rustdoc: Revert "fix bare urls split text"

[lolbinarycat]

This reverts commit d4e6b8f.

Reverted commit contains incorrect comments,
does not implement all the functionality it claims to, and lacks test coverage.

If someone wishes to re-add the functionality, it should be done as postprocessing after the regex match, as regex is not capable of handling arbitrary nesting of matched pairs. Additionally, future PRs attempting to reinstate this should frame it as a change to a fuzzy heuristic, not as a simple "fix".

r? @GuillaumeGomez

[GuillaumeGomez]

It was merged in #162381.

cc @notriddle so you're aware of this.

Thanks for noticing this issue @lolbinarycat !

[GuillaumeGomez]

r=me once CI pass

[notriddle]

incorrect comments

Which comments? This one?

    // pulldown-cmark can split a URL into multiple `Text` events while processing
    // characters such as `_` according to CommonMark's emphasis rules.
    // `TextMergeWithOffset` merges these events so we can check the complete URL.

That comment is true.

Given the URL from the test case, we can see it happen with the pulldown-cmark CLI. That's why the lint emits a bad suggestion in #162345

$ git clone https://github.com/pulldown-cmark/pulldown-cmark.git
$ cd pulldown-cmark
$ cargo run --release -- --events
https://en.wikipedia.org/wiki/Rust_(programming_language)
^D
0..58: Start(Paragraph)
0..34: Text(Borrowed("https://en.wikipedia.org/wiki/Rust"))
34..35: Text(Borrowed("_"))
35..57: Text(Borrowed("(programming_language)"))
0..58: End(Paragraph)
EOF

Or are you talking about this one?

        // Match URL characters and balanced parenthesized segments, without
        // consuming a trailing `)` that belongs to the surrounding prose.

That one is a bit confusing. The parens are balanced, since the regex won't match a lone paren, but they can't be nested, because regex can't represent recursive grammars.

It does, indeed, correctly handle a trailing ) that belongs to surrounding prose.

regex is not capable of handling arbitrary nesting of matched pairs

I knew that the improved regex only handled a single pair of nested parens. I also know that multiple levels of nesting are rare, and Wikipedia article URLs are common.

[notriddle]

In any case, if you would prefer a version of this code that handles nested parens, I can write it. I'll also add more test cases.

[notriddle]

@bors r=GuillaumeGomez,notriddle

Sorry about closing the PR. Wrong button.

[rust-bors]

📌 Commit 711a32b has been approved by GuillaumeGomez,notriddle

It is now in the queue for this repository.

[GuillaumeGomez]

In any case, if you would prefer a version of this code that handles nested parens, I can write it. I'll also add more test cases.

Personally I'd love to. Although not sure if regex is the right path for this. In any case, I trust your judgement, can't wait to see the improved version. =D