BLD: move metadata to setup.cfg

[fangchenli]

move metadata to setup.cfg to separate them from build logic.

[fangchenli]

We could modify PATH and uncomment this line to enable pep517. But I feel that may require some discussion.

[jreback]

great. can you build like this https://github.com/pandas-dev/pandas-release to make sure things look good (mainly testing that the sdist looks good).

[jreback]

cc @simonjayhawkins if you give this a quick test as well. shall we backport to 1.2.1?

[fangchenli]

great. can you build like this https://github.com/pandas-dev/pandas-release to make sure things look good (mainly testing that the sdist looks good).

Is uploading the sdist to testpypi the only way to test this? I couldn't get the version right. It won't allow me to upload the sdist.

[MarcoGorelli]

Nice! I'd suggesting adding https://github.com/asottile/setup-cfg-fmt as a pre-commit hook

  - repo: https://github.com/asottile/setup-cfg-fmt
    rev: v1.16.0
    hooks:
      - id: setup-cfg-fmt

[jreback]

@fangchenli you don't need to tag or upload

just build the sdist in master and then on your PR and compare that they have the same metadata

[simonjayhawkins]

cc @simonjayhawkins if you give this a quick test as well. shall we backport to 1.2.1?

pip-test and conda-test failures, 74 failed.

eg. FileNotFoundError: [Errno 2] No such file or directory: '/opt/conda/envs/pip-test/lib/python3.7/site-packages/pandas/tests/io/parser/data/unicode_series.csv'

[simonjayhawkins]

install from sdist and running pandas.test(extra_args=['-m not clipboard', '--skip-slow', '--skip-network', '--skip-db', '-n=2']) gives 87 failed which includes 13 failures seen at https://dev.azure.com/pandas-dev/pandas-wheels/_build/results?buildId=51595&view=logs&j=bb1c2637-64c6-57bd-9ea6-93823b2df951&t=9c62e7ff-391e-517d-ede1-af63bc6a25da&s=d654deb9-056d-50a2-1717-90c08683d50a

[jreback]

@simonjayhawkins @jorisvandenbossche this lgtm. looks like final comments have been addressed. merging tomorrow.

[simonjayhawkins]

@simonjayhawkins should we consider keeping it? (I don't know how useful it is, or eg only on master)

we should maybe have this or something similar on ci. At the moment we will catch build errors on the nightly build.

but lets keep that discussion out of this PR.

[jorisvandenbossche]

OK, sounds good

[simonjayhawkins]

@fangchenli maybe we should open issues for conditional numpy import #38852 (comment) and ci check #38852 (comment)

[jreback]

@fangchenli one more rebase and then will merge. ping on green.

[simonjayhawkins]

Thanks @fangchenli

[simonjayhawkins]

@fangchenli the wheels doubled in size around the same time that this was merged. Could it be this PR that is responsible? https://anaconda.org/scipy-wheels-nightly/pandas/files?page=32

[fangchenli]

@fangchenli the wheels doubled in size around the same time that this was merged. Could it be this PR that is responsible? https://anaconda.org/scipy-wheels-nightly/pandas/files?page=32

Looking into it.