Question on function of python-version patch

[jacquelinegarrahan]

The python-version patching is not clear when used with an existing environment.yml file.

I have an environment.yml file that looks like:

name: my-model-dev
channels:
  - conda-forge
dependencies:
  - python=3.9
  - click
  - pydantic
  - lume-model=1.4 

From existing documentation, it seems that the python provided should be patched with the python version provided in:

  - uses: conda-incubator/setup-miniconda@main
    with:
      mamba-version: "*"
      channels: conda-forge
      activate-environment: my-model-dev
      environment-file: environment.yml
      python-version: ${{ matrix.python-version }}

From the workflow, I can see that the patch workflow is being kicked off, but the python pin is not updated and I am unable to resolve my environment using the passed version. Is the environment file being updated only in the case of missing python? Are there ways to override pinned python?

[anstadnik]

I'm also interested in this

[jaimergp]

It should work if the environment contains a bare python requirement (no version), like:

name: my-model-dev
channels:
  - conda-forge
dependencies:
  - python  # changed
  - click
  - pydantic
  - lume-model=1.4 

[vthemelis]

What exactly are the semantics of the python-version argument in the action uses invocation?

Am I right to expect that

  - uses: conda-incubator/setup-miniconda@main
    with:
      python-version: 3.9

is going to install version 3.9 of Python?

[mattwthompson]

Yes, that will install Python 3.9 into the base environment. There evidently is a conflict if you also use an environment file that requires a different version.

As a matter of opinion, I personally recommend not specifying the Python version in the environment at all. In other words, use environment files like Jaime posted above, just with - python and nothing else.