PS/PDF API Change Proposal: (Re: [cairo] Semantics of transparent
objects)
Bill Spitzak
spitzak at d2.com
Fri Jan 20 14:40:59 PST 2006
Carl Worth wrote:
> On Fri, 20 Jan 2006 11:29:34 -0800, Bill Spitzak wrote:
>
>>If you make a COLOR window, and draw CLEAR into it, do you get
>>black?
>
>
> Yes.
>
>
>>If you make a COLOR_ALPHA and draw CLEAR into it, and then copy it to a
>>COLOR surface, do you get black?
>
>
> "Copy" meaning what?
>
> If you mean drawing with CAIRO_OPERATOR_SOURCE then, yes you will get
> black in that case.
Yes this is what I meant by "copy".
> But most of the time, people drawing one surface
> to another will be using CAIRO_OPERATOR_OVER, (which is the default)
> and the cleared portions will result in no change to the destination
> surface in this case.
Yes, but OVER produces the same output for all three of the current
COLOR_ALPHA, COLOR, and my proposed "always over white" COLOR.
I thought the original problem was that "CLEAR" produced unexpected
results on the PostScript output. It seems now there are three solutions:
1. From the above, it sounds like doing nothing (where PostScript uses a
COLOR surface and CLEAR produces black) is a perfectly acceptable
solution, in that it matches other COLOR surfaces perfectly.
2. Have PostScript use the COLOR_ALPHA model. This means CLEAR produces
white, and also things using the destination alpha work. This appears to
be the preferred solution, however I am wondering if the overhead of
supporting the seperate alpha channel is worth it. If PostScript is
defined to be a COLOR surface then a lot less possible composites need
to be emulated. Also preview in a COLOR window is still broken and will
show black where the PostScript output is white.
3. My proposal was to alter the COLOR model so that CLEAR produces
white, but that other operations that use destination alpha continue to
act as though the destination is 1 everywhere.
In any case all three of these do not require specifying the surface
type when opening the printer. Just have the printer default to whatever
it uses.
More information about the cairo
mailing list