Re: GnomeCanvasText behavioural change
- From: Keith Packard <keithp keithp com>
- To: Miles Lane <miles megapathdsl net>
- Cc: ERDI Gergo <cactus cactus rulez org>, Alex Larsson <alexl redhat com>, gnome-2-0-list <gnome-2-0-list gnome org>, David Turner <david turner freetype org>
- Cc: Keith Packard <keithp keithp com>
- Subject: Re: GnomeCanvasText behavioural change
- Date: Mon, 10 Dec 2001 12:04:16 -0800
On Mon, 2001-12-10 at 10:24, ERDI Gergo wrote:
> I'll never understand why one would want to scale text by rendering it
> into a bitmap and scaling that.
> What should be done is rendering it anti-aliased, at a _very large_ size,
> into a greyscale pixbuf, and then translating that. Or maybe render the
> text with size = requested_size * zoom_factor
I suggest that you should probably always use FreeType to rasterize the
glyphs as this will avoid any pixelized edges. It also uses an analytic
algorithm to compute pixel coverage which generates reasonable AA results
much faster than supersampling and filtering.
The question then is whether you want to scale the fonts before or after
hinting the glyphs -- if you scale beforehand, you'll get the best looking
glyphs at any particular size. If you scale afterwards, you'll get glyphs
that look the same at any particular zoom factor, but which will
invariably have fuzzy edges.
Xft lets you do either; it provides a pixel size and a separate affine
transformation matrix.
Keith Packard XFree86 Core Team Compaq Cambridge Research Lab
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]