Japanese TeX: if you're really macho...

[turnbull@example.com (Stephen J. Turnbull)]

Yow!

OK, for all you folks who wish you had a Japanese TeX you could
integrate with Ghostscript, here's the short synopsis.  I dunno when
I'll get around to writing up the whole thing in accurate detail with
config files.  Sorry, but I left the apartment today at 8:00am and
tomorrow will start just as early and go just as late....

(1) Get a recent Ghostscript, you want the Aladdin 4.03 version.  Stay 
    away from Aladdin 4.0[12]?, they have bugs in garbage collection
    that make kanji (and anything big, for that matter) painful.

	     file://ftp.cs.wisc.edu/ghost/aladdin/*.4.03*

or

   file://turnbull.sk.tsukuba.ac.jp/pub/linux/packages/Ghostscript/

This includes fonts and other stuff you don't need if you've got
Ghostscript already; there may be closer mirrors, too.

(2) Get the Wadalab fonts.

		 file://ftp.ipl.t.u-tokyo.ac.jp/Fonts/*
	     file://ftp.ipl.t.u-tokyo.ac.jp/Fonts/tools/*

or

   file://turnbull.sk.tsukuba.ac.jp/pub/linux/packages/Ghostscript/*

    These are *big*, like multiple megabytes per gzipped font.  You
    *need* the programs in the tools directory, but you can pick and
    choose the fonts.  You must get the symbol font, though; it's used
    in combination with all the others.

(3) Build Ghostscript and the Wadalab tools, especially 'wftodm'.
    Install the fonts in a reasonable place, $KANJI in the sequel.
    Don't forget the symbol fonts, wadalab-symbol-6.tar.gz.

(4) Get and build NTT-JTeX.

      file://turnbull.sk.tsukuba.ac.jp/pub/linux/packages/Text/

    - web* (there are five of these archives, you need them all)
    - lib* (there are two of these archives, you need them both)
    - jl*  (there are five of these archives, you need them all)
    - base.tar.gz
    - dvipsk-5.58f.tar.gz
    - xdvik-18f.tar.gz

    You don't really need either of the last two, but either one has
    the latest version of Karl Berry's kpathsea library that I've been
    able to find.  It makes life easier if all your utilities use the
    same library and config files, I've found; may as well be up to
    date.  I don't know to what extent xdvik can do Japanese; the 'k'
    is for "kpathsea," not "kanji."

    Find a nice place to be the root of your source tree.

    First untar the non-'j' versions of web* in the root.  This
    produces a subdirectory web2c-6.1, containing both the web sources 
    for TeX and friends, and the web2c converter.  Obviously the
    "*-add*" versions come later, as far as I can tell it doesn't
    matter whether you do web or web2c first.

    Next do the 'j' versions.  These produce a subdirectory
    web2c-6.1j, and it must be a sibling of web2c-6.1 because it's
    full of symlinks of the form "../web2c-6.1."

    If you are going to build xdvik and dvipsk, or just want the
    kpathsea library, put them here too.  They untar into nice
    versionified subdirectories.  All of these subdirectories have
    kpathsea subdirectories.  I believe it is safe to have only one
    real copy of the subdirectory which is shared among xdvik, dvipsk, 
    web2c, and web2c-j.  It is probably safest to have a real
    subdirectory for each of the programs, and then do

     cd web2c-6.1/kpathsea; ln -s ../../dvipsk-5.58f/kpathsea/* .

    in each one.  It is possible that the various programs autoconfig
    scripts will step on each other if you symlink the subdirs instead
    of their leaf files.  On the other hand, you may do fine with only
    one copy of kpathsea.a.  Haven't tried it yet....

    Before you build, you need to install the TeX library files.
    Chose a root (from here on, $ROOT) for the libraries, cd there,
    and untar the lib and jlib archives.  Then you need to cd down to
    $ROOT/lib/texmf/tex/latex2e, and untar base.tar.gz there.  (I
    think that's how it worked; you should do a tar tzf to make sure
    first, and maybe the "base" files needed to be in the source
    directory, for some reason.)

    Do ./configure --prefix $ROOT; make; make install in each of the
    source directories.

(5) TeXify the kanji fonts (the example here does mincho, "dm"):

    cd $KANJI; wftodm -FontBase dm min*.ps jis*.ps
    for $i in *.afm; do afm2tfm ${i%.afm} -v ${i%.afm} r${i%.afm}; done
    for $i in *.vpl; do vptovf $i ${i%vpl}vf ${i%vpl}tfm; done

    Move the tfms and vfs to an appropriate place (see the docs for
    dvipsk for this, but in general just about anywhere below
    $ROOT/lib/texmf/fonts/tfm and $ROOT/lib/texmf/fonts/vf
    respectively should do).

    The README.english for the Wadalab fonts says that if you do
    wftodm and then pstopk, the baselines of the kanji are wrong; they
    are shifted up a bit.  This does not appear to be true for the pfa
    fonts themselves, fortunately; the bottom of the hiragana are
    aligned with the baseline of the Helvetica and the kanji descend
    slightly more.

(6) Teach Ghostscript about the fonts.  It may be possible to do this
    using Fontpaths, but I don't understand them well.  The tedious
    way to do it is to edit the Fontmap file.  Basically, for evey
    single pfa file you need a line in the fontmap.  For the hiragana
    subfont, you find

    /dmjhira		($KANJI/dmjhira.pfa)	;

    All together you need about 30 of these per font.

(7) Inhibit dvipsk from trying to make the fonts.  For each of these
    fonts, you need to add a line to $ROOT/lib/dvips/psfonts.map:

    dmjhira dmjhira

(8) I have not yet messed around with jlatex2e styles to see if these
    automagically get included.  In jtex, the following code produces
    a few lines of output:

    \font\myroman = phvr8r at 28pt
    \font\mysmallroman = phvr8r at 14pt
    \makejfontdata{dm}{dm}{}{1}
    \jfont\mymincho = dm at 28pt

    \myroman This is Helvetica.\vskip 20pt

    \mysmallroman This is more Helvetica.\vskip 20pt

    \mymincho これはみんちょう瘢雹です\vskip 20pt
    \bye

    Note that jtex wants New-JIS input, although the DVI output is a
    sort of munged kuten I believe.

And there you have it....

If you can't afford/don't want to build all those binaries, check back 
with me in a bit.  It may be possible to use ASCII-JTeX this way; I
didn't understand dvipsk well enough when I was trying that
configuration.  The NTT-JTeX has the advantage, however, that only the 
subfonts you actually use get loaded; this can be done with
wftopfa-style fonts as well, but I don't trust the code that the
Wadalab project produced.  Older Ghostscripts (3.33 or later is
necessary I'm pretty sure) may also work.

And I still have the ambition of putting together a good package of
TeX with Ghostscript for Japanese processing....  Probably not an RPM, 
though, it would only be about 50MB in gzip form....  ;-)

Ciao.

Steve
-----------------------------------------------------------------
a word from the sponsor will appear below
-----------------------------------------------------------------
The TLUG mailing list is proudly sponsored by TWICS - Japan's First
Public-Access Internet System.  Now offering 20,000 yen/year flat
rate Internet access with no time charges.  Full line of corporate
Internet and intranet products are available.   info@example.com
Tel: 03-3351-5977   Fax: 03-3353-6096