Re: [jgrapht-users] Bug in removeAllEdges?

[John V. S. <js...@gm...>]

Trevor Harmon wrote:
> Well, that part I would still say is a bug. I could understand it if 
> outgoingEdgesOf, incomingEdgesOf, and edgesOf all ended up causing the 
> ConcurrentModificationException -- at least that would be consistent. 
> But that's not the case here. The behavior of an API shouldn't be left 
> to chance. :)

Go ahead and create a bug report for this inconsistency.  It may not get 
addressed for a while because the fix isn't trivial.

For a directed graph, edgesOf has to be computed on the fly by combining 
the incoming and outgoing edge sets of the vertex.  And that computation 
isn't something that's easy to express as a derived Set implementation, 
due to the requirement to filter out the duplicates for self-loops.  So 
the fix requires us to maintain a graph-level timestamp which can be 
used by ArrayUnenforcedSet to detect the ConcurrentModificationException.

One side-effect of the fix would be that any application code which is 
currently relying on the buffering behavior of edgesOf would break.  But 
in this case, a more consistent and less surprising API trumps 
backward-compatibility.

JVS