Support multiple env yaml specs

[jchorl]

Tested:

testmultiyml/env1.yml

channels:
  - conda-forge
dependencies:
  - python >=3.12

testmultiyml/env2.yml

channels:
  - conda-forge
dependencies:
  - pandas >=2.1.1

(mamba) mambauser@6eb769d94efc:/work$ ./build/micromamba/micromamba create -n multiymltest -f testmultiyml/env1.yml -f testmultiyml/env2.yml 
(mamba) mambauser@6eb769d94efc:/work$     micromamba activate multiymltest
(multiymltest) mambauser@6eb769d94efc:/work$ python3
Python 3.12.0 | packaged by conda-forge | (main, Oct  3 2023, 08:43:22) [GCC 12.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pandas as pd

Not sure if there's other edge cases worth considering or test cases that would be helpful.

[jonashaag]

Thanks for working on this!

This definitely needs more tests and coverage of edge cases.

Some things that come to mind, not an exhaustive list:

• conflicting specs
• conflicting channels (defaults vs nodefaults)
• conflicting channel priorities
• missing entries in one of the files
• passing non env.yml files as one of the files (should probably error)

[jchorl]

Thanks for working on this!

This definitely needs more tests and coverage of edge cases.

Some things that come to mind, not an exhaustive list:

* conflicting specs

* conflicting channels (defaults vs nodefaults)

* conflicting channel priorities

* missing entries in one of the files

* passing non env.yml files as one of the files (should probably error)

Thanks. These are great suggestions. I found a bit more time today to work on this.

General thought: It would be good to understand what new testing you feel is required just for multi-yaml-spec support. Happy to add more if you feel it makes sense. Thanks for the help thus far!

Re specific tests:

First, I modified the multi-spec test to properly handle multiple yamls.

conflicting channels (defaults vs nodefaults)

I did write a test for this (removed after): test_multiple_yaml_specs_conflicting_defaults.

I ran this through micromamba:

micromamba install -n base pandas -c nodefaults -c defaults

This actually works. This makes me think the check for conflicting defaults belongs at a lower layer, not multi-yaml-spec parsing. Thoughts?

conflicting channel priorities

I don't think yaml specs support channel_priority: conda/conda#8675

If you point me at docs or anything, happy to add a test here.

missing entries in one of the files

I wasn't sure what you meant here. I added test_multiple_yaml_specs_only_one_has_channels whereby only one yaml specifies the channels. The result is the union of all channels (so, only the one channel from the one yaml) but all the deps get concatted. This is expected/desirable to me. Did you mean to test for something else?

passing non env.yml files as one of the files (should probably error)

done! added a test too.

[jonashaag]

I don't like that we are duplicating the type checks and that we are mixing type checking and type "inference". Maybe it's cleaner to use a separate type check loop

[jonashaag]

Looks pretty good already. What happens if you specify conflicting specs in the files? eg. python=3.11 and python=3.10. Is the error message helpful?

[jchorl]

Looks pretty good already. What happens if you specify conflicting specs in the files? eg. python=3.11 and python=3.10. Is the error message helpful?

Tried:

(mamba) root@068e7d994f42:/work# printf 'dependencies: [pandas==1.5.3]\nchannels: [conda-forge]' > spec1.yml
(mamba) root@068e7d994f42:/work# printf 'dependencies: [pandas==2.1.3]\nchannels: [conda-forge]' > spec2.yml
(mamba) root@068e7d994f42:/work# ./build/micromamba/micromamba create -y -n joshtest -f spec1.yml -f spec2.yml 
conda-forge/linux-64                                        Using cache
conda-forge/noarch                                          Using cache
error    libmamba Could not solve for environment specs
    The following packages are incompatible
    ├─ pandas 1.5.3  is requested and can be installed;
    └─ pandas 2.1.3  is not installable because it conflicts with any installable versions previously reported.
critical libmamba Could not solve for environment specs

It's similar (but not identical!) to just passing cli specs:

(mamba) root@068e7d994f42:/work# ./build/micromamba/micromamba install pandas==1.5.3 pandas==2.1.3 -c conda-forge
conda-forge/linux-64                                        Using cache
conda-forge/noarch                                          Using cache

Pinned packages:
  - python 3.12.*

error    libmamba Could not solve for environment specs
    The following packages are incompatible
    ├─ pandas 1.5.3  is installable with the potential options
    │  ├─ pandas 1.5.3 would require
    │  │  └─ python >=3.10,<3.11.0a0 , which can be installed;
    │  ├─ pandas 1.5.3 would require
    │  │  └─ python >=3.11,<3.12.0a0 , which can be installed;
    │  ├─ pandas 1.5.3 would require
    │  │  └─ python >=3.8,<3.9.0a0 , which can be installed;
    │  └─ pandas 1.5.3 would require
    │     └─ python >=3.9,<3.10.0a0 , which can be installed;
    └─ pin-1 is not installable because it requires
       └─ python 3.12.* , which conflicts with any installable versions previously reported.
critical libmamba Could not solve for environment specs

[jchorl]

I think it should be ready for another look!

[jonashaag]

Nice thank you!

[AntoinePrv]

Sorry @jonashaag for the fail merge attempts. I was experimenting with this merge queue, but it seems not to be working well. Giving it a last try or I will remove the setting.

[AntoinePrv]

Thank you for your patience @jchorl