What LaTeX is

Typesetting

Every text printed or read on a screen has been typeset: someone, or something, has decided the exact place where each letter falls. Typesetting is the sum of those decisions: which typeface and which size are used, where each line ends and how the space between its words is distributed to justify it, which words are hyphenated and where, when two letters merge into a ligature, how the distance between certain pairs of letters is adjusted (kerning), where each page breaks, at what height the notes go and how the headings are arranged. For centuries it was the trade of the compositor, who set the lines with metal type; today a program carries it out, but the decisions remain the same.

A good paragraph goes unnoticed: the eye runs along the lines without stumbling over uneven gaps or absurd hyphenations. Achieving this is not trivial, however, because each decision constrains the following ones: the place where the first line ends determines the word with which the second begins, and so on to the end of the paragraph.

Academic texts demand this care to a greater degree than others. They contain formulas, whose typesetting obeys its own rules (the space around an equals sign is not the space around a comma, and a subscript is reduced by a fixed proportion); they refer to themselves constantly, with numbered sections, theorems, figures and equations that are cited by their number; they accumulate footnotes and bibliographic references that must follow a uniform style; they often mix several languages and several scripts; and they tend to be long, so that a slip repeated over a hundred pages cannot be corrected by hand. If this is so, it is wise to entrust the typesetting to a system that handles it with explicit rules and lets the author attend to what is being said.

Markup and the word processor

There are two ways of asking a program to typeset a text. In a word processor, such as Microsoft Word or LibreOffice Writer, the author sees on the screen the page as it will be printed, and modifies it directly: selects a word and puts it in italics, drags a margin, chooses a larger size for a heading. This mode is usually described by the phrase what you see is what you get.

In a markup language, by contrast, the author writes a plain text file in which marks are interspersed that say what each thing is or how it should be treated: that this word is emphasized, that this line is the heading of a section, that this passage is a quotation. A program then reads that file and produces the typeset document. Web pages are written this way, in HTML; LaTeX is a markup language designed to produce high-quality printed documents.

To be fair, the comparison must acknowledge what each mode does best. The word processor offers an immediate response, requires almost no prior learning, is the tool known to most of the people one collaborates with, and its tracked changes are the usual means of revision in many publishing houses and offices. The markup language offers, for its part, plain text files that any program will be able to open forty years from now, that take up little space and whose versions can be compared line by line; an explicit structure, from which the tables of contents, the numbering and the cross-references follow by themselves; a uniformity that does not depend on the author's discipline; and a typesetting of formulas and paragraphs that word processors do not attain. Its cost is equally clear: the marks have to be learned, the result is not seen until compiling, and a typing error can stop the compilation.

TeX and LaTeX

Donald Knuth and TeX

Donald Knuth, professor of computer science at Stanford University, had been publishing the volumes of The Art of Computer Programming since 1968. Around 1977, on receiving the proofs of a new edition of the second volume, set with the photographic technique that had replaced metal type, he judged them far inferior to those of the first edition, above all in the formulas. Instead of resigning himself to it, he decided to write a typesetting program himself, one that would restore to mathematics books the quality of the best traditional typography, and whose output would be identical on any computer. The first version of TeX appeared in 1978; Knuth rewrote it completely at the beginning of the following decade and also designed the METAFONT program, with which he drew the Computer Modern typefaces.

The name comes from Greek: its letters are the capitals tau, epsilon and chi, the beginning of the word τέχνη, which means both art and technique, and that is why the last letter is not pronounced like an x. Knuth has declared TeX finished and admits only bug fixes to it: since version 3, the version number approaches π, and each fix adds one more digit to it. This stability is one of its virtues, because a document written decades ago compiles today with the same result.

TeX, however, is a low-level language: its primitive instructions speak of boxes, of stretchable spaces and of penalties for breaking a line at one place or another. For an author who only wants to write an article, programming every heading and every note in those terms would be a disproportionate burden. TeX therefore allows the definition of macros, new instructions that expand into more elementary ones, and almost all of its real use goes through collections of macros.

Leslie Lamport and LaTeX: content separated from form

In the early 1980s, Leslie Lamport, a computer scientist known above all for his work on distributed systems, wrote on top of TeX a collection of macros that he called LaTeX, whose first version circulated in 1984; in 1986 he published the manual LaTeX: A Document Preparation System. Since 1994, the current version, called LaTeX2e, has been maintained by a team of volunteers, the LaTeX Project.

The central idea of LaTeX is that the author marks the content and not the form. One does not write “this line in bold and in a larger size”, but “this is the heading of a section”, with \section{Results}; one does not write “this word in italics”, but “this word is emphasized”, with \emph{hypothesis}. What a section heading looks like, whether it carries a number and how far it stands from the text, is decided by the document class, declared at the beginning, and by the instructions the author adds in the preamble. It follows that the appearance of the whole document is changed in a single place, and that the same text can be typeset according to the rules of two different journals without touching a line of the body.

ExampleThe structure is declared and the form is derived

In the following source there is no indication of size, typeface or number: it only says what each line is. The numbering of the section and the subsection, the size and weight of their headings and the italics of the emphasized word are decided by the document class.

LaTeX code

\section{Results}The data confirm the initial \emph{hypothesis}.​\subsection{The sample}The sample comprises twelve texts by three authors.

Result

If a section is later inserted before this one, its number changes by itself from one to two, and the same happens with the subsection and with every reference that cites them.

The engines: pdfLaTeX, XeLaTeX and LuaLaTeX

LaTeX is a set of macros, not a program: it needs an engine, a program of the TeX family that executes them and produces the document. Three are in use today, and all three understand the same LaTeX; they differ in how they read the text and in which typefaces they can use.

• pdfLaTeX uses the pdfTeX engine, which Hàn Thế Thành developed to produce a PDF file directly. It is fast and very widespread, and it is the default engine of Overleaf. Its typefaces, however, are those of the classic TeX format, with at most 256 characters each, so it does not load the typefaces installed on the system directly, and writing Greek, Cyrillic or other scripts requires special encodings and packages; without them, a Greek letter stops the compilation with an error.
• XeLaTeX uses the XeTeX engine, created by Jonathan Kew, which reads the text in Unicode, the standard that assigns a number to every character of every script, and loads any OpenType typeface of the system through the fontspec package.
• LuaLaTeX uses the LuaTeX engine, which offers the same as XeTeX as regards Unicode and typefaces, and incorporates in addition the Lua programming language, with which packages can intervene in the typesetting. Current distributions run it with the HarfBuzz library built in, which shapes the scripts that join or combine their letters, such as Arabic or Devanagari.

This course uses LuaLaTeX. The reasons follow from the above: it writes directly in Unicode, so that a quotation from Plato in polytonic Greek, with its accents and breathings, is typed as it is; it uses the system's typefaces, including those required by a university or a journal; it typesets Latin, Greek and English in a single document, each with its own hyphenation rules; and it supports, with the appropriate packages, scripts that run from right to left, such as Arabic, or that do not separate their words with spaces, such as Chinese and Japanese. Therefore a single engine serves for all the documents of a course addressed at once to the sciences and to the humanities. Its only cost compared with pdfLaTeX is a somewhat slower compilation, which in documents of ordinary size is barely noticeable.

ExampleGreek and Latin in a document compiled with LuaLaTeX

The following source loads fontspec, chooses a typeface that contains Greek letters and declares, through babel, English as the main language and Ancient Greek and Latin as secondary languages. The Greek words are written directly with their characters.

LaTeX code

\documentclass{article}\usepackage{fontspec}\usepackage[latin, english]{babel}\babelprovide[import]{ancientgreek}\setmainfont{Linux Libertine O}​\begin{document}Aristotle opens the \emph{Metaphysics} with this sentence:\foreignlanguage{ancientgreek}{πάντες ἄνθρωποι τοῦ εἰδέναι ὀρέγονται φύσει.}The Latin tradition handed it down thus:\foreignlanguage{latin}{omnes homines natura scire desiderant.}\end{document}

Result

Marking a passage with \foreignlanguage does not change the appearance of the letters: it indicates what language the passage is in, so that its hyphenation rules and its conventions apply. With pdfLaTeX, this same source would not compile: fontspec requires XeLaTeX or LuaLaTeX.

The source file, compilation and the PDF

A LaTeX document is a plain text file with the extension .tex. It consists of two parts. The preamble runs from the first line, \documentclass{article}, which declares the document class, to \begin{document}; in it the packages are loaded, with \usepackage, and the decisions of form are fixed. The body runs from \begin{document} to \end{document} and contains the text. Instructions begin with a backslash, and what they receive is written between braces.

ExampleA minimal complete document

This source is a complete document: it compiles without adding anything to it, and it will be the starting point of the documents of the course.

LaTeX code

\documentclass{article}\usepackage{fontspec}\usepackage[english]{babel}​\begin{document}\section{Introduction}Typesetting decides where each letter falls.\end{document}

Result

The section receives the number one without its having been written, and its heading uses the typeface and the size set by the article class.

The working cycle has three moments: the source is written, it is compiled and the result is read. Compiling is running the engine on the file, in this course the command lualatex, which reads the source and writes a PDF file with the same name. Next to it remain auxiliary files, among them one with the extension .log, the log, where the engine records errors and warnings, and another with the extension .aux, where it stores the numbers of the sections and the pages. That is why a table of contents or a cross-reference only appears correctly after a second compilation: in the first, the engine notes where each thing ended up; in the second, it uses that information. In practice the command is not typed by hand: the editor runs it with a button. In this course the work is done with Texmaker, on one's own computer, or with Overleaf, in the browser; the next lesson prepares that environment.

The source is not typeset as it is written, and three rules explain almost all the initial surprises. First, several consecutive spaces count the same as one, and a line break counts as a space: the engine distributes the space on its own, and the space the author writes only indicates that a word ends there. Second, one or more blank lines end the paragraph and begin another. Third, the sign % starts a comment: the engine ignores everything that follows up to the end of the line, including the line break itself.

ExampleSpaces, paragraphs and comments

The following source is written carelessly on purpose. The result has two paragraphs with uniform spaces, and the note that follows the comment sign does not appear.

LaTeX code

Repeated      spaces     are reduced to one, and this line breakis also a space.​A blank line opens a new paragraph. % This note is not printed.

Result

If this is so, to print the percent sign it is not enough to write it: it has to be written preceded by a backslash, like this: \%. The same happens with the other characters that LaTeX reserves for its own functions, such as braces or the dollar sign.

LaTeX in the academic disciplines

In mathematics, in physics and in a good part of computer science, LaTeX is the norm: journals offer their own document classes so that authors can submit their articles in the journal's format, and arXiv, the repository where the papers of those disciplines circulate before publication, accepts LaTeX sources and compiles them itself. The reason is the typesetting of formulas, which no other system matches with so little effort, and the automatic numbering of theorems and equations. Economics, statistics, engineering and chemistry also use it to a large extent.

It would be a mistake, however, to believe that LaTeX is the exclusive concern of the exact sciences. In philosophy, in classical philology and in history, its virtues are different, and no lesser. Critical editions, which accompany an ancient text with an apparatus that records the variants of the manuscripts in notes keyed to line numbers, have specialized LaTeX packages, such as reledmac, which number the lines, arrange several levels of notes and keep their references when the text changes. Polytonic Greek and Latin are written and hyphenated correctly, as we have seen. Footnotes, which in a humanities essay may run into the hundreds, are numbered and placed on their page by themselves. And the bibliography, with BibLaTeX, is written only once as a database and cited in whatever style is requested, be it the author-date style of the social sciences, the numeric style of the sciences or the footnote style of the humanities, by changing a single option.

The limits of LaTeX

LaTeX, however, has real costs, and it is wise to know them before adopting it. Its learning curve is steep: the first weeks require frequent looking up of how to do what in a word processor is done with a click. Its error messages, inherited from TeX, are often cryptic and sometimes point to a line later than the one containing the error; an upcoming lesson of this unit teaches how to read them. The composition of freely designed pages, such as a poster or an illustrated magazine, in which each element is placed by hand, is more laborious than in a page layout program. And not every recipient accepts it: quite a few humanities journals, publishers and offices require a word processor file, and revise with its tracked changes.

From which a simple criterion follows. A short letter, minutes, a form or a text that another person will revise in a word processor is better written in a word processor. A long document, with structure, formulas, citations, notes, several languages or a long life ahead of it, which will have to be corrected and republished for years, is better written in LaTeX.

Solved problems

The following problems do not ask you to recall what has been said, but to apply it: the first requires predicting exactly what a source produces, the second, deciding which engine and which typefaces a document requires, and the third, separating content from form in a source written without that care.

Typesetting   decides   where   each letter falls.A good paragraph goes unnoticed.​​Fully 50 % of readersnever notice the difference. This wo%rd stays whole.

{\Large\bfseries 2. Results}\\[6pt]As Aristotle states in the \textit{Metaphysics}, the desireto know is \textit{natural}; the Greek term is \textit{φύσει}.\\\vspace{12pt}{\large\bfseries 2.1. The sample}\\

Online exercises

In these exercises you write the code in the box and the browser checks it instantly: it checks the braces, the environments and what each exercise asks for, without compiling it. Once the check is passed, it is worth copying the code into Texmaker and compiling it, because only the compiled result shows what is really printed.

\documentclass{article}\usepackage{fontspec}\usepackage[spanish]{babel}​\begin{document}\section{Introducción}La composición tipográfica decide dónde cae cada letra.\end{document}

\section{Marcado y procesador de textos}En un lenguaje de marcado, el autor dice \emph{qué es} cada cosa.​\subsection{El procesador de textos}El autor ve la página tal como se imprimirá.

Proposed problems

Each problem has a hint, which is best opened only after having tried it, and an answer, which appears after the hint has been opened.

I have been writing with \LaTeX for a year.I use \LaTeX{} and \TeX\ every day.