On Thu, Jun 5, 2014 at 9:40 PM, Dirk Schulze <dschu...@adobe.com> wrote:
> > On Jun 6, 2014, at 6:27 AM, Robert O'Callahan <rob...@ocallahan.org> > wrote: > > > On Fri, Jun 6, 2014 at 4:22 PM, Dirk Schulze <dschu...@adobe.com> wrote: > > What about > > > > DOMMatrix(1,0,0,1,0,0) or > > DOMMatrix(1,0,0,0,0,1,0,0,0,0,1,0,0,0,0,1) > > > > Do we check the values and determine if the matrix is identity or not? > If we do, then authors could write DOMMatrix(other.a, other.b, other.c, > other.d, other.e, other.f) and check any kind of matrix after transforming > for identity. In this case, a real isIdentity check wouldn’t be worst IMO. > > > > I would lean towards just setting isIdentity to false for that case, but > I could go either way. If authors try really hard to shoot themselves in > the foot, they can. > > > > Rob > > Just as comparison: Gecko checks for IsIdentity 75 times (exclusive the > definitions in matrix and matrix4x4). Every time the values are simply > checked for 0 and 1. Means Gecko is shooting itself in the foot quite often > :P. (In WebKit it is about ~70 times as well.) The question is not that 'isIdentity' is bad. Benoit's issue was that checking for 'isIdentity' after doing transformations might cause jittery results (ie switch to true or false depending on the conditions). Quickle scanning mozilla's codebase, our current definition of 'isIdentity' would return the correct result in all cases. _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
