WebP enhancements

[wiredfool]

(promoted from the end of #204)

Possible/Future:

• Check Lossless read/write.
• Possibly build libwebp 0.3.0 on travisCI for testing.
• preallocate buffer in decode webp to prevent a buffer copy

[euangoddard]

Sounds good to me. Lossless decoding we get for free, I think, but encoding we'd need to specify in the extra options PIL core passes on save()

[liftoff]

I have just started experimenting with Pillow's WebP support... Setting the quality to 100 doesn't seem to perform lossless compression. Is that just not implemented yet or do I need to pass an additional argument to save()?

[wiredfool]

Probably not implemented, thought it would take a dive through the code and the webp interface to see for sure.

[liftoff]

I just submitted a pull request to add support for lossless WebP. I've never modified a Python module written in C before so please double-check my changes (barely know C--was mostly just guessing). Seems to work great though (if a bit slow).

[wiredfool]

Thanks for the pr. I think the format string is forcing everything to be lossless. Other than that, the c looks about right from a quick once over.

[liftoff]

I'm using a combination of lossless/lossy right now and I'm not seeing that effect (everything forced to be lossless). How are you calling save()?

[wiredfool]

That was an interpretation of what an object pointer into the lossless int would do. ( since it's not null). Sorry that I'm trying to review a pr from my pythonless iPad.

[euangoddard]

@liftoff - nice one. This is exactly the sort of thing I had in mind!