[NTG-context] towards some more consistency in regimes & unicode support

Mojca Miklavec mojca.miklavec.lists at gmail.com
Tue Sep 13 17:12:11 CEST 2005


Hello,

Sorry for a slightly longer mail. I wanted to send it to context-dev, 
but probably there's someone else besides Adam out there who could 
contribute (for example to re-chech Greek or Cyrillic section of Unicode 
or even add some missing Hebrew definitions for example). If someone 
thinks that it's more appropriate, please feel free to continue the 
discussion on context-dev.


I. in regi-utf it would be fine to add:

\defineregimesynonym[utf-8][utf]
\defineregimesynonym[utf8][utf]

II. After a long time I finally decided to write my first ruby script. I 
took UnicodeData.txt, adobe glyph list, enco-uc.tex, collected 
averything together, removed characters >FFFF (in case someone needs 
them they can trivially be added again, but I don't think that anyone is 
planning to name them shortly), did some manual corrections ... and here 
are the results:
     http://pub.mojca.org/tex/enco/contextlist/
     http://pub.mojca.org/tex/enco/contextbase/regi-temp.tex

The idea behind is that there is no "definite refence" to the ConTeXt 
glyph names, which means that every new regime that should be supported 
needs a lot of manual work and leads to many inconsistencies.

The file contextnames.txt contains the Unicode hexadecimal number, pdf 
name (from Adobe Glyph List), ConTeXt name and the Unicode name. This 
could then be a source of information when adding new regimes, writing 
unicode vectors (unic-*), mapping to font encodings, 
uppercasing/lowercasing information for font encoding and other files 
can now be derived directly from unicode and this list (unicode already 
contains information about upper/lowercase variants of the letters) ...

There is some more info missing, which should be either packed within 
the same file or in separate files:
- ConTeXt synonyms (like \Dcroat -> \Dstroke, ...)
- pdf synonyms (dbar -> dcroat), to help recognize the glyphs in .enc or 
.afm and automate support for it
- faking the characters (\ccaron -> \buildtextaccent\textcaron{C})
- unaccented version of the characters (\Aacute -> A, ...)
- other characters not present in unicode (Caron, Acute - these are 
accents for uppercase letters, ...)
- (I'm sure that I wanted to add some more points, but I don't remember 
any other right now)

When I wanted to add the names from unic-34.tex, I realized that we 
don't really need to have a command for "every single unicode character" 
(we certainly don't need to map math characters into that region), but 
if someone already has a file with unicode integrals, it costs nothing 
to give him those characters in output.
(Shortly: 0x2211, "N-ARY SUMMATION" should expand into $\sum$, but not 
the other way round)
I have to slightly change the syntax in the context glyph names file to 
note this difference and to be able to define math (and other) signs 
properly.

------------------------------------------------------------------------
III. Now I need some help - someone should help me revise the file 
contextname.txt (I prepared a HTML version of it): correct mistakes (if 
any are spotted), add new definitions, help to prepare a list of 
synonyms, a list of expansions (\buildtextaccent), ...
------------------------------------------------------------------------

Here are some points which I spotted, but can't fix them alone

1. Characters missing (needed by some regimes):

0020-007F section

037A GREEK YPOGEGRAMMENI
0384 GREEK TONOS
0385 GREEK DIALYTIKA TONOS
2015 HORIZONTAL BAR
2017 DOUBLE LOW LINE
20AA NEW SHEQEL SIGN
20AB DONG SIGN
20AF DRACHMA SIGN
2116 NUMERO SIGN
200E LEFT-TO-RIGHT MARK
200F RIGHT-TO-LEFT MARK

1Exx section

2. Greek - there are some name inconsistencies when compared to the 
unic-031 vector, but I don't know anything about old greek. I didn't 
check Cyrillic at all.

3. Punctuation and accents - mostly names for quotes and language 
dependency (lowerleftuppersixquote in comparison to lftdblquote ... or 
whatever they are called) (+ tricks, I already asked about quotes & 
hyphenation approximately a week ago).
I have problems understanding the difference between letter modifiers 
(U+02Cx) and usual accents (U+00Bx), "Combining Diacritical Marks" 
(U+03xx) should be supported somehow as well. I have no idea how to make 
U+0065 U+0301 (e + combining acute accent) into eacute.

4. should hungarumlaut be doubleacute and hungarumlaut only its synonym 
or the other way round?

5. tbar vs. tstroke: compare 0166 and 023E

6. cedilla/commaaccent dilema: there's a huge problem with "t with 
cedilla" (0162): "t with comma below" (021A) sould be used instead (at 
least this is stated in Unicode reference), but most regimes map a 
character to "t with cedilla" (0162), which seems stupid to me. Adobe 
glyph list therefore uses tcommaaccent for "t with cedilla", which looks 
like "t with comma accent", but is on the wrong place. lmr have both 
tcommaaccent and tcedilla. \tcedilla should be "t with cedilla" in my 
opinion and \tcommaaccent "t with comma accent". That currently isn't 
the case in ConTeXt unless something has changed recently.
There are many other letter wrongly named in Unicode ("with cedilla"), 
although they have a comma. I would suggest to name them 
\[gklnr]commaaccent and use \[gklnr]cedilla as a synonym (if needed at 
all for backward compatibility, otherwise it would be better to leave 
them out; there is no such letter with cedilla in unicode, if someone 
needs one, he can construct one trivially with \buildtextaccent)

7. there's "a-kind-of-bug-but-not-really-one" in enco-ans.tex. 
textcedilla maps to 184, which isn't defined in Antykwa for example 
(it's on place 24). It's more a "bug" in texnansi encoding, which has 
cedilla on two places, which is pretty stupid. But anyway:
\definecharacter textcedilla 24
would solve some problems (and hopefully not introduce new ones).

8. most letters are named
"c with cedilla" -> ccedilla
what about the names for "open o", "turned e", "long s", "turned r with 
hook"?
\openo or \oopen? \rturnedhook or \turnedrhook?

9. can latin letters and numbers be accessed somehow by name?

10. Adam prepared some dingbats support I think, this could be added here.

11. There's a showunicode pdf document on pragma-ade.com (at least I saw 
it once), but it's not listed on the overview.htm.

12. I don't know if anyone would ever need to switch from viscii regime 
to some other, but what would happen to the characters under 128 (some 
of them are redefined in viscii)? I'm affraid that there would remain 
Vietnamese leftovers in the lower part of the table.

13. If there are any other comments on the table and/or the script(s), 
please let me know.


IV. With the help of the prepared names list I processed definitions for 
regimes (taken from Unicode webpage) for ISO-8859-* and cp125* (others 
should be trivial). They are only preliminary, some (Hebrew, Thai, 
Arabic) probably don't make any sense yet, but could the rest be added 
to ConTeXt after someone checks if everything is OK? (iso88595, cp1251, 
il1, il2, il9, windows and viscii regimes already exist and should be 
compared for differences)
If possible in such a way that it wouldn't be necessary to include the 
regime definition file manually, but similarly as \usemodule[pre-polish] 
finds and processes the proper file, the \enableregime[xxx] should find 
the proper file and load it.

(And for those who made it till here - sorry again for that gigantic mail.)
	Mojca


More information about the ntg-context mailing list