Remove pyx dependencies from setup

[jbrockmendel]

See discussion in #17419

• closes #xxxx
• tests added / passed
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

[pep8speaks]

Hello @jbrockmendel! Thanks for updating the PR.

Cheers ! There are no PEP8 issues in this Pull Request. 🍻

Comment last updated on September 08, 2017 at 17:43 Hours UTC

[codecov]

Codecov Report

Merging #17478 into master will decrease coverage by 0.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master   #17478      +/-   ##
==========================================
- Coverage   91.15%   91.13%   -0.02%     
==========================================
  Files         163      163              
  Lines       49534    49534              
==========================================
- Hits        45153    45144       -9     
- Misses       4381     4390       +9

Flag	Coverage Δ
#multiple	88.92% <ø> (ø)	⬆️
#single	40.22% <ø> (-0.07%)	⬇️

Impacted Files	Coverage Δ
pandas/io/gbq.py	25% <0%> (-58.34%)	⬇️
pandas/core/frame.py	97.72% <0%> (-0.1%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d6df8ea...a06e178. Read the comment docs.

[gfyoung]

@jbrockmendel : The builds are passing, so can't complain, but just curious : is there anything further that can be done now that you've dropped these dependencies?

[jreback]

@jbrockmendel
does scripts/build_release_for_dist.sh work locally?

[jbrockmendel]

does scripts/build_release_for_dist.sh work locally?

Just tried it and it went through OK.

but just curious : is there anything further that can be done now that you've dropped these dependencies?

Not sure what you have in mind. This doesn't affect the actual dependencies, is just about the rules for declaring cython deps.

[jbrockmendel]

There are some pretty significant simplifications that can be made once timezone functions get moved out of tslib, and another big batch if/when NaT becomes standalone. Is that what you're referring to?

[jreback]

ok this looks fine. ping on green.

[jreback]

ok going to merge. I hope this doesn't bite us from something we are not testing, possibly on the development side.

[jbrockmendel]

Poking at this further has me confused. It is not clear to me what is the distinction is between the 'include' vs 'depends' vs 'pxdfiles'.

I saw that the ext_data entries that don't include a 'include' key get the default include of ['pandas/_libs/src/klib', 'pandas/_libs/src']. Many of the _libs extensions only need one or two of the files in _libs/src, so I tried specifying those explicitly and pinning include entries to [] in the hopes of trimming the build time. For the most part this led to compile-time errors complaining that there is no "datetime/np_datetime.h" file. Lots of reasonable explanations for that, no big deal.

More noteworthy is that removing "compat_helper.h" from lib_depends doesn't seem to matter. The only explanation I've come up with is that the ext_data['_libs.lib']['include'] gets set to the default value of common_include which is ['pandas/_libs/src/klib', 'pandas/_libs/src'] and since "compat_helper.h" is in the "pandas/_libs/src" directory, that is sufficient. But then why are there separate 'include' and 'depends' keys?

[jreback]

you should dig into the cython docs to see

[jbrockmendel]

I have. I've also used Google.

These args are getting passed to distutils.extension.Extension, so I don't think it is cython-specific. (Yes, I've looked through the distutils docs).