Closed brucemiller closed 9 months ago
I was just discussing with arXiv three of our existing successful Tikz conversions, and did a quick test with this PR - linking to the significantly older ar5iv editions:
So, focusing on the one pgfplots case that regressed: it indeed works with the latexml master branch. Isolating a minimal regressing example:
\documentclass{article}
\begin{filecontents}{coord.txt}
nc I_M_opt I_J I
1.000000 0.000000 0.000000 0.001190
2.000000 0.004133 0.003861 0.002870
3.000000 0.005510 0.007181 0.004562
\end{filecontents}
\usepackage{pgfplotstable}
\begin{document}
\begin{tikzpicture}
\begin{loglogaxis}[legend style={cells={align=left}}]
\pgfplotstableread{coord.txt}\mytable
\addplot[solid] table[x=nc,y=I] {\mytable};
\addlegendentry{legend}
\end{loglogaxis}
\end{tikzpicture}
\end{document}
Edit: can confirm this was fixed by 5298bf8 !
This PR provides a binding for
tikz-cd
, after fixing some infrastructure.The critical bug was that
ALIGN_STATE
which tracks{}
during scanning to enable or disable certain alignment tokens (&,\cr,...
) was being done too late (in Stomach rather than Gullet). It's easy to miss some corner cases, so some sandbox testing is warranted.This is a WiP since the resulting
svg
are displayed too low; there would seem to be one too many adjustments involvingpgf@pixminy
(?) , although other tikz pictures seem mostly correct.