How to convert and optimize an image to webp in one step with default settings?

[benoit74]

Currently, zimscraperlib makes a distinction between conversion (to another format) and optimization (compression).

This is not really convenient in the case I have at hand (gutenberg scraper) where I have any image I want to convert to WEBP and optimize with default zimscraperlib options/presets. Both in terms of code (two functions to call) and computing efficiency (probably more memory / disk to allocate, and maybe even more CPU power). In could even have adverse effects in terms of final results.

[cking100]

While going through the codebase, I noticed that the code already has a optimize_image() with a convert parameter. So when this parameter is True, the function even tho it is a single call, internally it does run convert first and then optimize it. So two function calls overall.
I think we can have two solutions here- either to make a standalone function convert_and_optimize() that converts the image and also optimizes it in a single pass or we can patch the optimize_image() function and make it single pass logic when the convert parameter is True. For the second solution, we won't have to make any changes in scrapers as I think ted2zim uses this, and we can use this function in gutenberg. What do you think? @benoit74

[benoit74]

I feel like making optimize_image wiser is what I would prefer, because I don't see good reasons to have two API to perform the same function.

In turn, it means that we needs to enhance optimize_xxx functions to accept a convert parameter as well and do the conversion when their input data is not in appropriate format.

Another part of the issue is that optimize_image is only accepting Path while I wanna operate on bytes. The solution is probably as well to extend optimize_image to also accept io.BytesIO as input (and hence add another parameter for expected output format).

Finally, if we could also add support for bytes in src (it makes no sense for dst), it would simplify the logic as well. But I might miss a downside.

[aaryansinhaa]

@benoit74 optimize_image already have convert as a parameter.

def optimize_image(
    src: pathlib.Path,
    dst: pathlib.Path,
    options: OptimizeOptions | None = None,
    *,
    delete_src: bool | None = False,
    convert: bool | str | None = False,
):

optimize_png, jpeg and webp already opterate on bytes, I have used it in an another issue(openzim/warc2zim#455), we just have to modify optimize_gif and optimize_image to operate on bytes. and others to support covert param. can i take this issue?

[cking100]

In turn, it means that we needs to enhance optimize_xxx functions to accept a convert parameter as well and do the conversion when their input data is not in appropriate format.

Agreed. It would be better to patch the optimize_image and other optimize_xxx functions to accept a convert parameter and perform conversion when needed.

Another part of the issue is that optimize_image is only accepting Path while I wanna operate on bytes. The solution is probably as well to extend optimize_image to also accept io.BytesIO as input (and hence add another parameter for expected output format).

Sounds good. We can extend the optimize_image and optimize_gif to accept bytes for src. Could you assign me? @benoit74

[benoit74]

I will let @cking100 work on the PR since, and @aaryansinhaa you are welcomed to review it.