Font is italic for unicode text

[abnerlee]

MathJax v3

For math like $$数学$$, the Chinese characters are italic, its character is rendered as

<mjx-utext variant="italic" style="font-size: 84%; padding: 0.893em 0px 0.238em; font-family: MJXZERO, serif; font-style: italic;">学</mjx-utext>

but in v2, the character are not in italic.

So, is there a reason of this spec change, and is there any options to control this?

[GoBigorGoHome]

@abnerlee Just upgraded to the latest Typora and noticed the change. The italic Chinese characters in my math formulas looks really bad (actually italic font shapes are just not for Chinese characters at all). Is it possible to change the font by solely configuring Typora, for example, using CSS?

A workaround is to wrap Chinese characters in math into \text{...}. But that's too much work for existing math formulas. I wish there is a way to change the font as I often include Chinese characters in math formulas for expository purpose. It is tedious for me to have to write \text{...}.

[abnerlee]

@GoBigorGoHome If you don't need italic in math at all, then you can use css

[data-variant='italic'] {
    font-style: normal;
}

to force ignore italic texts, but that is just workaround with side effects.

---

still requires feedback from MathJax side.

[dpvc]

In versions prior to 3.2, MathJax would place all characters that it doesn't think of as "letters" into <mo> elements. In version 3.2, that decision was made more nuanced so that the unicode block in which such characters occur is now taken into account to determine what MathML node to use. Now <mo> is used for symbols, while <mi> is used for letters and most language-specific characters. This improves the semantics of the underlying MathML (since these characters are not operators, as <mo> would indicate).

Since an <mi> with a single character is in italics, that is why your characters are now in italics. The expected encoding for your expression would be $$\text{数学}$$, since these two characters form a single word, just as you would use $$\text{mathematics}$$ rather than $$mathematics$$ (which would be typeset in italics with incorrect inter-character spacing). While it does take extra work to include \text{..}, that does properly indicate your intent, while leaving it out does not.

As a work-around, you could change the MathML node associated with the Chinese ranges to be <mtext> rather than <mi>, which would give you upright letters, but each in its own <mtext> element, which is still not correct, but at least will display as you prefer. To do that merge

MathJax = {
  startup: {
    ready() {
      const {RANGES} = MathJax._.core.MmlTree.OperatorDictionary;
      RANGES[28][3] = RANGES[30][3] = RANGES[33][3] = RANGES[47][3] = 'mtext';
      MathJax.startup.defaultReady();
    }
  }
};

into your current MathJax configuration.

[jtzwuwu]

In versions prior to 3.2, MathJax would place all characters that it doesn't think of as "letters" into <mo> elements. In version 3.2, that decision was made more nuanced so that the unicode block in which such characters occur is now taken into account to determine what MathML node to use. Now <mo> is used for symbols, while <mi> is used for letters and most language-specific characters. This improves the semantics of the underlying MathML (since these characters are not operators, as <mo> would indicate).

Since an <mi> with a single character is in italics, that is why your characters are now in italics. The expected encoding for your expression would be $$\text{数学}$$, since these two characters form a single word, just as you would use$$\test{mathematics}$$ rather than $$mathematics$$ (which would be typeset in italics with incorrect inter-character spacing). While it does take extra work to include \text{..}, that does properly indicate your intent, while leaving it out does not.

As a work-around, you could change the MathML node associated with the Chinese ranges to be <mtext> rather than <mi>, which would give you upright letters, but each in its own <mtext> element, which is still not correct, but at least will display as you prefer. To do that merge

MathJax = {
  startup: {
    ready() {
      const {RANGES} = MathJax._.core.MmlTree.OperatorDictionary;
      RANGES[28][3] = RANGES[30][3] = RANGES[33][3] = RANGES[47][3] = 'mtext';
      MathJax.startup.defaultReady();
    }
  }
};

into your current MathJax configuration.

Frankly speaking, the biggest problem is that almost no one uses italicized Chinese characters.

[abnerlee]

I saw English letters and geek letters like α are also be put into <mi>, like $$aaa$$, but they are not italics as default

[dpvc]

I saw English letters and geek letters like α are also be put into <mi>, like $$aaa$$, but they are not italics as default

They should be, and they are for me. Can you provide a minimal example page where they are not italic?

[abnerlee]

I see, so their italic are just font change?

[njlarsson]

Could I get an explanation for how to get upper-case greek letters to be displayed upright in math mode?

In my setup, for instance $Σ$ is displayed upright, just like $\Sigma$, in PDF documents produced via LaTeX (more specifically, lualatex), but they are slanted by MathJax in web pages. The way I understand the above, there is a fix in 3.2 that should allow it to be displayed upright, but what do I do to activate this? If the workaround by @dpvc from July 2021 is still applicable, how do I make it apply to greek upper-case?

(I write documents in Org Mode, and would like the maths to be displayed in the same way whether it’s exported to HTML or LaTeX.)

[dpvc]

@njlarsson, do you know the version of MathJax that you are using? (Use the "About MathJax" entry in the MathJax contextual menu to find out.) Older versions of org-mode used MathJax v2, but it looks like the current version of org-mode has been updated to use MathJax v3, so that should allow you to use v3.2, which should handle the Unicode characters properly for you.

[njlarsson]

@dpvc It’s 3.2. “Properly” may be different things for different people. I suppose some people want Σ to be slanted, and I don’t expect upright to be the default. But for instance, this small example,

<html>
    <head>
        <meta charset="UTF-8">
        <title>A web page</title>
    </head>
    <body>
        <script type="text/javascript" id="MathJax-script" async
                src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js">
        </script>
        
        LaTeX: \(\Sigma\), Unicode: \(Σ\).
    </body>
</html>

makes \Sigma come out upright but the Σ slanted:

I would like to know how to configure MathJax to make them both look like the first one.

[dpvc]

@njlarsson: OK, my fault. I was looking at 4.0.0-beta.3, which handles the Greek characters differently.

Here is a configuration for v3.2.2 that will make the upper-case Greek letters upright when entered as literal Unicode characters.

        MathJax = {
          startup: {
            ready() {
              const {RANGES} = MathJax._.core.MmlTree.OperatorDictionary;
              const {TEXCLASS} = MathJax._.core.MmlTree.MmlNode;
              RANGES.splice(4, 1,
                [0x0370, 0x0385, TEXCLASS.ORD, 'mi'],
                [0x0386, 0x3AB, TEXCLASS.ORD, 'mi', 'normal'],
                [0x03AC, 0x1A20, TEXCLASS.ORD, 'mi']
              );
              MathJax.startup.defaultReady();
            }
          }
        };

(at least all the plain ones; some explicitly accented ones are not fixed by this.)

See if that works for you.