a graph-based RDF data browser using

• Tulip 3.8.0, a graph analysis and visualization library bindings to the Qt application framework
• rdflib, an RDF library for Python

##development notes

###problem there is no browser that allows users to explore, scrutinize, and aggregate linked data.

###requirements

######need to have

• graph-based UI, showing nodes and edges as subjects/objects and predicates, respectively
• initialized with two persistent UI components
  • blank initial window for aggregated RDF
  • toolbar for changing interaction mode and entering URIs
• recursively explore chosen nodes and new URIs with subwindows showing returned RDF graphs
• select data in subwindows, triggering aggregation on the base graph
  • shortest path to aggregated nodes inherently selected, unless the aggregation is empty or disconnected
• seletively display values of literals when they arise

######nice to have

• visualized parameters for relatedness heuristic
  • egde weighting/coloring for common or important predicates
  • node colorizing and sizing for rediscovery, relatedness, etc.
• show/hide namespaced URI labels
• save base graph as rdf-xml file

###architecture

######MVC architecture

• model managed by rdflib
• view/controller managed by Tulip
  • the plan to use PyQt is no good - both tulip uses Qt and there is a conflict

######library choices

• Tulip 3.8.0, a graph analysis and visualization library
  • extensive python scripting API that is very well documented
  • tulipgui library for creation of standalone GUIs
  • not without bugs, but 3.8 is more stable than 4.x
    • 4.x do not support standalone gui applications
  • alternatives are Gephi
    • great for non-programmatic visualization, and is better maintained than Tulip
    • no scripting API to tap into and no library for standalone GUIs
  • and Graphvis
    • much older and more mature than Tulip, with better research documentation
    • documentation of scripting APIs not nearly as comprehensive as Tulip
• rdflib, an RDF library for Python
  • no real alternatives for python

######major components/data structures controller

• toolbar interface
  • now that tulip nodes will be used for the toolbar, command line input will be used for URI entry
    • since the point of this browser is to explore linked data, a single seed per session is reasonable
    • a single seed also makes for a single graph to serialize into a single document in the end. much simpler.
  • choose interaction mode: explore or select/aggregate
    • nodes forming the recursive exploration path are included in the aggregate to make it connected
    • all connecting edges between newly aggregated nodes are aggregated as well
    • interaction with a literal in either mode displays its value (in a popup? on terminal?)
  • quit
  • (eventually) save rdf-xml file
• observer object that does function calls to the other components based on event type
  • calls to model
    • exploration
      • URI string passed to model for dereferencing rdf document
    • (eventually) pass list of aggregated uris for serialization
• observer object that adds to aggregate graph in the event that the aggregate boolean property is set to true for a node or an edge
• parent tulip graph containing aggregate and all frontier graphs as subgraphs
  • each node and edge needs two string properties
    • content, for keeping track of the URI or literal that it represents
    • uriLiteral, so it knows whether its content is a URI or a literal value
  • each node and edge needs a boolean property
    • aggregated, to trigger the aggregate activity
  • for the purposes of maintaining consistency in node/edge properties, it seems easiest to keep frontier graphs as subgraphs of the parent, since node/edge properties can be dereferenced ubiquitously that way
    • the parent consequently holds all browsing history.
• translation function for converting RDF documents to tlp.Graph() structures
  • upon translation, calling the view is necessary to begin visualization

model

• history graph, containing all triples ever explored
• communication function for handling HTTP URI dereferencing

view

• window manager for tracking and placing exploration windows
• formatter handling visual properties
  • layout
  • colorization
  • node sizing
  • edge weighting
  • labels