Bare URLs support behind a new flag

[Hywan]

Hey!

Let's assume this test:

#[test]
fn foobar() {
    let original = r##"https://foo.bar/_/A and https://baz.qux/_/B"##;
    let expected = r##"<a href="https://foo.bar/_/A">https://foo.bar/_/A</a> and <a href="https://baz.qux/_/B">https://baz.qux/_/B</a>"##;

    test_markdown_html(original, expected, false, false, false);
}

This test is failing:

thread 'suite::spec::foobar' panicked at 'assertion failed: `(left == right)`
  left: `"<a href=\"https://foo.bar/_/A\">https://foo.bar/_/A</a> and <a href=\"https://baz.qux/_/B\">https://baz.qux/_/B</a>"`,
 right: `"<p>https://foo.bar/<em>/A and https://baz.qux/</em>/B</p>"`', tests/lib.rs:41:5

The first and second underscores in the links are considered as an emphasize. It seems like links have a lower priority than emphasizes.

Testing this here with Github markdown renderer: https://foo.bar/_/A and https://baz.qux/_/B; links are parsed correctly. I don't know what the spec says about priorities honestly, I believe it's implementation dependent.

[jplatte]

What I think other markdown parsers tend to do is only interpret _ as emphasis if the start is right before a word, and the end is right after a word. So the following should all not work (let's test what GitHub does):

• a_b c_d: a_b c_d
• a_ _b: a_ _b
• _a _b: _a _b
• a_ b_: b_ a_

… while _a b_ should produce emphasis: a b

[ehuss]

pulldown_cmark does not support detection of bare URLs. That is an extension that GitHub's parser has. On dingus, you can see that it parses the same as pulldown_cmark. The URLs need to be enclosed in angled brackets, like <https://foo.bar/_/A>.

[Martin1887]

Therefore, supporting bare URLs should be a new feature behind a flag, a non-urgent enhancement.

[jplatte]

But what about the more general thing about underscores within words being interpreted as emphasis? That's independent of link detection, as I showed above. Is that just a fundamental difference between commonmark and GFM, i.e. not subject to configuration?

[Martin1887]

Your example works in pulldown-cmark, that is the behaviour of CommonMark and pulldown-cmark gets the same than dingus:

[jplatte]

Oh I see, so in the original example it's the _s not being parts of words (there's /es to both sides in the URLs) which causes them to be interpreted as emphasis.

[Hywan]

Thanks for the details.