"user manual"
Contents
1. weight column a double is also understood by Zoomgraph to indicate the value used for various graph measures shortest path weighted betweenness etc It is notable that self loops are supported and are represented by listing the same node as the start and endpoint Multiple edges between two nodes are also supported Beyond these visual columns users may create whatever attributes they need These may be of string double or any valid SQL type Using these rules we can construct a more complex example Figure 3 representing a call graph for businesses in various cities In this example nodes sizes are declared by the user and each node contains an additional string indicating the city that node company resides in and the size of the company Edges contain a color definition as well as an integer column representing the number of calls between the nodes The visualization of this example is illustrated in Figure 3 DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION 3 2 Manipulating the Graph The Zoomgraph language provides a number of functions that take lists of primtive graph objects and acts on them For example the command nodecolor is followed by a new color and a list of nodes What is critical to users is the ability to easily reference the variables representing nodes and edges and to quickly generate sets of these While in certain cases this is achievable in other languages frequently the variable naming rules or graph representa2. Rogers red Since nodes are first class objects we can refer to them by name The second command sets the width of any edges between individuals in the finance department and individuals in Mexico DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION Beyond the command line Zoomgraph utilizes a zoomable user interface ZUI ZUIs are a class of graphical user interfaces that allow a user to zoom in and out of a workspace effectively introducing a third dimension to the more typically two dimensional navigation of computing environments Simple features such as centering on a selected node or zooming to focus on both end points of a clicked on edge greatly enhance navigability Our desire to implement a zoomable graph tool stemmed from a need to visualize large networks such as gene interaction networks or social communication networks in big organizations By large we mean graphs containing at least a few hundred nodes up to many thousands We sought to extend these ideas by creating a system that combines them in a smoothly animated real time environment with dynamic graph content Zoomgraph s ability to smoothly shift morph from one graph configuration to another enables the quick visualization of dynamic graphs which can then be saved as images or movies In addition to a number of built in functions Zoomgraph provides advanced statistical and mathematical network analysis by means of a connection to the R software
3. unique way There are many powerful mechanisms for handling dynamic graphs 8 that may be appropriate by require further study We have opted to provide the simplest solution initially A user may apply the layout algorithms previously mentioned to a graph utilizing the current node positions as a starting point rather than random By using an iterative layout algorithm such as the simple spring routine the user may specify the amount of change possible in the graph Users may also fix certain nodes to anchor them Sample animations are available on the Zoomgraph page at http www hpl hp com shl projects graphs movies html 4 2 Graph Analysis and R As much as possible we have avoided re implementing well known graph analysis algorithms While Zoomgraph does calculate certain basic statistics of the graph clustering coefficient density etc we have opted to use the R system for various tasks The RServe system 29 provides a Java based client library that communicates with a running R system over TCP IP The system provides a great deal of flexibility in the data that can be transmitted and received Zoomgraph will convert a graph into an R matrix and transmit the matrix to the R server At present we use a number of hard coded commands that perform specific functions on the matrix and return the results These are primarily centrality metrics betweenness information centraility closeness etc which are available in R s Social Network An
4. Huberman who has kindly allowed us to work on this project Lada Adamic one of our first users and the many implementers of the free software which we were able to use in this system These include the creators and developers of HSQLDB RePast Jazz as well as Simon Urbanek who kindly worked with us to get Rserve running REFERENCES 1 Adar E and J R Tyler Zoomgraph version 0 2 manual http www hpl hp com shl projects graphs doc zg manual htm 2 Auber D Tulip Proceedings of GD 01 Springer Verlag Lecture Notes in Computer Science vol 2265 2000 3 Batagelj V and A Mrvar Pajek Program for Large Analysis Connections 21 47 47 1998 4 Bederson B B and J D Hollan Pad A Zooming Graphical Interface for Exploring Alternate Interface Physics UIST 94 Marina del Ray CA Nov 2 4 1994 5 Bederson B B J Meyer and L Good Jazz An Extensible Zoomable User Interface Graphics Toolkit in Java UIST 2000 San Diego CA 2000 6 Brandes U M Eiglsperger I Herman M Himsolt and Marshall M S GraphML Progress Report Proceedings of GD 01 Springer Verlag Lecture Notes in Computer Science vol 2265 2001 7 Brandes U and D Wagner visone Analysis and Visualization of Social Networks Mathematics amp Visualization M Junger and P Mutzel eds Springer Verlag to appear Perlin K and D Fox Pad An Alternative Approach to the Computer Inte
5. Zoomgraph Eytan Adar Joshua R Tyler Information Dynamics Lab HP Laboratories 1501 Page Mill Road Palo Alto CA 94304 USA 1 650 857 1501 eytan jtyler hpl hp com ABSTRACT As graph models are applied to more widely varying fields researchers struggle with tools that are at times too general or too specific for their needs The Zoomgraph system attempts to address this issue by allowing users to define graphs with complex node and edge attributes and providing a command and scripting language to manipulate and explore these graphs The tool which is publicly available also includes a zoomable visualization system for rendered graphs that allows users to smoothly move between graph views Zoomgraph is a full featured graph analysis package with the ability to generate simple snapshots as well complex animations of dynamic graphs and to interface to external systems such as the R statistical system In this paper we will describe the Zoomgraph system and language and detail its novel features 1 INTRODUCTION As the popularity of graph models increase in domains as varied as social sciences organizational behavior physics and biological sciences the number of visualization options available to researchers has become almost overwhelming Researchers now must struggle to decide which tool best suited for his or her needs These tools are at times too general to handle the modeling of specific graph models or at times limited to one dom
6. ain Due to the nature of our own research involving biological social and computational networks we saw a need for a general purpose interactive tool that is flexible and reconfigurable enough to satisfy the demands of many different applications through one consistent interface Through our new language and interface Zoomgraph allows a user to define describe and visualize his or her data and is particularly intuitive to users with little programming experience yet provides advanced features based on a relational database model Nodes and edges in graphs naturally have varied attributes depending on the application Zoomgraph users can specify the types of data attributes and the possible relationships between them From this input Zoomgraph creates a database driven backend enabling a rich set of possible queries on the data Studying arbitrary graphs or networks within the database is made easy by Zoomgraph s simple command language With statements such as colornodes red Smith Rogers or edgewidth 1 dept finance country mexico users can quickly and simply explore and create informative graphs from their data For advanced users Zoomgraph can be driven by a SQL like relational query language and command line interface The language is supported both as an embedded interpreter as well as through scripts Both examples act on a social network The first example colors the nodes called Smith and
7. alysis SNA library The results of these functions are folded back into the Zoomgraph table as new columns Figure 4e demonstrates the simple way in which such a result is used DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION While certain commands are hard coded in Zoomgraph we also provide users with the ability to pass commands through to R and interact with the system as if it were local Users may simply type R commands and see the response directly inside the Zoomgraph interpreter In the future we hope to further integrate R and Zoomgraph to utilize functions such as clustering 5 CONCULSIONS AND FUTURE WORK While graph models are rapidly evolving to address many different domains and applications frequently the tools available to researchers have failed to evolve as quickly These tools can be too broad failing to take advantage of attributes of specific graph types or alternatively too narrow making their application to new domains difficult Both approaches limit the ability of a researcher to perform exploratory data analysis and visualization on graphs in a straightforward manner as researchers are unable to model their data fully Zoomgraph addresses this issue by allowing users to specify and manipulate the properties of graphs in a natural and general way For example a user in the social network domain who is studying corporate networks can construct a graph where nodes have attributes such as department and locations a
8. avior requires an additional infrastructure that provides access or storage of complex attributes of graph objects Another system Libsea 18 is a Java library that provides graph storage functions that was designed to handle graph objects with a wider set of attributes However it still does not provide visualization functions or a graph specific query language The database community and researchers leveraging database standards have also sought to address the storage and querying of graphs XML based approaches such as GraphML 6 and GXL 31 GML 16 etc allow for a set mechanism for describing and passing graphs between applications and provides the flexibility for additional attributes However such an approach still depends on the visualization program We have successfully loaded Zoomgraph with networks as large as 30 000 nodes and more is theoretically possible However more work needs be done to properly lay out these large graphs DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION to render and provide an interface for selection and updating of graph objects While using an XML standard provides access to evolved database technologies it also forces reliance on XML standards for selection and update i e the Web Consortium s XPath and XUpdate These standards are intended for very general applications and are thus unwieldy for the user wishing to manipulate graph objects in a natural Analysis way The Grap
9. hDB system 14 was designed Interpreter Components from the start to hold and query graph structures though it lacks visualization features Interface to R and thus depends on external drawing tools DBServer Systems such as Visone 7 and the commercial NetMiner 24 are powerful tools that utilize a graphical user interface for access SQL Database to graphs The Visone approach to object selection is to allow users to pick nodes and edges based on physical features color size shape etc and perform transformations or analyses based on the value of those features Such an approach however assumes that the data is initially partitioned in a way that the all attributes are directly convertible to a physical representation The NetMiner system provides slightly more sophistication by allowing additional features to be defined but limits the queries that may be applied to nodes to a simple Boolean scheme e g node attribute 1 equals x AND node attribute 2 equals y Fig 1 The basic Zoomgraph architecture We believe the system most similar to Zoomgraph is a new plug in 28 to the GGobi visualization package By tying into the R system a public implementation of the S language the system allows users to control and analyze graphs interactively using the S language While powerful this system uses the first class objects of the S language matrices vectors etc that naturally may be used to represent graphs but additi
10. hen creating a database users may opt to define the defaults or actual values for any of the visual attributes of graph objects as well as any additional columns At minimum a node definition must contain a name can simply be a numerical identifier and an edge must contain two endpoints corresponding to defined nodes A very simple graph definition file is illustrated in Figure 2 This file represents a graph with two nodes N1 and N2 and an undirected edge between them The nodedef and edgedef lines describe the schema for the input graph represented in basic SQL This approach was selected because the connection between the file and the database is obvious In the future we may move to a format such as GraphML as it gains acceptance In the current instantiation Zoomgraph is aware of the following visual properties e For all objects the column vis a bit indicates if the object should be displayed and color a string represents the color of the object e For nodes the columns xloc and yloc double values indicate the x and y location of the node fixed a bit indicates if the node can be moved size a double controls the size of the node and shape an int indicates the shape of the object from a set currently circles and squares e Edge properties include the width column a double and a directed column an integer indicating the directionality of the edge i e n1 n2 n1 fl n2 n1 fl n2 or an undirected link nl n2 A
11. ifying features of the endpoints For example to change the color of all edges between companies with 100 employees and companies with 200 employees only one in the above example B C a user would type edgecolor blue companysize 100 companysize 200 Zoomgraph commands exist to control all visual aspects of the graph including background colors rendering quality object colors sizes widths etc What view appears in the zoomable interface is controlled by a camera which can also be manipulated by the language For example the command center A B will zoom and pan the display to fit nodes A and B in the screen A full list of commands is available in the Zoomgraph user manual 1 but some examples are illustrated in Figure 4 For more complex commands beyond simple color manipulation it is notable that optional parameters are grouped between square brackets Zoomgraph currently contains a number of simple layout algorithms A spring based algorithm adapted from the Java libraries a random layout circular Kamada Kawai and Fruchterman Reingold layouts as implemented by 25 and a simple hierarchical layout The layout commands can be applied to specific sets of nodes and will not act on fixed nodes Users wishing to save an image of the graph currently can create jpeg and SVG format files 3 3 Scripting Frequently users may wish to perform the same operation many times or multiple times with minor modificatio
12. mber of milliseconds This allows one to automate visual analysis of a graph without worrying about the animation running too fast 3 4 Subgraphs One final important structure supported by the Zoomgraph language is the subgraph Subgraphs are hierarchically structured objects that encapsulate nodes edges and other subgraphs The contained objects can be arbitrarily selected from the graphs they do not need to be connected continuous etc Nodes Edges A A B Figure 5 demonstrates a possible subgraph structure The B subgraph sgA contains two nodes and an edge as well as pointers to two other subgraphs sgB and sgC Creation of such a subgraph Nodes Edges is simple in the language of Zoomgraph The initial creation is D D E accomplished through the command sg create sgA Adding E nodes and edges is through the same selection criteria previously described sg addnodes sgA A B and sg addedges sgA A B Adding subgraphs is equally simple addsg Fig 5 A simple subgraph structure The three subgraphs sgA sgB and Subgraphs reside in memory and are accessible through the sgC hold pointers to each other as name by which they were initially created It is possible to delete well as nodes and edges and copy these subgraphs and any component in them Programmatically iterators allow us to access all nodes A B D and E in Figure 5 and edges A B and D E Zoomgraphs commands are aware of when they are applied to
13. n To support this Zoomgraph allows users to save commands into a file To support repeated operation the Zoomgraph includes a foreach statement which will iterate over a set of objects nodes 3 If an edge is defined to be bi directional say Afl B it can still be accessed by A gt B in the Zoomgraph language Some further work is necessary to find if this makes semantic sense to users DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION nodedef gt name VARCHAR 256 XLOC DOUBLE YLOC DOUBLE grp VARCHAR 30 size DOUBLE shape TINYINT A 30 10 groupl1 2 1 E 50 55 group2 2 2 edgedef gt nl VARCHAR 256 n2 VAR CHAR 256 freq INTEGER width DOUBLE default 1 A B 10 1 D B 20 1 E G 4 1 a center the display center show the labels labelnodes b ot c hide some edges hideedges B E B F F G E G show everything showall d e calculate the betweenness on each node analysis betweenness color each node based on the score starting with light gray betweenness and going to black analysis colorize betweenness linear true startcolor lightGray endcolor black betweenness for low hide the disconnected nodes G color each group differently hidedis nodecolor black grp groupl nodecolor lightGray grp group2 E B E E F 4 hide some edges hideedges A B B D C B find the connected components subgraphs automatically cc_A_3 and cc_A_4 analy
14. nal of Visual Languages and Computing 7 1 33 55 1996 18 The LibSea Graph File Format and Java Graph Library CAIDA 2003 http www caida org tools visualization libsea 19 Krackhardt D J Blythe and C McGrath KrackPlot 3 0 User s Manual Analytic Techologies 1995 20 Mackinlay J D G G Robertson and S K Card The perspective wall Detail and context smoothly integrated CHI 91 New Orleans LA 1991 21 Marshall M S I Herman and G Malancon An object oriented design for graph visualization 2000 CWI Technical Report INS RO001 22 Misue K P Eades W Lai and K Sugiyama Layout Adjustment and the Mental Map Journal of Visual Languages and Computing 6 2 183 210 1995 23 Morse A Some principles for the effective display of data International Conference on Computer Graphics and Interactive Techniques Chicago IL 1979 24 NetMiner Cyram Co Ltd 2003 http www netminer com 25 RePast 2003 http repast sourceforge net 26 Robertson G G J D Mackinlay and S K Card Cone Trees animated 3D visualizations of hierarchical information CHI 91 New Orleans LA 1991 27 Sutherland I Sketchpad A man machine graphical communication system IFIPS Proceedings of the Spring Joint Computer Conference Detroit Michigan January 1963 28 Swayne D F B Andreas and D T Lang Exploratory Visual Analysis of Graphs in GGobi Worksho
15. nd edges with properties such as frequency of communication over e mail face to face etc Through Zoomgraph s interactive language as well as a zoomable user interface the researcher may manipulate and explore the visual representation of the graph based on these properties They can for example display only edges between two specific departments where communication frequency exceeds a certain threshold In a totally different context a biologist studying the discovery of protein protein interactions in yeast can specify function categories on the nodes and annotating edges with the type of experiment used to validate the interaction The Zoomgraph system implemented in portable Java attempts to provide a set of functions that are useful to applications requiring interactive or scripted manipulation of graphs In addition to its flexibility Zoomgraph ensures usability by abstracting many details of graph layout and building on conventional languages particularly SQL Beyond the novel zoomable interface which allows users to easily traverse complex graphs Zoomgraph includes a number of standard graph drawing algorithms as well as hierarchical and dynamic algorithms that take advantage of the database backend of the system Beyond visualization Zoomgraph provides analytical tools for generating graph and node statistics as well as a mechanism for interactively utilizing the R statistical package Zoomgraph is continually evolving as it suppor
16. onal steps are often required by the user for simple operations As we will show the Zoomgraph approach is to consider nodes and edges as first class objects and transform those into R matrices for analysis if needed Zoomgraph is not an all inclusive package Frequently the tools mentioned above provide unique features more advanced layout algorithms more efficient analysis algorithms etc that Zoomgraph lacks Where appropriate we have built in facility to exchange data with these applications currently Pajek and other simple comma separated formats However we believe that the approach we describe below combining a new language targeted specifically at graph objects and visualizing those structures extends the current state of the art 2 2 Zoomable Graph Visualization A zoomable interface is an attractive approach for presenting large graphs as it enables a user to see both the high level structure of the entire graph as well as zoom in and navigate specific subsections in detail Other approaches for visualizing large networks incorporate focus context techniques which allow the user to focus on some detail in the graph without losing the broader context Examples of these techniques include three dimensional constructions such as a graph projected onto hyperbolic space 17 DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION fisheye distortions 11 perspective based information walls 20 and cone trees 25 The
17. p on Distributed Statistical Computing DSC 2003 Vienna Austria March 20 22 2003 29 Urbanek S Rserve A Fast Way to Provide R Functionality to Applications Workshop on Distributed Statistical Computing DSC 2003 Vienna Austria March 20 22 2003 30 The Walrus Graph Visualization Tool CAIDA 2003 http www caida org tools visualization walrus 31 Winter A B Kullbach and V Riediger An Overview of the GXL Graph Exchange Language Software Visualization S Diehl ed Springer Verlag 2001 32 yWorks 2003 http www yworks de DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION
18. package Results of this analysis e g the betweenness centrality measure can be integrated into the Zoomgraph database enabling visualizations based on these results e g different colors depending on betweenness values 1 1 Paper Roadmap In Section 2 we will briefly discuss related systems Sections 3 and 4 will cover the interactive language used in Zoomgraph and some advanced features mentioned above We conclude with a discussion of ongoing work on the system 2 RELATED WORK 2 1 Graph Exploration and Manipulation In general graph visualization tends to be a second separate step from analysis Analysis packages such as Ucinet 10 are used for partitioning the graph and performing various calculations The output of these programs is then passed to external programs such as Pajek 3 Netdraw 9 GraphViz 12 or KrackPlot 19 to name a few While powerful in rendering and performing layout operations these packages by necessity tend to serve only the most common required features for graphs colors shapes etc Exploring the graph based on more complex node and edge properties requires going back and forth between the analysis and the visualization packages Some address this problem by generating a common API for the manipulation of graph objects in memory as well as visually 21 13 32 Unfortunately this approach requires compilation and does not allow the user to interact with the graph dynamically To support such beh
19. re is also a distinction to be made between the geometric zooming of a system like Zoomgraph Walrus 30 or Tulip 1 and a semantic zooming system in which the information content presented and level of detail changes as the user navigates in and out of a graph for a survey of these and other styles of graph visualization and navigation see 15 The use of ZUIs in computing dates back as far as Sutherland s Sketchpad 27 by virtue of its use of vector graphics rather than bitmaps More recently the Pad system and its descendants have explored the use of zoomable interfaces for shared online workspaces knowledge representation and many other sorts of information visualization applications One such descendant Piccolo provides Java libraries and APIs for developing zoomable interface applications and is the foundation of the zooming features in Zoomgraph specifically we use a previous version called Jazz 5 These systems aim to take advantage of our natural spatial and geographic ways of thinking 4 23 to help us understand and navigate volumes of information The general design principle of these systems is that objects are placed in an infinite two dimensional plane and the user is able to navigate and interact freely in these two dimensions as well as zoom in for more detail or zoom out for a broader view of the space 3 LANGUAGE FOR MANIPULATING AND EXPLORING GRAPHS The Zoomgraph system as illustrated in Figure 1 consi
20. resentation Zoomgraph objects are of two principal types nodes and edges A third type subgraphs function to store groups of nodes edges and other subgraphs gt R nodedef gt name VARCHAR 3 and will be discussed in Section 3 4 N1 Data in Zoomgraph is stored within a relational database in two tables one holding nodes and the other edges Tables contain a set of columns that N2 edgedef gt nl VARCHAR 3 n2 VARCHAR 3 have significance to Zoomgraph color shape Maene width etc for visualization purposes as well as any additional information the user supplied about the nodes and edges Each row representing either a Fig 2 A simple database definition containing 2 node or an edge is uniquely identified by an id field nodes and 1 edge DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION nodedef gt name VARCHAR 32 SIZE DOUBLE COMPANYSIZE DOUBLE CITY VARCHAR 32 A 300 30 new york B 100 10 san jose C 200 20 san jose edgedef gt nl VARCHAR 256 n2 VARCHAR 256 COLOR VARCHAR 32 DEFAULT green CALLS INT DEFAULT 0 A B black 40 B C lightGray 30 Fig 3 A more complex graph definition and the visualization it produces Rows are reconstituted by the DBServer into a Java object with most visual characteristics held in an object field for caching purposes Zoomgraph takes responsibility for maintaining consistency between the Java object and the underlying database W
21. rface ACM SIGGRAPH 93 New York NY 1993 8 Branke J Dynamic Graph Drawing Graph Drawing Models and Algorithms M Kaufmann and D Wager eds Springer 2002 pp 228 246 9 Borgatti S NetDraw 1 Analytic Technologies http www analytictech com downloadnd htm 10 Borgatti S M G Everett and L C Freeman UCINET Analytic Technologies http www analytictech com 11 Furnas G Generalized fisheye views In Proceedings of the ACM Conference on Human Factors in Computer Systems Boston MA April 1986 12 Gansner E R and S C North An open visualization system and its applications to software engineering Software Practice and Experience 30 11 1203 1233 2000 13 Graph Foundation Classes for Java IBM 1999 http www alphaworks ibm com tech gfc 14 G ting R H GraphDB Modeling and Querying Graphs in Databases VLDB 94 Santiago De Chile Chile Sep 12 15 1994 15 Herman I G Melancon and M S Marshall Graph Visualization and Navigation in Information Visualization A Survey IEEE Transactions on Visualization and Computer Graphics 6 1 24 43 2000 DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION 16 Himsolt M GML A Portable Graph File Format http www infosun fmi uni passau de Graphlet GML gml tr html 17 Lamping J and R Rao The Hyperbolic Browser A Focus context Technique for Visualizing Large Hierarchies Jour
22. sis concom color the nodes and edges of each subgraph nodecolor black cc_A_3 edgecolor black cc_B_4 f this generates two g color the edges based on the freq column starting from white going to black analysis colorize freq linear true startcolor white endcolor black table edge s reset the node and edge colors to light gray nodecolor lightGray edgecolor lightGray color the edges between the two groups black edgecolor black grp groupl1 grp h group2 Fig 4a h The application of various commands to the graph defined in a DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION edges or subgraphs and perform a specific task One simple application is to create an image file that highlights specific nodes in a set T foreach nodex Zee center name nodeX 33 nodecolor red nodexX 4 savejpg nodeX jpg Ds nodecolor blue nodex In this example line 1 tells Zoomgraph over which nodes to iterate all For each node in the set the variable nodeX will be replaced by a specific node name Line 2 causes the display to center on all edges to and from nodeX The color of nodeX is changed to red line 3 and a jpeg image is saved of that view line 4 Finally the color of the node is returned to the original blue line 5 For each loops may be nested to support more complex behaviors Users may also utilize the pause keyword followed by a nu
23. sts of a number of components At the bottom layer resides a SQL database which retains node and edge data Zoomgraph s DBServer transforms entries in the database into renderable objects These are passed through the main interpreter to a visualization UI component The embedded interpreter processes the user commands or scripts to manipulate the visualization and database as well as transfer information to and from other applications such as R The Zoomgraph interpreter has been optimized to handle the concept of nodes and edges While it was possible to provide some functionality through existing languages this frequently forces the user to think of graph elements as non first order objects For example a system that represents graphs as matrices requires edges to be addressed as an access to a two dimensional array where the index of desired nodes needs to be looked up Instead Zoomgraph provides a more natural interface to these objects An edge between the node abc and def is simply abc def in the Zoomgraph language If we would like to access the directed edge between the two nodes from abc to def we can simply say abc gt def Below we provide a few details on data representation within Zoomgraph setting the stage for a discussion on how the language is used to manipulate node and edge properties how scripting works and finally the built in support for grouping of nodes and edges into subgraphs 3 1 Data Types and Database Rep
24. subgraphs and will extract the appropriate sub object For example colornodes red sgA will color nodes A B D and Ered Similar rules apply to edges and nested subgraphs can be accessed by adding a to the name of the higher level subgraph sgA is the set sgB and sgC Zoomgraph will also generate a visual convex hull around the nodes of a subgraph It is notable that while users can manually create subgraphs various analysis commands will also generate groupings automatically For example finding connected components through the analysis concom will create a subgraph holding each connected component of the graph see Figure 4f The Zoomgraph manual 1 details the subgraph commands in more detail DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION 4 EXTENDED FEATURES 4 1 Graph States and Dynamic Layouts One advantageous property of utilizing a database backend in the Zoomgraph system is that it allows us to quickly switch betweem different graph properties For example in the study of communication patterns in an organization it is valuable to be able to maintain data for different weeks in order to track changes in communication patterns Zoomgraph supports the saving and uploading of new tables to represent different states and enables the smooth transition between them Through the command line users may save out the current graph state into a new table by typing savestate state_name where state_name can be an
25. tion restrict the flexibility that we would like to expose to users In Zoomgraph there are two primary ways to select nodes The first is to address them by name In the example in Figure 3 if the user wanted to change the color of nodes A and B to red they could simply type nodecolor red A B The other mechanism for node selection is through what is essentially a SQL WHERE clause To hide all nodes where the size is more than 10 and with the city san jose just node C in this example the user enters hidenodes size gt 10 AND city san jose The interpreter is aware of which type of selection is being used and automatically constructs a result set and acts on it For both nodes and edges means all objects of that type Edges can be accessed in a similar way Edge names are simply the two end nodes with a dash between them The order of nodes is not important so the command edgewidth 1 A B is equivalent to edgewidth 1 B A both set the width of the edge to pixel If more than one edge exists between the two nodes Zoomgraph assumes that all matching edges are the target of the command To choose a specific edge in this case a user may select an edge by its unique identifier Although this example is for an undirected graph Zoomgraph supports directed edges In this case an edge can be referenced by A gt B A lt B or A lt gt B depending on the graph definition Finally edges can be selected by spec
26. ts a number of our own internal graph based projects We have released Zoomgraph to the public and hope to continue making new versions available There are many features that one could envision for Zoomgraph For example it would be powerful to support the ability to select complex paths from graph Instead of simply finding edges between two connected nodes one may wish to list a chain of nodes with specific properties that should be connected Other potential additions include the ability to specify template objects and apply those to groups the use of clustering algorithms to generate groups based on different properties and the ability to simulate dynamic information message passing on graphs As we observe the users of Zoomgraph and our own usage we also hope to further refine the Zoomgraph language Building managing and visualizing large graphs and networks is a challenging and important problem which is evidenced by the proliferation of tools intended to address it Even as it currently stands Zoomgraph as a data driven zoomable and statistical analysis enabled software package allows users to explore many different applications and datasets with a minimal amount of transitional effort DRAFT PLEASE DO NOT DISTRIBUTE WITH PERMISSION 6 AVAILABILITY Zoomgraph both binaries and sources is available from our website at http www hpl hp com shl projects graphs 7 ACKNOWLEDGEMENTS The authors would like to thank Bernardo
27. y string A state is restored by the command loadstate state_name which causes the graph to shift back to the previously stored configuration colors locations etc While powerful this is not always consistent with the mental map model 22 as nodes can shift dramatically between states To handle this we have created the command morph which accepts a series of state names and transitions between them over a configurable amount of time Node and edge colors and locations smoothly transition during this animation The user has a wide variety of cinematographic controls over the animations Movies in QuickTime format can be generated with control over the frames per second the speed of the animation node highlighting camera control and how edges and nodes appear and disappear We have found the last to be critical for generating animations that are not too confusing For example certain layouts force disconnected nodes to the periphery of the screen As they are pulled back when they are connected long edges fill the display making tracking nodes difficult By offsetting the appearance disappearance can be similarly managed of edges in the animation cycle edges no longer fill the screen but appear only as the node reaches its final destination Camera control further helps as the display can follow specific nodes as they move or keep certain nodes or regions centered We have not currently addressed the issue of dynamic layouts in a
Download Pdf Manuals
Related Search