Arabic font rendering with ICU is clearly supported, as I can see the routines here:
and the news here:
Indic font rendering apparently needs its own special routines which are discussed in the context of OpenType fonts here:
Context: when gargi.ttf is configured based on this readme:
the result is the same as the graphics attached to the mapserver ticket.
Thus, enhanced support for OpenType Indic fonts is highly desirable within Mapnik.
[springmeyer] See Ticket #19 for context on why ICU can solve this problem.
[springmeyer] I tried applying the patch from http://lists.openstreetmap.org/pipermail/dev/2008-October/012005.html
and I've attached a graphic comparing the raw text in the postgres insert of sample indic data (from the linked mapserver ticket) to the rendered text. I'm not sure which is right and which is wrong, so feedback is welcome.
[springmeyer] data and test case script's to produce the previously attached image are here: http://mapnik-utils.googlecode.com/svn/trunk/tutorials/fonts/
[springmeyer] The link has now moved for the testcase to:
Town/city names in India can't be localised because of this bug . This bug affects almost every town/city in northern india where hindi language is predominantly used .
I have attached an example showing proper rendering of telugu . We used to have these issues with pango long back . They were solved by installing ttf-indic-fonts . Pango rendering issues are long gone and almost all major rendering engines render indic fonts .
[pavithran] ICU was rendering telugu fonts ( atleast to some extent) as per http://bugs.icu-project.org/trac/attachment/ticket/6113/gautami.pdf for the version 3.6 though there is a slight positioning bug http://bugs.icu-project.org/trac/ticket/6113
[jburgess777] I'm adding a comment at the request of pvai on IRC (pavithran), I think this information is known already but is not mentioned in this bug. I believe the underlying issue here is that Mapnik is rendering the string one character at a time in order to make the characters follow along the path. These Indic languages use ligatures where the rendering depends on multi-character sequences and can not be correctly rendered one character at a time, http://en.wikipedia.org/wiki/Typographic_ligature
For reference, the location which pavithran supplied can be seen at: http://www.openstreetmap.org/?mlat=17.25&mlon=80.15&zoom=12
[springmeyer] fyi, new shaping library: http://cgit.freedesktop.org/harfbuzz/
[springmeyer] fyi, make sure to see Tom's comment on ICU and Shaping: https://lists.berlios.de/pipermail/mapnik-devel/2010-September/001245.html
This issue came up in the Cartonama workshop a couple weeks ago. Added Kannada names to the map, and they didn't render properly. Working properly, this is a huge bonus for the open toolset in India.
I don't have the expertise to get involved, but could ask some folks in India if they might be able to contribute. Is the problem defined enough for that kind of help?
This bug should be fixed when we switch to HarfBuzz as our shaping engine. It shouldn't need any special handling.
good to here ... when will this take place approximately?
Except first results in a few weeks and final code in about 2 months (in git). I'm not sure when a version containing the new code will be released.
work with CJK support ?
Closing this ticket as the issue is solved in the harfbuzz branch. Refs #1428.