This guide is for developers who want to contribute code to the project, or who need to modify or debug it.
If you simply want to use MathLive in your project, see the Usage Guide.
- Getting Started: Setting up Your Development Environment
- Deploy / Publish
- Code Structure
- Language and Coding Style
- Bundling
- Browser Support
- Accessibility
- Architecture
The project uses NPM scripts for its
build system. The package.json file and the scripts/ directory contain the
definitions of the build scripts.
To get started developing:
- Install Node.js on your dev machine (this will also
install
npm). The LTS version is recommended. - In your shell, type:
$ git clone https://github.com/arnog/mathlive
$ cd mathlive
$ npm ciThe npm ci command installs in the mathlive/node_modules directory all the
Node modules necessary to build and test the MathLive SDK.
Depending on your system setup, you may need to run as admin, in which case use
sudo npm ci or equivalent.
If your version of npm or node is out of date, you will be prompted to
update.
Once the installation is successful, you can use the following commands:
# Make a local build and watch source file changes and rebuid.
# Run a local server (http://localhost:8080/examples/) to view
# the examples and do some simple debugging
$ npm start
# Make a local development build
# 1. Compile the `.css/.less` file to `build/*.css`
# 2. Compile the TypeScript source files and create
# a bundle with sourcemap in the `dist/` directory
$ npm run build
# Run test suite
$ npm test
# Calculate the code coverage and output to `coverage/`
$ npm test coverage
# Create a production build to `dist/`.
# The `dist/` folder will contain the `.js`, `.css` and
# font files necessary to use MathLive.
$ npm run build production
During development, keep npm start running. A build will be triggered when a
source file is updated. Note however that changes to .less files do no trigger
a rebuild. You'll need to stop and restart the npm start command.
Run the test suite with npm test and linter npm run lint to make sure your
changes are ready to submit, then push a PR to the main branch
After you push your changes to main, a Travis continuous integration task will
run on a remote server to make sure the build can be reproduced in a clean
environment.
If you are getting build errors after updating your repo, your node-modules/
directory may need to be updated. Run:
$ npm ciNote: only the owner of the project should deploy to NPM.
You will need a GH_PUBLIC_TOKEN env variable set up with a "Personal Access
Token" to publish the GitHub Release.
Go to GitHub > [user] > Settings > Developer Settings > Personal Access Tokens to create one and set it as a shell env global variable.
To publish a new version of the SDK:
# Increase the version number of the SDK
# and publish a GitHub release
$ npm version [major | minor | patch]
# Publish to NPM
$ npm publishThese commands will:
- Increment the version number of the SDK and create a corresponding git tag
- Update the CHANGELOG with the current version number
- Publish a GitHub release
- Publish to NPM
(See scripts/version.sh, which is invoked by npm version)
Note on versioning: Use the semver convention for versions:
npm version patch: bug fixes and other minor changes. Last number of the version is incremented, e.g.1.2.41→1.2.42npm version minor: new features which don't break existing features. Middle number of the version is incremented, e.g.1.2.42→1.3.0npm version major: changes which break backward compatibility of the API. Increment the first number, e.g.1.3.56→2.0.0
The MathLive SDK consists of the following key directories:
-
css/the stylesheets and fonts -
sounds/the default sound files -
src/core,src/core-atoms,src/core-definitionsthe core JavaScript code needed to render math. This module depends on thecss/module. -
src/editor,src/editor-mathfield,src/editor-modelthe JavaScript code needed for the editor. This module depends on thesrc/core*modules. -
src/addonssome optional modules that provide additional functionality -
the
dist/directory contains executable build artifacts. If a file named "DEVELOPMENT_BUILD" is present in the directory, the content of the directory is suitable only for development purposes, not for production. This means the files have not been minified and compiled, and include additional .map files referencing back to the source file for ease of debugging.
The content of the dist/ directory are entirely generated as part of the build
process. No other directory should contain intermediated files generated as part
of the build process.
MathLive is written in TypeScript.
The project uses the prettier tool to enforce a consistent formatting style
and naming conventions. The tool will be run automatically before commits. You
can also run it manually using npm run lint.
The code base attempts to follow these general guidelines:
- Consistency All code in the codebase should look as if it had been written by a single person. Don't write code for yourself, but for the many people who will read it later.
- Clarity before performance Write code that is easy to read, and avoid obscure constructs that may obfuscate the code to improve performance. For example, RegEx are crazy fast in all modern browsers, and trying to roll out your own pattern matching will result in more code and less performance. If you think something could be made faster, use http://jsben.ch/ to try out options in various browsers and compare the results. You might be surprised.
- Follow Postel's Law, the Robustness Principle "Be conservative in what you do, be liberal in what you accept from others". For example, functions that are invoked internally do not need to check that the input parameters are valid. However, public APIs should check the validity of parameters, and behave reasonably when they aren't.
The TypeScript code is compiled to JavaScript by the tsc compiler. When doing
a production build, the JavaScript is further minimized with terser, then
bundled into a single file with rollup. The CSS files are minimized with
postcss.
MathLive is designed for the modern web. Supporting older browsers complicates the effort involved in building new features, but it is also an insecure practice that should not be encouraged.
In this context, modern means the latest two releases of Chrome, Edge, Safari and Firefox. Both desktop and mobile are supported.
Note that the HTML quirks mode is not supported. This means that the host page
should use the strict mode, indicated by a <!doctype html> directive at the
top of the page.
Note that the HTML page should use the UTF-8 encoding. Use a server header or a
<meta charset="UTF-8"> tag in the page if necessary.
MathLive renders math using HTML and CSS. Digits, letters and math symbols are
displayed in <span> tags with the necessary CSS styling to display them in the
right place. In addition, rules (lines) such as the fraction line, are rendered
using CSS borders. In a few rare cases, SVG is used to render some decorations,
such as the annotations of the \enclose command.
The rendered math is not purely graphical, and as such can be accessed by screen readers.
However, in addition to the "visual" HTML+CSS representation that MathLive outputs, it can also generate alternate renditions, including:
- LaTeX: a string of LaTeX code equivalent to the formula.
- Spoken Text: a text representation of the formula as someone would speak
it, for example:
f(x) = x^2→ "f of x equals x squared" - Annotated Spoken Text: as above, but in addition prosody hints are inserted for a more natural rendition by text to speech systems (breathing pauses, variation in pitch, etc...).
Those alternate renditions are rendered as an ARIA-label, or as an element that is not visually rendered, but visible to screen readers.
Although MathLive works with screen readers, since math is its own language MathLive has its own built-in text to speech renderer. With the speech interface it is possible to:
- read the current group (numerator or subscript, for example)
- Mac:
Ctrl + Command + Down - Windows/Linux/ChromeOS:
Ctrl + Alt + Down
- Mac:
- read what's before or after the selection
- Mac:
Ctrl + Command + Left/Right - Windows/Linux/ChromeOS:
Ctrl + Alt + Left/Right
- Mac:
- read the parent of the current group
- Mac:
Ctrl + Command + Up - Windows/Linux/ChromeOS:
Ctrl + Alt + Up
- Mac:
- read the current selection
- Mac:
Ctrl + Command + Shift + Down - Windows/Linux/ChromeOS:
Ctrl + Alt + Shift + Down
- Mac:
With these convenient keyboard shortcuts, it is possible to aurally navigate and understand even complex formulas.
MathLive supports multiple modalities for input: in addition to pointer devices (mouse, trackpad, touch screen), MathLive has an extensive set of keyboard shortcuts that allow navigation and editing of the most complex formulas. Every operation is possible without the use of a pointing device.
Conversely, it is possible to enter commands and complex mathematical symbols using only a pointing device: the command bar can be invoked by tapping a round toggle button displayed to the right of the formula. The command bar offers large buttons that act as a virtual keyboard, but offer contextual operations depending on the current selection, and the content around it. Those buttons are easy to use on touch screens and for users of alternative pointing devices.
The core of MathLive is a rendering engine that generates HTML (and SVG) markup. This engine uses the TeX layout algorithms because of their quality. Given the same input, MathLive will render pixel for pixel (or very close to it) what TeX would have rendered.
To do so, it makes use of a web version of the fonts used by TeX and which are
included in the dist/fonts/ directory.
Although the rendering engine follows the TeX algorithms, MathLive also has an
in-memory data structure to represent a math expression while it is being edited
(a tree of Atoms).
MathLive is divided into two main components:
- Core: handles rendering of Latex to HTML markup
- Editor: handles the user interaction with the formula, using Core for the rendering.
Core takes a Latex string as input. A lexer converts the string into Token[]
which are then passed on to a parser. The parser uses the information from
modes-* to parse the tokens depending on the current mode (text, math,
etc...). The Latex commands are defined in definitions-*, and used by the
parser to properly interpret the commands it encounter and turn them into
Atom[].
An Atom is an elementary layout unit, for example a genfrac Atom can layout
a "generalized fraction", that is something with a numerator and denominator,
optionally a bar separating the two, and some optionally some opening and
closing fences. It is used by the \frac command, but also \choose, \pdiff
and others.
The Atom[] are then turned into Span[] which are virtual markup elements.
Eventually, the Span[] get rendered into HTML/SVG markup.
The lexer converts a string of TeX code into tokens that can be digested by the parser.
The parser turns a stream of tokens generated by the lexer into math atoms. Those atoms then can be rendered into spans, or back into LaTeX or into spoken text.
A span is an virtal DOM node that is used to represent an element displayed in a web page: a symbol such as x or =, an open brace, a line separating the numerator and denominator of a fraction, etc...
The basic layout strategy is to calculate the vertical placement of the spans and position them accordingly, while letting the HTML rendering engine position and display the horizontal items. When horizontal adjustments need to be made, such as additional space between items the CSS margin are adjusted.
Spans can be rendered to HTML markup with Span.toMarkup() before being
displayed on the page.
An atom is an object representing a mathematical symbol, for example x, 1, a
fraction, a delimiter, etc...
There are several different classes of Atom (subclass of the base Atom class).
Each class represents different layout algorithm (different ways of generating
spans in their render() method) as well as different ways to generate Latex to
represent the atom (in their toLatex() method)
It can be of one of the following classes:
- Atom: the base class is used for the simplest symbol, e.g.
x,1,\alpha - AccentAtom: a diacritic mark above a symbol
- ArrayAtom: "environments" in TeX parlance, a matrix, vector or other array-like structure
- BoxAtom: a decoration around a "nucleus", including a color background, lines, etc...
- DelimAtom and SizedDelimAtom delimiters and extensible delimiters
- see
src/core-atomsfor more.
The mathfield is the object handling the user interaction and driving the
rendering of the formula into the document.
It makes use of several subcomponents (virtual-keyboard, undo, etc...) to
handle specific aspects of the user interaction. It makes changes to the formula
by issuing basic editing commands such as insert(), delete() and modifying
the selection to the model.
The model keep track of the state of the formula, including its content (a
tree of Atom) and the selection and interacts with the core to turn the Atom
into Span and into markup.
The Model class encapsulates the operations that can be done to a tree of
atoms, including adding and removing content and keeping track of and modifying
an insertion point and selection.
The Mathfield class is a user interface element that captures the keyboard and
pointing device events, and presents an appropriate user experience.
It uses a Model to manipulate the in-memory representation of the math
expression being edited.

