Include pyx in sdist but don't install #1666

Merged
merged 6 commits into from May 22, 2014

Projects

None yet

4 participants

@rgommers
Member

Also a cleanup of setup.py included. cythonize.py adapted from scipy. See gh-1606 for discussion.

@jseabold jseabold commented on the diff May 12, 2014
@@ -510,5 +439,5 @@ def get_data_files():
cmdclass = cmdclass,
packages = packages,
package_data = package_data,
- include_package_data=True,
+ include_package_data=False, # True will install all files in repo
@jseabold
jseabold May 12, 2014 Member

Does this have any effect on #907?

@rgommers
rgommers May 12, 2014 Member

I can check a few ways by hand, but the TravisCI tests would fail if those didn't install right?

@josef-pkt
josef-pkt May 12, 2014 Member

It looks like I haven't removed the workaround (moving the files to another directory), so only the license and a text file would be missing.

@rgommers
rgommers May 14, 2014 Member

This was due to some typos in the name libqsturng, see last commit. Terrible name by the way......

@josef-pkt
josef-pkt May 14, 2014 Member

Very good, thanks for finding this.
I need to check the spelling of libqsturng several times when I try to write it, and still I make typos. I remember what it's supposed to stand for when I work on it and forget it 5 minutes later
I think it's studentized range statistic
and the q stands for quantiles.

@jseabold
jseabold May 14, 2014 Member

Yikes. I always thought it was someone's name...

@rgommers
Member

Is there a point to the empty setup.cfg file by the way, or can that be removed?

@josef-pkt
Member

My guess is it can be removed. I don't see any use for it. And I don't know why we have it.
Brief history check: I removed (renamed) it once, but then added it again.
I'm pretty sure this is there by accident.

@rgommers
Member

Then I'll take it out. I've already done :e se<TAB> in Vim several time and each time I open the .cfg by accident:)

@jseabold
Member

Would you mind also removing the tsa/setup.py as part of this PR? It's old cruft and is no longer used AFAICT.

@rgommers
Member

Sure, done.

@coveralls

Coverage Status

Coverage remained the same when pulling 98379ba on rgommers:sdist-build into b472807 on statsmodels:master.

@jseabold jseabold commented on the diff May 21, 2014
MANIFEST.in
recursive-include examples *
-#missed files: .npz, .npy
-include statsmodels/tsa/vector_ar/tests/results/vars_results.npz
-include statsmodels/iolib/tests/results/*
-include statsmodels/stats/tests/results/influence_lsdiag_R.json
-include statsmodels/sandbox/panel/test_data.txt
-include statsmodels/sandbox/regression/tests/*
-include statsmodels/stats/tests/results/influence_measures_R.csv
-include statsmodels/stats/libqsturng/tests/results/*
-include statsmodels/stats/libqsturng/tests/bootleg.dat
-include statsmodels/stats/libqsturng/tests/bootleg.csv
-include statsmodels/stats/libqsturng/CH.r
-include statsmodels/stats/libqsturng/LICENSE.txt
-include statsmodels/regression/tests/results/leverage_influence_ols_nostars.txt
@jseabold
jseabold May 21, 2014 Member

Can you explain these changes? Is it the case that files that files in MANIFEST.in are specified in addition to (not instead of) those specified in package_data? If not, don't we need to include these files in the sdist?

@rgommers
rgommers May 21, 2014 Member

Indeed. What's in package_data gets put into sdist output as well as installed. What's only in MANIFEST.in gets into sdist output but doesn't get installed.

@jseabold
Member

This looks good to me. Thanks. Would you mind rebasing? Then I'll hit the green button, so @chadfulton can work off these changes.

@josef-pkt
Member

Looks all fine to me.
I just tried with tox for python 3.3 and it works again. And spot checking the files, they look good (no pyx but all the qsturng files in the installed version, pyx are in sdist.)

Also, it didn't reuse the old .c files from the source checkout (inplace install). I only realized this because it told me my cython was too old.

@rgommers
Member

rebased.

@coveralls

Coverage Status

Coverage remained the same when pulling bf85b41 on rgommers:sdist-build into 4b94f8e on statsmodels:master.

@jseabold jseabold merged commit d27b085 into statsmodels:master May 22, 2014

1 check passed

continuous-integration/travis-ci The Travis CI build passed
Details
@jseabold
Member

Thanks. That intermittent hanging test is worrisome, but unrelated to this PR.

@josef-pkt
Member

Which intermittent hanging test? I haven't seen one in a few weeks.

@jseabold
Member

This PR had one for 2.7 and timed out. I don't know what test it was.

Sent from my mobile

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment