Enhancement: improve animated/multipage image resizing

[lovell]

Given Safari will soon be supporting WebP, I suspect a shift away from the use of animated GIF images towards the use of animated WebP images, which are significantly more efficient, is coming.

Operations like extract and rotate on individual pages/frames might be rather complex, but a possible new resizePages operation to resize all pages/frames in an image to a given width or height (these would have to be mutually exclusive, plus an optional kernel), would be relatively straightforward.

Update: this has been implement via #2789, which sets page heights automatically using the existing resize API.

[OJezu]

Why not load all pages by default, though? Is BC the only concern? I guess it can break "convert first frame of gif/tiff to jpeg" scenario.

[tpetry]

Why not simply apply all operations on all frames? I would need compositing functionality for animated gifs. Duplicating every method to apply for all frames would be a rather complex task.
It would be easier applying all operations on all frames and the ability to extract single frames or sequences and construct new animated images based on frames/sequences. So compositing images on a subset on frames can be implemented based on these primitives.

[salipitchero]

Hi,

was a decision ever made on this enhancement? A resize function for multiple frames would be really useful. Failing that, is it possible to do with the current sharp API, and could some docs be added instead for the mean time?

Thanks

[lovell]

The future possible enhancement in PR #2789 may remove the need to provide pageHeight.

[wchaws]

Hi, @lovell

How about expose arrayjoin API in sharp to support animated/multipage image resizing? I found a sample code at libvips/libvips#1167 (comment). It's quite simple. The only missing API is arrayjoin. I found this API is already existing in sharp source code.

sharp/src/libvips/cplusplus/vips-operators.cpp

Line 283 in 1ff84b2

VImage VImage::arrayjoin( std::vector<VImage> in, VOption *options )

PROPOSED API

sharp.arrayjoin([img0, img1, img2, ...], {})

[lovell]

@wchaws Please see #1580

[lovell]

This oft requested feature is now available in v0.30.0 - enjoy!

Thanks again, as always, to @kleisauke for doing the work to implement it.