As we can see in this question, Bengali (Bangla) support in polyglossia is incomplete.
The major weak areas are,
- Support for numerals.
- Support for other language related areas like dates.
For the first one, the current solutions are,
- As found here, defining own interfaces. The issues to be solved keep growing and one wants to give up at some point.
- As suggested here (near the end), redefine the meaning of
\@arabic(and may be\@alph). As rightly indicated there, this is a risky approach.
My preferred approach has been the second one. But I am already paying the price of using this approach. I see that I can not even use those label or ref change options in enumitem.
If I decide to contribute by adding Bengali (Bangla) support to polyglossia, where should I start?
I understand that this must be some kind of table. If I get pointed to the right ones, this task will be easier.
gloss-bengali.ldf? That's usually the place to start. Also, contact the author/maintainer. Welsh is much improved as the result of my discussions with him ;). It no longer contains characters which don't exist in the language, for example. (I don't use Polyglossia, but this has also improved Babel's file at least marginally.)datetimefile to override the standard one (which is wrong and includes characters which don't exist in the language - this problem pervades all TeX support for Welsh - I guess people all copied Babel's mess). But that's mainly because I use pdfTeX and so can't enjoy the new Polyglossia file. For Polyglossia, I just had to explain what was needed for it to be correct, basically..ldffile is the place to start. I find Bengali digits for section, page etc are defined now (tex.stackexchange.com/questions/246881/…) plus also the translated names of the document parts.