inline `$$x$$` math matches 2 times and returns nested mathjax

[rambip]

Description

In Tex, the $$x$$ format is supposed to be used only on newlines.
But in practice, a lot of parsers support $$x$$ at the end of lines.

The arithmetex modules here made the choice of treating inline $$x$$ as $x$, but there is a bug.
The matches occurs 2 times and we end up with a nested mathjax output.

Minimal Reproduction

foo = "hello $$x$$"
md = markdown.Markdown(extensions=[ArithmatexExtension()])
out = md.convert(foo)
print(out)

Output is:

<p>hello <span class="arithmatex"><span class="MathJax_Preview"><span class="arithmatex"><span class="MathJax_Preview">x</span><script type="math/tex">x</script></span></span><script type="math/tex"><span class="arithmatex"><span class="MathJax_Preview">x</span><script type="math/tex">x</script></span></script></span></p>

It should be:

<p>hello <span class="arithmatex"><span class="MathJax_Preview">x</span><script type="math/tex">x</script></span></p>

Note: same behaviour with and without preview.

Version(s) & System Info

• Operating System: Linux
• Python Version: 3.14
• Package Version: latest

[facelessuser]

@rambip You have not provided a minimal reproduction. But I assume something like this demonstrates what you are tallking about.

>>> markdown.markdown('a $$x$$ b', extensions=['pymdownx.arithmatex'])
'<p>a <span class="arithmatex"><span class="MathJax_Preview"><span class="arithmatex"><span class="MathJax_Preview">x</span><script type="math/tex">x</script></span></span><script type="math/tex"><span class="arithmatex"><span class="MathJax_Preview">x</span><script type="math/tex">x</script></span></script></span> b</p>'

It seems you've attempted a PR via #2806, but I disagree with the solution being attempted. Inline patterns should not consume more dollar signs.

The root cause of the issue is actually a bit different than assumed. Arithmatex correctly takes an inline pattern such as $$x$$ and parsers only the inner dollar signs: $<math stuff>$. But Python Markdown then captures the HTML and replaces it with a placeholder:

>>> import markdown
>>> markdown.markdown('a $$x$$ b', extensions=['pymdownx.arithmatex'])
match: b'$x$'
match: b'$\x02klzzwxh:0000\x03$'

That placeholder gets parsed as math content, and when the placeholders are restored as actual HTML, you have nested math.

This is just a battle of different steps in the Python Markdown processor stepping on each other.

A better solution, from my point of view, is to avoid parsing math content if they included the STX (\x02) or EOT (\x03) characters that are used to denote the start and end of an HTML placeholder.

RE_SMART_DOLLAR_INLINE = r'(?:(?<!\\)((?:\\{2})+)(?=\$)|(?<!\\)(\$)(?!\s)((?:\\.|[^\\$\x02\x03])+?)(?<!\s)(?:\$))'
RE_DOLLAR_INLINE = r'(?:(?<!\\)((?:\\{2})+)(?=\$)|(?<!\\)(\$)((?:\\.|[^\\$\x02\x03])+?)(?:\$))'
RE_BRACKET_INLINE = r'(?:(?<!\\)((?:\\{2})+?)(?=\\\()|(?<!\\)(\\\()((?:\\[^)]|[^\\\x02\x03])+?)(?:\\\)))'

>>> import markdown
>>> markdown.markdown('a $$x$$ b', extensions=['pymdownx.arithmatex'])
'<p>a $<span class="arithmatex"><span class="MathJax_Preview">x</span><script type="math/tex">x</script></span>$ b</p>'

[rambip]

@facelessuser sorry, I completely forgot to copy the value for my variable. I edited my post.

Ok, I understand better what happens. I was still a bit confused because I did not know how markdown orders the calls from the extensions.

I can totally close the PR and let you make your fix. In my opinion both are fine, but I'm very aware that I don't know a lot about how markdown extensions work.
A problem with your solution though, is that it's a breaking change. Projects such as marimo (here and there ) now assume it's the correct behaviour.
It's probably not a big deal, but worth mentioning in release notes somewhere.

Thank's for the quick response.

[facelessuser]

A problem with your solution though, is that it's a breaking change. Projects such as marimo (here and there ) now assume it's the correct behaviour.

No, this is not a breaking change; they were working around broken logic.

1. Documentation is clear that block syntax should not be used inline.
2. Double-nested logic is not expected and is a bug.
3. Multiple dollars should not be consumed for inline. Only one dollar sign should be consumed.

[rambip]

We mostly agree then. Just wanted to bring up the point.
Thank's

[facelessuser]

New tagged release: https://github.com/facelessuser/pymdown-extensions/releases/tag/10.19.1