Incomplete mime_type data for several ContentTypes

[Nicba1010]

I do no know if this is by design, but some ContentTypes seem to be missing their associated mime_type values.

e.g. the MKV content type specifies its mime_type as None instead of video/x-matroska

[reyammer]

Thanks for reporting! If a content type that is actually used by any of the model has any of the metadata set to None, that's not by design. But I just checked MKV, and it has the correct mime_type (see https://github.com/google/magika/blob/main/assets/content_types_kb.min.json)? where did you see its mime_type to None?

Are you maybe checking an old version of the metadata?

[Nicba1010]

Well I'm using the magika Python package (version 0.5.1) on Python 3.12.

This is the code I used to print all the MIME Types.

from magika.content_types import ContentTypesManager


def get_mime_type_list() -> list[str]:
    return list({
        str(content_type.mime_type)
        for name, content_type
        in ContentTypesManager().cts.items()
        if content_type.mime_type is not None
    })

if __name__ == '__main__':
    for mime_type in get_mime_type_list():
        print(mime_type)

Also second question, I see that on 0.6.0rc3 the ContentTypesManager is gone. I can't seem to find it to test this with 0.6.0rc3.

[reyammer]

[Sorry for late reply, was on holidays.]

Thanks for clarifying. 0.5.1 is indeed the old version; in the new update we should have proper metadata for all the content types supported by any of the models.

And yes, the ContentTypesManager class has been refactored out (and it was never supposed to be used by external clients). The same information is now in a private field _cts_infos (https://github.com/google/magika/blob/main/python/src/magika/magika.py#L98), stored in a better way (with enums and such), but again, we did not think about this as to be used by external clients.

One thing to keep in mind: there is a list of "content types a given model supports" vs. "the full list of content types we are aware of" (which is a super set of the former, and for which the metadata may not be ready for consumption).

What would be your use case for using it? If we were to add it, which kind of APIs would you like to have?

[Nicba1010]

All good, thanks for replying at all. I send it out to the frontend of my app because a user can type in a mime type into a field and this is how I verify it actually exists. Might be quite a niche use case though and there are probably better solutions.

[reyammer]

I see, interesting. In general I believe the knowledge base of types may useful for use cases like yours, so I see a good argument to expose it in some way or another. Maybe we could add some methods to the Magika class itself, something that would allow you to "get all content types" (getting a list of ContentTypeLabel (which is an enum), and then another method to query the metadata associated to each of them. Opened #826 to track this. Closing this one as I believe we clarified why things were missing; feel free to re-open and follow up if something is still not clear. Thanks!