Francesc Alted
Francesc Alted
Yeah, that was much needed; thanks for addressing this. Regarding the Zstd errors, well, it looks like somehow, the Python extension is not finding the zstd.h header, mmmhhh...
It looks like the importing machinery assumes a plain string, not a Unicode one. This is a bug indeed. You are welcome to provide a pull request to accelerate fixing...
Oh, on a second thought, I remember now that I deliberately left all the operators unimplemented for bcolz. The reason is that once 'a == b' and 'a - b'...
Hi. Could you please provide a self-contained example? Also, which machine architecture is your local and azure machines?
Yeah, using the `newer` utility that comes with distutils would be enough. See: https://github.com/PyTables/PyTables/blob/develop/setup.py#L660
The recent work on setup.py (see #239) relies on setuptools cython support, so it should do the correct thing. @nymoorland can you test if removing cython from: https://github.com/Blosc/bcolz/pull/239/files#diff-2eeaed663bd0d25b7e608891384b7298R158 would be...
Sure. Could you test the different approaches? In case one of them do what you want and passes the test suite, please file a pull request; I'll be glad to...
Well, everything can be made faster, but how much faster is PyTables in this situation?
Oh yes, that's a lot of difference, and that should be fixed. Pull requests are welcome!
Not really because PyTables also has to deal with chunked data, so the bottleneck must be in some other place. Doing a profile would be helpful.