Add support for file's builtin magic module

[felixonmars]

file's builtin magic module has a somewhat different API, and it's not
co-installable with pypi:python-magic as both has the same name. Adding
a fallback logic here makes mocket work with either one.

All related tests are passing here.

[mindflayer]

Hi @felixonmars, my main concern is about how to test the two solutions, since they have the same namespace import magic. Fixing the issue you opened (#119) does not seem the most difficult part of the change.

[mindflayer]

I was having a look at https://pypi.org/project/filemagic/ and I thought it was the module you are trying to support but I see it implements a slightly different API. Is that what you used?
If so I guess you wanted to use the flag magic.MAGIC_MIME_TYPE:

>>> magic.MAGIC_MIME_TYPE
16
>>> magic.MAGIC_MIME
1040

[felixonmars]

I was having a look at https://pypi.org/project/filemagic/ and I thought it was the module you are trying to support but I see it implements a slightly different API. Is that what you used?

Actually, it's yet another. The one I use is bundled inside upstream "file" project and unfortunately not available from PyPI: https://github.com/archlinux/svntogit-community/blob/packages/python-magic/trunk/PKGBUILD#L12

So projects like s3cmd has implemented a quite complex compatibility check for the three versions:
https://github.com/s3tools/s3cmd/blob/3d4a4caf91582c2a674deaa837a5e07d96510059/S3/S3.py#L55

[mindflayer]

The one I use is bundled inside upstream "file" project and unfortunately not available from PyPI

So the question is: how to test it in a CI environment if not available in PyPI?

[felixonmars]

So the question is: how to test it in a CI environment if not available in PyPI?

I see no easy answer either. Since the Arch repository version is the "file" one, and I always run tests during packaging, would you feel okay if I report back in a timely manner if the tests are not passing?

[mindflayer]

In that case I will have to exclude that block from mocket's test coverage. Of course I'll be more than happy to accept PRs or issues from you.
Could you try to see if the other flag I suggested is available in the library you are testing? I am curious to see if that's the reason for the two different results.

[felixonmars]

@mindflayer Indeed, that's the problem. I have updated this PR to use MAGIC_MIME_TYPE instead and reverted the test change, all tests are now passing.

[mindflayer]

OK, let me finish what I am doing for my job and I'll release a version of mocket with the changes you need for make it work on Arch.
By the way, I used to be an Arch user, so I am really happy to help.

[felixonmars]

Thanks a lot :)

[mindflayer]

@felixonmars Here you are: https://pypi.org/project/mocket/3.8.7/

[felixonmars]

Thanks!