Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Borislav Iordanov
    @bolerio
    Welcome to the HyperGraphDB chat channel!
    Pierre-David Bélanger
    @pierredavidbelanger
    Hi! I want to evaluate if hypergraphdb is well suited for my use case .. can I expose my use case and you guys could tell me if I should go forward and spent some time learning hypergraphdb ?
    Pierre-David Bélanger
    @pierredavidbelanger
    Ok, I guess I'll come back later.
    But in short: 2 processes: build time and run time
    build time: parse (huge) OSM data and insert vertices (OSM nodes) and edges (OSM ways) into hypergraphdb
    run time: query (dijkstra or A*) hypergraphdb to find fastest (or shortest) path from N locations to N locations, to build NxN matrix of time, where N can be somewhere 20 et 2000 locations
    Borislav Iordanov
    @bolerio
    @pierredavidbelanger Sadly , I didn't get notified by gitter that there was someone writing so I missed your question.
    I would ask you how huge is "huge". HyperGraphDB is generally for complex knowledge modeling, not for large-scale graph traversals as in path finding. So if you are looking for a conceptually simple problem like shortest paths for some GIS, HyperGraphDB is not the right choice. That said, if you do decide to try, I'd be interested to hear what comes out of your experiment!
    Pierre-David Bélanger
    @pierredavidbelanger
    Thank you for your (late but appreciated nonetheless) reply :)
    in one of my simplest use case, I would like to store something like 2 millions vertices and 1.5 millions edges
    and the average path finding query I will do should returns something like 100 vertices (with their edges)
    Pierre-David Bélanger
    @pierredavidbelanger
    this path finding query will be called N^2 times , where N is somewhere between 20 and 2000
    to build a NxN matrix
    this matrix will be rebuilt with different start/end vertex I couple of time per hour .. say 50 times
    I do not mind if the db population (creation) is slow, this process will runs async at night .. but I need the NxN matrix building (path finding query) to return fast
    Pierre-David Bélanger
    @pierredavidbelanger
    the solution I use right now is able to build me a 100x100 matrix in 5 seconds .. in a 2millions vertices graph
    Borislav Iordanov
    @bolerio
    2 million vertices and 1.5 edges is something HGDB should be able to handle
    I think that numbers should actually in a cache on any modern computer, so path finding should be as fast as with any graph representation
    shortest paths is implemented in a utility class, but since you are building a matrix of many paths, you might want to re-use results
    Borislav Iordanov
    @bolerio
    But again, my question stands: are you interested in HyperGraphDB as a graph that you could be traversing or as a graph-based knowledge representation? It is intended for the latter as much as I'd like to have it perform well for large-scale graph traversal algorithms as well...
    Pierre-David Bélanger
    @pierredavidbelanger
    i would only use it , indeed, for graph traversal