Here is a bug I experienced while writing python scripts, I was able to repeat with the online python interpreter. Here in its minimal form:
Take a graph (here I am using an graph imported from the Wang et. al. import plugin), manually select two incident edges from it (edges share a node). Invoke the inducedSubgraph method.
graph.inducedSubGraph(graph['viewSelection'])
The resulting subgraph then contains more than 3 nodes and more than 3 edges (although the result may vary) . Selecting more (incident) edges gets you even freakier results.
Any idea what's going on? Thanks to the experts.
My bad.
This is due to an old optim replacing a set by a vector.
It will be fixed soon.
Finally, the error was not in the inducedSubGraph optimized code,
which was perfectly legitimate, but it revealed a problem in a much older code.
I'm happy I could help renovate Tulip's dusty and deeper layers ... :-)