networkx silently ignores a misspelled weight= (Louvain, Dijkstra, modularity); igraph raises InternalError
In networkx, a weight attribute name that does not exist on the edges is silently treated as weight 1: with edges carrying 'wt', nx.dijkstra_path_length(G, 0, 1, weight='wieght') returns the hop count (1) instead of the weighted length, louvain_communities(G, weight='nonexistent', seed=1) returns exactly the unweighted partition, and modularity(..., weight='wieght') computes a value without warning. Weighted analyses therefore quietly become unweighted after a column rename or typo. python-igraph raises for the same mistake: InternalError: No edge attribute named "wieght" exists. -- Invalid value (community_leiden, community_multilevel, betweenness). Fix: assert the attribute exists before analysis, e.g. all('w' in d for *_, d in G.edges(data=True)), and keep attribute names in one constant. networkx 3.7, python-igraph 1.0.0, leidenalg 0.12.0, Python 3.12.