Things that make LaTeX definitely worth it, but that I have to look up every single time:

Compilers

if then
Multiple scripts in the same document XeLaTeX
Springer Nature template pdfLaTeX

Macros

Example macro with two arguments that creates a clean version of an unwieldy URL whilst retaining all the https://s and encoded mess in its clickable version:

\newcommand{\cleanurl}[2]{\href{#1}{\nolinkurl{#2}}}
% usage: \cleanurl{https://www.unwieldy.nope}{unwieldy.nope}

Resizing stuff to text/column width

Essentially

\resizebox{\textwidth}{!}{whatever}
% replace \textwidth with \columnwidth as needed
% prepend 0.N to \xxxwidth if the content should be resized to a fraction of the given width

but for tables, the box should be around the tabular and not wrap the entire table environment.

Adding a full-page segment to a two-column paper

\onecolumn

Preventing page breaks

Create an environment where page breaks are absolutely forbidden (the magic is from StackOverflow):

\newenvironment{nopagebreakswhatsoever}
    {\par\nobreak\vfil\penalty0\vfilneg
    \vtop\bgroup}
    {\par\xdef\tpd{\the\prevdepth}\egroup
    \prevdepth=\tpd}

Use it as:

\begin{nopagebreakswhatsoever}
    whatever should absolutely be all in the same page 
\end{nopagebreakswhatsoever}

Adjusting margins

\usepackage[margin=Xin]{geometry}

Multiline comments

(or something like that)

\iffalse
    whatever you don't wish LaTeX to compile
\fi

Glossed linguistic examples

Add

\usepackage{covington}

and optionally

\setglossoptions{
    fspreamble=\scshape\small, % style of the pre-gloss text
    fsi=\itshape,              % style of the main text
    fsii=\normalfont}          % style of the gloss

(there is also a bunch of other options, described in the awfully long incredibly thorough covington package docs; these are the ones I most commonly use)

Use as:

\begin{example}
    \label{utbildning}
    \digloss[preamble=Swedish]
        {En bra mening behöver inte vara en bra exempelmening .}
        {a good sentence need.PRES not be a good example.sentence .}
        {A good sentence is not necessarily a good example sentence. }
\end{example}

which will render as

The example above rendered in a PDF

\triglosses are also possible.

Ge`ez and Latin script in the same document

\usepackage{polyglossia}
\usepackage{microtype, newunicodechar}
\usepackage[sf, bf, big]{titlesec}
\defaultfontfeatures{Scale=MatchUppercase}
\setmainfont{Abyssinica SIL}[Scale=1]
\setsansfont{Libertinus Sans}
\setmainlanguage{english}
\setotherlanguage{amharic}
\newunicodechar{፡}{፡\ }
\newunicodechar{።}{\@{።} }
\newunicodechar{፣}{፣ }
\newunicodechar{፤}{፤ }
\newunicodechar{፥}{፥ }
\newunicodechar{፦}{፦ }
\newunicodechar{፧}{\@{፧} }
\newunicodechar{፨}{\@{፨} }
\newunicodechar{፠}{\@{፠} }
\newfontfamily{\amharicfont}{Abyssinica SIL}[
  Script=Ethiopic,
  Ligatures=Common,
  WordSpace = {0.1,30.0,1.0}]

Generating tables with Pandas

Nobody likes LaTeX tables (…right?) and a lot of people use GUI table generators like this. A much more powerful option that handles subcolumns and other annoyances flawlessly is to generate tables with Pandas. This is especially practical when the data to put in the table comes from a Python script.

Basic idea for basic (non-nested) tables: store your data as a DataFrame and run .to_latex() on it. A dictionary can be easily converted into a DataFrame:

df = pandas.DataFrame(a_dict) # dict -> DataFrame
df.to_latex()                 # DataFrame -> LaTeX string

Advanced idea for advanced (nested) tables: store your data as a DataFrame of DataFrames, do some mystery reindexing and run .to_latex() on the resulting data structure. Note that:

  • a DataFrame of DataFrames can be obtained by concatenating a dictionary of DataFrames:

      df_of_dfs = pandas.concat(dict_of_dfs, axis=1) # I don't remember what axis=1 does

  • mystery reindexing amounts to running .reindex():

      df_of_dfs.reindex(rows)

Working example that is not at all self-contained and I will generalize as soon as I have time:

def build_recap_table(specific_tables):
    rows = ["{} -- {}".format(lang.title(), subcorpus) for (lang,subcorpus) in langcorpora] + ["total"]
    cols = SPLITS + ["total"]
    subcols = ["original_essays", "reference_essays_1", "reference_essays_2"]#, "reference_essays_3", "reference_essays_4"]
    nested_data = {}
    for col in cols:
        col_data = {}
        for subcol in subcols:
            subcol_data = {}
            cross_sub_tot = 0
            for (lang, subc) in langcorpora:
                row = "{} -- {}".format(lang.title(), subc)
                if subcol in specific_tables[lang][subc]["texts"]:
                    n = specific_tables[lang][subc]["texts"][subcol][col]
                else:
                   n = 0
                subcol_data[row] = n
                cross_sub_tot += n
            subcol_data["total"] = cross_sub_tot
            col_data[subcol] = subcol_data
        nested_data[col] = pd.DataFrame(col_data)
    nested_df = pd.concat(nested_data, axis=1)
    nested_df.reindex(rows)
    return nested_df



recap_table = build_recap_table(specific_tables)
recap_table.to_latex()

This generates a table that looks more or less like this (modulo some text replacements e.g. to shorten the column names):

A large nested LaTeX table