Define a special verbatim commmand

Question

To learn more about Plain TeX, I try to define a special verbatim command \verb{code}, which puts code inside braces.

When there is no braces in code, It's easy. But when there are some balanced braces inside code, it seems hard to define this command.

Furthermore, there may be some \{ or \} inside code. How could I define this?

For example, I want \verb{a{b\}c\{d}e} to produce a{b\}c\{d}e.

Answer 1

A plain TeX solution without e-TeX.

1. The argument is read with verbatim category codes (12). Plain TeX provides a list of special characters in \dospecials.

2. The argument is surrounded by curly braces. Therefore the curly braces gets their usual category codes. Then the argument may contain balanced curly braces. However it is not possible to escape a curly brace with a backslash, because the backslash has category code 12.

3. The argument is read and stored in a macro \verbText.

4. \meaning prints the macro definition. The characters have the category code 12 (other) and the space character has its normal category code 10 (space).

5. Font \tentt has a special space character, but the space needs a different catcode. This is done by macro \ConvertSpacesToCatcodeOther.

Example file:

\def\verb{%
  \begingroup
  % verbatim catcodes
  \def\do##1{\catcode`##1=12 }%
  \dospecials
  % except curly braces
  \catcode`\{=1 %
  \catcode`\}=2 %
  \verbAux
}
\def\verbAux#1{%
  \def\verbText{#1}%
  \edef\verbText{%
    \expandafter\StripPrefix\meaning\verbText
  }%
  \tentt
  \FirstOfOne{\expandafter\ConvertSpacesToCatcodeOther\verbText} \NIL
  \endgroup
}
\def\StripPrefix#1>{}
\long\def\FirstOfOne#1{#1}
\def\ConvertSpacesToCatcodeOther#1 #2\NIL{%
  #1%
  \def\temp{#2}%
  \ifx\temp\empty
    \expandafter\Gobble
  \else
    \SpaceOther
    \expandafter\FirstOfOne
  \fi
  {\ConvertSpacesToCatcodeOther#2\NIL}%
}
\long\def\Gobble#1{}
\begingroup
  \lccode`\9=32 % space
\lowercase{\endgroup
  \def\SpaceOther{9}%
}

\verb{a{b\}c\{d}e  with spaces}

\bye

Escaping

Also the escaping of the curly braces with a backslash can be implemented. If the backslash has its escape category code (0), then command names are tokenized as usual. \meaning will add a space after command names that consists of characters with letter catcodes (11). \abc would become \abc␣. This can be avoided by changing the category codes for all letters:

\def\verb{%
  \begingroup
  % verbatim catcodes
  \def\do##1{\catcode`##1=12 }%
  \dospecials
  % except curly braces and backslash
  \catcode`\\=0 %
  \catcode`\{=1 %
  \catcode`\}=2 %
  \count255=`\A %
  \loop
    \catcode\count255=12 %
  \ifnum\count255<`\Z %
    \advance\count255 by 1 %  
  \repeat
  \count255=`\a
  \loop
    \catcode\count255=12 %
  \ifnum\count255<`\z %
    \advance\count255 by 1 %  
  \repeat
  \verbAux
}
\def\verbAux#1{%
  \def\verbText{#1}%
  \edef\verbText{%
    \expandafter\StripPrefix\meaning\verbText
  }%
  \tentt
  \FirstOfOne{\expandafter\ConvertSpacesToCatcodeOther\verbText} \NIL
  \endgroup
}
\def\StripPrefix#1>{}
\long\def\FirstOfOne#1{#1}
\def\ConvertSpacesToCatcodeOther#1 #2\NIL{%
  #1%
  \def\temp{#2}%
  \ifx\temp\empty
    \expandafter\Gobble
  \else
    \SpaceOther
    \expandafter\FirstOfOne
  \fi
  {\ConvertSpacesToCatcodeOther#2\NIL}%
}
\long\def\Gobble#1{}
\begingroup
  \lccode`\9=32 % space
\lowercase{\endgroup
  \def\SpaceOther{9}%
}

\verb{a{b\}c\{d}e  with spaces \abc\} \{}

\bye

Answer 2

Here is more simple solution:

\newcount\tmpnum
\def\catcodeletters{\tmpnum=64
  \loop \advance\tmpnum by1 
     \ifnum\tmpnum<128
        \ifnum\catcode\tmpnum=11 \catcode\tmpnum=12 \fi
     \repeat 
}
\def\verb{\bgroup\catcode`\%=12\catcode`\^=13\catcode`\ =12\catcode`\#=12
   \catcodeletters\verbA}
\def\verbA#1{\def\tmp{#1}\tt\expandafter\mm\meaning\tmp\endmm\tmp\egroup}
\def\mm#1->#2\endmm{\def\tmp{#2}}

\verb{a{b\}c\{d}e a  #sp_&ace}

\end

I've added the \catcodeletters to my macro in order to spaces after \foo don't arise. Thanks to Heiko.

Edit: My 12-lines macro was not chosen as an accepted answer but 50-line macro was. May be that the spacemarks replaced by real spaces was the core of the interest, but it was not specified in the question. This is a part of printing, no scanning verbatim. So it can be implemented after \tmp is scanned. You can change \verbA definition and you can add the \printverb macro, which does the space replacement by another way than in the previous answer.

\def\verbA#1{\def\tmp{#1}\expandafter\mm\meaning\tmp\endmm
   \tt\expandafter\printverb\tmp\end\egroup}

\def\printverb{\futurelet\next\printverbA}
\def\printverbA{%
  \ifx\next\end \def\next{\let\next}%
  \else \ifx\next\spacetoken \char32
            \def\next{\afterassignment\printverb\let\next= }%
        \else \next \def\next{\afterassignment\printverb\let\next}%
  \fi   \fi
  \next
} 
\edef\tmp{\let\noexpand\spacetoken= \space}\tmp

Now the \tt\char32 spacemarks are printed instead spaces.