Thumbor should process gifs in a faster way.

[maiagripp]

Expected behaviour

Thumbor should process gifs in a faster way.

Actual behaviour

Thumbor is taking too long to process some gifs. So we run some tests using gifsicle (our actual process gifs lib) and pil to compare time processing on both.

Conclusion

We thought, although Gifsicle indeed had a better performance in flips, pil had better processing times in all the others. We don't think the flip filter is heavily used, so the trade off pays itself.

[maiagripp]

@scorphus @guilhermef @kkopachev thoughts about that?

[guilhermef]

I agree with you @maiagripp, using Pil to process GIFs would definitely be an improvement.

[scorphus]

I seem to be missing something. Usage of gifsicle is optional and pil (or whatever engine is configured) is used whenever USE_GIFSICLE_ENGINE is set to None:

thumbor/thumbor/handlers/__init__.py

Line 340 in c30b208

or self.context.config.USE_GIFSICLE_ENGINE is None

Perhaps we should adapt that line and document the behavior.

[scorphus]

Also, this might be a good opportunity to support the cover "filter" when gifsicle isn't enabled.

[maiagripp]

Hey @scorphus,
indeed you are right.
But we ran some tests comparing how Thumbor process GIFs using PIL and the engine we made and still we got better results.
Here what we got:

Processing Time
Crop 
Thumbor                 | Engine
real	0m20.348s         real	0m13.656s

Flips
Thumbor                 | Engine
real	0m22.240s         real	0m11.058s
real	0m20.546s         real	0m12.156s

Resize
Thumbor                 | Engine
real	0m58.331s         real	0m17.397s


Cover
Thumbor                  | Engine
(doesn't work)             real	0m9.699s

Content-Length 
Thumbor                  | Engine
Content-Length: 97324349   Content-Length: 38657490

And the visual result is also better.
On the left side is Thumbor and on the right side is the engine

[marcelometal]

@maiagripp please, share your code changes =)

[maiagripp]

Here it is:
Code Gist

Could you take a look and share your thoughts?
@marcelometal @scorphus @guilhermef

[github-actions]

This issue is stale because it has been open 30 days with no activity. Remove the stale label or add a comment, or this issue will be closed in 15 days. You can always re-open if you still feel this is still an issue. Tag @heynemann for more information.

[github-actions]

This issue was closed because it has been stale for 15 days with no activity.