Unexpected vertical align change for some cells

[BenjaminGalliot]

Here is the MWE:

\documentclass{book}

\usepackage{tabularray}
\usepackage{xcolor}

\begin{document}

\begin{longtblr}{
  colspec = {X[c,m]X[c,m]X[c,m]},
  cell{2-3}{2-3} = {m,blue!50!white} % change here (m to b/t)
}
  \hline
  head 1 & head 2 & head 3 \\
  \hline
  {a \\ b} & {c \\ d} & e \\
  \hline
  a & {b \\ c \\ d} & e \\
  \hline
  a & {b \\ c \\ d} & {e \\ \phantom{ } \\ \phantom{ }} \\
  \hline
\end{longtblr}

% Same but without content in first column.
\begin{longtblr}{
  colspec = {X[c,m]X[c,m]X[c,m]},
  cell{2-3}{2-3} = {m,blue!50!white} % change here (m to b/t)
}
  \hline
  head 1 & head 2 & head 3 \\
  \hline
  & {c \\ d} & e \\
  \hline
  & {b \\ c \\ d} & e \\
  \hline
  & {b \\ c \\ d} & {e \\ \phantom{ } \\ \phantom{ }} \\
  \hline
\end{longtblr}

\end{document}

I don't know if I am misusing the functionality, but I would simply like to change the vertical alignment (among other things) for a range of cells, yet other cells seem to interfere:

• the first column is also affected when it shouldn't be;
• margins appear in colored cells due to the presence of the first column (removing its contents will change these margins in colored cells, as seen in second table).

By the way, I'm using \phantom{ } to align certain contents between cells, is it on purpose that \\ alone isn't enough?

Thanks for your help!

[muzimuzhi]

It's expected behavior, see #434 for explanation and an attempt adding valign=M.

[muzimuzhi]

By the way, I'm using \phantom{ } to align certain contents between cells, is it on purpose that \\ alone isn't enough?

Sort of yes. Anything that can start an hmode suffices, for example \leavevmode. This article on Overleaf gives some discussion about the error "There's no line here to end".

For now I don't know if automatically adding \mode_leave_vertical: at the beginning of every line of a cell is a good idea or not.

[BenjaminGalliot]

It's expected behavior, see #434 for explanation and an attempt adding valign=M.

Thanks, I added your specific code in my preamble (until it is pushed on a new version) and it works!