My personal library of Python utilities.
Add pytil as a dependency to your conda recipe. Some submodules have additional
dependencies; open up each submodule you use and add dependencies corresponding
to third party imports. You needn't constrain those dependencies (unless you
need a particular version of them yourself), pytil's conda run_constrained:
will constrain them for you.
Guide for developing pytil itself.
Install miniconda and then:
git checkout ...
cd $dir_containing_setup_py
conda create -n pytil 'python==3.8.*'
conda activate pytil
conda install pylint
pip install -e .
You still need to install dependencies. If they are the same as one of the
released conda pkgs you can use conda install --only-deps. Otherwise open
conda/meta.yaml and conda install anything listed in run. You can find the
conda channels to use in the conda build command in
.github/workflows/publish.yaml. In order to run the tests, also install the
test requirements from meta.yaml. Now you can run tests with pytest. (I
only list dependencies in conda recipe, not in setup.py, keeps it DRY at the
cost of having to install them manually to set up dev env).
The version needs to be adjusted in setup.py and pytil/__init__.py. Release
it with github releases, tag it as github suggests (v1.2.3) and list the
changes in its description. The GitHub repo will publish to anaconda when you
publish a new GitHub release.
This is just in case you want to try out the build before creating a release on github. The conda recipe runs pytest after the build.
First time setup:
conda create -n conda-build 'python==3.8.*'
conda activate conda-build
conda install conda-build
Building a pkg (to try it out before pushing to master):
cd $dir_containing_meta.yaml
conda-build . --channel ...
If for some reason you want to try it out, create a new conda env, install
dependencies manually (or conda install --only-deps with a previous version
if deps are the same) and finally conda install --use-local pytil.
$repo/tests is kept separate from $repo/pytil to prevent including them in
the conda package.