code-shoily / yog_ex / 3ab957fd258bc4a4c7d5aeaa2e7c1019db068efb

93%

push

github

fix: eigenvector centrality no longer silently returns zeros on DAGs

On directed acyclic graphs (DAGs), eigenvector centrality is mathematically
ill-defined: the adjacency matrix is nilpotent, so the principal eigenvalue
is 0 and the principal eigenvector is the zero vector.

Previously, iterate_eigenvector/7 would converge to the all-zeros vector
without any indication to the caller. This is a silent failure mode that
produces uninformative rankings on common graph shapes like citation
networks, dependency graphs, and hierarchies.

Now, when the L2 norm becomes 0.0 during power iteration, we:
1. Log a warning explaining the degeneracy
2. Return a uniform distribution as a fallback
3. Document the behavior and recommend pagerank/2 or katz/2 for DAGs

Closes #137

12 of 12 new or added lines in 1 file covered. (100.0%)

10380 of 11160 relevant lines covered (93.01%)

8490.34 hits per line