Graphs.transform*

When I’m creating my graph, I don’t know all the nodes/edges, so I use proxies like the ids of the nodes and the ids of the edges while registering the edges and nodes as they come in. Think of this as a stream being parsed where the node 1 says “I’m linked to 2 and 4”, but we haven’t parsed 2 nor 4 already.

When I’m done creating it, I’d like to have a copy of the graph but with the actual nodes and edges. So similarly to all the Maps.transform* methods, I’d like to have Graphs.transform* methods where applicable.

The only alternative I see is to create proxy items before creating the graph, but that’s a heavy burden on the user,

[jbduncan]

(@jrtom I just happened to discover the project when browsing the bazelbuild org page a few days ago (Bazel’s interested me for a while now), and in turn I got curious about the project’s implementation, so I skimmed over its internal files. I just happened to view a particular file, and in the process I found that particular comment.

Kind of a cool coincidence, I’d say. 😃)

[jrtom]

@Sadgrinner whoops, another good catch. 😃

The most obvious way to deal with that is to just brute-force it, something like (I haven’t checked the code below, just winging it):

static <T> ElementOrder<T> transformedNodeOrder(Graph<?> graph, @Nullable Comparator<T> comparator) {
  switch graph.nodeOrder() {
     case UNORDERED: return ElementOrder.<T>unordered();
     case INSERTION: return ElementOrder.<T>insertion();
     case SORTED: return ElementOrder.<T>sorted(checkNotNull(comparator));
     default: throw new IllegalStateException(...);
}

and then call transformedNodeOrder(originalGraph, comparator) instead of graph.nodeOrder().

I think that should do it. Needless to say, for cases where you don’t have a comparator you should probably overload this method so that you don’t have to supply one.