tlug: XFree86 + X-TrueType: Success!

["Stephen J. Turnbull" <turnbull@example.com>]

>>>>> "Matt" == Matt Gushee <matt@example.com> writes:

    Matt> Stephen J. Turnbull writes:

    >> Seriously, my experience with the Ryobi fonts is about
    >> 10-15 sec/instance in xfontsel.  My XEmacs usually instantiates
    >> about 50 fonts on startup, of which 12 are Japanese.

    Matt> You don't mean 50/12 font *families*, do you? If so, then
    Matt> that raises the question, 'What on earth do you *do* with
    Matt> all of those?' (you don't have to answer if it's
    Matt> embarrassing :-)

No.  I mean font _instances_.  Scaling mincho to 12pt does not provide
any speedup for scaling mincho to 14pt.  The 12 Japanese fonts come
from 3 sizes X 4 families (intended to correspond to roman, italic,
bold, and bold-italic).  I don't use them actually, but I want the
auto-face-derivation code to work with Japanese ;-)  (And someday
they'll be usefull when I start reading text/enriched Japanese.)  The
remainder are defined by various package customizations.  I
experimented with using variable-width fonts in comments with
font-lock, for example, and it works pretty well.

    Matt> Hmm. I don't know if I can come close to duplicating your
    Matt> setup. For one thing, I don't really understand yet how to
    Matt> put together a collection of XEmacs faces that is at once
    Matt> diverse, useful, sensible, and pleasing to the eye (what
    Matt> should I be reading about that, by the way?).

The source, unfortunately.  This is not well-documented to start with,
and the interface is in serious flux (and the new one is really buggy
wrt non-default-character-set fonts---we don't really know in
principle how to handle the issue of WTF "italic" means in Japanese or
Chinese text).  However, in most recent XEmacsen there is either a
Options | Browse Faces or a Options | Edit Faces menu item, which is
pretty self-explanatory.

    Matt> But you got me interested, so I ran a simple (simplistic?) 
    Matt> performance test ... lisp code attached below, of course. I
    Matt> have no idea how to compare my results to your situation,
    Matt> but FWIW, I started up X, started XEmacs (so none of the
    Matt> fonts in question would be loaded yet), and ran the test on
    Matt> a buffer containing 10 copies of the output of 'jchar'.

Hmm.  This takes minutes in my setup; I'll have to see what exactly
the problem is.  Maybe it's VFlib; I may very well have a
VFlib-enabled server, not just a TrueType-enabled server (I'm using
the TL-J stuff, I think, long story as to why I don't know---suffice
it to say, upgrading from TL 2.0-J to TL 3.0-J via TL 3.0-USA beta was 
not a good idea ;-).

BTW, as far as I can tell, all you need is one character to get the
whole font loaded....  The 10 copies of jchar output means that a lot
of the time (about 1.75 seconds, at a guess :-) is spent in the XEmacs 
redisplay code, and not in font-related processing at all.

And the recommended way for doing this kind of timing (assuming you
can spare the cycles :-) is to use the GNU time command with `xemacs
-batch -l file -e test-func'; or maybe use a shell loop to get
averages ;-)

-- 
University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
__________________________________________________________________________
__________________________________________________________________________
What are those two straight lines for?  "Free software rules."
-------------------------------------------------------------------
Next Nomikai: 14 January 1999, 19:30 Tengu TokyoEkiMae 03-3275-3691
*** it will will be Jan 14 (Thu), as Jan 15 (Fri) is a natl holiday
Next Technical Meeting: 13 February, 12:30               Place: TBD
-------------------------------------------------------------------
more info: http://tlug.linux.or.jp                     Sponsor: PHT