Bug 23243 – Need to add multLeft() to TransformationMatrix and WebKitCSSMatrix

• Status: RESOLVED
• Resolution: FIXED
• Component: Layout and Rendering

[Simon Fraser (smfr)]

TransformationMatrix::multiply() does a multRight with the supplied matrix. The other operations, like scale(), are applied to the left (as expected).

I think TransformationMatrix needs explicit multLeft and multRight (or pre- and post-multiply).

[Chris Marrin]

I don't think it's really necessary to have both variants. This is usually provided as an optimization, since the operation is self-modifying. Today, if you wanted to multiply b * a, but put the result in a, you would have to:

    TransformationMatrix tmp = b;
    a = tmp.multiply(a);

Which isn't that bad. Alternatively, we could change to a non-destructive multiply, which would ALWAYS be more inefficient, rather than just when you need to flip the operation order. I don't think our use of TransformationMatrix is extensive enough to warrant this.

[Simon Fraser (smfr)]

We have determined that SVGMatrix's multiply() is a multRight(), which also does not match its scale() etc, and indeed we think that TransformationMatrix should be more explicit about left vs. right multiply.

[Chris Marrin]

Now multiply() does a multRight() in TransformationMatrix, CSSMatrix and SVGMatrix. I think we should leave it that way for compatibility and add a multLeft() function (at least to TransformationMatrix and CSSMatrix).

[Jeff Schiller]

I'm ready and willing to hack - does anyone care to mentor me through this and Bug 16062?  Is there a good IRC channel to ask questions?

[Simon Fraser (smfr)]

Find me (smfr) on #webkit later today.

[Jeff Schiller]

In looking at the source, TransformationMatrix now has an explicit multLeft() and a multiply(), which does a multRight.  This was done by cmarrin for Bug 6868 and landed r40761.

I will fix SVGMatrix's incorrect matrix multiplication as part of Bug 16062.

Is there anything left to do for this bug?

[Jeff Schiller]

smfr: I think this bug can be closed, do you agree?

[Simon Fraser (smfr)]

Yes.