mathml_mathjax_js

Handle browser-dependence on MathML with MathJax

This package [mathml_mathjax_js](https://github.com/masasakano/mathml_mathjax_js) provides MathML-MathJax related JavaScript codes, specifically, small codes to automatically determine to load MathJax or not.

The original repository is hosted on Github: https://github.com/masasakano/mathml_mathjax_js
The JavaScript and sample HTML files are ported to https://masasakano.github.io/mathml_mathjax_js/ (Index)

Introduction

As of 2022 summer, the web-browser support for MathML is limited; in short only Firefox and Safari provide a reasonable support, but not Chrome, Opera, or Edge (or Internet Explorer) —— see Browser compatibility on the Mozilla website. So, if you write a HTML containing MathML, many people are unable to read it at least in an originally intended form.

A popular and probably the most powerful fallback to interpret MathML by browsers is MathJax, a JavaScript framework. However, MathJax should not be loaded for browsers that can natively interpret MathML. This package offers a JavaScript code to selectively load MathJax, that is, MathJax is loaded only when the browser is a visual browser and does not interpret MathML.

Predecessor of This Package

In mozilla.org, specifically in Section “Fallback”, there is a guideline of how to deal with the browser-dependence. Unfortunately, the document seems outdated as of 2022 August.

In short, the JavaScripts described in the Mozilla site no longer work, as they fail to load MathJax JavaScript. Indeed, some JavaScripts contain hard-coded URIs for MathJax JavaScript, the hosting URIs of which are outdated, presumably because of the shutdown of the cdn server on mathjax.org) in 2017.

The couple of pieces of the main JavaScript code introduced in the Mozilla site were obviously developed (or at least published) by Frédéric Wang. However, I have failed to locate the original repository for the set of minified code, including mpadded-min.js. In fact, the original repository of Mathzilla, a collection of MathML-related code developed by the same author, comes with a warning now (as of 2022 August):

Warning: This directory contains some obsolete or experimental content and is only kept for historical purpose.

(Some or all of) The JavaScripts in this package are based on the minified code introduced in the above-mentioned Mozilla document. The major point of modification is that the URIs for MathJax are updated and work at the time of writing. I note that the version number has to be explicitly given in this package, unlike the original minified code, which simply specifies the latest version, because the new hosting site https://cdnjs.cloudflare.com/ gives no option to specify the latest version but requires you to specify the version explicitly.

This package is licensed under the MPL 2.

Background

Historically, it is not easy to write mathematical expressions in a presentable way on websites. Brian Hayes’s article Writing Math on the Web published on American Scientist in 2009 summarises the situation at the time, for example.

MathML is probably the most natural way for the purpose, being compatible with HTML. It is a text form of mathematical expressions and hence is most easily processed by a computer, meaning perhaps optimum for indexing (for searching etc) and processing for non-visual environments, such as auditorial browsers. However, MathML does have drawbacks. In short, MathML is not human-friendly. It is hard to read and even harder to write, requiring a lot of effort to write even the smallest piece of mathematical expression. I guess most MathML documents on the Internet were generated, machine-converted from a different format. Also, the format is fairly lengthy and so needs a slightly more amount of data storage and data transfer and computer power to interpret, although the former is nothing compared with binary such as images and the latter is probably nothing, either, in the modern computer standard.

There are some fallback mechanisms for MathML so browsers that cannot natively interpret MathML can display it nicely. Also, there are alternative ways to present maths on websites. Here is a brief summary (though incomplete).

  1. Fallback for MathML
    1. mathml.css
      • Pros: Relatively simple CSS to interpret MathML to be provided with the HTML+MathML.
      • Cons: This works only for a limited (aka most basic) subset of MathML. This should not be provided for browsers that can interpret MathML. Therefore, the website must judge the user’s browser or instruct users’ choices of the browser (for example, separate pages for the same contents should be provided, depending on the browser).
    2. MathJax
      • Pros: Most powerful JavaScript fallback. It works in all major visual browsers. It can interpret a vast majority of MathML, though still incomplete (as of Version 3.2.2, the latest stable version in 2022 August), e.g., it does not interpret the rowspan attribute for the MathML tag mtd in mtable.
      • Cons: Like mathml.css, this should not be provided for browsers that can interpret MathML, although in most cases the browser-native and MathJax interpretations of MathML (by the browser that can natively interpret MathML) are pretty similar, except some unsupported cases by MathJax.
    3. KaTeX
      • Pros: Similar to MathJax, but processing is much lighter and faster.
      • Cons: The range of support of MathML is more limited than MathJax (at least as of 2022 August). Like mathml.css, this should not be provided for browsers that can interpret MathML. If one does, the chance of failure for the correct interpretation is higher than MathJax.
  2. Alternative way to present mathematical expressions
    1. PDF or epub
      • Pros: What a user sees is what the author intends.
      • Cons: Unfriendly for automated processing, including indexing and searching. Not suitable for non-visual environments, such as for auditorial or text browsers.
    2. conversion to images; famously, latex2html converts LaTeX format into HTML, in which all the mathematical expressions are converted into images.
      • Pros: The document is mostly computer-friendly, like searchable, except the mathematical expressions.
      • Cons: The same problem as for the PDF about mathematical expression.
    3. LaTeX (or AMS-TeX or even plain TeX) source code
      • Pros: A gold standard to write mathematical expressions in the text form (in ASCII, UTF-8, etc). One of the easiest to write (though the learning cost is fairly expensive). Pretty computer friendly, though LaTeX does not care the semantics.
      • Cons: Very hard to read, except for the simplest ones. Not presentable on websites as it is, partly because there is no official standard to specify the form in the website protocol and no browsers natively interpret LaTeX.
    4. On-the-spot interpretation of LaTeX-style expressions embedded in HTML
      • Popular ones include MathJax and KaTeX. See a list of alternatives. https://alternativeto.net/software/mathjax/
      • Pros: Maths expressions are pretty easy to write (providing the author is knowledgeable about LaTeX) and is easier to read than MathML. Indexing by search engines is in principle relatively easy, though it probably is not for users’ environments. Browsers do not need to do anything, except interpreting the JavaScript code.
      • Cons: As far as the existing frameworks are concerned, they are no use (or maybe arguably worse?) for non-visual or text-only (terminal-like) environments, though technically it is possible to develop systems to accommodate a wide variety of environments. No semantics are preserved.
    5. AsciiMath
      • Pros: Marginally simpler to write than LaTeX.
      • Cons: The native JavaScript interpreter works only on Firefox and Safari, though, alternatively, MathJax can also interpret it and works in all major visual browsers. Like LaTeX, (I think) it does not care the semantics. The learning cost to write is still high, though seemingly lower than LaTeX and though it is designed so that people knowledgeable about LaTeX can migrate fairly easily. Much less popular than LaTeX, which means less people can actually read it, though the simplest form is admittedly intuitive and can be interpreted by anyone regardless of their knowledge of AsciiMath.

How To Use the Code

Include one of the following inside (preferably) the <head> element in the HTML page. For the last one, an arbitrary (valid) MathJax version number can be specified.

 <script src="https://masasakano.github.io/mathml_mathjax_js/dist/mathml-mpadded-3-2-2.min.js"> </script>
 <script src="https://masasakano.github.io/mathml_mathjax_js/dist/mathml-mpadded-2-7-7.min.js"> </script>
 <script src="https://masasakano.github.io/mathml_mathjax_js/dist/mathml-mpadded.min.js" data-mathjax-version="3.1.0"> </script>

Their differences are potential versions of MathJax to be loaded if MathJax is found to be necessary.

The first one is the latest release one at the time of writing and the second one is (I think) the last release of MathJax version 2. The third one is a generic one, which loads the user-specified version of MathJax. Note that it assumes no other <script> tags in the HTML from which the JavaScript is called have a data-mathjax-version attribute.

Once this script has been run and if it finds a browser does not deal with MathML, it automatically loads MathJax and also inserts the following set of tags (with the version of the loaded MathJax) at the beginning of the <body>, which you can use in your own script/CSS.

  <div id="mathjax_is_loaded" itemscope><meta itemprop="version" content="3.2.2"></div>

Otherwise (in other words, if the browser is either Firefox or Safari as of 2022 August), the above-mentioned tag is not inserted and alters nothing in the existing contents. The tag-insertion is an extra feature which Frédéric Wang’s original code did not provide.

Note that this script has no way to know if the MathJax is loaded before or after this script is loaded and therefore non-existence of the tag does not always mean MathJax is not loaded in general.

Demo

dist/ directory contains a couple of MathML HTMLs, in the header of which a JavaScript of this package is specified.

MathML interpreted by (Left) Firefox and (Right) Chrome with MathJax

Limitations

The algorithm assumes a few points:

Development

In the developer environment, the following has to be available.

make

The basic command is make. In the master branch, make generates or updates minified (Common) JavaScript files (see below) and README.html both under the tmp/ directory, if any of the relevant files have been updated since the last update. The latter tmp/README.html is for local use only (for the convenience of development).

If you want to forcibly update tmp/README.html (from README.md), run make doc .

Routine testing (unit-testing) is carried out with make test at the top directory. It is far from sufficient. More browser-dependent test frameworks are desirable.

make in gh-pages branch

It has the following functions.

Minify and “dist”

The minified and combined versions of JavaScript are generated under tmp/ directory (in the master branch) with make executed at the top directory. The generated minified ones are compatible with classic JS, whereas the original ones under src directory are not.

In minification, the core function src/mathml-mpadded-core.js is combined with one of the wrappers, such as src/mathml-mpadded.js .

The generated JS files under tmp/ should be later copied into the dist/ directory in the gh-pages branch. The directory tmp/ is registered in both master and gh-pages branches (whereas none of the normal files but .gitignore under tmp/ is registered in either) and therefore the generated files are not lost during git switch. That is the idea for this separation. Note that the standard git diff is unusable in this case because XXX.min.js exists only in the gh-pages branch whereas the updated XXX.js exists only in the master branch.

The standard work-flow is as follows:

  1. In the master branch, edit the files as you like.
  2. make (and make test), which generates minified JSs.
  3. git commit (and maybe git push)
  4. git switch gh-pages (Move to gh-pages branch)
  5. make check (to see if there is any update required, i.e., if the corresponding file in the master branch has been updated — usually yes; this does the following checks but does not fix the problems)
    1. git diff master -- README.md
      • If any diff, run ``git checkout master README.md`
    2. for f in tmp/*.min.js; do diff ../dist/basename $f $f; done
      • If any diff, mv MODIFIED.min.js ../dist/ (n.b., this will usually overwrite an existing file.)
    3. git diff master -- dist/demo/*.html (should return none if no HTML is updated in the master branch.)
      • If any diff, git checkout master dist/demo/MY_UPDATED.html
  6. make (to update dist/README.md if need be (i.e., if JS and/or HTML are updated))
  7. git add -u and git commit (and maybe git push)

Publishing

The files under the directory dist/ are automatically published on https://masasakano.github.io/mathml_mathjax_js/ in a way they (JavaScript and HTML files) are available to be called directly from the website (see Index).

None of the generated files under the directory tmp/ (or dist/) is registered in the master branch, whereas the static files, specifically dist/demo/sample-*.html and .gitignore, are. The gh-pages branch contains those files under the directory dist/ in the repository (but none under tmp/), but almost none of the files in the other directories.

TODO

Licence and Disclaimer

The licence is MPL 2, as described in LICENSE.txt in the package.

The author owes no responsibility for the package. Use it at your own risk.

Acknowledgement

I thank Frédéric Wang and Mozilla team for the earlier development. I hope MathML will be fully supported by all browsers one day.

Author: Masa Sakano - https://github.com/masasakano/

This site is open source. Improve this page.