I recently finished my Ph.D thesis (yay!) and accumulated a few tips related to formatting technical writing in Latex.
Many people are familiar with the \hyphenation command, which tells Latex where it may hyphenate words it does not already know about to prevent lines from being too long.
Latex will not know about many technical words, or function names that are not "plain english."
In English writing, line breaks should occur at syllable boundaries only.
\hyphenation{cuda-Memcpy-Peer-Async bi-di-rec-tion-al}
You can also allow breaks at underscores, which is convenient for certain programming languages or libraries.
\renewcommand\_{\textunderscore\allowbreak}
You may then want to prevent line breaks in some places.
Use \mbox for that.
\mbox{\_\_device\_\_}
Latex will only break already-hypenated words at the hyphen position.
You can add more optional breaks with {\-}
A super-hypenated-latex-confusing compound adjective.
A su{\-}per-hy{\-}phen{\-}at{\-}ed...
The lstlisting environment is used to add and format code in Latex documents.
You can use minipage to prevent short lstlistings from being broken across pages.
You can also use the \noindent command to prevent the minipage from being indented if it starts a new paragraph
The
\noindent
\begin{minipage}{\linewidth}
\begin{lstlisting}
\end{lstlisting}
Use the siunitx package to automatically format numbers with SI units.
([binary-units] may not be necessary depending on your version.)
\usepackage[binary-units]{siunitx}
\SI{512}{\byte}
In newer versions of the siunitx package, \SI may be replaced with \qty.
Algorithm statements are often written in math mode, which treats consecutive letters as the product of variables and can lead to strange kerning.
You can fix with \mathit{}
\State $\mathit{word} \gets ...$
Latex can make it hard to manually format digit grouping and separators in numbers.
The number package makes this easy.
\usepackage[group-separator={,}]{siunitx}
\num{242000}
You can use aspell with the Latex filter (to reduce false positives).
The -t flag puts it in Latex mode.
aspell -t -c main.tex
Write one latex sentence on each line, so version control diffs are easier to follow. It feels a little unnatural at first.
This is one sentence.
This is another sentence.
Arxiv does not want a raw PDF, annoyingly.
You can defined a makefile target that will create a zip for you to upload.
You'll need to tweak this to get all the files uploaded you need.
Arxiv usually wants the pre-processed bib file that ends in bbl, so you need to run bibtex.
arxiv: main.tex main.bib ${FIGS}
pdflatex ${PAPER}.tex
bibtex ${PAPER}.aux
rm -f upload.zip
zip -r upload.zip main.tex main.bbl figures acmart.cls ACM-Reference*
Use resizebox for tables that are too wide or too tall (which is almost always the case, right???)
To match the text width:
\label{tab:related}
\resizebox{\textwidth}{!}{%
\begin{tabular}{...}
\end{tabular}
}%resizebox
\end{table*}
To match the text height:
\label{tab:related}
\resizebox{!}{\textheight}{%
\begin{tabular}{...}
\end{tabular}
}%resizebox
\end{table*}
I think there is too much wasted space between figures and captions in some templates.
You may be able to tweak this yourself with the caption package.
\usepackage{caption}
\captionsetup{skip=1pt}
You may wish to repeatedly apply formatting to a particular word.
You can define your own command for that.
The xpsace command will try to be smart about whether to put a space after the word.
\newcommand{\StreamData}{\textit{StreamData}\xspace}
Now \StreamData is in italics.
% notice the absence of \xspace
\newcommand{\DenseData}{\textit{DenseData}}
When I use \DenseData{} I may need to put the braces afterwords to get a space.
Use the pifont package, and then the \ding command
\usepackage{pifont}
This: \ding{202} will make a black circle with a white "1" in it.
You can then draw a matching object in your figure.
\ding{203} will make a 2, \ding{204} will make a 3, etc.