Replace character defined by ASCII code (by something else)

Question

I am trying to replace, in a given string, some characters. The characters to be replaced is given by its ASCII code. So for example, in abc, I want to replace character 97 (which is a, but I don't know this yet) by something else, let's say a dot (.).

Now, a, \char97 and \char\thenumber all look the same to me when printing them in a document, but replacing this respective argument in a string a only works with the first. Why is that, and how can I solve it? This is my MWE (unchanged from the original question):

\documentclass{book}
\usepackage{xstring}
\newcounter{number}
\setcounter{number}{97}
\newcommand{\replace}[1]{\StrSubstitute{#1}{a}{.}}
\begin{document}
a

\char97

\char\thenumber

\replace{a}

\expandafter\replace{\char97}

\expandafter\expandafter\expandafter\replace{\char\thenumber}

\end{document}

Answer 1

The TeX \char primitive is not expandable. This means that in the example the attempted expansion

\expandafter\replace\expandafter{\char97}

does nothing and is equivalent to

\replace{\char97}

which for hopefully-obvious reasons fails to match a literal a. What is needed is therefore some expandable way to turn a number into a char. Classical TeX offers us the \ifcase primitive, which is used in the general form

\ifcase<number> %
   <case 0>\or
   <case 1>\or
   <case 2>\or
   ...
   <case n>\else
   <no match>
\fi

Thus with appropriate set up it's possible to do a case-by-case conversion. Such an approach is taken by LaTeX's \@alph and \@Alph, which index lower-/upper-case letters, respectively, from 1:

\@alph{1}% => a

Clearly here some maths made be needed to get the offset right. Such an approach is expandable but is obviously tedious for long lists of chars.

It's possible to set up a 'selective' case statement that is expandable: there is one pre-built in for example expl3

\int_case:nnF { <number> }
  {
    {  97 } { a }
    { 101 } { e }
    { 105 } { i }
    { 111 } { o }
    { 117 } { u }
  }
  { No~match }

There is a performance hit over the primitive for short lists of continuous integer ranges but better input format and clarity, particularly for long lists. (I think there are other implementations of the same concept in addition to the expl3 one.)

The problem with any case-by-case approach is if you are looking beyond ASCII to the entire UTF-8 range. If LuaTeX is in use then there is an expandable primitive that will do the job

\Uchar97 % => a

That's obviously the easiest solution if you can be sure of the engine in use.

Answer 2

There is a standard \lowercase trick which creates a token with a given ASCII code. For example you need to replace all ASCII 97 characters from the string abcabcac by double ?. Then you can try:

\bgroup\lccode`X=97 \lowercase{\egroup \StrSubstitute{abcabcabc}{X}{??}}
output:  ??bc??bc??bc

This code creates the token with given ASCII code instead the X letter and then the \StrSubstitute macro is executed.

Of course, your processed string cannot include the X letter itself. If this is a problem then you can create the following \replace macro:

\def\replace#1{\bgroup \lccode`X=#1 \lowercase{\egroup \replaceA{X}}{abcabcabcXuv}}
\def\replaceA#1#2{\StrSubstitute{#2}{#1}{??}}

\replace{97}
output:  ??bc??bc??bcXuv
\replace{98}
output:  a??ca??ca??cXuv

Answer 3

What I wanted to achieve I achieved using \@Alph and \@alph, since my input ASCII codes are in the a-z, A-Z range. \@Alph etc. appear to expand to the 'real' letters. As multiple others have pointed out, \char is not expandable. I hope this may help someone else sometime.