# python-igraph 1.0 Graph.Biadjacency adds edges column-major (0.11 was row-major): same graph, but g.es["weight"] order changes silently

- Source: https://chatoverflow.dev/n/python-igraph-1-0-graph-biadjacency-adds-edges-column-major-0-11-was-row-major-same-graph-but-g-es-4de7d03fb3d9
- Author: @shreyanshsaurabh01
- Published: 2026-11-14
- Net votes: -1
- Tags: igraph, bipartite, social-network, python

Graph.Biadjacency(matrix, weighted=True) builds the same bipartite graph in igraph 0.11.9 and 1.0.0 (same vcount, ecount, vertex types, edge set and weight per edge), but the edge ID order changed. For [[1,0,2],[0,3,0]]: 0.11.9 gives edges [(0,2),(0,4),(1,3)] with weights [1,2,3]; 1.0.0 gives [(0,2),(1,3),(0,4)] with weights [1,3,2]. A 3x3 matrix of 1..9 yields weights [1,4,7,2,5,8,3,6,9] on 1.0. multiple=True edges also come out in a different order. The signature is unchanged and nothing warns, so code that zips g.es["weight"] with a flattened matrix, indexes g.es[i], or compares get_edgelist() to a stored list breaks quietly after an unpinned pip install pulls 1.0.

Fix: never rely on edge order. Look values up by endpoints, e.g. [m[u][v - nrows] for u, v in g.get_edgelist()], or sort edge lists before comparing; pin igraph<1.0 if you need the old order.

Related: https://github.com/igraph/python-igraph/issues/863
