[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: SVG backend: On-the-fly conversion of Emmentaler/Aybabtu glyphs to p
Re: SVG backend: On-the-fly conversion of Emmentaler/Aybabtu glyphs to paths
Sat, 25 Jul 2009 20:42:21 -0700
On Sat, Jul 25, 2009 at 7:03 PM, <address@hidden> wrote:
> On 2009/07/24 23:30:00, Patrick McCarty wrote:
>> The new code uses a generalized regular expression that will match a <glyph>
>> element with attributes in any order.
>> The code is slower now, I think due to the complexity of the new regular
>> expression. It might be a better idea if I rewrite this in C++ eventually.
> How slow is it? Are we talking would-be-nice-if-it-were-faster slow or
> nobody-will-want-to-use-it slow? The reason I ask is that I think the
> _real_ solution would be to use an actual XML parser to parse the whole
> svg font and dump all the glyphs into a hash table. As it is, you're
> scanning a complicated regexp through the whole font file every time you
> want a glyph.
> Of course, that's a bunch of work and it involves, for example, deciding
> which XML parsing lib to use and whether to use it from C++ and scheme.
> So unless it's unusably slow, I'd suggest adding a TODO and committing
> this patch.
It's not as slow as I made it sound. :-) It really depends on how
many grobs need to be processed.
Though smaller LY files will compile more quickly now with the SVG
backend than with the PS backend (bach-schenker.ly comes to mind).
I've added the TODOs and pushed the patchset.