Py5Graphics transparency on save different from Processing Python?

[villares]

I have tried this, and I get the transparent background I expected:

def setup():
    size(50, 50)
    sf = 10
    f = createGraphics(width * sf, height * sf)
    beginRecord(f) 
    f.scale(sf)
    circle(6, 6, 10)
    endRecord()
    f.save("out.png")

But in py5 I get a black background. Even if I try f.clear().

def setup():
    size(50, 50)
    sf = 10
    f = create_graphics(width * sf, height * sf)
    begin_record(f)
    # f.clear()  # doesn't help
    f.scale(sf)
    circle(6, 6, 10)
    end_record()
    f.save("out_py5.png")

Is this a known issue or is it by design? Am I doing something wrong?

[Maltiec]

you should set parameter drop_alpha=False:

f.save("out_py5.png", drop_alpha=False)

by default py5 drops the alpha channel for compatibility with file formats like jpg

[hx2A]

you should set parameter drop_alpha=False:

f.save("out_py5.png", drop_alpha=False)

by default py5 drops the alpha channel for compatibility with file formats like jpg

Excellent answer, @Maltiec !

py5 uses the Python library Pillow to save images. Pillow will throw an error if you try to save an image with an alpha channel to a format that does not support it. However, it might be possible to improve this. What if py5 would check the alpha channel before dropping it? If the alpha channel is nothing but 255 values, it adds no value. If it contains other values, it can not drop it, attempt to save with Pillow, and hope that the user picked a file format that supports alpha values.

[villares]

Great! Now I understood. Thanks @Maltiec

@hx2A, instead of looking at the alpha contents, why not do like Processing and look at the output formats only and drop for JPG and formats that don't have alpha?

Could we do something like this perhaps?

if drop_alpha is None:
    try:
        output_buffer.save(output_file_path, drop_alpha=False)
    except OSError:
        output_buffer.save(output_file_path, drop_alpha=True)
else:
    output_buffer.save(output_file_path, drop_alpha=drop_alpha)

[hx2A]

@hx2A, instead of looking at the alpha contents, why not do like Processing and look at the output formats only and drop for JPG and formats that don't have alpha?

Hmmm, that could work, but Pillow supports many image formats, many of which I am not familiar with. Plus, their documentation contains gems like this:

Pillow reads and writes WebP files. The specifics of Pillow’s capabilities with this format are currently undocumented.

It would be headache to try to figure out which support alpha channels and which don't.

Could we do something like this perhaps?

if drop_alpha is None:
    try:
        output_buffer.save(output_file_path, drop_alpha=False)
    except OSError:
        output_buffer.save(output_file_path, drop_alpha=True)
else:
    output_buffer.save(output_file_path, drop_alpha=drop_alpha)

Something like that could work. I'd probably emit a Python warning if an exception is thrown so that the user knows to set drop_alpha to True. If someone is saving many frames, they would not want an exception to be thrown and caught every single time (for performance reasons).

[villares]

Cheers @hx2A!

I was trying to show PNG export to someone today, and this didn't work as I expected:

extra_canvas.save('out.png', drop_alhpa=False)

I was going nuts, then I saw my mistake. Can you see it?
I messed the keyword argument name, but the thing is, it failed silently.

Is this a consequence of passing the **kwargs to a delegated Pillow save mechanism? Should we open an issue?

[hx2A]

The kwarg drop_alhpa is misspelled. The drop_alpha kwarg is a py5 kwarg, but if you misspell it, it will be passed along to Pillow instead. If Pillow does not raise an exception when passed unrecognized kwargs, it will be ignored. It is a design decision on their end to either raise an exception or ignore it. I would prefer an exception, and I think that's what matplotlib does? Not sure if I remember correctly. Anyhow, I would check their repo's issues to see if anyone brought this up already, and if not, open an issue to see what they think of it.