Added support for v3d and m44d

[Ambrosiussen]

[tiltX]

I just tried this patch and I got an error for m44d:

attribute_values = struct.unpack('d' * 32, exr_file.read(4 * 32))
                                       ^ this should still be 16 for a 4x4 matrix

Moreover, an EXR created by the latest Karma renderer (Houdini) failed due to an attribute of type "v2d" which I guess is just like v2i and v2f but twice the size?

[vvzen]

hi @Ambrosiussen ! sorry for the delay and thanks for the MR - I missed the notification for this one but lgtm! Do you mind linking the information that you used to write the implementation for the m44d ? @tiltX: I haven't checked the white paper, but I suppose that v2d is a vector of doubles ?

Thanks all!

[tiltX]

@tiltX: I haven't checked the white paper, but I suppose that v2d is a vector of doubles ?

Yes, it works if I add a case for v2d by duplicating the code for v2f with this adjustment for struct.unpack()

            elif attribute_type == b'v2d':
                vector_2d_value = struct.unpack('dd', exr_file.read(8 * 2))

Unfortunately I don't know what the proper workflow would be to add to @Ambrosiussen 's pull request. I haven't tried reading any of the test cases in the openexrs-images repository to see if they have such a v2d attribute. But both this addition and the fix for m44d above allowed me to parse EXR files written by the latest production build of Houdini's Karma renderer.

[vvzen]

Hey @tiltX ! the cleanest workflow would be to have @Ambrosiussen update his PR so that I can merge all changes in one place. However - if you'd like to contribute, I'd really appreciate if you could make a PR that adds a few of these EXRs generated from Houdini that have these metadata (or you could just send them via email if that's easier for you). I could add them to the test image suite that is used when running the tests. Thanks and happy holidays!