commonmark_x reader (wrongly) identifies lists inside math blocks

[AlbertLei]

Description. I am gradually converting from the (default) markdown reader to commonmark_x reader. I am confused about the following behavior of the commonmark_x reader, which seems like a bug to me.

Using the commonmark_x reader, pandoc will convert the following

$$
- a
$$

to:

\$\$

- a \$\$

Obviously, the desired output (i want) is the display math $-a$, which is the output of the default pandoc reader.

If one removes the space between - and a,
then commonmark_x reader will parse the previous input as a math block.

[jgm]

Tricky. Technically, I think this isn't a bug, because Math is an Inline construction, and in commonmark, indicators of block structure take precedence over indicators of inline structure. (That is, we parse block structure first, then parse paragraph blocks using the inline parser.) It's definitely surprising, though.

If you enable the tex_math_gfm then you can use gfm-style math blocks, which should avoid this problem:

```math
- a
```

[AlbertLei]

In case my previous example was too artificial,
here's a more real one.
When writing a long math expression, it's natural to insert linebreak before operators like -. eg:

$$
\begin{split}
a + b + c \\
  - e - f
\end{split}
$$

And the commonmark_x reader will parse it as a list.

Switching to gfm style math blocks could avoid this problem,
but i'm afraid many people (or at least me) are used to the $$ syntax and would stick with it.

If that breaks the commonmark tradition,
would it be reasonable to request an extension that lets $$ indicate a (math) block delimiter, like ```math does?
I used to think that's what tex_math_dollars did. Turns out i was wrong.

[jgm]

Math is an inline element (even display math). However, one thing we could do is add a block parser for display math that can't interrupt a paragraph. It would insert a paragraph containing just one display math element. (So, this would be used when you had a blank line in front.) This would allow your example to work, except when it is inside a paragraph.

[AlbertLei]

Thanks for the explanation. I mistakenly thought that display math were blocks. I don't think a block parser for display math would be very useful. For me, almost all my display math elements live inside a paragraph, and I can't think of an example from papers I've read where that wasn't the case. I think I'll just stick to the default pandoc markdown reader, which clearly does some dark magic to get around this problem and fits my mental model much better. On Apr 1, 2026 at 12:06 AM +0800, John MacFarlane ***@***.***>, wrote: Math is an inline element (even display math). However, one thing we could do is add a block parser for display math that can't interrupt a paragraph. It would insert a paragraph containing just one display math element. (So, this would be used when you had a blank line in front.) This would allow your example to work, except when it is inside a paragraph. — Reply to this email directly, view it on GitHub<#11559?email_source=notifications&email_token=AD7XMQTTSNBXHSPRMPYDS634TPUJBA5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNRTHE3TENBVUZZGKYLTN5XKMYLVORUG64VFMV3GK3TUVRTG633UMVZF6Y3MNFRWW#discussioncomment-16397245>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AD7XMQSWVS4ERRKDJVHNWZL4TPUJBAVCNFSM6AAAAACXHHOA6SVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMMZZG4ZDINI>. You are receiving this because you authored the thread.Message ID: ***@***.***>