akshar.dtx

% \iffalse meta-comment
%
% File: akshar.dtx
% ---------------------------------------------------------------------------
% Package:      akshar
% Author:       Vu Van Dung
% Description:  Support for syllables in the Devanagari script
% Repository:   https://github.com/joulev/akshar
% Bug tracker:  https://github.com/joulev/akshar/issues
% License:      The LaTeX Project Public License v1.3c or later
% ---------------------------------------------------------------------------
% This work may be distributed and/or modified under the conditions of the 
% LaTeX Project Public License, either version 1.3c of this license or (at 
% your option) any later version.
% 
% The latest version of this license is in
% 
%     http://www.latex-project.org/lppl.txt.
% 
% This work has the LPPL maintenance status `maintained'.
% 
% The Current Maintainer of this work is Vu Van Dung.
% 
% This work consists of the files akshar.dtx
%                                 akshar.ins
% and the derived file            akshar.sty.
% ---------------------------------------------------------------------------
% The manual uses the following fonts:
% * Siddhanta for serif font.
%   Source: https://sites.google.com/site/bayaryn
% * Biolinum for sans font.
%   Source: Package libertine (https://ctan.org/pkg/libertine)
% * Hack Nerd Font for monospace font.
%   Source: https://www.nerdfonts.com/font-downloads
% ---------------------------------------------------------------------------
% \fi
% \iffalse
%<*internal>
\iffalse
%</internal>
%<*readme>
Package:      akshar
Author:       Vu Van Dung
Version:      0.2  (06 Sep 2020)
Description:  Support for syllables in the Devanagari script
Repository:   https://github.com/joulev/akshar
Bug tracker:  https://github.com/joulev/akshar/issues
License:      The LaTeX Project Public License v1.3c or later. 
              See the source file akshar.dtx for more information.
%</readme>
%<*internal>
\fi
%</internal>
%<*driver|package>
\def\aksharPackageName{akshar}
\def\aksharPackageVersion{0.2}
\def\aksharPackageDate{2020/09/06}
\def\aksharPackageDescription{Support for syllables in the Devanagari script (JV)}
%</driver|package>
%<*driver>
\documentclass{l3doc}
\renewcommand\MacroLongFont{\ttfamily\small}
\usepackage[margin=1cm,left=7cm,bottom=1.5cm]{geometry}
\usepackage{fontspec}
\usepackage{biolinum}
\setmainfont{siddhanta}[Scale=0.9,
                        Script=Devanagari,
                        Extension=.ttf]
\setmonofont{HackNF}[Scale=0.85,
                     Extension=.ttf,
                     UprightFont=*_R,
                     BoldFont=*_B,
                     ItalicFont=*_I,
                     BoldItalicFont=*_BI]
\newfontfamily{\aksharDocFontFallback}{siddhanta}
  [Scale=0.9,Script=Devanagari,Extension=.ttf]
\usepackage[libertine]{newtxmath}
\renewcommand\meta[1]{$\langle${\itshape\rmfamily#1}$\rangle$}
\usepackage{newunicodechar}
\ExplSyntaxOn
\tl_map_inline:nn { ािीुूेैोौंःॢृॅँ़ॆॊॉॎॏ॓ॕॖॗॄऺऻ꣠꣡꣢꣣꣤꣥꣦꣧꣨꣩꣪꣫꣬꣭꣮꣯꣰꣱ꣿ्कनमस्कारमंळीममडळीममडमंमममडडमंळीममडममंममडडमंळीममड ॑ ॒ ॓ ॔}
  { \newunicodechar{#1}{{\iffontchar\font`#1\else\aksharDocFontFallback\fi#1}} }
\ExplSyntaxOff
\usepackage{akshar}
\usepackage[listings]{tcolorbox}
\definecolor{myred}{HTML}{C0392B}
\definecolor{myblue}{HTML}{2980B9}
\definecolor{mygreen}{HTML}{27AE60}
\newtcblisting{codeexample}[1][]{size=minimal,
  sidebyside gap=1cm,lefthand width=6.5cm,lower separated=false,
  fontupper=\raggedleft,colback=white,colframe=white,
  verbatim ignore percent,grow to left by=7cm,text side listing,
  listing options={basicstyle=\ttfamily\small,keywordstyle=\ttfamily\small,columns=fullflexible,
  numbers=left,numberstyle=\sffamily\tiny\color{gray},stepnumber=1,numbersep=5pt},#1}
\pgfkeys{/tcb/.cd,code only/.code={\pgfkeysalso{listing only,grow to left by=-5mm}}}
\hypersetup{linkcolor=mygreen,urlcolor=myblue,linktoc=page}
^^A Hacking the index a bit.
\IndexPrologue
  {
    \section*{Index}
    \markboth{Index}{Index}
    \addcontentsline{toc}{section}{Index}
    Underlined page numbers point to the definition, 
    all others indicate the places where it is used or described.
  }
\usepackage{sectsty}
\sectionfont{\LARGE\color{myred}}
\subsectionfont{\large\color{myblue}}
\usepackage{fontawesome5}
\let\bfseries\mdseries^^A For consistency
\parindent=0pt\relax
\IndexMin=100pt\relax
\EnableCrossrefs
\CodelineIndex
\RecordChanges
\begin{document}
\DocInput{\aksharPackageName.dtx}
\end{document}
%</driver>
% \fi
%
%
%
% \title{The \pkg{\aksharPackageName} package}
% \author{Vu Van Dung}
% \date^^A
%   {^^A
%     Version \aksharPackageVersion\ --- \aksharPackageDate\\[1ex]^^A
%     {\small\faIcon{link}\quad\url{https://ctan.org/pkg/akshar}}\\^^A
%     {\small\faIcon{github}\quad\url{https://github.com/joulev/akshar}}^^A
%   }
%
% \maketitle
%
% \begin{abstract}
%   This package provides tools to deal with special characters in a Devanagari 
%   string.
% \end{abstract}
%
% \tableofcontents
% \parskip=1ex
%
% \begin{documentation}
% \section{Introduction}
% When dealing with processing strings in the Devanagari script, normal \LaTeX\ 
% commands usually find some difficulties in distinguishing ``normal'' 
% characters, like क, and ``special'' characters, for example  ् or ी. Let's 
% consider this example code:
% \begin{codeexample}[]
%\ExplSyntaxOn
%\tl_set:Nn \l_tmpa_tl { की }
%\tl_count:N \l_tmpa_tl \c_space_token tokens.
%\ExplSyntaxOff
% \end{codeexample}
% The output is 2, but the number of characters in it is only one! The reason 
% is quite simple: the compiler treats ी as a normal character, and it shouldn't 
% do so.
%
% To tackle that, this package provides \pkg{expl3} functions to ``convert'' a 
% given string, written in the Devanagari script, to a sequence of token lists. 
% each of these token lists is a ``true'' Devanagari character. You can now do 
% anything you want with this sequence; and this package does provide some 
% front-end macros for some simple actions on the input string.
%
% \section{User manual}
%
% Due to the current implementation, \emph{all} of these macros and functions
% are \emph{not} expandable.
%
% \subsection{\LaTeXe\ macros}\label{latex2e-macros}
% \begin{function}{\aksharStrLen}
%   \begin{syntax}
%     \cs{aksharStrLen} \marg{token list}
%   \end{syntax}
% Return the number of Devanagari characters in the \meta{token list}.
% \end{function}
% \begin{codeexample}[]
%There are \aksharStrLen{ नमस्कार } characters in नमस्कार.\par
%\ExplSyntaxOn
%\pkg{expl3}~returns~\tl_count:n { नमस्कार },~which~is~wrong.
%\ExplSyntaxOff
% \end{codeexample}
% \begin{function}{\aksharStrHead}
%   \begin{syntax}
%     \cs{aksharStrHead} \marg{token list} \marg{n}
%   \end{syntax}
% Return the first character of the token list.
% \end{function}
% \begin{codeexample}[]
%\aksharStrHead { मंळीममड }
% \end{codeexample}
% \begin{function}{\aksharStrTail}
%   \begin{syntax}
%     \cs{aksharStrTail} \marg{token list} \marg{n}
%   \end{syntax}
% Return the last character of the token list.
% \end{function}
% \begin{codeexample}[]
%\aksharStrTail { ळीममडमं }
% \end{codeexample}
% \begin{function}{\aksharStrChar}
%   \begin{syntax}
%     \cs{aksharStrChar} \marg{token list} \marg{n}
%   \end{syntax}
% Return the $n$-th character of the token list.
% \end{function}
% \begin{codeexample}[]
%3rd character of नमस्कार is \aksharStrChar{ नमस्कार}{3}.\par
%\ExplSyntaxOn
%It~is~not~\tl_item:nn { नमस्कार } {3}.
%\ExplSyntaxOff
% \end{codeexample}
% \begin{function}{\aksharStrReplace, \aksharStrReplace*}
%   \begin{syntax}
%     \cs{aksharStrReplace} \marg{tl 1} \marg{tl 2} \marg{tl 3}
%   \end{syntax}
%   Replace all occurences of \meta{tl 2} in \meta{tl 1} with \meta{tl 3}, and 
%   leaves the modified \meta{tl 1} in the input stream.
%
%   The starred variant will replace only the first occurence of \meta{tl 2}, 
%   all others are left intact.
% \end{function}
% \begin{codeexample}[]
%\ExplSyntaxOn
%\pkg{expl3} ~ output:\par
%\tl_set:Nn \l_tmpa_tl { मममडडमंळीममड }
%\tl_replace_all:Nnn \l_tmpa_tl { म } { स्का }
%\tl_use:N \l_tmpa_tl\par
%\ExplSyntaxOff
%\cs{aksharStrReplace} output:\par
%\aksharStrReplace { मममडडमंळीममड } { म } { स्का }
% \end{codeexample}
% \begin{codeexample}[]
%\ExplSyntaxOn
%\pkg{expl3} ~ output:\par
%\tl_set:Nn \l_tmpa_tl { ममंममडडमंळीममड }
%\tl_replace_once:Nnn \l_tmpa_tl { मम } { स्का }
%\tl_use:N \l_tmpa_tl\par
%\ExplSyntaxOff
%\cs{aksharStrReplace*} output:\par
%\aksharStrReplace* { ममंममडडमंळीममड } { मम } { स्का }
% \end{codeexample}
% \begin{function}{\aksharStrRemove, \aksharStrRemove*}
%   \begin{syntax}
%     \cs{aksharStrRemove} \marg{tl 1} \marg{tl 2}
%   \end{syntax}
%   Remove all occurences of \meta{tl 2} in \meta{tl 1}, and leaves the modified 
%   \meta{tl 1} in the input stream.
%
%   The starred variant will remove only the first occurence of \meta{tl 2}, 
%   all others are left intact.
% \end{function}
% \begin{codeexample}[]
%\ExplSyntaxOn
%\pkg{expl3} ~ output:\par
%\tl_set:Nn \l_tmpa_tl { मममडडमंळीममड }
%\tl_remove_all:Nn \l_tmpa_tl { म }
%\tl_use:N \l_tmpa_tl\par
%\ExplSyntaxOff
%\cs{aksharStrRemove} output:\par
%\aksharStrRemove { मममडडमंळीममड } { म }
% \end{codeexample}
% \begin{codeexample}[]
%\ExplSyntaxOn
%\pkg{expl3} ~ output:\par
%\tl_set:Nn \l_tmpa_tl { ममंममडडमंळीममड }
%\tl_remove_once:Nn \l_tmpa_tl { मम }
%\tl_use:N \l_tmpa_tl\par
%\ExplSyntaxOff
%\cs{aksharStrRemove*} output:\par
%\aksharStrRemove* { ममंममडडमंळीममड } { मम }
% \end{codeexample}
% \subsection{\pkg{expl3} functions}
% This section assumes that you have a basic knowledge in \LaTeX3 programming.
% All macros in \ref{latex2e-macros} directly depend on the following function,
% so it is much more powerful than all features we have described above.
% \begin{function}{
%    \akshar_convert:Nn, \akshar_convert:cn, 
%    \akshar_convert:Nx, \akshar_convert:cx}
%   \begin{syntax}
%     \cs{akshar_convert:Nn} \meta{seq var} \marg{token list}
%   \end{syntax}
%   This function converts \meta{token list} to a sequence of characters, that 
%   sequence is stored in \meta{seq var}.
% \end{function}
% \begin{codeexample}[]
%\ExplSyntaxOn
%\akshar_convert:Nn \l_tmpa_seq { नमस्कार }
%\seq_use:Nnnn \l_tmpa_seq { ~and~ } { ,~ } { ,~and~ }
%\ExplSyntaxOff
% \end{codeexample}
% \end{documentation}
% 
% \StopEventually{\PrintIndex}
% 
% \begin{implementation}
% \section{Implementation}
%    \begin{macrocode}
%<@@=akshar>
%<*package>
%    \end{macrocode}
% Declare the package. By loading \pkg{fontspec}, \pkg{xparse}, and in turn, 
% \pkg{expl3}, are also loaded.
%    \begin{macrocode}
\RequirePackage{fontspec}
\ProvidesExplPackage {\aksharPackageName} 
  {\aksharPackageDate} {\aksharPackageVersion} {\aksharPackageDescription}
%    \end{macrocode}
% \subsection{Variable declarations}
% \begin{variable}{\c_@@_joining_tl, \c_@@_diacritics_tl}
% These variables store the special characters we need to take into account:
% \begin{itemize}
%   \item \cs{c_@@_joining_tl} is the ``connecting'' character  ्.
%   \item \cs{c_@@_diacritics_tl} is the list of all diacritics: ा,  ि,  ी,  ु, 
%    ू,  े,  ै,  ो,  ौ,  ं,  ः,  ॢ,  ृ,  ॅ,  ँ,  ़,  ॆ,  ॊ,  ॉ,  ॎ,  ॏ,  ॑,
%     ॒,  ॓,  ॔,  ॕ,  ॖ,  ॗ,  ॄ,  ऺ,  ऻ,  ꣠,  ꣡,  ꣢,  ꣣,  ꣤,  ꣥,  ꣦,  ꣧,  ꣨,  ꣩, 
%    ꣪,  ꣫,  ꣬,  ꣭,  ꣮,  ꣯,  ꣰,  ꣱,  ꣿ.
% \end{itemize}
%    \begin{macrocode}
\tl_const:Nn \c_@@_joining_tl { ्}
\tl_const:Nn \c_@@_diacritics_tl
  {
    ा,  ि,  ी,  ु,  ू,  े,  ै,  ो,  ौ,  ं,  ः,  ॢ,  ृ, 
    ॅ,  ँ,  ़,  ॆ,  ॊ,  ॉ,  ॎ,  ॏ,  ॑,  ॒,  ॓,  ॔,  ॕ,  ॖ,  ॗ, 
    ॄ,  ऺ,  ऻ,  ꣠,  ꣡,  ꣢,  ꣣,  ꣤,  ꣥,  ꣦,  ꣧,  ꣨,  ꣩,  ꣪, 
    ꣫,  ꣬,  ꣭,  ꣮,  ꣯,  ꣰,  ꣱,  ꣿ
  }
%    \end{macrocode}
% \end{variable}
% \begin{variable}{\l_@@_prev_joining_bool}
% When we get to a normal character, we need to know whether it is joined, i.e. 
% whether the previous character is the joining character. This boolean variable 
% takes care of that.
%    \begin{macrocode}
\bool_new:N \l_@@_prev_joining_bool
%    \end{macrocode}
% \end{variable}
% \begin{variable}{\l_@@_char_seq}
% This local sequence stores the output of the converter.
%    \begin{macrocode}
\seq_new:N \l_@@_char_seq
%    \end{macrocode}
% \end{variable}
% \begin{variable}{\c_@@_str_g_tl, \c_@@_str_seq_tl, \c_@@_str_comma_tl}
% Some self-descriptive constant variables.
%    \begin{macrocode}
\tl_const:Nx \c_@@_str_g_tl { \tl_to_str:n {g} }
\tl_const:Nx \c_@@_str_seq_tl { \tl_to_str:n {seq} }
\tl_const:Nx \c_@@_str_comma_tl { \tl_to_str:n {,} }
%    \end{macrocode}
% \end{variable}
% \begin{variable}{\l_@@_tmpa_tl, \l_@@_tmpb_tl, \l_@@_tmpa_seq, \l_@@_tmpb_seq,
%                  \l_@@_tmpc_seq, \l_@@_tmpd_seq, \l_@@_tmpe_seq,
%                  \l_@@_tmpa_int, \l_@@_tmpb_int}
% Some temporary variables.
%    \begin{macrocode}
\tl_new:N \l_@@_tmpa_tl
\tl_new:N \l_@@_tmpb_tl
\seq_new:N \l_@@_tmpa_seq
\seq_new:N \l_@@_tmpb_seq
\seq_new:N \l_@@_tmpc_seq
\seq_new:N \l_@@_tmpd_seq
\seq_new:N \l_@@_tmpe_seq
\int_new:N \l_@@_tmpa_int
\int_new:N \l_@@_tmpb_int
%    \end{macrocode}
% \end{variable}
% \subsection{Messages}
% In \cs{akshar_convert:Nn} and friends, the argument needs to be a sequence 
% variable. There will be an error if it isn't.
% ^^A I use underscore here because my editor has a good support for it.
%    \begin{macrocode}
\msg_new:nnnn { akshar } { err_not_a_sequence_variable }
  { #1 ~ is ~ not ~ a ~ valid ~ LaTeX3 ~ sequence ~ variable. }
  {
    You ~ have ~ requested ~ me ~ to ~ assign ~ some ~ value ~ to ~ 
    the ~ control ~ sequence ~ #1, ~ but ~ it ~ is ~ not ~ a ~ valid ~ 
    sequence ~ variable. ~ Read ~ the ~ documentation ~ of ~ expl3 ~ 
    for ~ more ~ information. ~ Proceed ~ and ~ I ~ will ~ pretend ~ 
    that ~ #1 ~ is ~ a ~ local ~ sequence ~ variable ~ (beware ~ that ~ 
    unexpected ~ behaviours ~ may ~ occur).
  }
%    \end{macrocode}
% In \cs{aksharStrChar}, we need to guard against accessing an `out-of-bound' 
% character (like trying to get the 8th character in a 5-character string.)
%    \begin{macrocode}
\msg_new:nnnn { akshar } { err_character_out_of_bound }
  { Character ~ index ~ out ~ of ~ bound. }
  {
    You ~ are ~ trying ~ to ~ get ~ the ~ #2 ~ character ~ of ~ the ~ 
    string ~ #1. ~ However ~ that ~ character ~ doesn't ~ exist. ~ 
    Make ~ sure ~ that ~ you ~ use ~ a ~ number ~ between ~ and ~ not ~ 
    including ~ 0 ~ and ~ #3, ~ so ~ that ~ I ~ can ~ return ~ a ~ 
    good ~ output. ~ Proceed ~ and ~ I ~ will ~ return ~ 
    \token_to_str:N \scan_stop:.
  }
%    \end{macrocode}
% In \cs{aksharStrHead} and \cs{aksharStrTail}, the string must not be blank.
%    \begin{macrocode}
\msg_new:nnnn { akshar } { err_string_empty }
  { The ~ input ~ string ~ is ~ empty. }
  {
    To ~ get ~ the ~ #1 ~ character ~ of ~ a ~ string, ~ that ~ string ~ 
    must ~ not ~ be ~ empty, ~ but ~ the ~ input ~ string ~ is ~ empty. 
    Make ~ sure ~ the ~ string ~ contains ~ something, ~ or ~ proceed ~ 
    and ~ I ~ will ~ use ~ \token_to_str:N \scan_stop:.
  }
%    \end{macrocode}
% \subsection{Utilities}
% \begin{macro}[TF,internal]{\tl_if_in:No}
% When we get to a character which is not the joining one, we need to know if 
% it is a diacritic. The current character is stored in a variable, so an 
% expanded variant is needed. We only need it to expand only \textbf{o}nce.
%    \begin{macrocode}
\prg_generate_conditional_variant:Nnn \tl_if_in:Nn { No } { TF }
%    \end{macrocode}
% \end{macro}
% \begin{macro}[internal]{\seq_set_split:Nxx}
% A variant we will need in \cs{@@_var_if_global}.
%    \begin{macrocode}
\cs_generate_variant:Nn \seq_set_split:Nnn { Nxx }
%    \end{macrocode}
% \end{macro}
% \begin{macro}[internal]{\msg_error:nnx, \msg_error:nnnxx}
% Some variants of \pkg{l3msg} functions that we will need when issuing error 
% messages.
%    \begin{macrocode}
\cs_generate_variant:Nn \msg_error:nnn { nnx }
\cs_generate_variant:Nn \msg_error:nnnnn { nnnxx }
%    \end{macrocode}
% \end{macro}
% \begin{macro}[TF]{\@@_tl_if_in_ncomma:NN}
% This conditional is essentially \cs{tl_if_in:Nn}, but if |#2| is a comma this
% conditional always return false.
%    \begin{macrocode}
\prg_new_conditional:Npnn \@@_tl_if_in_ncomma:NN #1 #2 { T, F, TF }
  {
    \tl_if_eq:NNTF \c_@@_str_comma_tl #2 { \prg_return_false: }
      { \tl_if_in:NoTF #1 {#2} { \prg_return_true: } { \prg_return_false: } }
  }
%    \end{macrocode}
% \end{macro}
% \begin{macro}[TF]{\@@_var_if_global:N}
% This conditional checks if |#1| is a global sequence variable or not. In other 
% words, it returns true iff |#1| is a control sequence in the format 
% |\g_|\meta{name}|_seq|. If it is not a sequence variable, this function will 
% (TODO) issue an error message.
%    \begin{macrocode}
\prg_new_conditional:Npnn \@@_var_if_global:N #1 { T, F, TF }
  {
    \bool_if:nTF 
      { \exp_last_unbraced:Nf \use_iii:nnn { \cs_split_function:N #1 } }
      {
        \msg_error:nnx { akshar } { err_not_a_sequence_variable } 
          { \token_to_str:N #1 }
        \prg_return_false:
      }
      {
        \seq_set_split:Nxx \l_@@_tmpb_seq { \token_to_str:N _ }
          { \exp_last_unbraced:Nf \use_i:nnn { \cs_split_function:N #1 } }
        \seq_get_left:NN  \l_@@_tmpb_seq \l_@@_tmpa_tl
        \seq_get_right:NN \l_@@_tmpb_seq \l_@@_tmpb_tl
        \tl_if_eq:NNTF \c_@@_str_seq_tl \l_@@_tmpb_tl
          {
            \tl_if_eq:NNTF \c_@@_str_g_tl \l_@@_tmpa_tl
              { \prg_return_true: } { \prg_return_false: }
          }
          {
            \msg_error:nnx { akshar } { err_not_a_sequence_variable } 
              { \token_to_str:N #1 }
            \prg_return_false:
          }
      }
  }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\@@_int_append_ordinal:n}
% Append st, nd, rd or th to interger |#1|. Will be needed in error messages.
%    \begin{macrocode}
\cs_new:Npn \@@_int_append_ordinal:n #1
  {
    #1
    \int_case:nnF { #1 }
      {
        { 11 } { th }
        { 12 } { th }
        { 13 } { th }
        { -11 } { th }
        { -12 } { th }
        { -13 } { th }
      }
      {
        \int_compare:nNnTF { #1 } > { -1 }
          {
            \int_case:nnF { #1 - 10 * (#1 / 10) }
              {
                { 1 } { st }
                { 2 } { nd }
                { 3 } { rd }
              } { th }
          }
          {
            \int_case:nnF { (- #1) - 10 * ((- #1) / 10) }
              {
                { 1 } { st }
                { 2 } { nd }
                { 3 } { rd }
              } { th }
          }
      }
  }
%    \end{macrocode}
% \end{macro}
% \subsection{The \cs{akshar_convert:Nn} function and its variants}
% \begin{macro}{\akshar_convert:Nn, \akshar_convert:cn,
%               \akshar_convert:Nx, \akshar_convert:cx}
% This converts |#2| to a sequence of \emph{true} Devanagari characters. The 
% sequence is set to |#1|, which should be a sequence variable.
%    \begin{macrocode}
\cs_new:Npn \akshar_convert:Nn #1 #2
  {
%    \end{macrocode}
% Clear anything stored in advance. We don't want different calls of the 
% function to conflict with each other.
%    \begin{macrocode}
    \seq_clear:N \l_@@_char_seq
    \bool_set_false:N \l_@@_prev_joining_bool
%    \end{macrocode}
% Loop through every token of the input.
%    \begin{macrocode}
    \tl_map_variable:NNn {#2} \l_@@_map_tl
      {
        \@@_tl_if_in_ncomma:NNTF
          \c_@@_diacritics_tl \l_@@_map_tl
          {
%    \end{macrocode}
% It is a diacritic. We append the current diacritic to the last item of the 
% sequence instead of pushing the diacritic to a new sequence item.
%    \begin{macrocode}
            \seq_pop_right:NN \l_@@_char_seq \l_@@_tmpa_tl
            \seq_put_right:Nx \l_@@_char_seq 
              { \l_@@_tmpa_tl \l_@@_map_tl }
          }
          {
            \tl_if_eq:NNTF \l_@@_map_tl \c_@@_joining_tl
              {
%    \end{macrocode}
% In this case, the character is the joining character,  ्. What we do is 
% similar to the above case, but \cs{l_@@_prev_joining_bool} is set to true so 
% that the next character is also appended to this item.
%    \begin{macrocode}
                \seq_pop_right:NN \l_@@_char_seq \l_@@_tmpa_tl
                \seq_put_right:Nx \l_@@_char_seq 
                  { \l_@@_tmpa_tl \l_@@_map_tl }
                \bool_set_true:N \l_@@_prev_joining_bool
              }
              {
%    \end{macrocode}
% Now the character is normal. We see if we can push to a new item or not. It 
% depends on the boolean variable.
%    \begin{macrocode}
                \bool_if:NTF \l_@@_prev_joining_bool
                  {
                    \seq_pop_right:NN \l_@@_char_seq \l_@@_tmpa_tl
                    \seq_put_right:Nx \l_@@_char_seq 
                      { \l_@@_tmpa_tl \l_@@_map_tl }
                    \bool_set_false:N \l_@@_prev_joining_bool
                  }
                  {
                    \seq_put_right:Nx 
                      \l_@@_char_seq { \l_@@_map_tl }
                  }
              }
          }
      }
%    \end{macrocode}
% Set |#1| to \cs{l_@@_char_seq}. The package automatically determines whether 
% the variable is a global one or a local one.
%    \begin{macrocode}
    \@@_var_if_global:NTF #1
      { \seq_gset_eq:NN #1 \l_@@_char_seq }
      { \seq_set_eq:NN #1 \l_@@_char_seq }
  }
%    \end{macrocode}
% Generate variants that might be helpful for some.
%    \begin{macrocode}
\cs_generate_variant:Nn \akshar_convert:Nn { cn, Nx, cx }
%    \end{macrocode}
% \end{macro}
% \subsection{Other internal functions}
% \begin{macro}{\@@_seq_push_seq:NN}
% Append sequence |#1| to the end of sequence |#2|. A simple loop will do.
%    \begin{macrocode}
\cs_new:Npn \@@_seq_push_seq:NN #1 #2
  { \seq_map_inline:Nn #2 { \seq_put_right:Nn #1 { ##1 } } }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\@@_replace:NnnnN}
% If |#5| is \cs{c_false_bool}, this function replaces \emph{all} occurences of 
% |#3| in |#2| by |#4| and stores the output sequence to |#1|. If |#5| is 
% \cs{c_true_bool}, the replacement only happens \emph{once}.
%
% The algorithm used in this function: We will use \cs{l_@@_tmpa_int} to store 
% the ``current position'' in the sequence of |#3|. At first it is set to |1|.
%
% We will store any subsequence of |#2| that may match |#3| to a temporary 
% sequence. If it doesn't match, we push this temporary sequence to the output, 
% but if it matches, |#4| is pushed instead.
%
% We loop over |#2|. For each of these loops, we need to make sure the 
% \cs{l_@@_tmpa_int}-th item must indeed appear in |#3|. So we need to compare 
% that with the length of |#3|.
% \begin{itemize}
%   \item If now \cs{l_@@_tmpa_int} is greater than the length of |#3|, the 
%   whole of |#3| has been matched somewhere, so we reinitialize the integer to 
%   |1| and push |#4| to the output.
%
%   Note that it is possible that the current character might be the start of 
%   another match, so we have to compare it to the first character of |#3|. 
%   If they are not the same, we may now push the current mapping character to 
%   the output and proceed; otherwise the current character is pushed to the 
%   temporary variable.
%   \item Otherwise, we compare the current loop character of |#2| with the 
%   \cs{l_@@_tmpa_int}-th character of |#3|.
%   \begin{itemize}
%     \item If they are the same, we still have a chance that it will match, so 
%     we increase the ``iterator'' \cs{l_@@_tmpa_int} by |1| and push the 
%     current mapping character to the temporary sequence.
%     \item If they are the same, the temporary sequence won't match. Let's push 
%     that sequence to the output and set the iterator back to |1|.
%
%     Note that now the iterator has changed. Who knows whether the current 
%     character may start a match? Let's compare it to the first character of 
%     |#3|, and do as in the case of \cs{l_@@_tmpa_int} is greater than the 
%     length of |#3|.
%   \end{itemize}
% \end{itemize}
% The complexity of this algorithm is $O(m\max(n,p))$, where $m,n,p$ are the 
% lengths of the sequences created from |#2|, |#3| and |#4|. As |#3| and |#4| 
% are generally short strings, this is (almost) linear to the length of the 
% original sequence |#2|.
%    \begin{macrocode}
\cs_new:Npn \@@_replace:NnnnN #1 #2 #3 #4 #5
  {
    \akshar_convert:Nn \l_@@_tmpc_seq {#2}
    \akshar_convert:Nn \l_@@_tmpd_seq {#3}
    \akshar_convert:Nn \l_@@_tmpe_seq {#4}
    \seq_clear:N \l_@@_tmpa_seq
    \seq_clear:N \l_@@_tmpb_seq
    \int_set:Nn \l_@@_tmpa_int { 1 }
    \int_set:Nn \l_@@_tmpb_int { 0 }
    \seq_map_variable:NNn \l_@@_tmpc_seq \l_@@_map_tl
      {
        \int_compare:nNnTF { \l_@@_tmpb_int } > { 0 }
          { \seq_put_right:NV \l_@@_tmpb_seq \l_@@_map_tl }
          {
            \int_compare:nNnTF 
              {\l_@@_tmpa_int} = {1 + \seq_count:N \l_@@_tmpd_seq}
              {
                \bool_if:NT {#5}
                  { \int_incr:N \l_@@_tmpb_int }
                \seq_clear:N \l_@@_tmpb_seq
                \@@_seq_push_seq:NN 
                  \l_@@_tmpa_seq \l_@@_tmpe_seq
                \int_set:Nn \l_@@_tmpa_int { 1 }
                \tl_set:Nx \l_@@_tmpa_tl 
                  { \seq_item:Nn \l_@@_tmpd_seq { 1 } }
                \tl_if_eq:NNTF \l_@@_map_tl \l_@@_tmpa_tl
                  {
                    \int_incr:N \l_@@_tmpa_int
                    \seq_put_right:NV \l_@@_tmpb_seq \l_@@_map_tl
                  }
                  {
                    \seq_put_right:NV \l_@@_tmpa_seq \l_@@_map_tl
                  }
              }
              {
                \tl_set:Nx \l_@@_tmpa_tl 
                  {
                    \seq_item:Nn \l_@@_tmpd_seq { \l_@@_tmpa_int }
                  }
                \tl_if_eq:NNTF \l_@@_map_tl \l_@@_tmpa_tl
                  {
                    \int_incr:N \l_@@_tmpa_int
                    \seq_put_right:NV \l_@@_tmpb_seq \l_@@_map_tl
                  }
                  {
                    \int_set:Nn \l_@@_tmpa_int { 1 }
                    \@@_seq_push_seq:NN 
                      \l_@@_tmpa_seq \l_@@_tmpb_seq
                    \seq_clear:N \l_@@_tmpb_seq
                    \tl_set:Nx \l_@@_tmpa_tl 
                      { \seq_item:Nn \l_@@_tmpd_seq { 1 } }
                    \tl_if_eq:NNTF \l_@@_map_tl \l_@@_tmpa_tl
                      {
                        \int_incr:N \l_@@_tmpa_int
                        \seq_put_right:NV
                          \l_@@_tmpb_seq \l_@@_map_tl
                      }
                      { 
                        \seq_put_right:NV
                          \l_@@_tmpa_seq \l_@@_map_tl
                      }
                  }
              }
          }
      }
    \@@_seq_push_seq:NN \l_@@_tmpa_seq \l_@@_tmpb_seq
    \@@_var_if_global:NTF #1
      { \seq_gset_eq:NN #1 \l_@@_tmpa_seq }
      { \seq_set_eq:NN #1 \l_@@_tmpa_seq }
  }
%    \end{macrocode}
% \end{macro}
% \subsection{Front-end \LaTeXe\ macros}
% \begin{macro}{\aksharStrLen}
% Expands to the length of the string.
%    \begin{macrocode}
\NewDocumentCommand \aksharStrLen {m}
  {
    \akshar_convert:Nn \l_@@_tmpa_seq {#1}
    \seq_count:N \l_@@_tmpa_seq
  }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\aksharStrChar}
% Returns the $n$-th character of the string.
%    \begin{macrocode}
\NewDocumentCommand \aksharStrChar {mm}
  {
    \akshar_convert:Nn \l_@@_tmpa_seq {#1}
    \bool_if:nTF
      {
        \int_compare_p:nNn {#2} > {0} &&
        \int_compare_p:nNn {#2} < {1 + \seq_count:N \l_@@_tmpa_seq}
      }
      { \seq_item:Nn \l_@@_tmpa_seq { #2 } }
      {
        \msg_error:nnnxx { akshar } { err_character_out_of_bound }
          { #1 } { \@@_int_append_ordinal:n { #2 } }
          { \int_eval:n { 1 + \seq_count:N \l_@@_tmpa_seq } }
        \scan_stop:
      }
  }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\aksharStrHead}
% Return the first character of the string.
%    \begin{macrocode}
\NewDocumentCommand \aksharStrHead {m}
  {
    \akshar_convert:Nn \l_@@_tmpa_seq {#1}
    \int_compare:nNnTF { \seq_count:N \l_@@_tmpa_seq } = {0}
      {
        \msg_error:nnn { akshar } { err_character_out_of_bound }
          { first }
        \scan_stop:
      }
      { \seq_item:Nn \l_@@_tmpa_seq { 1 } }
  }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\aksharStrTail}
% Return the last character of the string.
%    \begin{macrocode}
\NewDocumentCommand \aksharStrTail {m}
  {
    \akshar_convert:Nn \l_@@_tmpa_seq {#1}
    \int_compare:nNnTF { \seq_count:N \l_@@_tmpa_seq } = {0}
      {
        \msg_error:nnn { akshar } { err_character_out_of_bound }
          { last }
        \scan_stop:
      }
      { \seq_item:Nn \l_@@_tmpa_seq {\seq_count:N \l_@@_tmpa_seq} }
  }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\aksharStrReplace, \aksharStrReplace*}
% Replace occurences of |#3| of a string |#2| with another string |#4|.
%    \begin{macrocode}
\NewDocumentCommand \aksharStrReplace {smmm}
  {
    \IfBooleanTF {#1}
      {
        \@@_replace:NnnnN \l_@@_tmpa_seq
          {#2} {#3} {#4} \c_true_bool
      }
      {
        \@@_replace:NnnnN \l_@@_tmpa_seq
          {#2} {#3} {#4} \c_false_bool
      }
    \seq_use:Nn \l_@@_tmpa_seq {}
  }
%    \end{macrocode}
% \end{macro}
% \begin{macro}{\aksharStrRemove, \aksharStrRemove*}
% Remove occurences of |#3| in |#2|. This is just a special case of 
% \cs{aksharStrReplace}.
%    \begin{macrocode}
\NewDocumentCommand \aksharStrRemove {smm}
  {
    \IfBooleanTF {#1}
      {
        \@@_replace:NnnnN \l_@@_tmpa_seq
          {#2} {#3} {} \c_true_bool
      }
      {
        \@@_replace:NnnnN \l_@@_tmpa_seq
          {#2} {#3} {} \c_false_bool
      }
    \seq_use:Nn \l_@@_tmpa_seq {}
  }
%    \end{macrocode}
% \end{macro}
%    \begin{macrocode}
%</package>
%    \end{macrocode}
% \end{implementation}
%
% \Finale