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<'_>, msg: &'static str, range: Range<usize>| {
21        let maybe_sp = source_span_for_markdown_range(cx.tcx, dox, &range, &item.attrs.doc_strings);
22        let sp = maybe_sp.unwrap_or_else(|| item.attr_span(cx.tcx));
23        cx.tcx.node_span_lint(crate::lint::BARE_URLS, hir_id, sp, |lint| {
24            lint.primary_message(msg)
25                .note("bare URLs are not automatically turned into clickable links");
26            // The fallback of using the attribute span is suitable for
27            // highlighting where the error is, but not for placing the < and >
28            if let Some(sp) = maybe_sp {
29                lint.multipart_suggestion(
30                    "use an automatic link instead",
31                    vec![
32                        (sp.shrink_to_lo(), "<".to_string()),
33                        (sp.shrink_to_hi(), ">".to_string()),
34                    ],
35                    Applicability::MachineApplicable,
36                );
37            }
38        });
39    };
40
41    let mut p = Parser::new_ext(dox, main_body_opts()).into_offset_iter();
42
43    while let Some((event, range)) = p.next() {
44        match event {
45            Event::Text(s) => find_raw_urls(cx, &s, range, &report_diag),
46            // We don't want to check the text inside code blocks or links.
47            Event::Start(tag @ (Tag::CodeBlock(_) | Tag::Link { .. })) => {
48                for (event, _) in p.by_ref() {
49                    match event {
50                        Event::End(end)
51                            if mem::discriminant(&end) == mem::discriminant(&tag.to_end()) =>
52                        {
53                            break;
54                        }
55                        _ => {}
56                    }
57                }
58            }
59            _ => {}
60        }
61    }
62}
63
64static URL_REGEX: LazyLock<Regex> = LazyLock::new(|| {
65    Regex::new(concat!(
66        r"https?://",                          // url scheme
67        r"([-a-zA-Z0-9@:%._\+~#=]{2,256}\.)+", // one or more subdomains
68        r"[a-zA-Z]{2,63}",                     // root domain
69        r"\b([-a-zA-Z0-9@:%_\+.~#?&/=]*)"      // optional query or url fragments
70    ))
71    .expect("failed to build regex")
72});
73
74fn find_raw_urls(
75    cx: &DocContext<'_>,
76    text: &str,
77    range: Range<usize>,
78    f: &impl Fn(&DocContext<'_>, &'static str, Range<usize>),
79) {
80    trace!("looking for raw urls in {text}");
81    // For now, we only check "full" URLs (meaning, starting with "http://" or "https://").
82    for match_ in URL_REGEX.find_iter(text) {
83        let url_range = match_.range();
84        f(
85            cx,
86            "this URL is not a hyperlink",
87            Range { start: range.start + url_range.start, end: range.start + url_range.end },
88        );
89    }
90}
rustdoc/passes/lint/bare_urls.rs

rustdoc/passes/lint/
bare_urls.rs
