Closed marci543 closed 4 years ago
This is awesome thanks for doing this! I pulled your branch, rebuilt the images and I get all tests pass, not sure how to reproduce your issue. Maybe you've got a stale docker image in cache? I usually docker container prune
and then docker image prune
pretty frequently but you might want to take a less nuclear approach.
Edit: oh duh, it's failing in Travis, hmm ok i'll keep trying to reproduce I must be the one with the cached something.
I opened an issue for the problem: GraphBLAS/LAGraph#92
@marci543 looks like the tests are now passing, is this ready for review?
I would wait for the SuiteSparse release.
I only added the bitwise binary operators, but the unary ones, the monoids, and the semirings are still to go. A few names changed in one of the drafts which should be followed: https://github.com/DrTimothyAldenDavis/GraphBLAS/releases What do you think about adding other new features e.g. complex types? Is there a good Python equivalent?
CFFI maps common C complex types to python complex()
https://cffi.readthedocs.io/en/latest/ref.html#conversions
This is working already for LAGraphComplex, it's commented out here, I can't remember exactly why :shrug:
I wanted to prepare for the new SuiteSparse version while it is draft.
This cannot be done with the current minimal image until SS is released as a package. Adding an almost minimal image that compiles SS can be an option. Currently, I used the notebook image for testing.
A strange problem has arisen. Building with the new draft causes test of
Vector.to_dense
to fail because it returns vector filled with zeros. This is strange because the function wraps an LAGraph call, whose source code have not changed, since it is copied into this repository. https://travis-ci.com/github/szarnyasg/pygraphblas/builds/169510340#L4364-L4379@michelp, are there any special steps to take to keep the local header and the one in SuiteSparse consistent?