New Developments in Space Syntax Software Edited by Alasdair Turner ITU Faculty of Architecture Istanbul 2007 Contents Preface v Spatial Positioning Tool (SPOT) 1 H. Markhede, P. Miranda Carranza WebmapAtHome 7 N. S. Dalton Confeego: Tool Set for Spatial Configuration Studies 15 J. Gil, C. Stutz, A. Chiaradia Syntax2D: An Open Source Software Platform for Space Syntax Analysis 23 J. Wineman, J. Turner, S. Psarra, S. K. Jung, N. Senske Segmen: A Programmable Application Environment for Processing Axial Maps 27 S. Iida Place Syntax Tool — GIS Software for Analysing Geographic Accessibility with Axial Lines 35 A. St˚ahle, L. Marcus, A. Karlstr¨om UCL Depthmap 7: From Isovist Analysis to Generic Spatial Network Analysis 43 A. Turner iv Contents Preface These proceedings cover the Workshop on New Developments in Space Syntax Software held on 12thJune2007atthe6thInternationalSpaceSyntaxSymposium,IstanbulTechnicalUniversity. This is the first time that the New Developments workshop has its own proceedings. In the past the workshop has tended to focus on just one or two pieces of software, usually from the historic heart of space syntax research, the Bartlett at UCL. This has led to undue prominence of the Bartlettinafieldwherethereisanincreasingnumberofcontributionsfromtheworldover. These proceedingshavemanagedtocatchafewofthelatestprogramswrittentoperformspacesyntax analysis,oratleast,programstomeasurerelationshipsbetweenoccupiablespace,whateverform that may take. Thefirstpaper, MarkhedeandMiranda’sSPOTtool, focusesonjustthatphrase‘occupiable space’. They argue that space syntax tools have largely ignored the special status of occupied space by giving precedence to ‘all-to-all’ measures of space, such as integration, and so they have implemented a tool to examine the relationships between occupied spaces. Markhede and Miranda’s paper offers a new insight into how we approach space syntax. They are followed by a stalwart of space syntax tool creators, Dalton. Dalton presents a standard syntax tool, WebmapAtHome, and his argument here is not focused on new ways to approach space syntax, althoughWebmapAtHomehasnewmeasuresbackedupbyrecentresearchbyDalton,butrather whatthetoolitselfshouldbe. Dalton’sconvictionisthatatoolshouldbeself-contained,allowing the user to perform all the necessary operations, from drawing a system of axial lines through analysing them to statistical analysis of them within a single package. Furthermore, Dalton’s program is cross-platform (that is, it works on Macs as well as PCs). Thus, the user is not tied to any specific drawing or statistics package, nor, indeed, any particular computer. Gil et al present the opposite approach: an industry solution to the space syntax tool. Confeego’s very advantages are in the fact that it is locked into a specific program, MapInfo. MapInfo provides powerful mapping and visualisation of results, so Gil et al’s program merely needs to do the analytic work. In addition, Confeego offers interoperability with other systems, such as JMP for statistics support, meaning that a user who is familiar with the basic tools can easily access additional space syntax features. Wineman et al address another shibboleth in their paper, the fact that until now almost all space syntax software has not been ‘open source’. Their tool, Syntax2D aims to remedy this sit- uation. Indeed,theyargue,itisimperativethatprogramsareopensourcesothatthealgorithms they use are open to inspection. If space syntax is to be thought of as scientific, then making an experiment repeatable entails making the algorithm itself readable and repeatable. Moreover, public institutions such as universities should not be in the business of hiding information be- hindcompiledcode(wherethe‘sourcecode’,whichisreadablebyanexpert,hasbeenoptimised to run on a computer and can no longer be understood), but instead making that information availabletothepublic. ItisanargumenteasilymadefromarichuniversityintheUSA,butitis somethingthatUKuniversities,andperhapsothersintherestoftheworld,needtorelearn: how to survive financially yet exist as a public institution. Wineman et al also remind us of the need to cite software. This gives credit to those who do not necessarily report their work in papers, rather within the code they write; but in addition, it serves another important purpose relating the issue of repeatability. A program may develop over time and the algorithms it employs may change as it does so; therefore, if an experiment is to be repeatable, a scientist must be able to vi Preface find the program that ran the experiment in the first place. Consequently, I would like to add to Wineman et al’s demand to reference the original authors of software, and demand that the version of the software used is also explicitly referenced. While Wineman et al argue for academic accountability, Iida takes on its counterpart, aca- demic rigour, with his Segmen tool. Segmen is an old-style ‘command line’ program (meaning that the user must type in instructions for it to work), but it has been put together with such painstakingcarethatitsmeasurescanbeconsideredabenchmarkforotherprograms. Iidamar- ries this careful approach with extensibility: the experienced user can themselves rewrite many partswithinhisprograminordertoinvestigatenewmeasuresordifferentalgorithmicapproaches. Segmen, as its name suggests, primarily exists to analyse segmented systems, where axial lines are chopped at their intersections, leading to new measures from the axial basis. St˚ahle et al’s Place Syntax tool also extends axial analysis, but in another direction: that of including axial steps as part of the wider investigation of accessibility. The Place Syntax tool, like Markhede andMiranda’sSPOTtool, turnsaroundhowweconceivespacesyntaxtotheaccessibilityofthe occupied space, and its relationship to other spaces, and other social factors contained within thosespaces. Itismybeliefthatthesereinventionsofspacesyntaxareimperativetoitssurvival asadiscipline, andthusmyownpaperonDepthmaphighlightsthefactthatweshouldcontinu- ally revitalise our software through research, whilst retaining the ability to perform the original space syntax measures in an easy and approachable manner for new and future researchers. ThisworkshopcouldnothavetakenplacewithoutthededicationoftheSpaceSyntaxSympo- siumorganisers. IwouldliketothankinparticulartheSymposiumchair,ProfAy¸seSemaKubat for her support, the workshop assistants, H. Serdar Kaya and Mehmet Topc¸u, and Yasemin I˙nce Gu¨ney who organised printing these proceedings. A.T. London, May 2007 Spatial Positioning Tool (SPOT) Henrik Markhede and Pablo Miranda Carranza School of Architecture, KTH, Stockholm, Sweden Abstract. The Spatial Positioning tool (SPOT) is an isovist-based spatial analysis software, and is writteninJavaworkingasastand-aloneprogram. SPOTdiffersfromregularSpacesyntaxsoftwareasit canproduceintegrationgraphsandintervisibilitygraphsfromaselectionofpositions. Theconceptofthe software originates from a series of field studies on building interiors highly influenced by organizations and social groups. We have developed SPOT as a prototype. Basic SPOT operations use selections of positions and creations of isovist sets. The sets are color-coded and layered; the layers can be activated andvisiblebybeingturnedonoroff. Atthispoint,therearetwographsproducedinSPOT,theisovist overlap graph that shows intervisibility between overlapping isovist fields and the network integration analysisbuiltonvisibilityrelations. Theprogramaimstobeusedasafastandinteractivesketchtoolas well as a precise analysis tool. Data, images, and diagrams can be exported for use in conjunction with other CAD or illustration programs. The first stage of development is to have a functioning prototype withtheimplementationofallthebasicalgorithmsandaminimalbasicfunctionalityinrespecttouser interaction. 1 Introduction Regular space syntax tools (such as Depthmap) create graphs of the occupiable space (Turner, 2004); however, the Spatial Positioning Tool (SPOT) creates graphs of the occupied. Therefore, SPOT is not strictly a tool for spatial analysis; it analyzes how organizational entities occupy space in relation to each other. The main difference from the regular space syntax graph is that SPOT produces graphs of the distribution in space where the former is an analysis of the distribution of space (Turner and Penn, 1999). This shift is relevant when analyzing spaces occupied by a specific organization. There are several basic specifications for building the software: the possibility to create an uneven distribution of isovists; layer systems with color-coding; and calculation of integration measures. During the process, new functions have been developed and other concepts of mea- sures were added. Building the software has in itself been an investigation of how to create representations of social and spatial relations. In addition, we will use SPOT as a platform for further investigations of socio-spatial representations. The software was produced during a one-month period and work as a prototype, for academic use only. To produce graphs with SPOT, we use isovists to represent space because the phenomena studied here are related to visibility. An isovist, having its origins in analysis of sightlines, is an attempttorepresentwhatcanbeseenfromonepositionorareainspace(360degreeorientation) as represented by a two-dimensional slice of visibility (Benedikt, 1979). In the SPOT graph is both the origin of the isovist as well as the field part of the analysis both together and on their own. The isovist on its own can also be interpreted in different ways: position, internalized to a seeing subject, what is seen from one position, and externalized and interpreted as a field seen fromtheoriginoftheisovist. Thisdoublerelationshipwithintheisovistisaboutwhoorwhatis exposed and what who or what is exposed to. It tells us something about the spatial strategies 2 SPOT used to form and inhibit social relations through space. ThispaperprovidesatechnicaldescriptionofSPOTandexampleofhowagraphisproduced. As the software is not complete, we will also discuss further development. 2 Platform and technical specification SPOTiswritteninJava,andthusitisplatformindependent;itrunsinWindows,MacandUnix machines. BecauseitiswritteninJava,andthereforeitisstrictlyObjectOriented,itisintended to be modular and easy to re-use, allowing the specific modules for calculating isovists, their graphs and topological relations, to be used in different contexts and in combination with other Java packages. The intention of SPOT is to build a platform that will allow future expansions and research, not merely to provide a solution to a specific problem. SPOT also implements a simple user interface, and it is the intention keeping it simple and easy to use, if necessary to divide functionality in to different programs (which its modularity allows) instead of adding features to the existing basic program. The calculation process in SPOT starts by reading a file of dxf format which contains a de- scription of the two dimensional geometry we want to analyse (sections or plans, for example), and then it breaks that geometry in two ‘wall’ components, made of simple lines. This geomet- rical description is stored into an ‘Environment’ class which all isovist objects share and can interrogate. ThemostimportantcomponentoftheprogramistheIsovistclass, whichusestheoperations availablethroughthejava.geom. Areaclass,partofthestandardJavalibraries,tomanipulate2D polygons. The calculation of an isovist is done quite simply, by generating an initial rectangular polygon that covers the total area of the specified drawing, and by subtracting the polygons of the ‘shadow’ areas produced by the position of the isovist. Because of the speed of these calculations isovists can be added, deleted, moved in real time or organised in layers which can be turned on or off. Below is shown a part of the code regarding the process of calculation: Isovist(Environment enviro, float cx, float cy) { this.enviro = enviro; vpt = new Point2D.Float(cx, cy); isopol = new Area(enviro.perimeter); rangex = enviro.perimeter.width; rangey = enviro.perimeter.height; maxrange = rangex > rangey ? rangex : rangey; int nw = enviro.walls.size(); for (int i = 0; i < nw; i++) { Area shadow = calcShadow(enviro.walls.elementAt(i)); isopol.subtract(shadow); } } The code snippet defines the constructor method of the Isovist class, showing the process of calculation. The constructor takes 3 arguments: an object of type Environment, which is the class that stores the description of the 2d geometry from which to calculate the isovists, and the xandy ofthecentreoftheisovist. TheEnvironmentclasshasafieldconsistingofarectangular polygon called ‘perimeter’, which is taken as the initial polygonal area of the isovist (an Area H. Markhede, P. Miranda Carranza 3 Figure 1: Screenshot showing Spot windows with imported dxf line representation, and two examples of positioning isovists. object called isopol). Then the constructor iterates through all the ‘walls’ of the Environment object (enviro) and subtracts their shadow areas from its polygonal area. The resulting Area object is the polygonal boundary of the isovist. A number of calculations can also be performed in these isovists; a particularly interesting one is the difference between the point from which it is calculated and the centroid of its area. This shows a certain directionality of the space from where it is being perceived, which seems to intuitively relate to a certain directionality feeling of that space. More in depth studies on how this measure may relate to actual empirical data needs to be done, but at this stage it suggests a possible interesting measure, combined with other geometrical properties of isovists. The interface is based on windows with four menus: file, edit, view, and help. Under file is the open and store functions. Under edit is the operative commands, add and delete isovist, delete all, and the layer manager. In the layer manager it is possible to create and delete a layer and assign colours to the set of isovists within the layer. Layers can also be turned on and off. Undertheviewmenuitispossibletochoosewhichinformationtoviewintheworkarea. Graph featurescanbeturnedonandoff. Underthehelpmenuisthispapertobefound. Whencreating the graphs in SPOT, one can move around, delete, or add positions and see the graphs change in real-time. The main function of the program is to import line drawings and position isovists within the line drawing. The isovists are positioned by using the pointer and clicking within the drawing area. Whendoingthisisanisovistfieldexpanded, limitedbytheimportedlinedrawingandthe drawing area box which gives an outer limit for the isovists. Theapplicationofseveralisovistswithdifferentpositionsbutwithoverlappingfieldsgivesrise to a differentiation in colour among the overlaps due to the gamma transparency. There are two kinds of graphs produced in this version of SPOT. Intervisibility which is an overlapping isovist fields graph, and network graph showing relative asymmetry (RA) integration. The graph of overlapping isovist fields is very crude and is not yet possible to calculate with any space syntax integration measure. The gamma structure of the isovist field creates a visual effect giving the range of the graph. The range depends on the amount of layered isovists, where there are many overlapping isovist the graph becomes darker and where its fever it becomes lighter. The graph 4 SPOT Figure2: Anexampleofintervisibilitygraphandnetworkintegrationgraph, generatedfromthe same positions. can be said to show the visual situation created by the selected positions. When choosing the show centroid command (figure 3) is an arrow within the isovist lit. The arrow goes from the isovist’s starting point to the centre of the isovist. This feature is added only for further evaluation and is not backed up with any systematic studies. All graphs are related to layers that can be turned on and off. Each isovist position carries informationaboutitsrelationwithotherisovistpositions. Whenusingtheshowgraphcommand, a visibility network between the isovist positions is secleted. Each node of the network shows an RA (relative asymmetry) value and a circle; the size of the circle depends on the RA value. Thereisalsoalineshownbetweenthosepositionsseeingeachother(figure3). Thethickness of the line indicates the distance between the positions, following a ten grade scale depending uponthesizeofthelinedrawing. TheRAdescribestheintegrationofanodebyavaluebetween or equal to 0 and 1, where all low values describe high interactions. RA is calculated by the formula RA=2(MD−1)/(k−2). When using the move isovist command, it is possible to click and drag the isovist to any position in the drawing area. The isovist, centroid, and network graph changes in real-time and along the transportation between the positions. The isovists can be divided into different sets that can be put in different layers. All layers have an assigned color that is also seen on the isovists. Each layer can be turned on and off. When a layer is turned on its set of isovists automatically becomes a part of the graph. Drawinginterchangeformat(DXF)is the onlydrawingformat importable. TheDXFformat is a tagged data representation of all the information contained in an AutoCAD drawing file. Tagged data means that each data element in the file is preceded by an integer number that is called a group code. A group code’s value indicates what type of data element follows. Virtually all user-specified information in a drawing file can be represented in DXF format (Autodesk, 2006). The program was built during a one-month period and finished during the production of this paper. There are some minor bugs regarding importing and exporting data and sometimes in the order of appearance of the layers and graphs; however, it works for academic use and will be available when this paper is published.