Creating
230. r int i 0 i lt insertionPoint i by buf append lt 0 gt buf append intString fetum signum 8 0 0 intSting return buf toString Java method public String java math BigDecimal toString 5 private String intString d Figure 5 30 The default view of a method The view also contains all variables that the method references and all methods that may be called providing easy navigation to them Different visualization relations for this view can give different forms of the code and Figure 5 31 shows this and in particular the effect of adding colour to a Nassi Shneiderman Diagram The graphical structure of these diagrams is specified entirely using linear constraints described in Section 4 6 1 It is even possible to modify the visualization relation in the Legend to turn one of these views into any other and Figure 5 32 shows how text can be transformed into a coloured Nassi Shneiderman Diagram At each step a different notation has been selected in the Legend thus transforming the view There are some block constructs that Nassi and Shneiderman could not anticipate in 1973 A synchronized block is shown in Figure 5 33 and a try catch finally block is shown in Figure 5 34 There are many other possible notations Switch statements with many options tend to make the diagram very wide and this is shown in Figure 5 35 a A modification to this notation to rectify this is shown in Fi
231. r 1 Introduction Computer software has revolutionized the way we work and the software industry is worth trillions of dollars world wide Software systems are the most complex artificial objects ever constructed and managing their complexity is difficult Software is also very difficult to write understand and debug In response software tools are rapidly evolving to create software more quickly and reliably One approach to the problems of comprehension and complexity is through computer graphics and visualization This is software visualization SV which encompasses all applications of graphics to software development including static program visualization visual programming manipulating graphical representations of code and visualization of run time behaviour The rationale for SV is that programmers and managers need to understand a software system often long after it was written for debugging and updating It is widely accepted that code maintenance is the most expensive part of software engineering 94 Even small algorithms can be difficult to understand debug or communicate to students Computer graphics offers a powerful means of communicating that information through visualization The Year 2000 Bug highlights the need for SV A programmer trying to debug poorly doc umented legacy code would wish to know not just the overall structure of the program but the dependencies of the various functions in the system which varia
232. r contains Self Child contains Self Child A basic directory tree is shown in Figure 5 7 a Directory trees can also show files as shown in Figure 5 7 b and the hierarchy can be represented in different ways as shown by Figure 5 7 Drag and drop has been implemented to move files and directories CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 87 Jrectory tree under Momes camg100 Work SVT bin samples Javaicomiice E Vax Files under shomes camg1 00 Work SVT in samples Javacomiice File Edit Window Options Content Form Action Debug He File Edit Window Options Content Form Action Directory tree under homes camg100 Work SVT bin samples Java comfice Files under homes camg100 Work SVT bin samples Java com ice hones camg100 Hork SWT bin samples Java con ice tar hones camg100 Hork SWT bin samples Java com ice a Fish eye directories b Fish eye files EE a Directory tree under Momes camg100 Work SVTbin samples Java co Bal Directory tree under Momes camgl00 Work SVTmin samples Java com AN doc Directory tree under homes camg100 Work SVT bin samples Java com ice Directory tree under homes camg100AWork SVT bin samples Java com ice Files under homes camg100 Work SVT bin samples Java comfice hones camg100 Hork SWT bin samples Java com ice hones camg100 Hork SWT bin samples Java com ice w e Contained directories f Contained files Fi
233. r has moved over an object Object with mouse buttons Buttons 4 7 1 Disambiguating Actions It will often be the case that there is more than one object beneath the mouse cursor perhaps because the objects are nested or because the line of sight beneath the mouse cursor intersects several objects These two cases are illustrated in Figure 4 19 Figure 4 19 Ambiguous mouse input The mouse cursor indicates more than one object so the system must decide which the user intended The rule for selecting actions is to allow only one visual object to be subject to interaction so the user s input generates just a single action term When objects are nested the innermost object in the nesting is given precedence The action 3 predicate is queried to determine if an action exists for that object If not its containing object is queried until an object with an action is found In the 3 D case the mouse position will indicate a line intersecting the scene rather than just a single point as in the 2 D case The camera position indicates the view frustum and is stored as a position vector and three orthogonal unit vectors forming a coordinate frame for the viewer The mouse position is indicated by a vector perpendicular to the plane of the viewer extending along the z axis of the viewer This line is transformed from the viewer s coordinate frame to the scene graph s coordinate frame and objects in the scene graph are tested for inters
234. raint MidpointConstraint FurtherConstraint Figure 4 18 A class hierarchy to implement constraint satisfaction by value propagation Initially all constraint values are set to NaN a special floating point value used to indicate that the value has not been set Then an anchor in the visual object is selected arbitrarily and the SatisfyConstraints method is called for each of its three axes SatisfyConstraints applies each constraint attached to the value in turn If the values attached to that constraint are changed then SatisfyConstraints is called for the changed value In this way the network of constraints is traversed until every constraint is satisfied Each constraint ensures that its values are monotonically increasing which prevents instability of the algorithm and a tolerance must be used to account for numerical rounding errors It is possible for constraints to be unsatisfiable such as y x 3 and a y 4 This algorithm will not detect such inconsistencies but a maximum iteration count can ensure that the algorithm terminates Each time a constraint is applied a counter is decremented and if it reaches zero the algorithm terminates a warning message is printed and the current incomplete solution is used There are many more advanced types of constraint system 47 but even these simple con straints can express very varied visual languages The architecture of SVT would allow more advanced geom
235. ram database 5 5 and the run time data 85 7 B 1 If the existing views prove to be deficient a new view can be defined to augment the system It provides code visualization at the statement level 85 6 3 B 1 1 The control flow between statements and between methods is shown completely 85 6 3 85 6 2 B 1 1 1 Vmax has complete access to the run time data displayable in a wide variety of formats B 1 2 Vmax is weak on data flow because Java is a control orientated language The data flow within a method is not shown although the information is available to compute it The data dependencies between methods and variables are shown however 5 6 2 B 1 2 1 Vmax cannot visualize algorithms because it only has access to raw source code However it can be structured into a flow chart which would help visualize the algorithm 85 6 3 B 2 Expressions in statements can also be visualized if that is really desired 5 6 3 B 2 1 The control flow is shown within a method and also between methods However the actual control flow of an executing program is not displayable Showing run time control flow is not possible because analysis of run time data is done off line 5 7 A possibility would be to interface Prolog to the Java Virtual Machine Debug Interface 68 to perform live debugging B 2 1 1 Prolog has complete access to the Java s parse tree and can compute any information about the program because it is Turing powerf
236. re members of the class Class This includes members inherited from supertypes class_method Class Method Not extensible Returns methods Method which are methods in the class Class This includes methods inherited from supertypes class_path Dir Not extensible Returns the directory Dir of the Java source code May return multiple solutions on backtracking conditional_statement Statement Condition CText Then Dynamic Statement is an if then statement with condition expression Condition text node of the condition CText and body statement Then conditional_statement Statement Condition CText Then Else Dynamic Statement is an if then else statement with condition expression Condition text node of the condition CText which takes the statement Then if true and False if false 172 B 3 VMAX PREDICATE SUMMARY constructor Method Not extensible Indicates that Method is a constructor contains_statement Statement1 Statement2 Not extensible Statement Statement1 contains statement Statement2 for example within a block continue_statement Statement Dynamic Statement is a continue statement continue_statement Statement Label Dynamic Statement is a continue statement to label Label declaration Declaration Not extensible Declaration is a Java declaration either a class or an interface member 2 gives the members of the declaration and identifier 2 short identifier 2 and long_identifier 2 return descri
237. res favourably to existing SV systems in many taxonometric criteria including automation scope information content graphical output form specifica tion tailorability navigation granularity and elision control The performance and scalability of the new approach is very reasonable We conclude that Prolog provides a formal and high level specification language that is suitable for specifying all aspects of a SV system Contents 1 Introduction 3 1 1 1 2 1 3 1 4 1 5 1 6 Restless e ee AA Contribution 2 kc ee RE ERE Ae GG eR ae Ee eA Notation and Typography 2 2 0 e Dissertation OVErvieW e ca oso aR nana le ol AAA A eo ge Je ks Previous Work 2 1 2 2 2 3 2 4 2 5 2 6 2 7 2 8 Basic Definitions i saisi ea OLRM Ae ee Pe be ee ed Benefits of Software Visualization 2 0 2 2 0 eee ee eee Taxonomies of Software Visualization 002002 ee eee Software Visualization Systems o 0 0000 000 2 4 1 Development Tools 2 0 002 eee eee eee 2 4 2 Visual Programming Systems 0 0 2 4 3 Code Visualization Systems 000 2 4 4 Algorithm Animation Systems 2 4 5 Run time Visualization Systems 0 0 224 6 ATChiteCtires jf a o a oe A ce CR Ae She Beet Wed wea RA Specification a 4 4 4 2406 2 ASD das AA wR A 2 5 1 Visual Language Theory o o e 2 59 22 Numerica
238. resentation Abstraction Structural Represenation Synthesised Rep Predefinition Annotation Specification Declaration Manipulation Program Visualization Simple Objects Composite Objects Visual Events Multiple Worlds Graphical Vocabulary Through Controls Interaction Through the Image Interpretation of graphics Analytical Presentation Presentation Explanatory Presentation Orchestration Figure 2 3 Roman and Cox s taxonomy of program visualization Roman and Cox s Scope category contains Price s Scope and Content categories By Abstraction Roman and Cox mean the degree to which the code has been transformed This overlaps Price s Content and Form categories Roman and Cox s Specification category is contained within Price s Method category Roman and Cox s Interface category describes both rendering and interaction so contains Price s Form and Interaction categories Roman and Cox s Presentation category denotes the semantics of the visualization and is contained within Price s Form category Roman and Cox s taxonomy is awkward here because it sep arates Graphical vocabulary and Presentation Roman and Cox do not consider Effective ness and purpose in their taxonomy when this is surely
239. ress c Delete action a The Actions menu a Nothing Display identifier in status bar Display short identifier in status bar Navigate to the object Display in the zoom view Hand mouse cursor The action menu The navigate menu Navigate to previous view The action menu The navigate menu spider cursor Arrow mouse cursor b The Reactions menu Figure 4 23 The pop up menus to change interactive behaviour action menu Object Action ActionText where Object is the object being acted upon Action is the action term and ActionText is the text appearing in the actions menu Object could also be a pair of objects such as A B for a drag action Reactions are declared in the Reactions menu by reaction menu Object Reaction ReactionText where Reaction is the reaction term and ReactionText is the text that appears in the menu The condition on the action is declared by user_action Reaction The dynamic predicate user action Action ActionContext Reaction stores this mapping for the action context ActionContext The user defined actions are imple mented by the clause CHAPTER 4 SEMANTIC VISUALIZATION TOOL 75 action Action ActionContext Reaction user_action Action ActionContext Reaction user_action Reaction For example action menu X click X off on off 1 Single middle button click reaction menu X delete file X Delete file user _action de
240. ry Oxford University Press 1980 101 E Tufte Visual Display of Quantitative Information Cheshire Press Conneticut 1983 102 J D Ullman Implementation of logic query languages for databases ACM Transactions on Database Systems pages 289 321 1985 103 J van Leeuven ed Handbook of Theoretical Computer Science vol 2 Elsevier Ams terdam New York 1990 104 G M Vose and G Williams LabVIEW Laboratory virtual instrument engineering work bench Byte pages 84 92 September 1986 105 C Ware and G Franck Viewing a graph in a virtual reality is three times as good as a 2D diagram In Proceedings 1996 IEEE Symposium on Visual Languages September 3 6 Boulder Colorado pages 182 183 IEEE Computer Society Press 1994 106 J Werneke The Inventor Mentor programming object oriented 3D graphics Addison Wesley 1994 107 K N Whitley Visual programming languages and the empirical evidence for and against Journal of Visual Languages and Computing 8 1 109 142 1997 Appendix A Definitions This appendix contains mathematical definitions that have not been included else where Definition 16 Atoms Let A be the set of atoms An atom is an indivisible unit of information such as a name or a number and the set A is infinite For example a b c charles harry 3 and 3 14 are all atoms Definition 17 Terms The set of terms T is the smallest set closed under ac A fEA ti tn ET n gt 1 aET
241. s Prolog has its own internal database which acts as the program database and because Java s language rules can be expressed as Prolog clauses 5 6 Java Visualization 5 6 1 Visualizing Files Classes and Packages Clicking on Java source code file will load and navigate to that file The default view for a file shows the main class or interface in the file shown in Figure 5 9 Another view of a file shows the overall structure of the file shown in Figure 5 10 which puts the whole text file in the editing window Each of the items in Figure 5 10 can be navigated to so for example clicking on the package navigates to the package clicking on the import declarations navigates to the classes indicated and clicking on a class navigates to that class The class view shows the class members and methods excluding those inherited from su pertypes and interfaces and includes information about the members such as whether they are public native or abstract Each symbol in the view can be navigated to The sizes of the methods are indicated The text of the class appears in the editing window This view is declared view Decl java_decl Decl Class contents java declaration Decl 92 5 6 JAVA VISUALIZATION Ef Vmax Java class com ice gallery AboutDialog 8 ho AboutDialog Frame parent void actionPerformed ActionEvent evt o p void establishDialogContents String RCS_ID String RCS_REV String messageStrin
242. s containment and 138 6 1 CLASSIFYING VMAX height bars 5 8 1 C 2 1 2 The visualizations are static and animation can only be shown as a sequence of frames 5 8 Animation constraints could be added to the viewer so that the output in the view is animated A modification to the view generator could automatically generate smooth paths between objects with the same identifier in sequential frames More primitive approaches such as Tango 95 are more versatile because they allow animation to be programmed explicitly C 2 2 Sound was not incorporated into Vmax because there have been no demonstrated benefits of using sound in program visualization 25 Sound constraints could be provided that are generated from the information content of a view or Prolog could call the sound library directly Zeus 15 is stronger in this area because it has an interface to a sound library C 2 3 Vmax s control over granularity is far more extensive than in previous systems A range of granularity from individual expressions 5 6 3 to whole class 5 6 1 and package hierarchies 5 4 can be viewed while other systems are often limited to one task and one level of granu larity 79 C 3 Vmax s elision control is also very advanced because different views of the same object can hide or reveal data 4 7 4 and the legend can be modified to hide or reveal information 4 8 2 Elision is selected by the user In many other systems eli
243. s direct manipulation Kk KK Kk KKK KKK Kk KKK KK KKK expert software development kkx xx x Table 6 1 The classification of Vmax in Price Baecker and Smalls taxonomy 79 CHAPTER 6 DISCUSSION 137 and any type of run time data 5 7 Vmax can be used for any visualization task A 1 4 Vmax has no speciality because of its broad scope A 1 4 1 Vmax is very scalable because its query mechanism filters away extraneous data that can clutter the views and if views fill up their contents are shrunk 34 4 The limitations are due to the size of the program database This can fill up the physical memory and so the performance of Prolog degrades 86 2 2 A 2 Vmax can read 6 320 lines of Java per second as measured on a 400MHz Pentium PC with 128MB of memory With this setup a project of 115 000 lines of code has been visualized 6 2 2 Vmax is inadequate for projects of several million lines of code due mainly to the internal representation of the program database and the comparatively slow execution speed of the Prolog analyzer Program analysis in C or C would be much faster A 2 1 Similar constraints are placed on the run time data 19 500 lines of run time data are read per second One line of run time data is equivalent to one integer 6 2 2 A 2 2 6 1 2 Content The information content of the views is extensive due to the query mechanism that Prolog pro vides 3 3 and the completeness of the prog
244. s FP Ez H Ez The constraint network in the shadow_box 2 example is shown in Figure 4 17 Each anchor has been assigned a letter and the number of constraints between anchors is shown Figure 4 17 The constraint network for the shadow box The object component contain Child a 5 3 b 5 3 c 5 3 d 5 3 generates the following constraints for the anchor expressions e a 5 fi b 5 Y ats hy dy 5 e a2 3 fo b2 3 go 3 ha d2 3 e3 as 0 fz b3 0 g3 c3 0 hg d3 0 and the following constraints for the container e gt goty fi gt ete q h fi fr 92 fi 3 hg 3 2 93 83 ha CHAPTER 4 SEMANTIC VISUALIZATION TOOL 69 where the dimensions of the contained object are x y z The remaining anchor expressions in the definition of shadow_box 2 generate the following constraints a 4 ji a 4 m di 4 ky b1 0 la amp 0 jo C 4 m da 4 k b24 4 i3 0 js c amp 3 0 m3 d3 0 ka b3 0 lL b1 4 l b 4 lg b3 0 The constraints are implemented by the class hierarchy shown in Figure 4 18 Each graphical component that adds anchor constraints contains the constraints as member variables so the con straint network is constructed automatically when the scene graph is created For example the class Voffset contains three OffsetConstraints one for each axis PropValue PropConstraint OffsetConst
245. s of action that are available shown in Figure 4 23 a and selecting one changes the action relation and updates the view Delete action removes the entry from the table Clicking on the reaction text pops up a menu showing the different types of reaction that are available shown in Figure 4 23 b and selecting one changes the action relation and updates the view These modifications are not currently saved for future use Options are declared in the Actions menu by 74 4 7 INTERACTION vmax Display of the action relation DG File Edit Window Options Content Form Action Debug Help Action Mouse move Mouse background move Mouse linger Mouse move Single left button click Single middle button click Single right button click Single background left button click Single background middle button click Single background right button click Reaction Hand mouse cursor A spider cursor Display in the zoom view Display identifier in status bar Navigate to the object The action menu The navigate menu Navigate to previous view The action menu The navigate menu Display ofthe action relation Ready a Figure 4 22 The Actions view for displaying and modifying the action relation Single left button click Single middle button click Single right button click Double left button click Double middle button click Double right button click F Mouse move Mouse linger Key press a Key press b Key p
246. s taxonomy 79 provides a suitable framework for such discussion It is given in Sec tion 6 1 This provides a means of comparison for particular aspects of the system but does not provide an overall measure of its worth The suitability of a particular tool is entirely task dependant 6 2 1 Benefits Although Vmax provides very flexible graphical output and has very fine control over the infor mation presented in its views it has not been shown whether this is of benefit There are several studies that show that different notations perform better in different circum stances 107 Green 37 argues a notation is never absolutely good therefore but only in relation to certain tasks Because programming is such a varied task maximum flexibility in the graphical representation would be required for the visualization system to be effective because specific notations may not be appropriate in all circumstances Information control or elision is an important aspect of information systems and in particular SV systems because of the different kinds of programming task and areas of interest So called spaghetti plate syndrome 17 is a major problem for control oriented languages because of the visual complexity of the output The Deutsch limit 67 places a limit on the information density of graphical displays However there is also evidence that visual notations become more effective when the problem size grows 22 78 Whitley 1
247. scape ad plugins README registry spell netscape movemail libnullplugin dynMotif so e e g a libjsd so bookmark htm README Netscape ad pell vreg XKeysymDB Netscape ad plugins README registry LICENSE XKeysymDB vreg XKeysymDB Byer Directory homes camg100 local netscape Directory homes camg100 local netscape Directory homes camg100 local netscape ee AMOO Y MEA OEA Y rones cang100 10001 netscape plugins B a File sizes b File ages c Files by age Figure 5 4 Alternative views of a directory Using SVT s default action context a single left mouse button click on an object in the scene navigates to it selecting the default view for that object Moving the mouse cursor over an object displays its identifier in the status bar and if the Preview window is open the file or directory beneath the mouse cursor is displayed in it Directories are automatically read in as navigation proceeds Directory contents Sub directories Directory tree File tree Java files in package Home Figure 5 5 The Content menu for a directory 86 5 4 DIRECTORY VISUALIZATION iva files under shomes camg100AVork SVTbin samples Java comiice sqiclient US File Edit Window Options Content Form Action Debug Help O UpdatePanel TablesPanel j TableDefsPanel SQLClientHiandler SQLClient QueryTablePanel QueryPanel J MySQLClient MySQLAdapter MainPanel MainF rame JOBCAdapter j HostinfoPanel
248. selected to tailor the output e A type system for information artefacts 38 and graphical artefacts has been implemented e Visual languages can be specified in Prolog e New methods of interaction have been demonstrated including a browser interface to nav igate around a program database a legend and a preview window URL http www cl cam ac uk Research Rainbow vmax CHAPTER 1 INTRODUCTION 15 Est Union Ot Care Forw Acta Bg a a a a a vate inj oiiae cha pico o po cha po int n a el priate in jo long Sings eto Sing anotherSting e g pate cha java lang Shing value private int java larg Sting count E Coti falses ara method public booleani java lang Sting oquas Object anObjac rite cer v eee A Ss a Method control flow b Method block structure Este EAE ini Vmax References within class com ruleofeight find GnuNaliveFind JER Filo Edit Window Options Content Form ction Tebug Help eae PA 9 d o T ra vl ol ol oo ul ul ul E veld vat yd vasca to Reteenoes un class com lesa a crutamver nc Subtypes of java lang Throwable Pes Ready J c Class hierarchy d Cross referencing i Vmax Examine value tree left right 0X File Edit Window Options Content Form Action Debug BinNode value left null BinNode value left null BinNode vaus N
249. sform the graph of the image structure Attributed multiset grammars consist of rewrite rules on collections of symbols with attributes One of the most general and powerful types of attributed multiset grammar are Constraint Mul tiset Grammars 65 Geometric constraints on the attributes are tagged to the grammar rules Such grammars can be implemented in Prolog by using constraint logic programming 48 to specify the geometric constraints Semantics of Diagrams The semantics of diagrams can be described by logical or algebraic formalisms As well as using logic to reason about geometry description logics 28 can be used to assign meaning to the 34 2 6 INFORMATION VISUALIZATION structure of the diagram Description logics are less powerful than full first order logic and are therefore more tractable Algebraic specification 103 maps the application domain into typed abstract data structures An image is mapped into a data type representing concepts or semantics Algebraic rules define a functional mappings between the domains 2 5 2 Numerical Data Numerical data has particular significance in scientific visualization and has uses in software visualization Numerical only methods are insufficient for heterogeneous visualization and vi sual languages Hibbard et al 43 develop a scalar mapping technique for mapping numerical data in a program to continuous quantities in a display including time A system called VIS AD implem
250. sion is achieved by editing text C 3 1 Vmax has very good control over multiple views because the user can create as many viewing windows as desired 4 4 and display any kind of data there The views in different windows cannot be synchronized however C 4 The run time data is not synchronized with the program because it is displayed off line 5 8 It is not clear whether on line or off line viewing of run time data is better because on line allows the data state of the program to be modified live and asynchronous viewing requires more storage but allows much better temporal control and allows static views of dynamic events 5 8 such as in Vogue 51 Asynchronous data display interrupts the running program much less Vmax could be run in synchronous mode by interfacing to a remote Java Virtual Machine via its debug interface 68 C 5 6 1 4 Method Views can be specified on many levels Defining the information in a view is done via textual specification in Prolog 3 3 and also using visual specification 85 9 Choosing the presen tation style of the view is done through pull down menus 84 7 4 and selecting the data to visualize is done via direct interaction with the view 84 7 3 5 7 1 D 1 The system displays many different types of intelligence For example it uses automatic lay out 4 5 automatically finds suitable views for an object 4 7 3 automatically finds suitable visualization contexts for
251. source code and analyzing the parse tree using language rules from the Java Language Specification 35 These rules can be expressed quite naturally in Prolog Program analysis itself is well understood forming the basis of all compilers Java source code is loaded on demand and the program database the database that stores information about the program is stored in dynamic predicates in memory and is therefore not persistent The program database could be stored externally to make it persistent but that would have been slower and slightly more laborious to implement The predicate ensure_java_loaded File is invoked to load the given file File This predicate is called from ensure_fresh 1 when an attempt is make to navigate to a Java source file or from the CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 89 Load all Java files option is selected from the Action menu of a directory ensure java Joaded 1 loads the given file into a text buffer and invokes Lex and Yacc to return the parse tree of the Java file The Lex and Yacc scripts for the Java syntax can be derived from the LALR 1 grammar in the Java Language Specification 35 Alternatively a compiler front end could be interfaced directly to Prolog The parse tree is in the form node Type TextNode C 1 Ca Cn where Type is an integer identifying the node TextNode is the address of the text node and C C are terms of the same form Leaf nodes of the form node Type TextNode
252. t visual primitive Visual Primitive visual component Visual Component visual primitive Component Primitive and in particular lists of visual components can be declared by visual component H T H visual component H T T This means that the specifications for a labeled icon visual primitive labeled icon lId Label Type Primitive icon icon Id Type Primitive text text Id Label Primitive v_stack Id icon Id text Id Primitive visual_primitive labeled_icon Id Label Type Primitive visual_primitive icon icon Id Type Primitive visual_primitive text text Id Label Primitive visual_primitive v_stack Id icon Id text Id Primitive visual_component labeled_icon Id Label Type Component Component icon icon Id Type Component text text Id Label Component v_stack Id icon Id text Id visual_component labeled_icon Id Label Type 64 4 6 EXPANDING THE GRAPHICAL VOCABULARY icon icon Id Type text text Id Label v_stack Id icon Id text Id l are all equivalent The latter offers the most succinct method of specification and there fore all the visuals provided by SVT have been defined using only visual component 2 and ob ject component 3 described in Section 4 6 1 In this example visual_component coloured_line A B Colour line A B line_colour A B Colour Jy visual component r
253. t ArrayLength integer N values N L values 0 gt uN values N H T gt N gt 0 M is N 1 value H values M object nullobject gt Object null object object Id gt ObjectNo integer ld object object Id Class Fields gt ClassName string Class ObjectNo integer Id Fields integer F fields F Fields fields 0 gt uN fields N H T gt N gt 0 M is N 1 field H fields M T field field Name Value gt FieldName string Name value Value 116 5 8 RUN TIME DATA VISUALIZATION The implemented DCG reads lexemes slightly differently using the token 2 predicate and also adds this data to Prolog s database by adding facts to the dynamic predicates trace event S and trace value 3 Each input event is stored trace event Input File EventNo Time Value where Input is a unique identifier for the input number File is the identifier of the trace file EventNo is the event identifier Time is the time stamp of the event and Value is the unique identifier of the input value Each input value is stored trace value Value String Term where Value is the unique identifier of the value String is a string that describes the value such as myarray 5 x and Term is a term that stores the value Term can be array Values An array with a list of values Values float Float A float or double with value Float integer Int
254. t Let Tro C T be the set of reaction contexts Letr Tro be the reaction context ays R Definition 15 Reaction relation Let be a relation between TR and Tro The reaction relation implements the response in Prolog The predicate reaction 2 reaction Reaction ReactionContext is used where Reaction is the term denoting the reaction and ReactionContext is the reaction context Reactions are implemented by adding rules to the reaction 2 predicate that have side effects to perform the required task The reaction term from action 3 is passed immediately to ae A R reaction 2 for execution Interaction is the combined relation gt The action and reaction contexts and terms denoting actions and reactions T 4 and Tp are implementation defined The action terms used by SVT are given in Section B 7 52 3 9 CHAPTER SUMMARY Example 22 The reaction relation reaction navigate to ViewContext r new_view_context ViewContext implements a reaction to navigate_to ViewContext in reaction context r that changes the view context and redraws the view Example 23 The reaction relation reaction delete file File r filename File Filename system_call rm Filename implements a reaction to delete_file File in reaction context r Example 24 The reaction relation reaction Reaction call call Reaction executes the reaction term as a predicate 3 9 Chapter Summary Visual
255. t may be represented by visual containment but if the relationships in the view content do not form a tree then a visual object could be contained in two others and an error occurs Another visualization relation that better suited the data would be required Each member in To is assigned a type by the type relation c and each member in Ty is assigned a type by the type relation y Definition 7 Content types Let Tr be the terms that represent types The type relation is a relation between Tc and Tr Definition 8 Visual types The type relation y is a relation between Ty and Tr View content c may be mapped onto a visual g only if they have the same type Formally 2A and B cannot be the identical otherwise this set would reduce to circle A red line A A line_colour A A green which was not observed CHAPTER 3 A MODEL OF INFORMATION VISUALIZATION 47 cug gt HETr ciotAg yt 3 1 This ensures that information is not lost from the content when it is rendered and that the visual representation of the content is suitable The type relations can be implemented in Prolog with the predicates content_type 2 for and visual_type 2 for y content_type Content Type visual_type Visual Type which indicate the type of the visuals or content The format of the type term Type is com pletely unrestricted Each datum can be characterized by the number of references it makes to other entities These references a
256. t of the current view in ViewContext navigate_as Object ViewContext Not extensible Changes the current view context to visual ize Object or if that fails use ViewContext navigate_to ViewContext Not extensible Changes the view context to ViewContext This will also ensure that the contents of the view are current by calling ensure fresh 1 new_view ViewContext VisualContext Not extensible Changes the view context to View Context and the visualization context to VisualContext This is a direct call that does not ensure that data is loaded new_view_context ViewContext Not extensible Changes the view context to ViewContext This is a direct call that does not ensure the validity of the view context or automatically loads data new_visual_context VisualContext Not extensible Changes the visualization context to Visu alContext and redraws the view new_zoom_view ViewContext VisualContext Not extensible Changes the preview window to have view context ViewContext and visualization context VisualContext The preview window is updated popup frame ViewContext VisualContext ActionContext ReactionContext Not extensible Creates a viewing window with no text pane status bar or pull down menus The initial view shown in the main viewing area has view context ViewContext visualization context VisualContext action context ActionContext and reaction context ReactionContext refresh_view Not extensible Redisplays the
257. t tools In this dissertation the visualization process is decomposed into a series of formal mappings which provides a formal foundation and allows separate aspects of visu alization to be specified independently The first mapping specifies the information content of each view The second mapping specifies a graphical representation of the information and a third mapping specifies the graphical components that make up the graphical representation By combining different mappings completely dif ferent views can be generated The approach has been implemented in Prolog to provide a very high level speci fication language for information visualization and a knowledge engineering envi ronment that allows data queries to tailor the information in a view The output is generated by a graphical constraint solver that assembles the graphical components into a scene This system provides a framework for SV called Vmax Source code and run time data are analyzed by Prolog to provide access to information about the program structure and run time data for a wide range of highly interconnected browsable views Different views and means of visualization can be selected from menus An automatic legend describes each view and can be interactively modified to cus tomize how data is presented A text window for editing source code is synchronized with the graphical view Vmax is a complete Java development environment and end user SV system Vmax compa
258. tain identifiers specified in the graphical constraints that are used to identify objects beneath the mouse cursor Example 19 The following actions are used in SVT click Object Buttons The mouse has been clicked Times times over object Times Object with the mouse buttons Buttons drag From To Buttons A mouse drag from object From to an object To with mouse buttons Buttons key Object Key A key Key has been pressed over object Object linger Object The mouse has stopped over object Object menu Menu Text The menu Menu has item Text selected move Object Buttons The mouse has been moved over object Object with mouse buttons Buttons Example 20 The action relation action click Person on _ _ l _ a navigate_to personal_details Person person Person declares that the reaction to a mouse press with the left mouse button over a person Person in action context a is navigate_to personal details Person The underscores are wild cards or unnamed variables Example 21 The action relation action menu actions File Delete file a delete file File file File declares that the reaction to the menu selection in action context a is delete file File if File is a file 3 8 2 Reactions Reactions execute the system s response in Prolog To allow different implementations of reac tions each implementation is given a reaction context to identify it Definition 14 Reaction contex
259. ted is so heterogeneous and the graphical output can be so varied What kind of theory could underlie such an open ended task Current theories of visualization or visual lan guages do not adequately model the semantics of graphics or are too limited so existing implementations are not based on formal approaches The work in this chapter presents such a theoretical model that formally models the relationship between graphics and its semantics It is named a semantic model because in this model the semantics of graphical views are always explicit This in turn leads to a decoupling of the information content in a view and its graphical representation and allows arbitrary queries to determine the content of a view An important aspect of any theory is its applicability and the purpose of this theory is to provide an extremely concise and high level specification method This will be used as the core of a visualization engine that is described in Chapter 4 Chapter 5 applies these specification techniques to software visualization 3 1 Introduction Software visualization is laborious to implement because the current tools are not sufficiently automatic and flexible and lack effective scripting 79 Computer software contains very rela tional structured and heterogeneous data and visualization tools are ill equipped to manipulate this kind of data because theory on visualization and has not been fully developed to handle arbitrary data struc
260. tem Pressing Help on the menu bar activates the help system This navigates to the view help main This view displays the help text in the text window and indicates the sub topics in the graphical window Clicking to one of the help topics navigates to that help topic In this way SVT s views are used to display help The predicate help_topic Topic SubTopic declares help sub topics and the predicate help to pic Topic Title Text declares the help text where Topic is a term identifying the topic SubTopic is the sub topic Title is the help topic title and Text is the help text for the topic The help topic for a particular view ViewContext could have the view context help ViewContext providing 132 5 10 THE HELP SYSTEM Open directory Move back Move forwards Execute command New view fo X Show log l Enter the name of the view Empty log New clause Add clause Add view O Recalculate cross references Calculator a b my_new view has view content content if true my_new_view has visual context vs c d D has wew context my_new_view and description my_new_view e Figure 5 58 Creating a new view CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 133 view specific help 5 11 Chapter Summary Vmax is a programmers editor that integrates many areas of software visualization As well as providing overviews of large program structures such as p
261. tement The predicate java method_contains_statement Method Statement uses this to return the statements Statement contained within a method Method java statement_entry point 2 returns the entry point first statement executed within a compound statement java statement_exit point 2 returns the last statement executed and there may be more than one for example due to branching java method _entry_point 2 gives the first statement executed in a method and java method exit point 3 gives the last statement and again there may be multiple solutions available by backtracking The predicate java statement_follows Method Statementl Statement2 Transition returns all pairs of sequentially executed statements Statement and Statement2 in a method the transition type Transition being case X default else for sequence then and while to indicate the different circumstances in which the statements follow Exception handling is not indicated by this predi cate The predicate java sub_expression Exprl Expr2 returns the sub expressions Expr2 of an expression Exprl java sub_expression decl Statement Decls1 Expr Decls2 returns the sub expressions Expr of a statement Statement and a list of declarations Decls2 that apply to the sub expression The predicate java method_sub_expression Method Expr Context uses this to find the sub expressions of a method and returns the context Context in which the expression used which includes the local dec
262. temperatures while the view relation temperature range Min Max view content temperature range Min Max inside range X Y sea _temperature X Y Temp Min lt Temp Max gt Temp view content temperature range Min Max outside range X Y sea temperature X Y Temp Temp lt Min Temp gt Max returns the sets of points inside and outside the given range Min Max Example 9 The view relation view content conjunction A B C view _content A C view content B C gives the intersection of A and B with view context conjunction A B while the view relation disjunction A B defined view content disjunction A B C view _content A C view content disjunction A B C view _content B C gives the union of A and B 3 4 Specifying Visual Content The graphical description of the view or visual content is a set of terms describing the graphical structure of the view It is derived from the view content by a mapping as shown in Figure 3 1 Each term in the visual content represents one graphical property of the display The visual content is defined Definition 4 Visual Content Let Ty C T be the set of terms that represent visuals Let V C Ty be the visual content A mapping is specified from the view content C to the visual content V This mapping is called a visualization relation and can be regarded as the visual language that defines the protocol for communication ege
263. tensible Method is a method which may be abstract or native method Method Type Modifiers Params Throws BlockStmt Statements Dynamic Method is a method with return type Type modifiers Modifiers parameters Params as a list of vari ables throws Throws as a list of class names BlockStmt as the identifier of the block body and Statements as the list of statements in the method body method_calls Method1 Method2 Not extensible The method Method calls the method Method2 method_refs Method Variable Not extensible Method Method references the variable Vari able method_contains_statement Method Statement Not extensible Java method Method contains the statement Statement method_entry_point Method Statement Not extensible Statement Statement is the first exe cutable statement in method Method method_sub_expression Method Expression Context Not extensible Method Method has a sub expression Expression in context Context which is the declarations that apply to the expression name_of Name QualifiedName Object Dynamic Object Object has fully qualified name QualifiedName and short name Name This mirrors has _name 2 but is indexed differently 174 B 3 VMAX PREDICATE SUMMARY nonstatic_initializer Id Block Statements Dynamic The non static initializer Id has a block Block of statements Statements package File Package Dynamic Java file File is in package Package e g package java lang would
264. the calculator buttons calculator digit D P D 7 P 20 D 8 P 21 D D 4 P 24 D 5 P 25 D D 1 P 28 D 2 P 29 D D 0 P 33 calculator function F P F MR P 1 F MC F M P 4 F 1 x F cos P 7 Fstan 9 P 223 6 P 26 a3 P 30 P 2 F M P 3 P 5 F sin P 6 P 8 F 1n P 9 APPENDIX C FURTHER EXAMPLES 191 F exp P 10 F x P 11 F x y P 12 F P 23 F P 27 F P 31 F P 35 F P 32 F P 34 F C P 51 F 2 P 52 F sqrt P 53 F P 54 calculator button calculator digit D D P Calculator digit D P calculator button calculator function F F P calculator function F P o Declare the calculator view view calculator 0 0 1 true 0 Calculator visual context calculator _ calculator view content calculator _ calculator button Id Text Pos calculator button Id Text Pos view content calculator R _ result result R o Declare the visualization relation visual_context_description calculator Calculator visual_content calculator calculator visual_content result Id Text calculator right_text Id Text visual_content calculator _button Id Text Pos calculator calculator button Id Text Pos o Declare visuals for the calculator visual right_text Id Text Id str Text Right aligned text Symbol right _text Symbol
265. the visualization is essentially the problem of computer graphics and includes rendering graph drawing and spatial constraint solving Psychological factors are rarely applied to SV mainly because it of the absence of techniques to empirically measure the effectiveness of SV systems 79 This does pose problems for ob jective and meaningful comparison between systems Much of the SV literature completely dis regards psychological factors and merely regards visualization as a conversion from a program to a graphical representation for example 84 Psychological experiments must be performed under very controlled conditions and software visualization is too large a domain to measure directly The result is that psychology has so far offered few concrete ideas to SV that could not be derived using intuition and indeed visualizations and visual languages can and generally are devised on an ad hoc basis and are only informally evaluated There are psychology studies that can contribute to creating more effective visualizations Rogowitz et al 82 look at how we can exploit preattentive perception and colour luminance mappings to create visualizations that can be searched more quickly and convey continuous or discreet data with greater veridicality Perceived brightness is proportional to a power of lu minance hence gamma correction so where luminance is used to convey continuous data a corrective power should be applied and examine other s
266. ther ta X Y is father to X Y a Prolog X is a parent of Y if X is mother to Y or X is father to Y Kis father to E b English c Query Mis an ancestor or d Xie parent or AAD is an ancestor of d Block structure is an ancestor of is an ancestor of e Tree Figure 5 53 Different visualizations of a clause CHAPTER 5 A TOOL FOR SOFTWARE VISUALIZATION 129 To display a clause each sub term of the clause is given an identifier Id of the form sub term Parent Position where Parent is the identifier of its parent term and Position is the posi tion of the sub term in its parent compound term Each term is represented in the view content as term Id Functor Args where Functor is the name of the functor and Args is a possibly empty list of identifiers that are the child terms of the compound The terms 2 2 and 2 are treated separately The block structured notation shown in Figure 5 53 d is similar to VPL 60 and uses the horizontal axis for conjunction and the vertical axis for disjunction A different view similar to the Transparent Prolog Machine TPM 30 is shown in Figure 5 54 A red horizontal line indicates conjunction Hl vmax Al Clauses TPM EA File Edit Window Options Content Form Action Debug Help AS ieee All Clauses TPM Ready A Figure 5 54 Visualizing predicates in the style of TPM 5 9 1 Editing Prolog Clauses can be edit
267. thods should be used SV systems that need to be highly tailored provide many different views and need to analyze and reason about programs could benefit from using Prolog as its visualization mechanism In this case the specification techniques in Chapters 3 and 4 could be used A constraint based layout mechanism is then needed to combine different graphical notations CHAPTER 7 CONCLUSIONS 153 7 3 Specifying SV Systems The way SV systems are specified is important to be able to construct systems quickly and effectively A special purpose specification language and environment can potentially reduce the costs of implementation An implementation is more likely to be error free if the syntax of the language matches the semantics of the problem 94 Previous systems have had only limited success with specification because they use a very low level approach The declarative specification languages may be no better than C for graphical layout Pavane 86 and Leonardo 20 are not per se better visualization systems than Polka 96 or Zeus 15 In previous declarative systems numbers from run time data are used to specify the screen coordinates of graphical objects on the screen This approach is too limited for code visualization and arbitrary data structures This work has made specification more powerful and flexible by subdividing the visualization process and generating high level graphical constraints In this way the specification is at
268. timuli for conveying continuous infor mation Perceived colour varies discretely with hue wavelength so this should be exploited for discrete data but compensated for in continuous data Experiments in preattentive vision show how we can subconsciously pick out certain features of images e g coloured spots which can be exploited when searching large datasets in constant time versus a linear searching time for attentive conscious searching Preattentive effects can also disrupt the interpretation process by distracting the subject from the important features and should be exploited to attract or direct attention to the important features of the visualization quickly Rogowitz and Treinish 83 propose a visualization system that considers perceptual rules to devise a visual representation for data They argue that this is required because traditional visualizers may introduce unwanted artefacts into the image that obfuscate the meaning of the visualization as described in 82 They also claim that visualization is easier to implement and produces better and more meaningful representations when visualization rules based upon perception are used to generate visualizations than high level tools or direct coding Current SV systems use high level tools and direct coding which produce inflexible visualizations and offer little guidance on the effectiveness of the visualization 2 6 2 Representations Some systems use a three dimensional repr
269. tion 6 Visualization context Let v T be a visualization context Let gt be the visualization relation for visualization context v The visualization context can be changed to select different visualization relations This rela tion can be implemented by the Prolog predicate visual_content ViewContent VisualizationContext VisualContent The first argument specifies the member of the view content The second argument specifies the visualization context The third argument specifies the visual content For a given visualiza tion context this predicates implements a binary relation between C and V The set of available visuals Ty is implementation defined Example 10 The visualization relation with visualization context v visual_content married X Y v next _to X Y visual content mother X Y v line X Y green visual content father X Y v line X Y blue represents married X Y by next_to X Y mother X Y by line X Y green and father X Y by line X Y blue Example 11 The visualization relations v1 and v2 declare two different views of the house wiring visual_content socket S L vl socket_drawing S L visual content wire A B live vl graph _ edge A B black visual content wire A B neutral vl graph edge A B red visual content wire A B earth vl graph edge A B green visual content socket S L v2 icon S L socket visual content wire A B live
270. tion rich vs information poor It emerged that there are eleven major clusters of representations graphs tables graphical tables time charts networks structure diagrams process diagrams maps cartograms icons and pictures The work offers many ideas and visual representations or combinations of them that could be applied to SV and gives the principle characteristics of each representation Tufte 101 details methods for scientific visualization of quantitative information He gives a comprehensive collection of hundreds of graphical representations with critique This book offers a rich source of ideas and much more useful rule of thumb approaches for effective visu alization than attempting to rationalize visualization on a formal basis Many visualizations of software are quantitative Bertin 9 wrote one of the first texts on the semiotics or meaning of graphical representations This is an informal treatment of diagrams Bertin regards order form and proportion to be the building blocks of any graphical representation Bertin mainly considers visualizing numerical tables Green and Benyon 38 regard entities attributes and relationships as the fundamen tal components of information display and discuss again informally the relationship between graphical properties graphical artefacts and the semantic properties information artefacts conveyed by the display 2 6 3 Architectures The AVE system 34 implemen
271. to entities attributes and relationships between graphical objects Bertin 9 regards order form and proportion as fundamental to visualization which are loosely observed as examples of attributes entities and associations Extra information may be conveyed by a visual For example a labeled icon could convey a text label while a height bar could convey numerical data The type of this data would be conveyed in the type of the visual The types assigned to visuals and content Tr are implementation defined The particular type representation used in the following examples and Vmax forms a list of properties of the type An entity in represented as Id and a reference as Id Strings have type str S and numbers have type num S Thus any type containing a Id is an object one Id is an attribute and two or more Id is an association Example 14 Some examples of content types man M aman M M height P H Phas height H P num H is_tall P Pis tall P socket S Loc a wall socket S str Loc wire A B Type a wire A B 48 3 6 SPECIFYING GRAPHICAL CONSTRAINTS Example 15 Some examples of visual types box X a box X circle X Colour a coloured circle X labeled_icon X D Type a labeled icon EX str D line X Y Colour a coloured line A B highlight X highlighting FX object_size X Size object s size X num Size Example 16 The types for highlight X and is tall X ar
272. to other tools that could return it information when requested The pro gram database could be stored externally in another tool 6 5 17 Interface to the Java Virtual Machine Debug Interface Run time information is displayed off line A graphical interface to the Java Virtual Machine Debug Interface 68 could provide access to live program information which could browse the run time data and show the control flow through the program CHAPTER 6 DISCUSSION 149 6 5 18 Improve Efficiency There are many areas of SVT and Vmax that could be speeded up For example there is a lot of dynamic memory allocation to construct each view Such data could be constructed on a stack An entire parse tree data structure is constructed during parsing which is unnecessary 6 6 Chapter Summary A classification of Vmax in a taxonomy of SV systems 79 shows Vmax to be strong in many areas of SV in spite of having such broad scope The general specification technique has been shown to be appropriate for all aspects of SV Unlike many other SV systems Vmax is very scalable and is practical to use in medium sized projects of 100 000 lines of code All aspects of software can be visualized and it is unusual for a system to be strong in both program and data visualization or to offer any flexibility in program visualization Its constraint based layout provides a much more flexible range of output The query based mechanism provides good elision control and de
273. to reset structure size resize position and render each object SVT implements a range of graphical output which can be extended by deriving new classes from the constraint hierarchy The graphical vocabulary can be expanded to provide more varied graphical output New visuals can be defined as combinations of existing visuals and visual objects can be defined in terms of graphical components and anchors Anchor expressions provide a concise means of specifying constraints that allow objects to be resized correctly Visual objects are laid out by satisfying the set of constraints created by the object components and the anchor expressions Interaction is specified entirely in Prolog using the simple model described in Chapter 3 Care must be taken to interact with the intended object which is taken to be the innermost object in a nesting Menus are also specified using this method and can be completely customized by the data Navigation is achieved by changing the view context which can be done by clicking on an object with a view context from a pull down menu or from a pop up menu The visualization 80 4 11 CHAPTER SUMMARY MU sava lang string fe x Figure 4 29 The Preview window context is also changed from a pull down menu There is a view to display and edit the current action relation The legend window automatically describes all the visual notation in the view and can also be used to modify the visualizati
274. ts a flexible mapping from data to diagram using a data driven approach where the data itself determines how it is to be visualized and AVE is particularly applicable to semantic networks 28 where the database contains heterogeneous data The data relations are mapped onto graphical relations on the screen such as interconnection or adjacency by analyzing the structure of the data Queries select subgraphs of the database which are subject to structural analysis Graphical relations are then assigned to the data relations using a ranking system and the diagram is generated to render the data with the assigned graphical structure Figure 2 15 a shows the graphical representations it chooses for the various data structures and Figure 2 15 b shows the overall architecture The drawback with AVE is that the ranking system is hard coded and it has only a limited number of different layout techniques to apply 2 6 4 Graphical Layout Graph layout remains a difficult area of computer science It is well known that optimizing graph layouts is generally NP hard and in particular minimizing edge crossings is NP complete 33 so much work has been done in developing heuristics for graph drawing Graph drawing is an important tool for SV where relational data may be presented in a node and link fashion There are many ways to present graphs including using polylines straight lines as orthogonal drawings as rooted trees as free trees in planar for
275. ts is far more wide ranging than in other SV systems Good elision control The user can control what to display or not to display Granularity The program can be viewed at many levels of detail Tailorability The view content and visual presentation can be changed to suit the user Scalability Quite large code bases and complex data structures can be visualized 151 152 7 2 IMPLEMENTING SV SYSTEMS Extensibility It is comparatively easy to add new views to the system using the specifica tion language Automatic legend Users can understand and modify the graphical representation User interface Navigation and browsing is achieved by single mouse clicks graphical bookmarks and the preview window e Number of views Vmax has 70 different types of view which is an order of magnitude greater than in other systems e Automation Minimum user input is required to select and view the data Visual program structures are created automatically from source code Integration into a complete development environment e Good temporal control over run time data It can be stepped and animated forwards and backwards 7 1 2 Weaknesses of Vmax Limitations in program size 115 000 lines of code is quite reasonable but is not as good as some professional analysis tools Limitations in input speed The parser and program analysis is slower than some other source code analyzers written in C e Lack of synchronization with the ru
276. tures In this chapter a theory of visualization is developed that forms the basis of a generic vi sualization tool that will be used to visualize software Foundations are a necessary first step and provide a common language a framework for implementation and the capability for formal reasoning While there are many automatic tools for numerical and scientific visualization heteroge neous data visualization has proved to be much more difficult to model and specify 44 Existing approaches to heterogeneous data visualization rely on low level programming using imperative languages Even declarative specification approaches such as Pavane 86 and Leonardo 20 explicitly specify graphical positions on the screen and deal only with numerical quantities The benefits of declarative or logic specification languages are not fully realized when such a low level approach is used and logic languages such as Prolog are inferior to imperative languages for implementing general graphical algorithms because of their slow speed and lack of efficient data structures In this chapter these basic declarative approaches are improved Instead of directly mapping 39 ss INTRODUCTION numerical data to screen coordinates the visualization process is further subdivided as shown in Figure 3 1 Flexible mappings l 01111010101 sample 9 125 person harry 1 line di harry 01010100001 age harry 15 person diana label harr
277. ty selectively hiding information and increases the number of graphical views available to programmers 5 1 Introduction Vmax is a text editor with graphical views for visualization and browsing The graphical views can visualize almost every program object in a variety of formats and most objects in the views can be navigated to by a single mouse click Run time data can also be gathered displayed and animated Vmax merely provides the data to visualize but SVT provides the framework for visualization and navigation The visualization windows can contain any views generated by SVT and display views of the filing system the program database Java source code and run time information The visualiza tion window provides a navigation and browsing facility by providing many different views of the program database including views of individual methods or variables Views are set up to link to related views so for example a view of a Java method has links to the variables it refer ences and the functions it calls It is possible with a single mouse action to navigate to the class that defines the variable other methods that reference the variable methods that call the current method the reachable functions or any number of user defined views In designing Vmax the aim was to anticipate what information a programmer needs at any given time to provide no more and no less to anticipate where a programmer wants to navigate and to minimize t
278. tyle data Style Name widget_data Style Status visual_content dialog window window Font select main vertical main options buttons Buttons button yes button Ok button no_ button Cancel button apply button Apply horizontal buttons yes button apply button no button Options vertical fontlist group font_group Font fontlist vertical stylelist group style group Style stylelist horizontal options font_group style group Midis visual_content font_option ld Text Status dialog labeled choice Id Text Status fontlist visual content style option Id Text Status dialog labeled_selection Id Text Status stylelist action move B _ svt indicate action selector_button B choice _button B action click B _ _ svt toggle selector B selector button B action click B _ _ svt Action button B Action action click B _ _ svt choice select B C choice _button B C The calculator in Figure 6 2 was created using similar methods The calculator supports key board input In comparing two calculator applications in Figure 6 3 the Tcl Tk 74 imple mentation in Figure 6 3 a obtained as a Tclet from Scriptics was 127 lines while the Prolog implementation in Figure 6 3 b was 58 lines long URL http www scriptics com products tcltk plugin calculator html 144 6 3 EVALUATING SVT o X lt
279. uages and Computing 1 97 123 March 1990 72 fay I Nassi and B Shneiderman Flowchart techniques for structured programming SIG PLAN Notices 8 12 26 1973 73 L O Brien Issues of programming Computer Language 10 1 45 52 January 1993 74 J K Ousterhout Tcl and the Tk toolkit Addison Wesley 1994 75 E C N Pereira and D H D Warren Definite clause grammars for language analysis a survey of the formalism and a comparison with augmented transition networks Artificial Intelligence 13 231 278 1980 76 Visual development tools Personal Computer World pages 124 144 March 1997 77 T Pietrzykowski S Matwin and T Muldner The programming language PROGRAPH Yet another application of graphics In Proceedings of Graphics Interface 83 pages 143 145 Edmonton Alberta May 1983 BIBLIOGRAPHY 161 78 J M Polich and S H Schwartz The effect of problem size in deductive problem solving Memory and Cognition 2 683 686 1974 79 B A Price R M Baecker and I S Small A principled taxonomy of software visualiza tion Journal of Visual Languages and Computing 4 211 266 1993 80 S Reeves and M Clarke Logic for Computer Science Addison Wesley 1990 81 J Rekers and A Schiirr A graph grammar approach to graphical parsing In Proc 11th Intl Symposium on Visual Languages VL 95 pages 195 202 1995 82 B E Rogowitz D T Ling and W A Kellogg Task dependence veridicality
280. ual For example visual component text _shadow Id Text text text Id Text shadow_box Id text Id IDIS creates a visual text_shadow 2 which displays text within a shadowed box A coloured back ground can be added to shadow_box 2 by object_component shadow_box_colour Box Colour Box p polygon a b d c p colour Colour visual_component shadow_box_colour Box Child Colour 66 4 6 EXPANDING THE GRAPHICAL VOCABULARY shadow_box Box Child shadow_box_ colour Box Colour visual component blue text _box Box Text text_box Box Text shadow_box colour Box paleblue 1 to define the new visuals shadow_box colour 2 shadow box colour 3 and blue text box 2 In the shadow_box 2 example the identifiers for the object components are p1 and p2 for the polygons and a b c and d for the anchors The predicate component primitive 3 combines these names with the identifier of the object so the polygons have identifiers vc Box p1 and vc Box p2 and the anchors have identifiers anchor Box a anchor Box b anchor Box c and anchor Box d This subtlety is necessary to allow the cross referencing described in Section 4 5 2 to work correctly Otherwise the name a would be a global identifier in the scene and two different shadow_boxes would share the same anchor and the output would be incorrect Name mangling puts anchors and object components in different name spaces and co
281. ualization relation The menu of alternative graphical forms is found by determining the type Type of content Content and finding compatible visuals with the same type The query content Content Type _ visual Visual Type Description _ _ CHAPTER 4 SEMANTIC VISUALIZATION TOOL 77 F A vertical block NSD finally block NSD coloured finally block Block in curly braces Text finally block 3 D block 3 D steps Right aligned list Horizontal block Horizontal bottom block Numbered horizontal block Vertical block Vertical left justified block Horizontal boxes Nothing Figure 4 25 A menu providing alternative graphical forms is used to build the menu Selecting a visual New Visual from this menu modifies visual content 3 by retract visual content Content VisualContext OldVisual assert visual content Content VisualContext NewVisual The predicates content _type 2 and visual type 2 described in Section 3 5 are defined content type Content Type content Content Type _ visual _type Visual Type visual Visual Type _ _ _ 4 9 The Bookmarks Selecting the Bookmarks option from the View menu creates a row of five small windows as shown in Figure 4 27 These windows can be used as place holders or bookmarks to store views for later retrieval and show a miniature version of the stored view A single left mouse click on the bookmark will retrieve 1t to the main viewing window A right mo
282. ucture Inherited methods structure Public members Subtypes of class Supertypes of class Method calls within class cached Method calls within class References within class cached References within class Class references Declarations of class Places class thrown Home Figure 5 11 The Content menu for a class or interface The view of a class including all of its inherited members is shown in Figure 5 12 Variations of this view show only methods variables or public inherited members Figure 5 13 Figure 5 14 shows the supertypes and superinterfaces of a class and these views are combined to show members in supertypes in Figure 5 15 A view showing the subtypes of a class is shown in Figure 5 16 and a view giving the imple mentations for an interface is shown in Figure 5 17 There are views showing all type declarations 94 5 6 JAVA VISUALIZATION Figure 5 18 and all interfaces Figure 5 19 The complete class hierarchy is shown in Figure 5 20 As mentioned previously each item in these views can be navigated to by clicking on it and navigating to a particular view for that object is achieved by right clicking on the object to pop up its Content menu Vmax Inherited members in com ice gallery Gallery Als File Edit Window Options Content Form Action Debug oh oe mo Mo Class getClass int hashCode boolean equals Object obj Object clone String toString ain p als void notify void
283. ue Class Fields in list field Name Value2 Fields trace value content Value2 Content Integer and floating point values can be visualized as shown in Figure 5 46 Arrays can be visualized as shown in Figure 5 47 and structures as in Figure 5 48 The name of the structure and fields are displayed with the data itself Very large data structures may clutter the view so it 1s possible to show a structure to a certain depth as shown in Figure 5 49 Browsing to members in the view would reveal more structure 7 E a Text b Horizontal c Horizontal with value d Horizontal offset e Vertical f 3 D bar Figure 5 46 Different ways of visualizing integer and floating point values By using the legend a textual representation of a value can be transformed into a graphical one as shown in Figure 5 50 At each step a different graphical notation is changed in the Legend thus transforming the view It should be clear that any combination of graphical representations can be combined to tailor the view More visuals could be added and specialized views can be declared in Prolog if the provided views prove inadequate 122 5 8 RUN TIME DATA VISUALIZATION 9 0040 01 e WWW N a Text b Horizontal eh tay ay a MA AS e c Horizontal numbered d Vertical 111121213 3 3 4 4 5 6 7 8 9 12 e Boxed Figure 5 47 Different ways of visualizing arrays CHAPTER 5
284. ul 5 5 It does not have complete access to run time data at all times although any individual data can be completely monitored at trace points 5 7 B 3 Vmax affects the performance of the running program much less than systems where the view is synchronized with the data because the viewer does not pause the program for animation and stepping 5 7 B 3 1 Vmax gathers data about the program before and after it is run 85 5 5 7 B 4 The run time data can be stepped and animated but static views of sequences of events can also be generated 85 8 B 4 1 Run time visualization occurs after the program has been run 35 8 B 4 2 6 1 3 Form The output of Vmax is on a colour monitor C 1 The constraint hierarchy implements many different kinds of graphical component and layout algorithm providing a wide range of output 84 5 B 6 There are limits in the capabilities of the graphical layout algorithms but a range of visual languages have been implemented 5 6 3 85 9 C 2 1 Colour is used widely to convey information and indeed any numerical quantity or attribute can be used to colour any graphical component including bitmaps C 2 1 1 The views in Vmax do not generally make use of three dimensions The algorithms for graph and container layout 4 5 5 are two dimensional but they could be adapted to produce 3 D layouts The constraint based object layout is fully three dimensional 84 6 1 allowing 3 D object
285. ure 4 24 Legends explaining the symbols in a view The legend communicates the visualization relation to the user Whenever the view context or visualization context is changed the legend automatically updates itself to the new view The legend is displayed when the user selects the Legend option from the View pull down menu When multiple main windows are being used the window with the most recent change updates the legend The legend can also be resized and deleted via the window manager 76 4 8 THE LEGEND 4 8 1 Displaying the Legend The legend is generated like any other view and has the view context legend ViewContext Vi sualContext where ViewContext and VisualContext are the view and visualization contexts of the main view This view queries the view content using view_content ViewContext Content to determine the content Content of the object view It then forms a set of terms with distinct functors which prevents duplication of symbols in the legend The descriptive text for each entry in the legend is declared content Content Type Description which specifies the type Type and descriptive text Description for view content Content The legend queries this predicate to determine the text in the legend entry The graphical symbol that appears in the legend is declared visual Visual Type Description SymbolId SymbolVisual This specifies that when the visual Visual is displayed in the legend the symbol is Sym
286. use click will pop up the menu shown in Figure 4 28 The Bookmark option sets the view and visualization context of the bookmark to those of the main view The Retrieve option sets the view context and visualization context of the main view to those of the bookmark The Swap option exchanges the view and visualization contexts of the bookmark and the main view The Refresh option updates the bookmarked view if data has changed 4 10 The Preview Window Selecting the Preview option from the View window makes a Preview window appear as shown in Figure 4 29 This window can be used to display any view and SVT s default action context provides a preview function that displays the object beneath the mouse cursor This gives a preview of what navigating to that object would show without the user having to commit to the navigation or having to press any mouse buttons The preview window is declared 4 10 Bey Vmax Ai auses OOOO OUUU UUO File Edit Window Options Content Form Action Debug Help M count fi coun Nis M m count BZ is mother to charles charles is father to harry inal is mother to hary is a parent of A is an ancestor of Y pc e is an ancestor of Y x is a parent of x is a parent of Y 4 X is father to y X is mother to A i it of is svat of Ml All Clauses eed a The initial view List query and List query or Prolog 3 Horiz
287. ut is an action The system s response is the reaction An action followed by a reaction is called an interaction This simple model can be used as a basis for specification 3 8 1 Actions Interaction is modeled as a relation between actions and reactions called an action relation Each action and each reaction is represented by a single term Definition 12 Action relation LetT4 C T be the set of terms that denote actions Let Tp C A A A T be the set of terms that denote reactions A relation between T4 and Tp is called an action relation It may be necessary to have more than one action relation For example different types of view may have different interactive behaviour or different users may have different preferences for interaction Therefore each action relation has a term called an action context to identify it wife r A i Definition 13 Action context Let a T be an action context Then is the action rela tion with action context a CHAPTER 3 A MODEL OF INFORMATION VISUALIZATION 51 The predicate action 3 action Action ActionContext Reaction A implements the binary relation between Action and Reaction for a given action context ActionContext Interactive behaviour is implemented by adding clauses to the predicate action 3 Every time a user input occurs such as a mouse move the predicate action 3 is called to determine the reaction The graphical objects in the view con
288. ve Not extensible Specifies the graphical constraints Primitive that an object component Component generates in a visual object with identifier Id view Object ViewContext Description Extensible Declares that the view context ViewCon text can be used to visualize entity Object and the view description is Description If a visualization is not tied to a particular entity then Object is the atom view_content ViewContext ViewContent Extensible Specifies the query 2 that declares what the content ViewContent is for a particular view context ViewContext visual_component Visual Component Extensible Declares that Component is a visual that is added to the view when the visual Visual is Component may be a list of visuals visual_content ViewContent VisualizationContext VisualContent Extensible Dynamic Sp V ecifies the relation gt that declares how view content ViewContent is mapped to visual content VisualContent for a particular visualization context VisualizationContext visual_context ViewContext VisualContext Extensible Declares that the visualization con text VisualContext is suitable for displaying the view context ViewContext visual_primitive Visual Primitive Not extensible Specifies the graphical constraints Primi tive that are generated by a visual Visual B 2 3 Legend Predicates content Content Type Description Extensible The content term Content has type Type and
289. verConne i Supertypes of com ruleofeight objserv ServerConnection java lang Runnable dj Figure 5 14 The supertypes and interfaces of a class EE Vmax Inherited members in com ruleofeight objserv ServerConnection SX EE Vmax inherited members in com ruleofeight objserv ServerConnection Eile Edit Window Options Content Form Action Debug Help File Edit Window Options Content Form Action Debug Help WA o PrintWrter clientOut Pert5Util regex Shige a ny String CRLF ai FA vond proviicSoracot omona araen a All members Inherited variab com ruleofeight objserv Server Connection File Edit Window Options Content Form Action Debug ServerConnection Connection Hashtable properties long counter long myCounter a Socket client a String CRLF PerlSUtil regex Print Writer clientOut Inherited variables in com ruleofeight objserv ServerConnection Inherited methods in com ruleofeight objserv ServerConnection con ruleofeight objserv ServerCannection protected com ruleofeight sf Connection Connection c Variables d Methods Figure 5 15 A class view giving members of supertypes 96 5 6 JAVA VISUALIZATION E Vmax Subtypes of java lang Thread File Edit Window Options Content Form Action Debug Help java lang Thread com ruleofeight sf Connection com ruleofeight objsery ServerConnection com ruleofeight sf
290. view generation depends on the constraints and the amount of compu CHAPTER 6 DISCUSSION 141 tation required to derive the view content A typical view may contain 100 or more constraints beyond which the view starts to get cluttered These figures are of course approximate because they depend on hardware and the specific data used Vmax offers a responsive interface and is practical to be used for medium sized projects of around 100 000 lines of code Source code of 115 000 lines has been successfully visualized This performance could be significantly improved by a more efficient implementation The performance was limited by Prolog s memory management which filled up the computer s mem ory as more facts were added to its database suggesting that specially written storage routines should be used for the program data 6 2 3 Extensibility Vmax is readily extensible by adding more views and program analyses in Prolog files and load ing them on start up The graphical output can be readily modified and extended by defining more visuals New languages can be added by adding lexical analyzers and parsers to the executable and the Prolog to analyze them which could coexist with the existing source code analyzers 6 3 Evaluating SVT 6 3 1 The Specification Language Specifying SVT is very different to specifying other visualization systems because there is so much more to specify There are views view content visual content visua
291. ws Positions within the scene could be specified using four dimensions and linear interpolation between points in space time could provide animation in the style of Pavane 85 6 5 5 Sound Program auralization 25 could potentially communicate more information than graphics alone Perhaps sound could form part of the view This could be specified by sound constraints gener ated from the information in the view A sound constraint would not affect the graphical output but would call a sound library A visual could combine aural constraints with visual constraints Moving the mouse over an object could also generate sound In this case the interaction with the object would result in a reaction that calls the sound library directly from Prolog 6 5 6 Automatic Choice of Graphical Layout The visualization relation must be specified manually by declaring how information is mapped to visuals Systems such as AVE 34 are data driven and determine a suitable mapping from the structure of the data Such technology could be integrated into SVT although it is not clear whether this feature would be useful The visualization context could be selected automatically depending on the form of the data 6 5 7 Improve the Graphical User Interface The specification method for menus is fairly primitive allowing only separators and bullet items This could be made more sophisticated by adding sub menus and implementing tool bars using CHAPTER 6 DISC
292. xpression that initializes the variable Methods are stored java method Method Type Modifiers Parameters Throws BlockStmt Statements where Type is the return type of the method Modifiers are the modifiers of the method Parameters is the list of parameters of the method which is a list of variable identifiers Throws is a list of class names that the method can throw BlockStmt is the block of statements Statements that make up the method body Constructors are treated as methods but the predicate java constructor Method indicates that the method is really a constructor Static initializers to classes are declared java static initializer Id Block Statements and non static initializers as java nonstatic initializer Id Block State ments where Block is a block of statements Statements The first version of Vmax visualized C Its grammar is not LALR 1 so a DCG was used 90 5 5 ANALYZING JAVA SOURCE CODE Statements are stored java statement Statement where Statement is the unique identifier of the statement Each block Block of statements is stored java block Block Statements where Statements is a list of statement identifiers A block statement is stored java block_statement Stat ement Block Statements where Block is the block of statements Statements that make up the block statement An expression statement is stored java expression statement Statement Ex pression where Expression is the term representing the
293. xt A B C D Declares a text component with string Text and identifier Id with corner anchors A B C and D Id thickness Thickness Attributes the thickness Thickness in point values to the line Id Id transparency Value Attributes the transparency Value between 0 and 1 to the object com ponent Id xalign P1 P2 Aligns two anchors P1 and P2 in the x axis yalign P1 P2 Aligns two anchors P1 and P2 in the y axis zalign P1 P2 Aligns two anchors P1 and P2 in the z axis B 6 Graphical Constraints Defined by SVT alignment Object Align Visual object Object is given alignment Align which is read by con taining objects such as tables Align may be e bottom Object is aligned to the bottom of the container e fill Object is expanded in both x and y axis to fill the container e left Object is aligned to the left of the container e right Object is aligned to the right of the container e top Object is aligned to the top of the container e xfill Object is expanded in the x axis to fill the container e yfill Object is expanded in the y axis to fill the container colour Object R G B Attributes visual object Object with the RGB colour value R G B contains A B Specifies that visual object A contains visual object B fill edge Id A B An edge between two visual objects A and B This type of edge is laid out as a hierarchical list graph Id X Y A graph Id with axes labelled X and Y graph_point Id X
294. xt VisContext can be acquired by issuing the query view_content ViewContext Content visual_content Content VisContext Visual visual_primitive Visual Constraint The graphical constraints on the right hand side of Figure 3 2 or 3 3 are passed to a graphical constraint solver as described in Section 4 5 Figure 3 4 shows the final output 50 3 8 SPECIFYING INTERACTION icon icon harry person A text text harry harry v_stack harry icon harry text harry 4 icon icon charles person text text charles charles 1 v_stack charles icon charles text charles i icon icon elizabeth person text text elizabeth elizabeth v_stack elizabeth icon elizabeth text elizabeth family_tree icon icon diana person k text text diana diana v_stack diana icon diana text diana le edge edge charles harry charles harry colour edge charles harry blue edge edge elizabeth charles elizabeth charles i colour edge elizabeth charles green edge edge diana harry diana harry i colour edge diana harry green View Combined Graphical Context Relation Constraints Q v G G X Figure 3 3 The combined relation BeN L e elizabeth diana F charles harry Figure 3 4 The family tree output by SVT 3 8 Specifying Interaction When the image has been generated the user can interact directly with the objects in the image for example with a mouse The user inp
295. y 15 10101001100 woman diana mother diana title Title muaa Title 11111000000 won player1 father charles E Title Input Data View Visual Graphical Output Source Content Content Constraints Image Figure 3 1 A semantic model of visualization The output is generated from a set of graphical constraints which include spatial associations such as adjacency containment or connections attributes such as colours and objects such as polygons and text This is a much more high level approach to specification and is completely generic Deriving the graphical constraints for a view is split into several steps The first step is data gathering from the input shown in Figure 3 1 to provide a data source in some database The information content in a view or view content is acquired by issuing a query to the database to construct a set of data A one to one mapping maps the information content to a graphical representation or visual content A third mapping defines the graphical constraints that compose each visual in the visual content We define a visual to be a set of graphical constraints that represents some meaningful graphical notation There are many advantages with this approach e Screen coordinates do not need to be specified but are derived automatically by layout algorithms Previous systems have only provided direct mapping to the display and are therefore less flexible in their graphical output e Itis possible to change th
296. y different levels of detail The elision control of most SV systems is very poor and there are no facilities for filtering away irrelevant information or for the user to specify what is or isn t of interest Existing SV tools or development environments have a very limited number of views of pro gram objects or output data In fact there are dozens of different ways of representing every pro gram object each conveying different information or using different visualization techniques 11 12 1 1 AIMS Many SV systems only have fixed output and changing the output requires manual program ming which hinders end users from tailoring their views The mapping from data to graphical form is fixed by programming or specification languages in SV systems but techniques for al lowing users to change this mapping interactively would allow end users to customize the views themselves Interaction with SV systems is also very limited Most views are not interactive so the objects in the view cannot be interacted with For example it might be useful to display information about objects beneath the mouse cursor or to provide a graphical display of whatever lies beneath the mouse cursor It might also be useful to display an automatic legend of each view because views containing many different types of graphical structures can be confusing Better navigation techniques would allow programmers to browse software and find information more quickly Th
297. y superior to text programmers can under stand complex software systems more easily thus decreasing development and debugging effort and improving documentation and code reuse Given that software is a trillion dollar industry even small gains should be considered important There are many claims of the benefits of the visual environments for software development CHAPTER 2 PREVIOUS WORK 19 and programming Blackwell 10 looks at the benefits computer scientists perceive of visual programming which also apply to software visualization He analyzes these claims in detail and finds a general consensus among computer scientists although some claims are indeed con troversial or counter empirical evidence Blackwell found that while few of these claims could be absolute most had a logical basis or some supporting evidence but the claims of superlativism of being natural and intuitive of utilizing more fully human cognitive resources and of increased information content were un sound However there are still many potential advantages of using imagery over text including increased productivity concretizing abstract concepts program structure being more easily con veyed pictorially assisting the programmer s mental model of software and resemblance to the real world thus utilizing our instinctive manipulative skills 90 There is some notable opposition to visual languages which is reasonable in the absence of much empirical eviden

Software Visualization in Prolog

Contents

Download Pdf Manuals

Related Search

Related Contents