This is a library plus a test harness for collecting algorithms that use the GraphBLAS. For test coverage reports, see https://graphblas.org/LAGraph/ . Documentation: https://lagraph.readthedocs.org
Merged in newer version of dev, hence the huge number of commits
Renamed instances of mapping (i.e. mapping of old labels to new labels) to newlabels to be more clear what the intent of this result is
Changes to LAGraph_Coarsen_Matching:
Changed type of S-matrix to GrB_FP64 to accommodate non-integer edge weights (before, this was causing non-integer edge weights to be truncated to integers)
Fixed bug where the parent mapping was not being returned when preserve_mapping was false.
Changes to LG_check_coarsen:
Finished implementation of naive coarsening method - this will work for all sorts of coarsenings, not just matching-based coarsenings
Changes to test_Coarsen_Matching:
Finished initial version of matching-based coarsening test, added sample test. A key remaining piece is to verify that the parent mapping produced by LAGraph_CoarsenMatching is produced by a valid matching
Changes to LAGraph_Incidence_Matrix:
In the final GrB_Matrix_build step, we don't need a binop for duplicates (there should be no duplicate entries, and if there are we want it to produce an error). So, we can set this to NULL. A similar thing was done in LAGraph_Coarsen_Matching when rebuilding the A matrix to be INT64.
General changes:
dev, hence the huge number of commitsmapping(i.e. mapping of old labels to new labels) tonewlabelsto be more clear what the intent of this result isChanges to
LAGraph_Coarsen_Matching:GrB_FP64to accommodate non-integer edge weights (before, this was causing non-integer edge weights to be truncated to integers)preserve_mappingwas false.Changes to
LG_check_coarsen:Changes to
test_Coarsen_Matching:LAGraph_CoarsenMatchingis produced by a valid matchingChanges to
LAGraph_Incidence_Matrix:GrB_Matrix_buildstep, we don't need a binop for duplicates (there should be no duplicate entries, and if there are we want it to produce an error). So, we can set this toNULL. A similar thing was done inLAGraph_Coarsen_Matchingwhen rebuilding the A matrix to beINT64.