Regression: Starling XML animations with extra numbers not handled properly

[Jsfasdf250]

The bug where animation names with numbers in addition to the frame index are mishandled was fixed in commit 454f714, but it now occurs again. Example name attributes of SubTextures in an affected TextureAtlas:

Anim10001
Anim10002
...
Anim20001
Anim20002
...

I believe that the earliest versions would combine all of these frames, then the behavior was changed so that Anim1 and Anim2 would be processed separately, and now I believe only SubTextures prefixed with Anim1 would be included.

[MeguminBOT]

Anim10001
Anim20001
Anim30001

Patterns works just fine for me to extract.
Tested using Pico Speaker from Week7 in Psych Engine.

If you have a starling sheet with the issue you are describing, please attach it.

[Jsfasdf250]

banban.xml

[MeguminBOT]

This is malformed by starlings standards atleast when exporting with Adobe Animate

<SubTexture name="banban10008" x="292" y="0" width="303" height="374" frameX="0" frameY="-8" frameWidth="303" frameHeight="383"/>
<SubTexture name="banban20009" x="595" y="0" width="227" height="377" frameX="-53" frameY="0" frameWidth="303" frameHeight="383"/>

"2xxxx" animation should start at 20000.
I'll see if I can add a workaround to take this into account though.

[Jsfasdf250]

NoI was able to tweak the strip_trailing_digits method to choose the number of digits to strip, then appending the arguments 4, 4, to the usages in the extract_tab_widget and the sprite_processor files. This fixes the problem of frames being incorrectly combined for SubTexture names which use four digits for frame indexes and a fifth for separate animations, but it may be better to be able to edit this setting in the preferences or automatically detect how many digits are used for frame indexes.

def strip_trailing_digits(name: str, min: int=0, max: int="") -> str:
    """Remove trailing frame numbers and optional file extension.

    Strips any number of trailing digits, preceding underscores/spaces,
    and any trailing underscores or whitespace.

    Args:
        name: Sprite or frame name (e.g., ``"idle_0001"`` or ``"run 5.png"``).
        min: Minimum number of digits to strip.
        max: Maximum number of digits to strip.

    Returns:
        Base animation name with trailing digits removed.
    """

    return (
        re.sub(
            fr"[_\s]*(?:\d{{{min},{max}}})?(?:\.(?:png|jpe?g|gif|webp|bmp|tiff?|tga|avif|dds))?$",
            "",
            name,
            flags=re.IGNORECASE,
        )
        .rstrip("_")
        .rstrip()
    )

[MeguminBOT]

NoI was able to tweak the strip_trailing_digits method to choose the number of digits to strip, then appending the arguments 4, 4, to the usages in the extract_tab_widget and the sprite_processor files. This fixes the problem of frames being incorrectly combined for SubTexture names which use four digits for frame indexes and a fifth for separate animations, but it may be better to be able to edit this setting in the preferences or automatically detect how many digits are used for frame indexes.

def strip_trailing_digits(name: str, min: int=0, max: int="") -> str:
    """Remove trailing frame numbers and optional file extension.

    Strips any number of trailing digits, preceding underscores/spaces,
    and any trailing underscores or whitespace.

    Args:
        name: Sprite or frame name (e.g., ``"idle_0001"`` or ``"run 5.png"``).
        min: Minimum number of digits to strip.
        max: Maximum number of digits to strip.

    Returns:
        Base animation name with trailing digits removed.
    """

    return (
        re.sub(
            fr"[_\s]*(?:\d{{{min},{max}}})?(?:\.(?:png|jpe?g|gif|webp|bmp|tiff?|tga|avif|dds))?$",
            "",
            name,
            flags=re.IGNORECASE,
        )
        .rstrip("_")
        .rstrip()
    )

Yeah might need delegated to a setting, since currently the extractor behaves exactly like we want it to, it's just outliers like this that need some tweaking.
I'll mess around with it when I have time, currently busy with my job and some music projects.

[MeguminBOT]

We could detect whether the 5 digit index starts with e.g "2" but doesn't start at 0000.
That would prevent breaking existing logic I believe.