[scml documentation (deprecated - use reference implementations instead)]

I've been writing Spriter integration into my own engine, and I think I've discovered and solved an error in the reference.

The issue lies in the applying of a parent's transform onto child bones:

SpatialInfo unmapFromParent(SpatialInfo parentInfo)

 

The current implementation of this function doesn't support non-uniform scaling values, and doesn't support the "flipping" of sprites via a negative scaling. I've found a fix on this forum for the latter, but the former I couldn't find any solution for. I've managed to solve the issue myself, so I would recommend anyone implementing the reference to use this version instead:

 

SpatialInfo unmapFromParent(SpatialInfo parentInfo)
{
    SpatialInfo unmappedObj = this;

    if (parentInfo.scaleX * parentInfo.scaleY < 0)
    {
        unmappedObj.angle = (360 - unmappedObj.angle) + parentInfo.angle;
    }
    else
    {
        unmappedObj.angle = unmappedObj.angle + parentInfo.angle;
    }

    unmappedObj.scaleX *= parentInfo.scaleX;
    unmappedObj.scaleY *= parentInfo.scaleY;
    unmappedObj.a *= parentInfo.a;

    if (x != 0 || y != 0)
    {
        // Negate angle if the sprite is "flipped"
        float angle = parentInfo.scaleX * parentInfo.scaleY < 0 ? -parentInfo.angle : parentInfo.angle;                                          
        float s = sin(toRadians(angle));
        float c = cos(toRadians(angle));

        unmappedObj.x = parentInfo.scaleX * ((x * c) - (y * s));
        unmappedObj.y = parentInfo.scaleY * ((x * s) + (y * c));

        unmappedObj.x += parentInfo.x;
        unmappedObj.y += parentInfo.y;
    }
    else 
    {
        // Mandatory optimization for future features           
        unmappedObj.x = parentInfo.x;
        unmappedObj.y = parentInfo.y;
    }

    return unmappedObj;
}

 

Hope this saves someone a lot of time and frustration.

 

This solution sadly doesn't preserve the Sprite structure when rotating and using non-uniform scaling at the same time. However, I've tested every other case and it works. I would imagine a perfect solution would scale the root bone around the axis of rotation, and leave the rest of the child rotations untouched, but I haven't gotten it to work yet.