Exclude test files from source/binary distributions

[seisman]

Running make package can create source and binary distributions. Here are the files/directories included in the source tarball:

$ tree -L 2
.
├── AUTHORSHIP.md
├── AUTHORS.md
├── CITATION.cff
├── CODE_OF_CONDUCT.md
├── LICENSE.txt
├── MANIFEST.in
├── PKG-INFO
├── pygmt
│   ├── accessors.py
│   ├── clib
│   ├── datasets
│   ├── exceptions.py
│   ├── figure.py
│   ├── helpers
│   ├── __init__.py
│   ├── io.py
│   ├── session_management.py
│   ├── sphinx_gallery.py
│   ├── src
│   └── tests
├── pygmt.egg-info
│   ├── dependency_links.txt
│   ├── PKG-INFO
│   ├── requires.txt
│   ├── SOURCES.txt
│   └── top_level.txt
├── pyproject.toml
├── README.rst
└── setup.cfg

8 directories, 22 files

1. AUTHORSHIP.md is excluded in MANIFEST.in. Why is it still here?
2. Exclude CODE_OF_CONDUCT.md from the tarball?
3. Exclude the whole pygmt/tests directory? The reasons are (1): tests are 536 KB and others are 900 KB. (2) Users can't run these tests due to the missing DVC files.

[weiji14]

AUTHORSHIP.md is excluded in MANIFEST.in. Why is it still here?

Might be a case of pypa/setuptools#3260? We switched from setup.py to pyproject.toml in #1848 around Aug 2022, but most of the 'exclude' lines in MANIFEST.in was last modified in #999 around Mar 2021, and we might not have checked things properly. I tried for a bit with some settings, but couldn't figure out how to exclude those files...

[weiji14]

Yeah, I just tried downloading the sdist tar.gz files for PyGMT, and v0.4.1 (released 7 Aug 2021) was the last time when AUTHORSHIP.md was excluded. All the releases from v0.5.0 (released 29 Oct 2021) onwards to v0.10.0 contain AUTHORSHIP.md 😅 Might not be a setup.py/pyproject.toml thing, but some setuptools bug.

[seisman]

It's likely that AUTHORSHIP.md is a special name and is always included:

make package
python -m build
* Creating venv isolated environment...
* Installing packages in isolated environment... (setuptools>=64, setuptools_scm[toml]>=6.2)
* Getting dependencies for sdist...
running egg_info
writing pygmt.egg-info/PKG-INFO
writing dependency_links to pygmt.egg-info/dependency_links.txt
writing requirements to pygmt.egg-info/requires.txt
writing top-level names to pygmt.egg-info/top_level.txt
reading manifest template 'MANIFEST.in'
adding license file 'LICENSE.txt'
adding license file 'AUTHORS.md'
adding license file 'AUTHORSHIP.md'
writing manifest file 'pygmt.egg-info/SOURCES.txt'
* Building sdist...
running sdist
running egg_info
writing pygmt.egg-info/PKG-INFO
writing dependency_links to pygmt.egg-info/dependency_links.txt
writing requirements to pygmt.egg-info/requires.txt
writing top-level names to pygmt.egg-info/top_level.txt
reading manifest template 'MANIFEST.in'
adding license file 'LICENSE.txt'
adding license file 'AUTHORS.md'
adding license file 'AUTHORSHIP.md'
writing manifest file 'pygmt.egg-info/SOURCES.txt'
running check
creating pygmt-0.10.1.dev203+g24e032439
creating pygmt-0.10.1.dev203+g24e032439/pygmt
creating pygmt-0.10.1.dev203+g24e032439/pygmt.egg-info
creating pygmt-0.10.1.dev203+g24e032439/pygmt/clib
creating pygmt-0.10.1.dev203+g24e032439/pygmt/datasets
creating pygmt-0.10.1.dev203+g24e032439/pygmt/helpers
creating pygmt-0.10.1.dev203+g24e032439/pygmt/src
creating pygmt-0.10.1.dev203+g24e032439/pygmt/tests
creating pygmt-0.10.1.dev203+g24e032439/pygmt/tests/data
copying files to pygmt-0.10.1.dev203+g24e032439...
copying AUTHORS.md -> pygmt-0.10.1.dev203+g24e032439
copying AUTHORSHIP.md -> pygmt-0.10.1.dev203+g24e032439
copying CITATION.cff -> pygmt-0.10.1.dev203+g24e032439
copying CODE_OF_CONDUCT.md -> pygmt-0.10.1.dev203+g24e032439
...

[seisman]

Making following changes to MANIFEST.in:

diff --git a/MANIFEST.in b/MANIFEST.in
index 88e203e94..da8835b2c 100644
--- a/MANIFEST.in
+++ b/MANIFEST.in
@@ -3,12 +3,13 @@ prune .github*
 prune ci
 prune doc*
 prune examples*
+prune pygmt/tests*
 exclude .dvcignore
 exclude .gitignore
 exclude .readthedocs.yaml
 exclude AUTHORSHIP.md
+exclude CODE_OF_CONDUCT.md
 exclude CONTRIBUTING.md
 exclude Makefile
 exclude environment.yml
 exclude requirements.txt
-exclude pygmt/tests/baseline/*.dvc

Then the source tarball contains following files/directories:

tree -a -L 2
.
├── AUTHORS.md
├── AUTHORSHIP.md
├── CITATION.cff
├── LICENSE.txt
├── MANIFEST.in
├── PKG-INFO
├── README.rst
├── pygmt
│   ├── __init__.py
│   ├── accessors.py
│   ├── clib
│   ├── datasets
│   ├── exceptions.py
│   ├── figure.py
│   ├── helpers
│   ├── io.py
│   ├── session_management.py
│   ├── sphinx_gallery.py
│   └── src
├── pygmt.egg-info
│   ├── PKG-INFO
│   ├── SOURCES.txt
│   ├── dependency_links.txt
│   ├── requires.txt
│   └── top_level.txt
├── pyproject.toml
└── setup.cfg

The new source tarball is 178 KB, compared to 244 KB for the old source tarball.

[seisman]

AUTHORSHIP.md is excluded in MANIFEST.in. Why is it still here?

Might be a case of pypa/setuptools#3260? We switched from setup.py to pyproject.toml in #1848 around Aug 2022, but most of the 'exclude' lines in MANIFEST.in was last modified in #999 around Mar 2021, and we might not have checked things properly. I tried for a bit with some settings, but couldn't figure out how to exclude those files...

Likely related to https://github.com/pypa/setuptools/blob/85fd45b7d10441f006cbbdee2b718aa741db5246/NEWS.rst#breaking-changes-12.

[weiji14]

AUTHORSHIP.md is excluded in MANIFEST.in. Why is it still here?

Might be a case of pypa/setuptools#3260? We switched from setup.py to pyproject.toml in #1848 around Aug 2022, but most of the 'exclude' lines in MANIFEST.in was last modified in #999 around Mar 2021, and we might not have checked things properly. I tried for a bit with some settings, but couldn't figure out how to exclude those files...

Likely related to https://github.com/pypa/setuptools/blob/85fd45b7d10441f006cbbdee2b718aa741db5246/NEWS.rst#breaking-changes-12.

Ah nice, so need to add this to pyproject.toml:

[tool.setuptools]
platforms = ["Any"]
include-package-data = true
+license-files = ["LICENSE.txt"]