Lawrence Mitchell
Lawrence Mitchell
Plausibly, at the cost of making the `to_arrow` functions in cudf aware of things, we could do this by unwrapping the wrapped result of `to_arrow().to_pandas()` (replacing it with `to_arrow().to_pandas()._fsproxy_slow`).
> > Plausibly, at the cost of making the to_arrow functions in cudf aware of things > > If this means making `cudf` aware of `cudf.pandas` proxy types, I'm strongly...
> Took a pass through the tests. I don't have much to say on them content-wise, they all look fine to me and it seems like you have good coverage....
> @wence- Is this the kind of change you had in mind? I looked over the list of files you put in #14527 and made changes for most of them....
> I haven't gotten to explore the inductive tests that you're seeking yet. Are there any places in the cudf code base that do this kind of thing generating nested...
> Couple of small issues left to be resolved here: > > * Need to figure out basically `to_bytes` for dicts that we can later create a test hash out...
> In the same way, I can't get the libcudf murmurhash3_x86_32 to give me the same answer as other external libraries, where the other external libraries seem to all agree...
> > In the same way, I can't get the libcudf murmurhash3_x86_32 to give me the same answer as other external libraries, where the other external libraries seem to all...