On 11-Feb-2015 22:44, chandrakant dhutadmal wrote:

Hi All. We have submitted requests for bringing out localized Inkscape for Indian languages through various bugs filed.  List of all such bugs filed can be looked at https://bugs.launchpad.net/~cpdhutadmal. We have also submitted translations for those languages. Would like to know from the community if we have considered making builds for Indic languages. If so, when are we going to do it. If not, can we have a discussions surrounding the topic ?

I have spent some time working on related issues, for instance:

https://bugs.launchpad.net/inkscape/+bug/1282968

There are two parts to your request:

1. Translating the user interface into those languages. This should not be technically difficult but will require a lot of work by the translators.

2. Having the program properly handle those languages. The problem here is that Inkscape only partially uses Pango to go from unicode strings to the final glyph positions you see on screen. The code above Pango was written with European languages in mind and does not always do the right thing when it is used for Indic languages, where characters can do some (from our point of view) peculiar things, such as combining characters to change the look and positions of glyphs - all within logical blocks which do not have a real equivalent in European languages. (The closest thing we have is syllables.) Assuming the right fonts are installed, and after all Indic text has been entered from the keyboard or read out of an SVG, it will probably look about right on screen. However, editing that text will almost certainly not behave in the manner you would expect.

To resolve the second problem somebody would need to complete the "Pangofication" of Inkscape, which would be about on par with a kidney transplant, in terms of how major a rework it would be. I'm not sure that we have a developer with the necessary language expertise. It isn't me. I have spent some time in that code, and with a bit of work cleaned it up so that it now works for one R->L language (Hebrew, which doesn't have the run on characters one sees in Arabic) and combinations of R->L and L->R (Hebrew again) in the same string. But I am not a linguist, nor do I have time to do the amount of work which would be required. The ideal person to do this sort of work would be one of the Pango developers. Another option would be to convert to Harfbuzz, which does similar things, and would be at least as much work. There may be some open source program around with the right Pango bits already implemented which we could "borrow", but again, pasting that in would be a big job, and getting any of it wrong is going to severely and adversely affect the behavior of Inkscape's text handling.

On the plus side, if we did finally shift the entire burden onto Pango as a bonus we should automatically get correct handling of all Pango supported languages. Which would be good, since the Pango project has people who know what pretty much all of the languages are supposed to do when entered or edited, and we, for the most part, do not.

As an aside, my libTERE, which reassembles formatted text from little chunks back into Inkscape formatted and editable strings, makes all sorts of assumptions which are only valid for simple L->R and R->L languages. libTERE is the code that lets an Inkscape formatted string be exported to EMF, and read back in again, and end up as exactly the same sort of editable object. (Without it the text would consist of a series of independent text blocks, one for each type of format.) However, when it fails, it is supposed to just leave the original fragments, so trying and failing should not make things worse than they already are.

Regards,

David Mathog mathog@...1176... Manager, Sequence Analysis Facility, Biology Division, Caltech