On Tue, 24 Mar 2015 22:36:17 +0100, Dirk Schulze <[email protected]>
wrote:
On Mar 24, 2015, at 10:27 PM, Simon Pieters <[email protected]> wrote:
Most interfaces in the Geometry spec have a dictionary type so you can
use a direct JS object as an argument instead of requiring the object
to implement the relevant interface. DOMMatrix doesn't have a
dictionary though. It seems to me that it should.
Since DOMMatrix has both a-f and m11-m44 attributes, the dictionary
will have to support members for all those as well and accept objects
where all are set.
It probably makes sense to have a dictionary member for is2D, to
preserve it being false even though the members indicate a 2d matrix.
The constructor could throw TypeError if it was set to true but the
other members indicate a 3d matrix.
I don't see any reason to have a dictionary member for isIdentity.
The a-f members and the corresponding mXX members can't have default
values in the IDL, but can be defaulted in prose in the algorithm. If
e.g. a and m11 are set to different values we could throw TypeError.
dictionary DOMMatrixInit {
unrestricted double a; // 1
unrestricted double b; // 0
unrestricted double c; // 0
unrestricted double d; // 1
unrestricted double e; // 0
unrestricted double f; // 0
unrestricted double m11; // 1
unrestricted double m12; // 0
unrestricted double m13 = 0;
unrestricted double m14 = 0;
unrestricted double m21; // 0
unrestricted double m22; // 1
unrestricted double m23 = 0;
unrestricted double m24 = 0;
unrestricted double m31 = 0;
unrestricted double m32 = 0;
unrestricted double m33 = 1;
unrestricted double m34 = 0;
unrestricted double m41; // 0
unrestricted double m42; // 0
unrestricted double m43 = 0;
unrestricted double m44 = 1;
boolean is2D;
};
Does this seem reasonable?
We didn’t want to add more things than there were requests for
initially. I think this is the only reason why we didn’t add it in the
first place.
OK, thanks. The current situation is inconsistent, which seems bad for Web
developers.
--
Simon Pieters
Opera Software
