Search code examples
orientdbtraversalvertex

OrientDB Edge From To Confusion

I'm new using Graph Databases and I'm a little confused about the concept of From / To or In/Out in OrientDB.

Which is the best approach to define which Vertex is From and which is To? Or Both for example when is a relation of contact in social network.

Solution

  • In my opinion, the most part of this confusion comes from Apache TinkerPop notation for edges. OrientDB historically adopted TinkerPop as its graph API (some things are changing here, so in v 3.0 there will be a native document/graph API that does not depend on Apache TinkerPop API, but we will still maintain the support with TinkerPop 3.x) so the notation is the same.

    Apache TinkerPop uses OUT/IN to define both edges connected to vertices and vertices connected to edges:

    • given a vertex, you have outgoing edges referenced by OUT and incoming edges referenced by IN (so far so good)
    • given an edge, it has the starting point (the vertex it comes from) identified by OUT (ouch!) and the end point (the vertex it goes to) identified by IN

    And here come the problems: for a beginner, it's very natural to consider an edge traversal as an out + out:

    vertex -out-> edge -out-> anotherVertex

    but unfortunately it's WRONG!

    The right way to do a straight edge traversal is out + in:

    vertex -out-> edge -in-> anotherVertex

    You can also traverse the edge backwards with an in + out.

    As I wrote before, IMHO it's extremely unintuitive, at the beginning I also had problems with this. IMHO the best notation for edge connection would have been FROM/TO, instead of OUT/IN, but it's a standard now, so we cannot do much.

    The only thing I can tell you is that it becomes natural with a little practice.

    A separate consideration is for social networks and in general for edges that are not intended to have a direction (in a social network the direction of edges is not important, the fact that I'm your friend also implied that you are my friend). Apache TinkerPop does not have a concept of undirected edges, so you just have to use directed edges and ignore the direction when traversing them (eg. using the both() operator or BOTH direction)