Prepare first upload to PyPI #107
Conversation
Codecov Report
@@ Coverage Diff @@
## master #107 +/- ##
==========================================
+ Coverage 37.03% 37.06% +0.03%
==========================================
Files 19 20 +1
Lines 1801 1802 +1
==========================================
+ Hits 667 668 +1
Misses 1134 1134
Continue to review full report at Codecov.
|
setup.py
Outdated
| @@ -42,7 +43,7 @@ def readme(): | |||
| packages=find_packages(), | |||
| package_data={"protopipe": ["aux/example_config_files/protopipe/analysis.yaml"]}, | |||
| include_package_data=True, | |||
| install_requires=["ctapipe"], | |||
| install_requires=["ctapipe==0.9.1", "jupyterlab", "pyirf", "vitables"], | |||
There was a problem hiding this comment.
is jupyterlab and vitables really required? or only to generate docs? install_requires should only be things that are required to run, not useful things in the user's environment (which would go into environment.yml, or similar). I think the only requirements should be ctapipe and pyirf.
There was a problem hiding this comment.
This is for a pure pip installation, so after pip install protopipe the user should have all the necessary tools to use protopipe and like this it doesn't really need even conda.
For now "use protopipe" means also benchmarking and checking data, so for me these 2 tools are necessary for the package as a whole.
There was a problem hiding this comment.
I think that is the incorrect way to do it. Pip is not meant for making environments, only for installing required dependencies. For environments you need to use pipenv or poetry, or just provide a requirements.txt file that is the output of pip freeze. In the package itself (in setup.py) you should only list the bare minimum to allow it to work. Otherwise you will run into dependency resolution problems later.
So here you should have only ctapipe and pyirf as dependencies.
For reference: https://packaging.python.org/discussions/install-requires-vs-requirements/:
Whereas install_requires requirements are minimal, requirements files often contain an exhaustive listing of pinned versions for the purpose of achieving repeatable installations of a complete environment.
setup.py
Outdated
| @@ -42,7 +43,7 @@ def readme(): | |||
| packages=find_packages(), | |||
| package_data={"protopipe": ["aux/example_config_files/protopipe/analysis.yaml"]}, | |||
| include_package_data=True, | |||
| install_requires=["ctapipe"], | |||
| install_requires=["ctapipe==0.9.1", "jupyterlab", "pyirf", "vitables"], | |||
There was a problem hiding this comment.
I think that is the incorrect way to do it. Pip is not meant for making environments, only for installing required dependencies. For environments you need to use pipenv or poetry, or just provide a requirements.txt file that is the output of pip freeze. In the package itself (in setup.py) you should only list the bare minimum to allow it to work. Otherwise you will run into dependency resolution problems later.
So here you should have only ctapipe and pyirf as dependencies.
For reference: https://packaging.python.org/discussions/install-requires-vs-requirements/:
Whereas install_requires requirements are minimal, requirements files often contain an exhaustive listing of pinned versions for the purpose of achieving repeatable installations of a complete environment.
This PR prepares the first release that will be uploaded to PyPI as
v0.4.0.post1by updating thesetup.pyin a way that a purepipinstallation will be possible withtestsanddocsextra requirements.It is a requirement for PR #105.
It also comes with the following modifications:
.zenodo.jsonwith changes that hopefully will unlock the DOI generation which got stuck with v0.4.0README.gitignorefor dummy files generated bysetup.pyand related to versioning