Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!zaphod.mps.ohio-state.edu!wuarchive!mit-eddie!uw-beaver!ubc-cs!jalbert
From: jalbert@cs.ubc.ca (Francois Jalbert)
Newsgroups: comp.text.tex
Subject: Japanese TeX
Summary: Advice sought
Message-ID: <10036@ubc-cs.UUCP>
Date: 12 Oct 90 06:12:45 GMT
Sender: news@cs.ubc.ca
Organization: University of British Columbia, Vancouver, B.C., Canada
Lines: 57

Hallo TeXperts. I have been working for a while on my own Japanese TeX
system, but before I invest more time, I thought I would mention what I am
doing and how I am doing it to all. Perhaps I am repeating another person's
work. Perhaps some of you have some advice to give me.

The biggest problem seems to be the large number of symbols. I first decided
I would limit myself to a few thousands, but which ones? The answer came
with the simple Japanese vi editor for MS-DOS machines called MOKE. There
is in there a file called JIS24 which contains about 7802 24 by 24 pixel
resolution japanese symbols. JIS stands for Japanese Industrial Standard and
could act for me as some sort of extended ASCII table. I decided to limit
myself to these symbols. I don't know where that file JIS24 comes from. 
The documentation seems to imply it was derived from some X-Window file. 
Any info regarding that and possible copyright violation is welcome.

I quickly wrote a few utilities with my Turbo-Pascal 5.0 which allowed me
to browse through JIS24, dump it all on my printer (+/- 50 pages), and also
manipulate the information for each individual symbol.

I decided to try the following approach. Write a small picture environment
for each Japanese symbol. The picture will be a simple 24 by 24 matrix with
circle*{1} put at the right places. If one assumes 10 such Japanese characters
per inch, that gives us a density of 240DPI. Of course, that is not "true"
240DPI since the characters don't have a continuous boundary, but it might
be enough for my simple needs.

So I wrote a utility which translated automatically JIS24 into a large number
of small .tex files, each containing the right pattern of circle*{1}. I could
then use LaTeX as usual, and here and there use commands like \jap{3056} to
get Japanese character #3056 to appear. That works fine. It looks great with
my screen previewer, and so so with my lousy dot matrix printer. My current
complaints are the size of the .dvi generated (typically 20 times the size of
the main .tex document), the amount of memory used (big emTeX blows up with
half a page of Japanese text), and the large CPU time required.

I thought it ought to be possible to use METAFONT to generate fonts of small
matrices of dots. After all, METAFONT must have primitive operators to draw
outlines of symbols. There is probably some sort of circle*{1} operator in
there. I could automatize the creation of these .mf files in the same way I did
it for my .tex files. That's no problem. Does anybody have examples of such
fonts? I could just change the size to 24 by 24 and the dot patterns.

The problem now is the number of different fonts needed. At 128 Japanese
symbols per font, I need around 40--50 fonts which might be potentially all
needed in a given document. Is that too much for LaTeX? Is it possible to
load a font, grab a character, and then discard the font. That would slow
things down, but would allow me to at least process the document. Each font
could be numbered like JAP23.TFM, and it would be "easy" from something like
\jap{3056} to deduce the font number and the symbol offset.  

Anyway, I sure would appreciate any advice or information anyone could have
for me. I want to avoid \specials since postscript dependant. I also know
quality won't be great, please no flames regarding the spirit of TeX being
violated. If this works fine, I may look at generating better fonts. But right
now, I just want a bare bone system running.

A million thanks in advance. Franky, hacker at large.