Vyas Ramasubramani
Vyas Ramasubramani
> For pyarrow, I believe the user always has to start with a constructor: there's no `from_capsule()`...it is more like `pa.array()` and `pa.table()` can accept capsules (so pyarrow always knows...
> Generally that works right now for pyarrow, but it certainly means you can't distinguish them at the moment if you have some functionality that could take both an array...
> @vyasr I'm currently traveling for a conference, but I'll try to give this a look over the weekend or next week. Sorry I know I've pinged you a few...
I'll get this PR updated soon based on the discussions above.
I think this PR is ready to go now (pending reviews). Thanks to everyone for the fruitful discussions! There is plenty of follow-up work to be done that came out...
Running ``` (rapids) coder _ ~/cudf $ GTEST_CUDF_RMM_MODE=cuda LIBCUDF_MEMCHECK_ENABLED=1 compute-sanitizer --tool memcheck ./cpp/build/latest/gtests/INTEROP_TEST ========= ERROR SUMMARY: 0 errors ```
My general approach to pylibcudf to this point has been to keep it as a minimal, faithful export of libcudf algorithms to Python without adding much in the way of...
My claim is that pylibcudf doesn't care, but users may want to produce arrow objects with metadata and so interop is the only place where there should be a principled...
Thanks for the contribution @NeilGeorge1! Could you add some tests of some different cases? I know you won't be able to run them on your machine without a GPU, but...