Normalize installed package names

[blueninja416]

Currently if a package is somehow named with an underscore, it will not be recognized as installed by ipydeps. Changing

def process_pip_freeze_output(pkgs) -> list:
    pkgs = str(pkgs, encoding='utf8').split('\n')
    pkgs = [p for p in pkgs if len(p) > 0 and '==' in p]
    pkgs = [get_freeze_package_name(p) for p in pkgs]
    return pkgs

to

def process_pip_freeze_output(pkgs) -> list:
    pkgs = str(pkgs, encoding='utf8').split('\n')
    pkgs = [p for p in pkgs if len(p) > 0 and '==' in p]
    pkgs = [get_freeze_package_name(p) for p in pkgs]
    pkgs = normalize_package_names(pkgs)
    return pkgs

fixes this issue.

[Universe-sudojam]

Could you give an example that shows the issue? I am trying to recreate the issue on my end but not seeing it.

[blueninja416]

I can't find the exact package that caused the issue at the moment, but somehow the package name included an underscore when it shouldn't have.

[Universe-sudojam]

Hmm, if it has random underscore in it, wouldnt the normalize function make it into dash which wouldnt solve the underlying issue?

[blueninja416]

The packages in the install list are normalized, but the packages in the pip freeze output isn't, so if you normalize the pip freeze output, it will match the names of the packages to install.

[Universe-sudojam]

Ohhh, got it! I was having an issue like that when working on #55 where couple of stdlib packages for Python 3.10+ had underscore one of them is pydocs_data. Were you thinking of those?

[blueninja416]

I don't believe it was that specific package, but it could've been something like that. Either way, it sounds like the same issue with the underscores