==0.1.* doesn't match 0.1+upstream.2

[habnabit]

As in the title, ==0.1.* doesn't match the version 0.1+upstream.2. I discovered this as a result of specifying a requirement ~=0.1.0, and getting 0.1 instead of 0.1+upstream.2. I can produce a reproducing example if necessary.

[dstufft]

Some notes from when I was testing this:

>> import packaging.specifiers
>>> packaging.specifiers.SpecifierSet("~=0.1.0").contains("0.1+upstream.2")
False
>>> packaging.specifiers.SpecifierSet(">=0.1.0,==0.1.*").contains("0.1+upstream.2")
False
>>> packaging.specifiers.SpecifierSet(">=0.1.0").contains("0.1+upstream.2")
True
>>> packaging.specifiers.SpecifierSet("==0.1.*").contains("0.1+upstream.2")
False
>>> packaging.specifiers.SpecifierSet("==0.*").contains("0.1+upstream.2")
True
>>> packaging.specifiers.SpecifierSet("==0.1.*").contains("0.1+upstream.2")
False
>>> packaging.specifiers.SpecifierSet("==0.1.0.*").contains("0.1+upstream.2")
False

It appears as if we do this correctly when ==0.* is the version specifier, but not when ==0.1.* is the version specifier.

We might also want to double check PEP 440 to make sure that this particular use case is clearly defined.

[dstufft]

Oh, and for completeness sake:

>>> packaging.specifiers.SpecifierSet("==0.1.*").contains("0.1")
True

[habnabit]

While I'm thinking about this, it would also be nice to be able to have requirements be able to match against upstream versions. >=0.1.0+upstream.2 ignores the bit after the +, and effectively works the same as >=0.1.0.

[dstufft]

Double checking PEP 440, I see why that doesn't work:

PEP 440 says:
Local version identifiers are NOT permitted in this version specifier.

I don't recall exactly why we did that, I think it was likely to prevent (or discourage) people from uploading things to PyPI that had a dependency on a particular local version. In fact, the PEP disallows local versions like that in all specifiers except for == (without the .*) and ===. The "ignoring" behavior you're seeing is because we fall back onto a legacy implementation if something isn't a valid PEP 440 specifier which I think probably has that behavior.

[dstufft]

Oh, to be clear, changing the behavior of >= will require at a minimum, discussion on distutils-sig, but the behavior of ~= is a bug.

[xavfernandez]

The ~= behavior should be fixed by #47