Could the problem be because math fonts use mode=base instead of
mode=node?  Or is it because in math fonts, the digits and/or full-stop
have different codes?
—MHB

On Mon, Aug 1, 2016 at 9:46 AM, Mohammad Hossein Bateni <bateni@gmail.com>
wrote:

> Dear list,
>
> I want to add a customized font feature to the math font, to automatically
> change the decimal point (unicode 0x2E) to the Arabic/Farsi version
> (unicode 0x66B).  This will go well with the mapping from Latin digits to
> Farsi digits (done in a sample with mathdigits fallback).
>
> For instance, I want to code in $123.45$ and get a result like ۱۲۳٫۴۵.
> However, I do not want to simply map 0x2E to 0x66B, because then "full
> stop" outside a number will also change.  In particular, I'd like to have
> the mapping only when full-stop is preceded and succeeded by a digit.  This
> will, among other things, leave the end-of-equation periods unchanged.
>
> Here is how I tried to achieve it.
>
>
> \startluacode
> digits = { 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39 }
> -- 0x66B instead of 0x2F gives an error!
> replacements = { [0x002E] = { 0x2F } }
> fonts.handlers.otf.addfeature {
>   name = "mpf",
>   type = "chainsubstitution",
>   lookups = { {
>     type = "multiple",
>     data = replacements,
>   } },
>   data = {
>     rules = {
>       {
>         current = { { 0x002E } },
>         after   = { digits },
>         before  = { digits },
>         lookups = { 1 },
>       }
>     }
>   }
> }
> \stopluacode
>
> \definefontfeature[mathperiod][mpf=yes]
>
> \resetfontfallback [mathdigits]
> %\definefontfallback[mathdigits][dejavusansmono][digitsextendedarabicindic][check=yes,force=yes,offset=digitsnormal]
>
> \definefontsynonym
> [MathRoman][name:modern][features={math\mathsizesuffix,mathperiod},goodies=lm-math,fallbacks=mathdigits]
>
> \starttext
> $123.45.$
>
> 123.45.
>
> \addff{mathperiod}
> 123.45.
> \stoptext
>
>
> Though this font feature works outside math, it does not seem to have any
> effect within the math font.  Why is that?  Can we add font feature to the
> math font?  Are the character codes different?
>
> Notice that I could not get it to work, so I commented out the fallback
> for Persian digits, to make things simpler.  To get the proof of concept, I
> am trying to change period to forward slash (see below).
>
> Is the syntax I use for having two features in \defontfontsynonym
> correct?  Actually, I did try with only mathperiod and it does not work,
> either.
>
> When I change 0x2F (slash) to 0x066B in "replacements", I get an error as
> follows.
>
>
> fonts           > otf chain > feature 'mpf', type 'gsub_contextchain',
> chain lookup 'mpf': rule 1 matches at char U+0002E (period) for (1,1,1)
> chars, lookuptype 'gsub_multiple'
>
> error: ...ext/tex/texmf-context/tex/context/base/mkiv/font-ots.lua:1232:
> attempt to index field '?' (a nil value)
>
> tex error       > tex error on line 1 in file
> /Users/bateni/farsitex/context/bostan/examples/mathfa/1.tex: ?
>
> \page ->\par
>              \dosingleempty \page_breaks_process
> \strc_pagenumbers_flush_final_page ...\else \page
>
> [\p_strc_pagenumbers_page ...
> \dostoptext ...\strc_pagenumbers_flush_final_page
>                                                   \page \the
> \everystoptext ...
> l.1 \dostoptext
>
> \clf_stoptext ->\luafunction \16>>clf_stoptext
>
> l.46 \stoptext
>
>
>
> Best,
> MHB
>