Re: comments on Matrix

I'd like to throw another voice in against throwing exceptions. The faulty
assumption here being that if a matrix fails to invert this is worth
stopping the entire app for.

Realistically most authors will blindly invert and never give a second
glance to the result. If invert throws this will result in many apps simply
halting, with the only way to get it back being a refresh. If invert
returns NaNs or similar most apps will have a bit of graphical data
flicker, become lit incorrectly, or possibly disappear. Depending on the
structure of the app this may naturally resolve itself within a frame or
two. While you can argue that a throw will make invert issues easier to
debug I feel that most authors would rather have a suboptimal user
experience than no user experience.

If you absolutely MUST retain compatibility with this aspect of SVGMatrix
then so be it, but I would highly favor including a non-throwing variant
and promoting that one as the preferred call.

--Brandon Jones


On Thu, Mar 21, 2013 at 11:59 AM, Dirk Schulze <dschulze@adobe.com> wrote:

>
> On Mar 21, 2013, at 11:52 AM, "Tab Atkins Jr." <jackalmage@gmail.com>
> wrote:
>
> > On Thu, Mar 21, 2013 at 11:47 AM, Rik Cabanier <cabanier@gmail.com>
> wrote:
> >> On Thu, Mar 21, 2013 at 11:36 AM, James Robinson <jamesr@google.com>
> wrote:
> >>> On Thu, Mar 21, 2013 at 11:33 AM, Rik Cabanier <cabanier@gmail.com>
> wrote:
> >>>> I'll formally object to an exception being thrown here.
> >>>>
> >>>> SVG matrix throws an exception and since this is a drop-in
> replacement,
> >>>> Matrix needs to throw one too :-(
> >>>
> >>> This is a clear example of SVG making a (bad) mistake that we should
> not
> >>> inflict on the rest of the web platform.
> >>
> >> Are you objecting to any exceptions being thrown?
> >> There are exceptions in all authors facing APIs.
> >
> > No, he objects for the same reason I am - a matrix being singular is
> > not a particularly exceptional situation
>
> A singular matrix does not mean the same as NaN and Inf values in the
> matrix.
>
> However, you need to take a look when you use inverse() for transformation
> matrices (again, that is the focus here). They are used to transform
> between coordinate spaces. And then a returning singular matrix does not
> make sense.
>
> Dirk
>
> > , and whether an
> > implementation considers a matrix singular will depend in some cases
> > on the exact algorithm and FP implementation being used.  This is,
> > thus, a *terrible* example of a place to throw an exception, as
> > authors won't expect it, and it'll vary between platforms for no
> > important reason.
> >
> > ~TJ
> >
>
>