not sure how to fix this. in my environment.yml file, it did not work when I specified it as @martin-g did. It only worked when I allowed pip to install the package.

i'd gladly change it, but i'm unsure how to track down and resolve the conflict.

Which version has been installed by pip ? Maybe we can try with exact version instead of a range ?!

Which version has been installed by pip ? Maybe we can try with exact version instead of a range ?!

version 2.1.1. So i had thought that specfiying khmer >=2.1 would be sufficient.

https://pypi.org/project/khmer/

just checked the bioconda page for khmer...it appears that 2.1.1 is not a version there.

looks like maybe the problem is the version of python it is dependent on:

depends python: >=3.6,<3.7.0a0

primerforge would need at least v3.9+ since it uses typehinting which wasn't formally supported until 3.9. I would expect to get a lot of errors saying like "cannot subset type tuple" because a function's signature line had something like def f(n:int) -> tuple[str,int]:. I am not entirely sure what version of python allows this, but i know with certainty that 3.7 would complain.

would including a line like pip install khmer in a build.sh file be compliant with bioconda syntax? or is this still a bad practice?

I think we have a few options

1. Try building khmer 2.1.1 with latest python pinnings
2. try rebuilding khmer 2.1 with latest python pinnings
3. (do not recommend) pingeon-hole primerforge to only work with python 3.6 (which might not even be supported by primerforge)

would including a line like pip install khmer in a build.sh file be compliant with bioconda syntax? or is this still a bad practice?

You are correct, this wouldn't be allowed

@rpetit3 thanks for the input. Here are my thoughts.

I think we have a few options

1. Try building khmer 2.1.1 with latest python pinnings
2. try rebuilding khmer 2.1 with latest python pinnings
3. (do not recommend) pingeon-hole primerforge to only work with python 3.6 (which might not even be supported by primerforge)

TL;DR

Option 3 is probably a bad choice; where can I find resources for doing options 1 or 2?

full response

I think option 1 is the best option, but khmer 2.1.1 does not have a conda recipe. So maybe that means option 2 is better? To be honest, I don't really know how we would go about rebuilding a recipe with the latest python pinnings. Is there a reference available that I could use to figure this out?

Option 3 is not good because I know for sure that the inclusion of type hinting without importing typing is not supported below python 3.9. If I had said like from typing import List, Dict, Tuple then it may have been compatible with lower versions. Obviously I would prefer not to rewrite the entire code base just to make a functioning conda installation especially when my pip installer works just fine.

As a note, khmer does not work with python 3.12 because SafeConfigParser was replaced with ConfigParser. But the pip installation does work with python >=3,9,<3.12.

additional note

would including a line like pip install khmer in a build.sh file be compliant with bioconda syntax? or is this still a bad practice?

You are correct, this wouldn't be allowed

Ok thanks for clearing that up. I had started working on this as a solution (this is what I was trying to do in my build.sh file, but I will discard this work and move things back to how they were.

this recipe should work once the recipe for khmer is working with higher versions of python3. I am trying to do this in PR #49359

updated the version, but this recipe is still broken since the recipe for khmer is incompatible with the python3 versions required for primerForge

Thank you for using CodeRabbit. We offer it for free to the OSS community and would appreciate your support in helping us grow. If you find it useful, would you consider giving us a shout-out on your favorite social media?