incomplete type stubs in 1.11 release

[David Hotham]

> Module "pkginfo.distribution" has no attribute "NewMetadataVersion"

... says mypy.

not sure whether you are writing the stubs by hand or generating them or what, but it looks as though things added in the 1.11.0 release didn't make it through

[Tres Seaver (tseaver)]

Thanks for the report! I'm not sure how to repro: running 'mypy --strict' reports no errors for me.

[Tres Seaver (tseaver)]

OK, the check should be 'stubcheck pkginfo.distribution'. I'll need to work out automating the check under 'tox'.

[Tres Seaver (tseaver)]

https://pypi.org/project/pkginfo/1.11.1/

[David Hotham (absence-the-absence)]

thanks! I probably could have made clearer that I was seeing mypy report errors on a project that uses pkginfo, rather than on pkginfo directly.

I think it is rather unusual to have separate stub files, most projects that I see annotate the code directly. I am mildly curious about whether you have reasons for preferring this approach - or perhaps it just happened to turn out that way.

Anyway, can confirm that mypy is happy again with pkginfo 1.11.1

[Tres Seaver (tseaver)]

> I think it is rather unusual to have separate stub files, most projects that I see annotate the code
> directly. I am mildly curious about whether you have reasons for preferring this approach
> - or perhaps it just happened to turn out that way.

I find that inlining the type declarations detracts quite a bit from the readability of the code: I've worked in large codebases using them, and much prefer to have them out-of-line.

Because 'mypy' was happy running against the out-of-date stubs when run inside the 'pkginfo' project,
I hadn't noticed the stubs were stale. The change I added to 'tox.ini' should keep me from letting
them drift again in the future.