Documentation issues : flags not that simple

[HuskyNator]

I'm trying to save an array of pixels but can't seem to do so.
I can't access the pixel data directly (mutate it), can't seem to load pixel data directly (eg using loadFromMemory.
I tried using packed (gapless) data and allPixelsAtOnce on an initialized Image, but it only seems readable.

[p0nce]

What would have help you achieving this more readily?

[p0nce]

@HuskyNator perhaps there is a documentation issue we could fix?

[HuskyNator]

It'd be nice to be able to use loadFromMemory using a plain pixel array?
I feel like I'm hacking my way around this by using an assignment like allPixelsAtOnce()[] = pixelsArray[].
Apart from that, it wasn't really clear whether this was possible from the docs. (Though I got it to work :) )

[p0nce]

Ah yes, loadFromMemory() is to load en encoded-image.
You are supposed to use image.createViewFromData() and then you have an Image which is a reference to that data (that you can .clone).

Maybe a new call that do that internally would be cool, please suggest a name maybe? I'm thinking of loadFromPixels or loadFromData.

[HuskyNator]

Ah I see. I totally missed that comment! Totally my bad :)

(EDIT: from p0nce: I'm sorry I accidentally replied in your comment!!!)

In what way would the two be different? I'd normally say making it doable using loadFromMemory, but that might be confusing/complex. I think loadFromPixels or loadFromPixelData works well. :)

Now I see how loadFromMemory can be a trap as an identifier.

Also slightly confused about the existence of LoadFlags in this case.
They're not used by createViewFromData right?

No because by creating from pixel data you cannot force some layout from load only (unlike, say, an image decoder could load a rgb8 image as rgba8 and save some time).

But you can change type with:
image.convertTo(PixelType type);
and change layout constraints with:
image.setLayout(LayoutConstraints constraints);

If your memory pixel already fulfilled some of the layout constraints, the constraints should be accepted without reallocating or cloning those pixels.

Honestly quite confused about the difference between Layout, PixelFormat and LoadFlags.
They seem to overlap a bit?

• PixelType is how the pixels are arranged, number of channels and bits. You can change it with convertToXXX.
• LayoutConstraints are how these pixels need to be arranged in memory with excess bytes outside the scanlines or vertical flip, but that doesn't change the content (interior) of the image. You can change it with setLayout. Layout constraints are "ad-hoc" inferred in order to not reallocate if some constraints can be followed already. For example, if you load an image whose width is 104, then the constraints "width must be multiple of 4" will be for free if you where to apply it.
• LoadFlags are an optimization, a way to apply setLayout and convertTo at load time, in order to save on reallocations. Indeed some image decoders can take advantage of that to convert scanlines as they are decoded. Similar to how stb_image can force a channel count at load, if we don't have that then gamut would be slower in some cases.