Are csvsimple and tabularray incompatible?

Question

Is there an incompatibility between csvsimple and the new package tabularray? The following MWE throws an error. (Compile with XeLaTeX or LuaLaTeX)

\documentclass{article}
\usepackage[ngerman,french]{babel}
\usepackage{fontspec}
\usepackage{tabularray} %needs package ninecolors.sty
\usepackage{booktabs}
\usepackage{csvsimple}
\begin{filecontents}{data.csv}
frA,frB,deA
un cogne,arg.,ein Bulle
morfler,arg.,eins abkriegen
\end{filecontents}
\begin{document}
This is what I'd like to obtain, with environment \verb|tblr|:
\noindent
\begin{tblr}{colspec={X[2,l]X[4,l]},
width=0.5\linewidth,
hlines,
}
\textit{un cogne} (arg.)& \textit{un policier ou un gendarme}\ 
\textit{morfler} (arg.)& {\selectlanguage{ngerman}\textit{eins abkriegen}}\
\end{tblr}
\bigskip
A normal tabular seems to work (but I'd miss the fonctions of tabularray...)
(not only the X column type, also the alignment facilities):
% \csvreader
\noindent
\begin{tabular}{p{4cm}p{8cm}}\toprule
\csvreader[
  late after line=\\midrule,
  late after last line=\\bottomrule,
  head to column names,
  ]{data.csv}{}{%
  \textit{\frA} (\frB) & \textit{\deA}
}
\end{tabular}
\bigskip
But this throws an error:
\noindent
\begin{tblr}{
  colspec={X[2,l]X[4,l]},
  width=0.5\linewidth,
  hlines,
}
\csvreader[
  late after line=\\midrule,
  late after last line=\\bottomrule,
  head to column names,
  ]{data.csv}{}{%
  \textit{\frA} (\frB) & \textit{\deA}
}
\end{tblr}
\end{document}

I'm not surprised at all, tabularray creates tabulars in a quite different way, if you want cvsimple to support it, you should make a feature request. — Ulrike Fischer, Jul 02 '21 at 16:18
A known limitation of tabularray, see https://github.com/lvjr/tabularray/issues/23. — muzimuzhi Z, Jul 02 '21 at 20:41

muzimuzhi Z · Accepted Answer · 2021-07-07T02:24:09.377

Update:

With csvsimple v2.1.0 (2021/07/06), csvsimple-l3 gets compatible with tabularray.

You should use \usepackage[l3]{csvsimple} or \usepackage{csvsimple-l3} to load csvsimple-l3, rather than the superseded csvsimple-legacy.

\begin{filecontents*}{data.csv}
frA,frB,deA
un cogne,arg.,ein Bulle
morfler,arg.,eins abkriegen
\end{filecontents*}
\documentclass{article}
\usepackage[ngerman,french]{babel}
\usepackage{fontspec}
\usepackage{tabularray} %needs package ninecolors.sty
\usepackage{booktabs}
\usepackage[l3]{csvsimple} % or "\usepackage{csvsimple-l3}"
\begin{document}
\subsection*{Desired output}
This is what I'd like to obtain, with environment \verb|tblr|:
\begin{tblr}{
  colspec={X[2,l]X[4,l]},
  width=0.5\linewidth,
  hlines,
}
  \textit{un cogne} (arg.) & \textit{un policier ou un gendarme}\ 
  \textit{morfler} (arg.)  & {\selectlanguage{ngerman}\textit{eins abkriegen}}\
\end{tblr}
\subsection*{With \texttt{tabular} env}
A normal tabular seems to work (but I'd miss the functions of tabularray...)
(not only the X column type, also the alignment facilities):
\begin{tabular}{p{4cm}p{8cm}}\toprule
\csvreader[
  late after line=\\midrule,
  late after last line=\\bottomrule,
  head to column names,
  ]{data.csv}{}{%
  \textit{\frA} (\frB) & \textit{\deA}
}
\end{tabular}
\subsection*{With \texttt{tblr} env}
% required by tabularray
\NewTableCommand\toprule{\hline[0.08em]}
\NewTableCommand\midrule{\hline[0.05em]}
\NewTableCommand\bottomrule{\hline[0.08em]}
\csvreader[
  tabularray={
      colspec={X[2,l]X[4,l]},
      width=0.5\linewidth,
%      hlines
  },
  table head=\toprule,
  late after line=\\midrule,
  late after last line=\\bottomrule,
  head to column names,
]{data.csv}{}{%
  \noexpand\textit{\frA} (\frB) & \noexpand\textit{\deA}
}
\end{document}

Output is the same as below.

Original:

Here is a first try on the csvsimple side. The idea is to make \csvloop (in the taste of csvsimple-legacy) firstly store the whole tabular staff in a macro (here \csv@output), instead of directly printing them. Then user can use \csvoutput to print the table.

Restricted by how tabularray is implemented, every \hline-like staff, \\, and & should be directly seen in \csv@output.

\begin{filecontents*}{data.csv}
frA,frB,deA
un cogne,arg.,ein Bulle
morfler,arg.,eins abkriegen
\end{filecontents*}
\documentclass{article}
\usepackage[ngerman,french]{babel}
\usepackage{fontspec}
\usepackage{tabularray} %needs package ninecolors.sty
\usepackage{booktabs}
\usepackage{csvsimple}
%%%
%%% patch for csvsimple-lagency
%%%
\usepackage{xpatch}
\makeatletter
\newif\ifcsv@storeonly
\csvset{
  % new keys
  table env/.store in=\csv@tableenv,
  store only/.is if=csv@storeonly,
  % redefinition
  tabular/.style={
    @table={\csv@pretable\begin{\csv@tableenv}{#1}\csv@tablehead}{\csv@tablefoot\end{\csv@tableenv}\csv@posttable},
    late after line=\
  },
  % init new keys
  table env=tabular,
  store only=false
}
% store-or-print patch for \csvloop
\long\def\csvloop#1{%
  % reset
  \global\let\csv@output@empty
  \global\let@endloophook\csv@empty%
  \global\let\csvlinetotablerow\csv@assemble@csvlinetotablerow%
  % options
  \csvset{default,every csv,#1}%
  \csv@preprocss%
  \csv@set@catcodes%
  \csv@print@or@ostore\csv@prereading%
  \csv@print@or@ostore\csv@table@begin%
  \setcounter{csvinputline}{0}%
  % start reading
  \openin\csv@file=\csv@input@filename\relax%
  \ifeof\csv@file%
    \csv@error{File '\csv@input@filename' not existent, not readable, or empty!}{}%
  \else%
    % the head line
    \csv@opt@processheadline%
  \fi%
  %
  \setcounter{csvrow}{0}%
  \gdef\csv@do@preline{%
    \csv@print@or@ostore\csv@prefirstline%
    \gdef\csv@do@preline{\csv@print@or@ostore\csv@preline}%
  }%
  \gdef\csv@do@postline{%
    \csv@print@or@ostore\csv@postfirstline%
    \gdef\csv@do@postline{\csv@print@or@ostore\csv@postline}%
  }%
  \gdef\csv@do@@latepostline{%
    \csv@print@or@ostore\csv@latepostfirstline%
    \gdef\csv@do@latepostline{\csv@print@or@ostore\csv@latepostline}%
  }%
  \gdef\csv@do@latepostline{%
    \csv@print@or@ostore\csv@lateposthead%
    \global\let\csv@do@latepostline=\csv@do@@latepostline%
  }%
  % command for the reading loop
  \gdef\csv@iterate{%
    \let\csv@usage=\csv@empty%
    \csvreadnext%
    \ifeof\csv@file%
      \global\let\csv@next=\csv@empty%
    \else%
      \global\let\csv@next=\csv@iterate%
      \ifx\csv@par\csvline\relax%
        \else%
        \csv@escanline{\csvline}%
        % check and decide
        \csv@opt@checkcolumncount%
      \fi%
    \fi%
    % do or do not
    \csv@usage%
    \csv@next}%
  \ifeof\csv@file%
    \global\let\csv@next=\csv@empty%
  \else%
    \global\let\csv@next=\csv@iterate%
  \fi%
  \csv@next%
  \closein\csv@file%
  @endloophook%
  \csv@print@or@ostore\csv@latepostlastline%
  \csv@print@or@ostore\csv@table@end%
  \csv@print@or@ostore\csv@postreading%
  \csv@reset@catcodes%
}
\newcommand\csvoutput{%
  % expose tabular-like env name
  \xpatchcmd@self\csv@output\csv@tableenv
  \xpatchcmd@self\csv@output\csv@tableenv
  % expose hrule-like commands
  \xpatchcmd@self\csv@output\csv@tablehead
  \xpatchcmd@self\csv@output\csv@tablefoot
  % output
  \csv@output
}
% helper, \xpatchcmd#1{#2}{<one-step expansion of #2>}{}{\fail}
\newcommand\xpatchcmd@self[2]{%
  \expanded{\unexpanded{\xpatchcmd#1{#2}}{\unexpanded\expandafter{#2}}}{}{\fail}%
}
% store-or-print patch for \csv@do@linecommand
\xpatchcmd\csv@do@linecommand
  {\csv@body}
  {\csv@print@or@estore\csv@body}
  {}{\fail}
% expand once
\def\csv@print@or@ostore#1{\csv@print@or@store@meta
  {\expandafter\g@addto@macro\expandafter\csv@output\expandafter{#1}}
  {#1}}
% fully expansion
\def\csv@print@or@estore#1{\csv@print@or@store@meta
  {\expanded{\noexpand\g@addto@macro\noexpand\csv@output{#1}}}
  {#1}}
\def\csv@print@or@store@meta{%
  \ifcsv@storeonly
    \expandafter@firstoftwo
  \else
    \expandafter@secondoftwo
  \fi
}
\makeatother
%%%
%%% end of patch
%%%
\begin{document}
\subsection*{Desired output}
This is what I'd like to obtain, with environment \verb|tblr|:
\begin{tblr}{
  colspec={X[2,l]X[4,l]},
  width=0.5\linewidth,
  hlines,
}
  \textit{un cogne} (arg.) & \textit{un policier ou un gendarme}\ 
  \textit{morfler} (arg.)  & {\selectlanguage{ngerman}\textit{eins abkriegen}}\
\end{tblr}
\subsection*{With \texttt{tabular} env}
A normal tabular seems to work (but I'd miss the functions of tabularray...)
(not only the X column type, also the alignment facilities):
\begin{tabular}{p{4cm}p{8cm}}\toprule
\csvreader[
  late after line=\\midrule,
  late after last line=\\bottomrule,
  head to column names,
  ]{data.csv}{}{%
  \textit{\frA} (\frB) & \textit{\deA}
}
\end{tabular}
\subsection*{With \texttt{tblr} env}
\csvreader[
  table env=tblr,
  tabular={
      colspec={X[2,l]X[4,l]},
      width=0.5\linewidth,
%      hlines
  },
  store only,
  table head=\toprule,
  late after line=\\midrule,
  late after last line=\\bottomrule,
  head to column names,
]{data.csv}{}{%
  \noexpand\textit{\frA} (\frB) & \noexpand\textit{\deA}
}
\NewTableCommand\toprule{\hline[0.08em]}
\NewTableCommand\midrule{\hline[0.05em]}
\NewTableCommand\bottomrule{\hline[0.08em]}
\csvoutput
\end{document}

I'm impressed, and happy that I don't have to understand this all :) Who should get the feature request, the author of tabularray or the author of csvsimple? — benjamin, Jul 05 '21 at 08:27
Sorry I forgot to mention that a feature request has been opened in https://github.com/T-F-S/csvsimple/issues/13 . — muzimuzhi Z, Jul 05 '21 at 16:55
@benjamin Hi, csvsimple has updated to add support for tabularray, and my answer is updated as well. — muzimuzhi Z, Jul 06 '21 at 23:23

Are csvsimple and tabularray incompatible?

1 Answers1

Linked