{\TeX\ documentation and \acro{CI} constraints for NumPy}
{Rohit Goswami}
{\TeX\ and \LaTeX\ have been used for offline documentation of software
packages and are supported by several auto-documenting systems including
\code{doxygen}, \code{sphinx} and \code{f2py}. Often, documentation
markup languages like Re\acro{ST} or Markdown will support a subset of
\TeX\ commands for various output formats (e.g., MathJax\slash KaTex for
\HTML).

With the rise of virtual machines for continuous integration, along with
a renewed focus on documenting code, the time taken for compiling
offline documentation (typically \PDF\ files) from \TeX\ sources has
become a bottleneck, and some projects (e.g., SciPy) have discontinued
the generation of \PDF\ files altogether. Alternatives have been
suggested, e.g. offline \HTML, web-\PDF{}s, etc. and will be covered
briefly.

In this talk, the main challenges and their mitigation strategies will
be discussed including Sphinx \LaTeX\ generation, styling, methods to
reduce documentation size and automated file-splitting with the aim of
preventing more projects from moving away from \TeX-based \PDF{}s. The
focus will be on the NumPy \TeX\ \acro{CI} documentation workflow, but
will be generally applicable to most Python projects.}