There will be problems when using verb, font, xelatex and CJK character in the begin of cell at the same time.

[note286]

Not Working MWE

% xelatex
\documentclass{ctexart}
\usepackage{tabularray}
\begin{document}
\begin{tblr}{colspec={c},row{1}={font=\bfseries},verb=true}
测test\\
\verb|a\b\b|\\
\end{tblr}
\end{document}

(c:/texlive/2022/texmf-dist/tex/latex/tabularray/tabularray.sty) (./1.aux)
! Undefined control sequence.
l.1 \bfseries\selectfont测test
                             \empty
l.7 \end
        {tblr}
?

Working MWE

change xelatex to pdflatex

% pdflatex
\documentclass{ctexart}
\usepackage{tabularray}
\begin{document}
\begin{tblr}{colspec={c},row{1}={font=\bfseries},verb=true}
测test\\
\verb|a\b\b|\\
\end{tblr}
\end{document}

remove the CJK character in the begin of cell

% xelatex
\documentclass{ctexart}
\usepackage{tabularray}
\begin{document}
\begin{tblr}{colspec={c},row{1}={font=\bfseries},verb=true}
te测st测\\
\verb|a\b\b|\\
\end{tblr}
\end{document}

change font key to cmd key

% xelatex
\documentclass{ctexart}
\usepackage{tabularray}
\begin{document}
\begin{tblr}{colspec={c},row{1}={cmd=\bfseries},verb=true}
测test\\
\verb|a\b\b|\\
\end{tblr}
\end{document}

do not use verb

% xelatex
\documentclass{ctexart}
\usepackage{tabularray}
\begin{document}
\begin{tblr}{colspec={c},row{1}={font=\bfseries}}
测test\\
% \verb|a\b\b|\\
\end{tblr}
\end{document}

[muzimuzhi]

A re-tokenization problem. XeLaTeX (and also LuaLaTeX) treats (most) CJK characters the same as a-zA-Z, thus after rescanning the table content, XeTeX thinks \selectfont测 is a single control sequence, rather than a control sequence \selectfont followed by a character 测。

Some workarounds:

% !TeX TS-program = xelatex
\documentclass{ctexart}
\usepackage{tabularray}
\usepackage{codehigh}
\usepackage{fvextra}

\begin{document}
% "{}" trick
\begin{tblr}{colspec={c},row{1}={font=\bfseries},verb=true,hlines,vlines}
  {}测test\\
  \verb|a\b\b|\\
\end{tblr}

% \fakeverb from codehigh package, which shares the same author with tabularray
\begin{tblr}{colspec={c},row{1}={font=\bfseries},hlines,vlines}
  测test\\
  \fakeverb{a\\b\\b}\\
\end{tblr}

% \EscVerb from fvextra package
\begin{tblr}{colspec={c},row{1}={font=\bfseries},hlines,vlines}
  测test\\
  \EscVerb{a\\b\\b}\\
\end{tblr}
\end{document}

Similar though not identical re-tokenization issue

• & and % in tblr cause an error despite verb option #358

[lvjr]

This is a duplicate of issue #294. I am planning to mark verb option as obsolete in a future version and remove it eventually some years later.

[lvjr]

In fact \fakeverb is simpler than \EscVerb because you can write \fakeverb{a\b\b}.

[muzimuzhi]

btw, the l3regex-based implementations are so slow.

[Sophanatprime]

Another implement would be to use \tl_analysis_map_inline:n, which is pretty fast:

\tl_const:Nn \c__cus_verbatimize_macros_tl { \#\%\\\ \{\}\^ }
\NewDocumentCommand \FakeVerb { m }
  {
    \group_begin:
    \ttfamily \frenchspacing
    \tl_analysis_map_inline:nn {#1}
      { 
        \int_compare:nNnTF {##2} = { -\c_one_int }
          {
            \exp_args:NNo \tl_if_in:NnTF \c__cus_verbatimize_macros_tl {##1} 
                { \exp_after:wN \cs_to_str:N ##1 } { \exp_after:wN \token_to_str:N ##1 }
          }
          { \exp_after:wN \token_to_str:N ##1 }
      }
    \group_end:
  }

[lvjr]

Yes, l3regex is quite slow. And I have to improve the speeds of codehigh and tabularray packages by using less l3regex and less loops. As a first step I need to write some benchmark tests.

[lvjr]

Since version 2023A, verb option was obsolete. Please use \fakeverb command from codehigh package instead.