Apache OpenOffice (AOO) Bugzilla – Full Text Issue Listing |
Summary: | font rendering problem of stacking diacritics | ||
---|---|---|---|
Product: | gsl | Reporter: | bibiko <bibiko> |
Component: | code | Assignee: | AOO issues mailing list <issues> |
Status: | ACCEPTED --- | QA Contact: | |
Severity: | Trivial | ||
Priority: | P3 | CC: | issues, lohmaier, moyogo, rajeev_jsv |
Version: | OOo 2.0 | Keywords: | oooqa |
Target Milestone: | AOO PleaseHelp | ||
Hardware: | All | ||
OS: | All | ||
Issue Type: | TASK | Latest Confirmation in: | --- |
Developer Difficulty: | --- | ||
Attachments: |
Description
bibiko
2004-11-12 11:35:10 UTC
Created attachment 19207 [details]
font rendering problem of stacking diacritics
cp->hdu: please push on your pile Interesting. -Is there an application on your system that doesn't have the problem? - If yes, do you happen to know which version of the font you have? - Which Version of the usp10.dll (which often can be found in c:\windows\system32\usp10.dll) is installed? Well, almost all applications have the same problem. Textedit on MAC and sometimes Notepad can handle these problems better. On my PC the usp10.dll is the version 1.0409.2600.1106 (xpsp1.020828-1920). I tried almost all fonts (Mac/PC) and each of them have this problem. I know that this is a BIG problem and there are some projects dealing with this issue by Microsoft, Mac and Graphite from SIL, OpenType etc. If I using MSWORD I'm able to stack some diacrits with the help of the character map and the direct input of the unicode hex code but I realized that sometimes it works sometimes not?! I tried out the latest OO and there some advancements. The question is which part of the puzzle is responsible for the correct rendering, the font, the application, something between? But if I see that Textedit or Notepad or sometimes WORD stack the diacritics correctly, then my question is why does not stack it all time? I'm very interested to solve this problem because I have to write a lot of these stacks (I'm a computer linguist). On the other hand there are some other problems of stacking. E.g. Tibetan characters in Arial Unicode. Some characters should be stacked but nothing happens. Although it is correct saved as a combining character within the font and the unicode.org says the same?! Maybe it is worth to have a look at the page of graphite project http://scripts.sil.org/cms/scripts/page.php?site_id=nrsi&cat_id=RenderingGraphite. Cheers Hans I found an other version of the usp10.dll in my Office enviroment: 1.0405.2416.1 (lab06_N.010104-1344) Hans SIL's Graphite has a lot of advantages compared to OpenType but most font
vendors still seem to prefer OpentType...
> The question is which part of the puzzle is responsible for the correct
> rendering, the font, the application, something between?
In the OpenType world the font's GSUB table is responsible for deciding how to
stack diacritics and the library responsible for Opentype processing is
responsible for the actual stacking. Of course the application has to request
the OpenType service (for Latin scripts this isn't done yet by OOo). So the
summary is that the font, the OT processor and the application have to work
together to make a script look good...
- Version 0.84 of Arial Unicode doesn't have GPOS information for Latin script
- USP from XP SP1 doesn't support OpenType processing of Latin script
- OOo doesn't yet request OpenType services for Latin script
I'm just writing or better modifying a MS WORD macro to input diacritcs and IPA symbols via AutoText and toolbars. Maybe it would be very very useful to have such an input methode as a built in function for OO as well. You can find it under http://www.kubrussel.ac.be/onderwijs/letterenwijs/tlk/vakken/1etb_usefulinfo.htm go to last item and click at 'DOWNLOAD'. On the other hand, I don't know whether this is the right place to announce, there is a JAVA based input methode editor available under http://sourceforge.net/projects/jgim/ It is based on Yudit coming from UNIX and have a lot of advantages. Unfortunately I can't use this tool with OO!! all the best hans The problem that is tracked in this issue is that for latin text complex text layout is not yet performed. HDU->BIBIKO: We prefer to keep issues isolated. Reading your last comment I suggest to create an enhancement request issue to include something like UNIQODER and another issue for the input method with jgim. I think the UNIQODER request should go to the Writer component, the jgim request to the gsl component. As I have the same issue on Linux platform, I have created a TTF which can be downloaded from ftp://debian.linux.org.tw/pub/3Anoppix/people/arne/uming2.ttf (14MB). This font is originally a CJK Unicode font, but I also use it to test other issues. As I need the combining character U+0358 combining dot right above (will be in Unicode 4.1.0) and no application currently recognizes this codepoint as combining character, I though, start with oo.o. :) Actually I noticed, that no combining sequence works correctly in oo.o. I tried with the m85 build. Now comes my font in place: 1. I added anchor points to the lowercase a-z and also to the combining marks. In fontforge, which I use to create the font, the combining seqence is displayed correctly, as you can see in the attached screenshot. It seems that the freetype library supports this feature. 2. For the squence 'o U+0358' I created a kerning pair. This also looks quite correct in fontforge. However, the anchor solution is preferred, because I can have a finer control where exactly the combining mark should attach. I hope this helps. (BTW. The M$ Times New Roman font does not have any GPOS information on my system) Cheers Arne Created attachment 23763 [details]
Screenshot of fontforge displaying combining marks correctly with anchors
I've left a comment on Issue #16032 that should probaly come here, so I'll copy it here. There is an OpenType TTF font that has OpenType features for Latin script. If you want to do tests with it. The GPLed font is here http://home.sus.mcgill.ca/~moyogo/lingala/fonts/Junicode-Regular.ttf The output with a patched Pango is here http://bugzilla.gnome.org/attachment.cgi?id=53890&action=view With this file http://home.sus.mcgill.ca/~moyogo/lingala/fonts/text.utf8 or that one http://bugzilla.gnome.org/attachment.cgi?id=32984&action=view . I wasn't able to reproduce this with the OOo2 I have by default on Ubuntu. The OpenType features kern, mkmk, mark for GPOS and ccmp, liga and clig are required for IPA but more importantly for languages that must use precomposed characters with diacritics. FYI: SIL's Graphite is available for OOo 2.0 (they have both a precompiled version (linux only) and the sources on their site. As mentioned before, the drawback is that you need graphite-enabled fonts. But the results are great. Created attachment 32587 [details]
IPA diacritics with Doulos SIL and graphite-enabled OOo
Reset assigne to the default "issues@openoffice.apache.org". |