Enhancements

• Support NetworkX 3.2, which made many improvements to dispatching (#68, #77, #78)
• Add support for Python 3.12 (#79)
• Drop support for Python 3.8 (#77)

New algorithms (#67)

• algorithms.efficiency_measures.efficiency
• algorithms.isomorphism.isomorph.fast_could_be_isomorphic
• algorithms.isomorphism.isomorph.faster_could_be_isomorphic
• algorithms.lowest_common_ancestors.lowest_common_ancestor
• algorithms.operators.unary.complement
• algorithms.operators.unary.reverse
• algorithms.shortest_paths.weighted.bellman_ford_path_length
• linalg.bethehessianmatrix.bethe_hessian_matrix
• linalg.graphmatrix.adjacency_matrix
• linalg.laplacianmatrix.laplacian_matrix
• linalg.laplacianmatrix.normalized_laplacian_matrix
• linalg.modularitymatrix.directed_modularity_matrix
• linalg.modularitymatrix.modularity_matrix

Enhancements

• Add support for Python 3.11 (#61)
• Add logo and benchmarks to README (#60)

New algorithms (#51, #61, #62, #64)

• Components
  • is_connected
  • is_weakly_connected
  • node_connected_component
• Generators
  • ego_graph
• Link Analysis
  • google_matrix
• Operators
  • compose
  • difference
  • disjoint_union
  • full_join
  • intersection
  • symmetric_difference
  • union
• Shortest Paths
  • all_pairs_shortest_path_length
  • bellman_ford_path
  • floyd_warshall_numpy
  • negative_edge_cycle
  • single_source_shortest_path_length
  • single_target_shortest_path_length
• Traversal
  • bfs_layers
  • descendants_at_distance

Note: this is a re-release of 2023.2.0, because 2023.2.0 didn't build and upload to PyPI

• Add scripts to run benchmarks and download data (#39)
• Add floyd_warshall algorithm for all-pairs shortest path (#42)
• Add floyd_warshall_predecessor_and_distance (#43)
• Add all_pairs_bellman_ford_path_length and single_source_bellman_ford_path_length (#44)
• Add NodeNodeMap class and matrix_to_nodenodemap and matrix_to_vectornodemap methods (#43)
  • These can replace matrix_to_dicts
• Add fill_value to NodeMap (#43)
• Allow NodeMap values to be interpreted as keys (#43)
• Add more cached properties (#44)
  • min_diagonal (and other {monoid_name}_diagonal)
  • has_negative_diagonal
  • has_negative_edges- and has_negative_edges+
  • is_iso
  • iso_value
• Add normalize_chunksize and partition utility functions to help run algorithms chunkwise (#47)
• Misc. maintenance (#41, #45, #46)

No functional changes from 2023.2.0

Pre-release to test automatic upload to PyPI, which didn't work for 2023.2.0.

• Add scripts to run benchmarks and download data (#39)
• Add floyd_warshall algorithm for all-pairs shortest path (#42)
• Add floyd_warshall_predecessor_and_distance (#43)
• Add all_pairs_bellman_ford_path_length and single_source_bellman_ford_path_length (#44)
• Add NodeNodeMap class and matrix_to_nodenodemap and matrix_to_vectornodemap methods (#43)
  • These can replace matrix_to_dicts
• Add fill_value to NodeMap (#43)
• Allow NodeMap values to be interpreted as keys (#43)
• Add more cached properties (#44)
  • min_diagonal (and other {monoid_name}_diagonal)
  • has_negative_diagonal
  • has_negative_edges- and has_negative_edges+
  • is_iso
  • iso_value
• Add normalize_chunksize and partition utility functions to help run algorithms chunkwise (#47)
• Misc. maintenance (#41, #45, #46)

• Update README to show installation instructions and basic usage (#34)
  • Including use as a NetworkX plugin!
• Create Graph from Matrix as Graph(A) instead of Graph.from_graphblas(A)(#35)
• Update more places to use to_coo and from_coo instead of to_values and from_values (#32)
• Use faster algorithm for s_metric (#38)
• Add environment.yml to create developer environment (#37)
• Add DOI from Zenodo to README (#33)

• Add plugin code to be a backend for NetworkX 3.0 to dispatch to
• Add graphblas_algorithms.nxapi that "looks" like the NetworkX API
• Add lots of algorithms and other nice things!

• Updated to use python-graphblas instead of grblas
• Added average_clustering, clustering, transitivity, and triangles for undirected, unweighted graphs
• Added git pre-commit hooks
• Added some helpful utility functions to make things easier/cleaner

First release!

graphblas-algorithms is just getting started. It only has PageRank:

• graphblas_algorithms.pagerank matches NetworkX API and passes all NetworkX PageRank tests.
• graphblas_algorithms.link_analysis.pagerank_core is a fast, GraphBLAS-only implementation that is used by the former.
  • This is the implementation to bring to benchmarking shootout.

This project is in alpha and may undergo significant changes.