Incompatibility between links wrapped in asterisks, pedantic, and GFM

[simov]

Describe the bug

I have a named link, that I want wrapped in asterisks, essentially making it italic.

To Reproduce

- italic
  - [*named link*][some-url]
  - *[named link][some-url]*
  - [_named link_][some-url]
  - _[named link][some-url]_
- bold
  - [**named link**][some-url]
  - **[named link][some-url]**
  - [__named link__][some-url]
  - __[named link][some-url]__
- bold italic
  - [***named link***][some-url]
  - ***[named link][some-url]***
  - [___named link___][some-url]
  - ___[named link][some-url]___
  - [*__named link__*][some-url]
  - [__*named link*__][some-url]
  - __*[named link][some-url]*__
- code
  - [`named link`][some-url]
- code italic
  - *[`named link`][some-url]*
  - [*`named link`*][some-url]
  - _[`named link`][some-url]_
  - [_`named link`_][some-url]
- code bold
  - **[`named link`][some-url]**
  - [**`named link`**][some-url]
  - __[`named link`][some-url]__
  - [__`named link`__][some-url]
- code bold italic
  - [***`named link`***][some-url]
  - ***[`named link`][some-url]***
  - [___`named link`___][some-url]
  - ___[`named link`][some-url]___
  - [*__`named link`__*][some-url]
  - [__*`named link`*__][some-url]
  - __*[`named link`][some-url]*__

The offending lines are:

- *[named link][some-url]*
- ***[named link][some-url]***
- __*[named link][some-url]*__
- *[`named link`][some-url]*
- ***[`named link`][some-url]***
- __*[`named link`][some-url]*__

Enabling the pedantic option fixes this issue, but at the same time it breaks lots of other stuff, like GFM tables and strikethrough. Also note that I only tested with named links, but it may be broken for other types of links as well.

Expected behavior

Well, I'm looking at the exact same example using Remark - another popular parser, that I think was forked out of Marked back then, and everything renders correctly.

Let me know if I need to provider more information.

[styfle]

Thanks this looks like a bug.

Compare the output from marked with the output from commonmark.

The workaround is to use <em> or <strong> instead of asterisks.

[simov]

@styfle I think wrapping up your links with asterisks as pretty common, because that's the only way the editors can apply the correct style to that link.

I only use HTML tags in my markdown documents as a last resort.

[styfle]

Agreed, this is a bug 👍

Would you like to submit a PR to fix it? 😃

[simov]

Well, you got me! I'm not familiar with this library, but I have a browser extension that can be used as a constant source of annoyances for you. 😄

[UziTech]

related to #1284

[x13machine]

It appears to be something wrong with this regex https://github.com/markedjs/marked/blob/master/lib/marked.js#L545 I'll see if I can fix it today.

[UziTech]

@x13machine The problem is that an em can't end inside a link so it will be difficult to find an em with just a regular expression.

so *[link*](url)* should be

<p><em><a href="url">link*</a></em></p>

not

<p><em>[link</em>](url)*</p>

but *[link*](url* should be

<p><em>[link</em>](url*</p>

so after finding an ending * or _ you will have to figure out if it is in a valid link, and if it is find the next ending * or _

[x13machine]

so the regex should replace with a function?

[UziTech]

It should be something like the way we check for nested parentheses in a link #1414. The regex should grab more than it needs if the end is in square brackets or parentheses then the lexer will run it through a function and find the actual end.

[x13machine]

Why does group 3 match _[link_](url)_ correctly, while group 5 not match *[link*](url)* correctly? I going to refresh up on regex.