To glue the horizontal lines of composite Bengali characters under special formatting

Question

NB. I guess the question has a little dependency on Language (Bengali). I would try to explain every font in transliterated format.

I like to use a character a below/above another character b. In my use-case a is either a - (hyphen) or । (stop character in Bengali). So for an example word "কুহুকেকা" (transliteration: "KuHuKeKa")

(please note the horizontal line on the top is continuous through each of the four (composite) characters); if I want to a - below the third character কে (Ke) Or add a । above the second character হু ('Hu') of the word,

the horizontal line is not remaining continuous at all - ideally it should look just as the first image but with the later modification.

First I implemented in classy way with amsmath package

\newcommand{\abelowb}[1]{$\underset{\hbox{\text{-}}}{\text{#1}}$}
\newcommand{\aoverb}[1]{$\overset{\raise0em\hbox{\text{#1}}}{\text{।}}$}

After a close observation I realised this problem arises only when ে ৈ ো and ৌ vowel modifiers (having ে common in appearance), are adjacent or in the b eg. মাতৃদেবো (MaTriDebo) .

Later I suspected the switch in mathmode and textmode may creating this disturbance. So I tried to handcraft the commands like the following:

\newcommand{\myabelowb}[1]{\sbox1{-}\sbox0{#1}#1\raisebox{\dimexpr-
0.5\baselineskip\relax}{\kern\dimexpr-0.5\wd0-0.5\wd1\relax-}\kern\dimexpr0.5\wd0-
0.5\wd1\relax}
\newcommand{\myaoverb}[1]{\sbox1{।}\sbox0{#1}#1\raisebox{\dimexpr0.6\baselineskip\relax}
{\kern\dimexpr-0.5\wd0-0.5\wd1\relax।}\kern\dimexpr0.5\wd0-0.5\wd1\relax}

With these I could fix some issue partially:

1. \myaoverb is giving the same problem using command কু\myaoverb{।}{হু}কেকা [Case 1]
2. \myabelowb is working for this word কুহু\myabelowb{কে}কা [Case 2a], but not here with command কু\myabelowb{হু}কেকা [Case 2b]
3. amsmath is maintaining a distance from b to a by some proportional way. My handcrafted commands cannot do that, eg.

---

What I Observed:

• If there were a concept of XeLaTeX cursor; once the cursor finds ে ৈ ো or ৌ -- under normal cases, it adds extra joining horizontal line to the left of this so that continuity preserves. Otherwise, once some a comes around ে ৈ ো or ৌ XeLaTeX forgets to go back to the last observed valid character - Case 1 or Case 2b - so that it can glue the horizontal line.

What I want: I want to have some commands that can alley by the problem of discontinuous horizontal line and also can maintain proportional distance from b.

---

An MWE:

\documentclass{article}
\usepackage{fontspec,amsmath}
\usepackage{polyglossia}
\setdefaultlanguage[numerals=Bengali,changecounternumbering=true]{bengali}
\setmainfont[Script=Bengali]{Noto Serif Bengali}
\newfontfamily\latinfont[Script=Latin]{Noto Serif}
\setotherlanguages{latin}
\newcommand{\abelowb}[1]{$\underset{\hbox{\text{-}}}{\text{#1}}$}
\newcommand{\aoverb}[1]{$\overset{\raise0em\hbox{\text{#1}}}{\text{।}}$}
\newcommand{\myabelowb}[1]{\sbox1{-}\sbox0{#1}#1\raisebox{\dimexpr-0.5\baselineskip\relax}{\kern\dimexpr-0.5\wd0-0.5\wd1\relax-}\kern\dimexpr0.5\wd0-0.5\wd1\relax}
\newcommand{\myaoverb}[1]{\sbox1{।}\sbox0{#1}#1\raisebox{\dimexpr0.6\baselineskip\relax}{\kern\dimexpr-0.5\wd0-0.5\wd1\relax।}\kern\dimexpr0.5\wd0-0.5\wd1\relax}
\begin{document}
    কুহু\abelowb{কে}কা\
    কু\aoverb{।}{হু}কেকা\
    কুহু\myabelowb{কে}কা\
    কু\myaoverb{।}{হু}কেকা\
    কু\myabelowb{হু}কেকা
\noindent
\abelowb{কু} \begin{latin} $\longleftarrow$ using \verb|\abelowb{}| \end{latin}

\myabelowb{কু} \begin{latin} $\longleftarrow$ using \verb|\myabelowb{}| \end{latin}

\end{document}

Answer 1

Use a Bengali font with Unicode support for the Vedic tone accents, such as recent versions of Noto Serif Bengali (I tested with version 2.003) and Noto Sans Bengali.

\documentclass{article}
\tracinglostchars=3 % Make it an error when the font does not have a character!
\usepackage{fontspec,amsmath}
\usepackage{polyglossia}
\usepackage{setspace} % for \doublespacing
\setdefaultlanguage[numerals=Bengali,changecounternumbering=true]{bengali}
\setmainfont[Script=Bengali]{NotoSerifBengali}
\newfontfamily\latinfont[Script=Latin]{Noto Serif}
\setotherlanguages{latin}
\newcommand\udatta[1]{#1^^^^0951}
\newcommand\anudatta[1]{#1^^^^0952}
\doublespacing
\begin{document}
    \noindent কুহু\anudatta{কে}কা\
    কু\udatta{হু}কেকা\
    কুহু\anudatta{কে}কা\
    কু\udatta{হু}কেকা\
    কু\anudatta{হু}কেকা
\noindent
\anudatta{কু} \begin{latin} $\longleftarrow$ using \verb|\anudatta{}| \end{latin}

\end{document}

You might want to increase the line spacing in order to leave room for the udattas. You definitely want to add a \tracinglostchars= command, since without one, TeX will silently ignore if the current font does not have a character it should display (and write a warning in the middle of the .log file). Set this to 3 to make this an error, or 2 to at least print a warning message to the console.

Answer 2

If I understand correctly (which I may not, as I can not read the script) what you really want is a Bengali font with a combining _ under accent and a combining over |

I could not find these but perhaps just because I was searching for the wrong terms, but to show how it might work I can show an over accent U+0981 BENGALI SIGN CANDRABINDU and an under accent U+09CD BENGALI SIGN VIRAMA which can be added without disturbing the spacing of the surrounding characters

\documentclass{article}
\usepackage{fontspec,amsmath}
\usepackage{polyglossia}
\setdefaultlanguage[numerals=Bengali,changecounternumbering=true]{bengali}
\setmainfont[Script=Bengali]{Noto Serif Bengali}
\newfontfamily\latinfont[Script=Latin]{Noto Serif}
\setotherlanguages{latin}
\begin{document}
\def\xxx{^^^^09cd}
\def\yyy{^^^^0981}
  \noindent
    কুহু\yyy কেকা\
    কুহ\xxx কেকা
\end{document}

sorry this is not so useful, but perhaps other fonts have more combining characters that could be used