use pyproject.toml [build-system] and setuptools_scm

[mathause]

I did not plan to do this now but as setuptool v64 causes problems with versioneer I went for it (instead of pinning setuptools). I do have a bit of experience with this setup by now, so I feel fairly confident this "just works". Maybe requires a re-install in your development environment.

@znicholls FYI.

[mathause]

So much to "I know what I am doing" 😂

[mathause]

Ok falling back to pkg_ressources for python 3.7 - we cannot conditionally requires pkg_resources for python 3.7 in conda...

[codecov-commenter]

Codecov Report

❗ No coverage uploaded for pull request base (main@f5f7be1). Click here to learn what that means.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #188   +/-   ##
=======================================
  Coverage        ?   79.28%           
=======================================
  Files           ?       31           
  Lines           ?     1400           
  Branches        ?        0           
=======================================
  Hits            ?     1110           
  Misses          ?      290           
  Partials        ?        0           

Flag	Coverage Δ
unittests	79.28% <0.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[znicholls]

In other packages I had to follow this pattern https://github.com/openscm/openscm-runner/blob/27eb7baddb686862c7996dbccabf34cea934fcb7/docs/source/conf.py#L19

I can't remember why exactly, but it was something to do with being installed or not being installed when building docs

[mathause]

Oh - I wasn't aware that it's not (no longer?) installed per default in RTD. In xarray and regionmask we install it manually using pip (but I thought this is because it installs it with eager updates [making it impossible to pin packages]).

In any case RTD currently reports "mesmer 0.0.0 documentation" - I'll try your pattern to fix this in a new PR.

Related - for regionmask it reported a "dirty" version even for releases. This happens because RTD changes some files before installing the package. I used the following to work around this: regionmask/regionmask/pull/348

[mathause]

We currently do an install on RTD using "setuptools", i.e. i.e., python ./setup.py install, which is deprecated and may therefore not work together with pyproject.toml.

see #190 and

mesmer/.readthedocs.yml

Line 10 in d71bbe1

- method: setuptools

[znicholls]

Out of curiosity , what is this file?

[mathause]

I only just got this recently. If you define <filename> export-subst in a .gitattributes file, git will expand several placeholders when adding this file to an archive (from https://git-scm.com/docs/gitattributes). These placeholders can be the commit hash etc.

It currently still points to the wrong file (#189). With #189 merged:

mkdir archive
git archive HEAD > archive/HEAD.tar.gz
cd archive
tar -xvf HEAD.tar.gz
cat .git_archival.txt

returns

node: d933fa6f0835203e8310ca68b7307ebc34e63c43
node-date: 2022-08-15T11:05:04+02:00
describe-name: v0.8.3-39-gd933fa6
ref-names: HEAD -> fix_gitattributes_versioneer, origin/fix_gitattributes_versioneer

I think git archive is used to create the files in https://github.com/MESMER-group/mesmer/releases/tag/v0.8.3. Versioneer does it slightly differntly by writing to mesmer/_version.py export-subst.