> On the other hand, might there exist a corner case where a > user loads a GraphML file into python-igraph, but the resulting graph > object is not what they were expecting because of this silently > ignoring double `id` attributes? Well, I think that if the user explicitly provided a vertex attribute named "id" in the GraphML file (using <data> tags), then he/she surely wants to see these IDs in the imported graph and not the ones that were used as internal identifiers in the <node> tags. So I see no problem here even if the two ID sets differ -- after all, the <node id="..."> identifiers are meant to be used within the GraphML file only anyway.
I will add a patch soon that removes this warning from igraph, so the behaviour will be as follows: - When reading a GraphML file, the <node id="..."> attributes will be converted to a "real" vertex attribute named "id" if such an attribute does not exist. If it exists, the identifiers in the <node> tags will silently be discarded. - When writing a GraphML file, the "id" vertex attribute will be stored using a <data> tag in each node, and the "id" attributes of the <node> tags will be made up by igraph to ensure that they are unique. Best, T. _______________________________________________ igraph-help mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/igraph-help
