How do I handle a closing brace in a 'character scanning' macro?

Question

Problem

I need to define a command that grabs all numerical digits following it and leaves the rest alone. Eventually I came up with this:

First Solution

\RequirePackage{xstring} % for \IfInteger

\newcommand*{\num}{%
    \def\@tmpnum{}%
    \@num%
}

\newcommand*{\@num}[1]{%
    \IfInteger{#1}{%
        \edef\@tmpnum{\@tmpnum#1}\@num%
    }{%
        [I saw the number \@tmpnum!]#1%
    }%
}

The \@num macro scans ahead by recursively calling itself. Every digit it encounters is appended to \@tmpnum. The first non-digit it sees ends the sequence and is put back.

It works as intended in simple cases:

\num1234...

outputs:

[I saw the number 1234!]...

How to Handle \par - Second Solution

It soon occurred to me that \IfInteger can't handle a \par, so... second solution:

\newcommand{\@num}[1]{%
    \def\@end{%
        [I saw the number \@tmpnum!]#1%
    }%
    \ifdefequal{#1}{\par}\@end{%
        \IfInteger{#1}{%
            \edef\@tmpnum{\@tmpnum#1}\@num%
        }\@end%
    }%
}

How to Handle a Closing Brace?

This one I can't figure out. What if I want to accept something like this:

\emph{\num42}

Whatever I try, I get some error about having too many closing braces. I guess I'm putting it back when it was already scanned. But that doesn't really help me.

• Can I test for a closing brace? Is there some other way to handle this?
• Are other such problems waiting for me?

Answer 1

To scan a character at a time you need to use \futurelet (or its LaTeX wrapper \@ifnextchar) not a macro argument #1 then it is safe for {}. But for numbers I would just do this:

\documentclass{article}

\begin{document}

\makeatletter
\def\num{\afterassignment\xnum\count@}


\def\xnum{the number was [\the\count@]}

\num44

\emph{\num99}

\end{document}

---

There is nothing special about \count@ here. It is just a handy scratch count register that is declared in LaTeX and plain. You could go \newcount\mycount and then use \mycount instead. The point is that when assigning a value to a count register TeX automatically parses for a number and puts the value in the register. \afterassignment then gives you a hook to gain control after the assignment has been done and do something with the value.

Answer 2

I do not know the purpose and format of the "number". There are many variants what can be considered a number:

1. Explicit number consisting of digits 0 up to 9.

2. TeX number, that includes count registers, internal numbers given by primitives (\inputlineno, \value{page}, \numexpr), even dimensions (\textwidth) can be interpreted as numbers.

This solution assumes an explicit number, because the questions speaks of "grabbing numerical digits". Also the tokens afterwords should be kept and not removed. When scanning a number TeX sometimes gobbles a space or expands the tokens afterwards in some situations.

The test relies on \futurelet to inspect the next token.

\documentclass{article}

\makeatletter
\newcommand*{\num}{%
    \let\@tmpnum\@empty
    \futurelet\@tmptok\scan@getnum
}
\newcommand*{\scan@getnum}{%
  \let\scan@next\scan@addnum
  \ifx\@tmptok0%
  \else\ifx\@tmptok1%
  \else\ifx\@tmptok2%
  \else\ifx\@tmptok3%
  \else\ifx\@tmptok4%
  \else\ifx\@tmptok5%
  \else\ifx\@tmptok6%
  \else\ifx\@tmptok7%
  \else\ifx\@tmptok8%
  \else\ifx\@tmptok9%
  \else
    \def\scan@next{%
      [I saw the number \@tmpnum!]%
    }%
  \fi\fi\fi\fi\fi\fi\fi\fi\fi\fi
  \scan@next
}
\newcommand*{\scan@addnum}[1]{%
  \edef\@tmpnum{\@tmpnum#1}%
  \futurelet\@tmptok\scan@getnum
}
\makeatother

\begin{document}

\num1234...

\num56 78 with spaces

\num9\par

\emph{\num42}

\end{document}

Answer 3

I just discovered a feature of etextools, based on \futurelet, that seems to be made for this purpose:

\newcommand*{\num}{%
    \futuredef[0123456789]{\@nn}{[I saw the number \@nn!]}%
}

\futuredef takes a list of tokens that it accepts and it stops scanning when it sees anything else.