search

conda-forge / qnorm-feedstock

A conda-smithy repository for qnorm.
BSD 3-Clause "New" or "Revised" License
0 stars 2 forks source link

Why the difference in dependencies between conda and upstream packages? #20

Open DManowitz opened 4 months ago

DManowitz commented 4 months ago

Comment:

@Maarten-vd-Sande Since you seem to be maintaining both the upstream package and this feedstock, I wanted to ask why the 2 have different dependencies. In particular, in the upstream package, you specify python >3.6 and numpy (without qualification). However, in this feedstock, python >=3.8 and numpy <=1.23.5 are dependencies.

Maarten-vd-Sande commented 4 months ago

Hi @DManowitz,

I honestly can't recall! Are you having issues? Feel free to start a PR to match and correct these dependencies, I'll approve it. I am currently out of academia so I do not have the time to spend time on this myself.

Cheers!

DManowitz commented 4 months ago

@Maarten-vd-Sande, I tested out installing qnorm from pip in 3 environments: a python 3.7 environment, a python 3.8 / numpy 1.24.3 environment, and a python 3.10 / numpy 2.0 environment, and in all cases I was at least able to import qnorm and its top-level modules in all 3 environments, so I will put in a PR that loosens the dependency restrictions in this feedstock.

I also tested out the first 2 examples you provide in the base qnorm repo in the python 3.10 / numpy 2 / pandas 2.2.2 environment, and they both worked (aside from a new pandas warning about deprecated behavior).

Maarten-vd-Sande commented 4 months ago

@DManowitz thanks for digging into this. I've approved your PR :pray: .

The build is still failing though

DManowitz commented 4 months ago

The build is still failing though

It appears to be an issue with restrictions on setuptools with python 3.12. I don't suppose you recall why setuptools is restricted to v60.*, do you?

If not, I'll remove the upper bound restriction and see what happens.