Closed h-vetinari closed 3 years ago
Hi! This is the friendly automated conda-forge-linting service.
I just wanted to let you know that I linted all conda-recipes in your PR (recipe
) and found it was in an excellent condition.
Ah, I see, we were just repackaging the upstream jars (which don't exist yet due to the issue noted in the OP). Should we build this from source @conda-forge/pyspark?
CC @conda-forge/core
OK, it looks like this is running into StackOverflows no matter what. 😑
@h-vetinari Instead of pypi we can also switch back to the apache-mirror tarballs, see e.g. this old commit https://github.com/conda-forge/pyspark-feedstock/commit/8cc5d21df1450e2369904c86d06be4f6d622c2b2#diff-f3725a55bf339595bf865fec73bda8ac99f283b0810c205442021f29c06eea9aL11
Thanks for the tip @dbast! I ended up using the official tarball from apache (also mentioned here). Still not enthused about the binary repackaging, but it's at least as good as what we had previously.
I'm going to let upstream know that this is more or less ready (might be reflected in the release announcements since PyPI not available yet), it would IMO be a good chance for conda-forge to present itself to a community that's not necessarily familiar with it. 🙃
PS. pip check
is a bit overly strict about the py4j dependency. I'll try to get this fixed in ~https://github.com/conda-forge/py4j-feedstock/pull/20~ https://github.com/conda-forge/py4j-feedstock/pull/21, but IMO this check could also just be skipped.
@h-vetinari Thanks for the work! FYI the file size issue is resolved and PySpark 3.2.0 is available on PyPI now: https://pypi.org/project/pyspark/3.2.0/ Do we still need this one?
FYI the file size issue is resolved and PySpark 3.2.0 is available on PyPI now: https://pypi.org/project/pyspark/3.2.0/
Good to hear!
Do we still need this one?
conda(-forge) & PyPI use different binaries & distribution channels, so yes, for conda it is still needed.
https://github.com/conda-forge/py4j-feedstock/pull/21 has been merged now, so once that package become available in about ~1h, this PR should be ready. I'm not a position to merge it myself, but I've already pinged the people who can.
@conda-forge/pyspark, this should be ready 🙃
Upstream is blocked https://github.com/apache/spark-website/pull/361 on PyPI, which is why the update bot didn't find the tag yet either. Aside from regularly just trying to build this for conda-forge, it would be nice to have a "python deployment story" for spark for their release announcement, so 🤞 this works out quickly.