rustdoc/passes/lint/
bare_urls.rs

1//! Detects links that are not linkified, e.g., in Markdown such as `Go to https://example.com/.`
2//! Suggests wrapping the link with angle brackets: `Go to <https://example.com/>.` to linkify it.
3
4use core::ops::Range;
5use std::mem;
6use std::sync::LazyLock;
7
8use pulldown_cmark::{Event, Parser, Tag};
9use regex::Regex;
10use rustc_errors::Applicability;
11use rustc_hir::HirId;
12use rustc_resolve::rustdoc::source_span_for_markdown_range;
13use tracing::trace;
14
15use crate::clean::*;
16use crate::core::DocContext;
17use crate::html::markdown::main_body_opts;
18
19pub(super) fn visit_item(cx: &DocContext<'_>, item: &Item, hir_id: HirId, dox: &str) {
20    let report_diag = |cx: &DocContext<'_>,
21                       msg: &'static str,
22                       range: Range<usize>,
23                       without_brackets: Option<&str>| {
24        let maybe_sp = source_span_for_markdown_range(cx.tcx, dox, &range, &item.attrs.doc_strings)
25            .map(|(sp, _)| sp);
26        let sp = maybe_sp.unwrap_or_else(|| item.attr_span(cx.tcx));
27        cx.tcx.node_span_lint(crate::lint::BARE_URLS, hir_id, sp, |lint| {
28            lint.primary_message(msg)
29                .note("bare URLs are not automatically turned into clickable links");
30            // The fallback of using the attribute span is suitable for
31            // highlighting where the error is, but not for placing the < and >
32            if let Some(sp) = maybe_sp {
33                if let Some(without_brackets) = without_brackets {
34                    lint.multipart_suggestion(
35                        "use an automatic link instead",
36                        vec![(sp, format!("<{without_brackets}>"))],
37                        Applicability::MachineApplicable,
38                    );
39                } else {
40                    lint.multipart_suggestion(
41                        "use an automatic link instead",
42                        vec![
43                            (sp.shrink_to_lo(), "<".to_string()),
44                            (sp.shrink_to_hi(), ">".to_string()),
45                        ],
46                        Applicability::MachineApplicable,
47                    );
48                }
49            }
50        });
51    };
52
53    let mut p = Parser::new_ext(dox, main_body_opts()).into_offset_iter();
54
55    while let Some((event, range)) = p.next() {
56        match event {
57            Event::Text(s) => find_raw_urls(cx, dox, &s, range, &report_diag),
58            // We don't want to check the text inside code blocks or links.
59            Event::Start(tag @ (Tag::CodeBlock(_) | Tag::Link { .. })) => {
60                for (event, _) in p.by_ref() {
61                    match event {
62                        Event::End(end)
63                            if mem::discriminant(&end) == mem::discriminant(&tag.to_end()) =>
64                        {
65                            break;
66                        }
67                        _ => {}
68                    }
69                }
70            }
71            _ => {}
72        }
73    }
74}
75
76static URL_REGEX: LazyLock<Regex> = LazyLock::new(|| {
77    Regex::new(concat!(
78        r"https?://",                          // url scheme
79        r"([-a-zA-Z0-9@:%._\+~#=]{2,256}\.)+", // one or more subdomains
80        r"[a-zA-Z]{2,63}",                     // root domain
81        r"\b([-a-zA-Z0-9@:%_\+.~#?&/=]*)",     // optional query or url fragments
82    ))
83    .expect("failed to build regex")
84});
85
86fn find_raw_urls(
87    cx: &DocContext<'_>,
88    dox: &str,
89    text: &str,
90    range: Range<usize>,
91    f: &impl Fn(&DocContext<'_>, &'static str, Range<usize>, Option<&str>),
92) {
93    trace!("looking for raw urls in {text}");
94    // For now, we only check "full" URLs (meaning, starting with "http://" or "https://").
95    for match_ in URL_REGEX.find_iter(text) {
96        let mut url_range = match_.range();
97        url_range.start += range.start;
98        url_range.end += range.start;
99        let mut without_brackets = None;
100        // If the link is contained inside `[]`, then we need to replace the brackets and
101        // not just add `<>`.
102        if dox[..url_range.start].ends_with('[')
103            && url_range.end <= dox.len()
104            && dox[url_range.end..].starts_with(']')
105        {
106            url_range.start -= 1;
107            url_range.end += 1;
108            without_brackets = Some(match_.as_str());
109        }
110        f(cx, "this URL is not a hyperlink", url_range, without_brackets);
111    }
112}