Such questions are better asked in the mailing lists or sourceforge trackers of the JGraphT project (see http://jgrapht.org). A quick answer, though: in the latest trunk there is org.jgrapht.ext.ComponentAttributeProvider which could be used; another (and possibly preferred) solution is to use special classes for edges an vertices which store the attributes.

This is John Sichi (perfecthash over at the jgrapht project).

I’m glad you are working on this; it’s a fairly frequent request from JGraphT users.

One suggestion I have is to take an optional Graph object as input (in addition to returning it as output). This is what we do for the GraphGenerator interface. The reason is that it allows the caller to choose the desired concrete graph implementation, which can be very useful when dealing with custom graph classes. If appropriate, the reader can verify that the graph is empty to start with (but in some cases a reader could also be used to augment an existing graph, potentially attaching to existing vertices by reference). And in your example where the graph type is declared in the file, then the reader can verify that the input graph (if supplied) matches. If a graph is passed, then no edge factory is required since the graph already supplies its own.

We can edit the comments on ComponentAttributeProvider to indicate that when used with a reader, the returned map is read/write (rather than read-only as for a writer).

JVS