TeX capacity exceeded while parsing a date string

Question

I'm trying to parse a date (say, a due date) in a DD-MMM-YYYY format, e.g., 06-May-2012 and create macros \dueday, \duemonth, and \dueyear. I use xstring's commands to extract the substrings. Then I use xstring's \IfStrEqCase to generate the numeric \duemonth:

\documentclass{article}

\usepackage{xstring}

\begin{document}
\newcommand{\duedate}{06-May-2012}

\newcommand{\dueday}{\StrBefore{\duedate}{-}}
\newcommand{\duemonthname}{\StrBetween[1,2]{\duedate}{-}{-}}
\newcommand{\dueyear}{\StrBehind[2]{\duedate}{-}}

Reconstruction: \dueday-\duemonthname-\dueyear.

%\renewcommand{\duemonthname}{May}   % This appears to solve the problem.

\IfStrEqCase{\duemonthname}{%
    {Jan}{\newcommand{\duemonth}{1}}%
    {Feb}{\newcommand{\duemonth}{2}}%
    {Mar}{\newcommand{\duemonth}{3}}%
    {Apr}{\newcommand{\duemonth}{4}}%
    {May}{\newcommand{\duemonth}{5}}%
    {Jun}{\newcommand{\duemonth}{6}}%
    {Jul}{\newcommand{\duemonth}{7}}%
    {Aug}{\newcommand{\duemonth}{8}}%
    {Sep}{\newcommand{\duemonth}{9}}%
    {Oct}{\newcommand{\duemonth}{10}}%
    {Nov}{\newcommand{\duemonth}{11}}%
    {Dec}{\newcommand{\duemonth}{12}}%
    }

\duemonth   % This should output 5.

\end{document}

which fails with an error message TeX capacity exceeded. The \IfStrEqCase doesn't seem to be the problem, because if I issue \renewcommand{\duemonthname}{May} right before it, everything is fine. This suggests that the error is due to splitting commands, but their outputs are what I expect them to be (at least their "visible" outputs are). What is causing the error then?

Answer 1

No, you don't need xstring for that. TeX's syntax is sufficient.

\def in TeX use a parameter template to extract any format of argument. The syntax is (c.f. TeX for the Impatient):

\def \foo parameter text {replacement text}

Thus you can directly use

\def\parsedate #1-#2-#3\stopmark{Use #1 and #2 and #3 as you wish}

to get the different arguments in a date. This is even simpler (and more effective) than xstring package.

To get a number from the name of a month, you can define a sequence of macros. For example, define

\def\theFeb{2}

then you can use \theFeb to get the number. What's more, \csname ...\endcsname allows you to use \csname the#1\endcsname to get \theFeb when #1 is Feb.

\parsedate above needs a \stopmark (it can be any thing) to denote the end of the argument, so it's convenient to define a new macro to use it:

\def\parsedateHelper #1-#2-#3\stopmark{Use #1 and #2 and #3 as you wish}
\def\parsedate#1{\parsedateHelper #1\stopmark}
% use as \parsedate{06-May-2012}

You may also meet some expansion issues when you use a macro as the argument of \parsedate, so it is better to store the argument into another macro and use it with help of \expandafter. It is a bit tricky:

\def\parsedate#1{%
  \edef\savedargument{#1}% #1 is expanded by \edef
  \expandafter\parsedateHelper\savedargument\stopmark}

---

Make everything together:

\documentclass{article}

\def\parsedate#1{\edef\temp{#1}%
  \expandafter\parsedateX\temp\relax}
\def\parsedateX #1-#2-#3\relax{%
  \def\dueday{#1}%
  \edef\duemonth{\csname the#2\endcsname}%
  \def\dueyear{#3}}
\def\theJan{1}
\def\theFeb{2}
\def\theMar{3}
\def\theApr{4}
\def\theMay{5}
\def\theJun{6}
\def\theJul{7}
\def\theAug{8}
\def\theSep{9}
\def\theOct{10}
\def\theNov{11}
\def\theDec{12}


\begin{document}
\parsedate{06-May-2012}
\dueyear/\duemonth/\dueday

\def\duedate{01-Aug-2003}
\parsedate{\duedate}
\dueyear/\duemonth/\dueday

\end{document}

Answer 2

The mandatory LaTeX3 solution:

\documentclass{article}
\usepackage{xparse}
\ExplSyntaxOn
\NewDocumentCommand{\parsedate}{ >{ \SplitArgument { 2 } { - } } m }
 {
  \parsedate_main:nnn #1
 }
\cs_new_protected:Npn \parsedate_main:nnn #1 #2 #3
 {
  \cs_gset:Npn \dueday { #1 }
  \cs_gset:Npx \duemonth { \prop_get:Nn \g_parsedate_months_prop { #2 } }
  \cs_gset:Npn \dueyear { #3 }
 }
\prop_new:N \g_parsedate_months_prop
\prop_gput:Nnn \g_parsedate_months_prop { Jan } { 01 }
\prop_gput:Nnn \g_parsedate_months_prop { Feb } { 02 }
\prop_gput:Nnn \g_parsedate_months_prop { Mar } { 03 }
\prop_gput:Nnn \g_parsedate_months_prop { Apr } { 04 }
\prop_gput:Nnn \g_parsedate_months_prop { May } { 05 }
\prop_gput:Nnn \g_parsedate_months_prop { Jun } { 06 }
\prop_gput:Nnn \g_parsedate_months_prop { Jul } { 07 }
\prop_gput:Nnn \g_parsedate_months_prop { Aug } { 08 }
\prop_gput:Nnn \g_parsedate_months_prop { Sep } { 09 }
\prop_gput:Nnn \g_parsedate_months_prop { Oct } { 10 }
\prop_gput:Nnn \g_parsedate_months_prop { Nov } { 11 }
\prop_gput:Nnn \g_parsedate_months_prop { Dec } { 12 }
\ExplSyntaxOff

\begin{document}
\parsedate{06-May-2012}
\dueday/\duemonth/\dueyear
\end{document}

This will print

06/05/2012

The argument is given as 06-May-2012, but the \SplitArgument facility presents it to \parsedate_main:nnn already in the form {06}{May}{2012}. Thus it's easy to extract the day and the year. For the month we use a property list: we associate to each month its number and retrieve this one with \prop_get:Nn.

Some words of explanation

An argument passed to a command defined with \NewDocumentCommand can be "processed" before being presented to the macros doing the actual work. In this case we ask that the argument is split into three parts at the - with

\SplitArgument{2}{-}

(it means that two - are expected). So what's denoted with #1 in the body of the definition is to be thought of as {chunk1}{chunk2}{chunk3}. This is passed to \parsedate_main:nnn which indeed has three arguments.

Its duty is to assign a meaning to \dueday, \duemonth and \dueyear. The first and third are easy: just (globally) define them as parameterless macros.

For \duemonth we have to translate the month abbreviation into a number. The easiest method would be with \prg_case_str:nnn:

\cs_gset:Npx \duemonth
  {
   \prg_case_str:nnn { #2 }
    {
     { Jan } { 01 }
     { Feb } { 02 }
     ...
     { Dec } { 12 }
    }{}
   }

but the approach with property lists has an advantage: one might define an entire set of property lists for various languages and decide at runtime which one to rename to \g_parsedate_month_prop for being used in \parsedate_main:nnn, thus adapting easily the same macro to a multilingual environment.

So \duemonth is defined with \prop_get:Nn, in an "expanded definition": the entire

\prop_get:Nn \g_parsedate_months_prop { #2 }

is expanded because we use \cs_gset:Npx (the final x means "expansion", it's the good old \xdef). In the case #2 is May, the property list is looked at for the value corresponding to the property May, which is 05, so this becomes the replacement text of \duemonth.

Answer 3

You make the common mistake of placing the \StrBefore etc. macros in the macro definition. These operations are not expandable and can't be used as input to other string macros. Instead use the trailing optional argument to store the extracted string into a macro. This way you don't re-extract the string every time you are using the macros and you can use them in other string manipulation macros.

\documentclass{article}

\usepackage{xstring}

\begin{document}
\newcommand{\duedate}{06-May-2012}

\StrBefore{\duedate}{-}[\dueday]
\StrBetween[1,2]{\duedate}{-}{-}[\duemonthname]
\StrBehind[2]{\duedate}{-}[\dueyear]

Reconstruction: \dueday-\duemonthname-\dueyear.

%\renewcommand{\duemonthname}{May}   % This appears to solve the problem.

\IfStrEqCase{\duemonthname}{%
    {Jan}{\newcommand{\duemonth}{1}}%
    {Feb}{\newcommand{\duemonth}{2}}%
    {Mar}{\newcommand{\duemonth}{3}}%
    {Apr}{\newcommand{\duemonth}{4}}%
    {May}{\newcommand{\duemonth}{5}}%
    {Jun}{\newcommand{\duemonth}{6}}%
    {Jul}{\newcommand{\duemonth}{7}}%
    {Aug}{\newcommand{\duemonth}{8}}%
    {Sep}{\newcommand{\duemonth}{9}}%
    {Oct}{\newcommand{\duemonth}{10}}%
    {Nov}{\newcommand{\duemonth}{11}}%
    {Dec}{\newcommand{\duemonth}{12}}%
}

\duemonth   % This should output 5.

\end{document}