bachotex2006-bernd-raichle-pearl4.tex

File contents

%%% Bernd Raichle: check if defined, no side effects

% LaTeX and other macro packages include a test if a control sequence is
% already defined using \csname...\endcsname. In principle the following
% definition is used:

\def\@ifundefined#1#2#3{%
  \expandafter\ifx\csname#1\endcsname\relax#2\else#3\fi}

% The use of \csname...\endcsname has the side effect that an undefined control
% sequence will be defined as \relax and a simple test using
% \ifx<control sequence>\undefined ... will fail.

% To avoid this side effect, you can change the definition
% above introducing a group ``around'' \csname:

\def\@ifundefined#1#2#3{%
  \begingroup \expandafter\expandafter\expandafter \endgroup
  \expandafter\ifx\csname#1\endcsname\relax#2\else#3\fi}

% Note: This definition is not fully expandable, thus it will fail in
% expansion-only contexts where the original definition will work.
%
% If you are using eTeX, the new \ifcsname ...\endcsname
% primitive can be used instead.

%%%%

% During the presentation of this pearl at the BachoTeX 2006 pearls session,
% two alternatives to this trick has been proposed. Worthy to note, that the
% macro above use more \expandafter's then actually needed. Once we say

\def\@ifundefined#1#2#3{%
  \begingroup \expandafter
  \ifx\csname#1\endcsname\relax\endgroup#2\else\endgroup#3\fi}

% we get exactly the same effect in a bit more efficient way. But there is one
% more sticky problem in all the constructions above; every control sequence
% defined as \relax (and \relax itself) is treated as undefined. The solution
% seems to be

\def\@ifundefined#1#2#3{%
  \begingroup \expandafter \endgroup \expandafter
  \ifx\csname#1\endcsname\undefined#2\else#3\fi}

% Here we expand \csname...\endcsname before \endgroup, but the condition is
% fixed outside the group. Thus, \undefined instead of \relax can be used for
% comparison.

\end