1 //! Markdown formatting for rustdoc.
3 //! This module implements markdown formatting through the pulldown-cmark library.
6 //! #![feature(rustc_private)]
8 //! extern crate rustc_span;
10 //! use rustc_span::edition::Edition;
11 //! use rustdoc::html::markdown::{HeadingOffset, IdMap, Markdown, ErrorCodes};
13 //! let s = "My *markdown* _text_";
14 //! let mut id_map = IdMap::new();
15 //! let md = Markdown {
19 //! error_codes: ErrorCodes::Yes,
20 //! edition: Edition::Edition2015,
21 //! playground: &None,
22 //! heading_offset: HeadingOffset::H2,
24 //! let html = md.into_string();
25 //! // ... something using html
28 use rustc_data_structures::fx::FxHashMap;
29 use rustc_hir::def_id::DefId;
31 use rustc_middle::ty::TyCtxt;
32 use rustc_span::edition::Edition;
35 use once_cell::sync::Lazy;
37 use std::cell::RefCell;
38 use std::collections::VecDeque;
39 use std::default::Default;
41 use std::ops::{ControlFlow, Range};
44 use crate::clean::RenderedLink;
46 use crate::html::escape::Escape;
47 use crate::html::format::Buffer;
48 use crate::html::highlight;
49 use crate::html::length_limit::HtmlWithLimit;
50 use crate::html::toc::TocBuilder;
53 html, BrokenLink, CodeBlockKind, CowStr, Event, LinkType, Options, Parser, Tag,
59 const MAX_HEADER_LEVEL: u32 = 6;
61 /// Options for rendering Markdown in the main body of documentation.
62 pub(crate) fn main_body_opts() -> Options {
63 Options::ENABLE_TABLES
64 | Options::ENABLE_FOOTNOTES
65 | Options::ENABLE_STRIKETHROUGH
66 | Options::ENABLE_TASKLISTS
67 | Options::ENABLE_SMART_PUNCTUATION
70 /// Options for rendering Markdown in summaries (e.g., in search results).
71 pub(crate) fn summary_opts() -> Options {
72 Options::ENABLE_TABLES
73 | Options::ENABLE_FOOTNOTES
74 | Options::ENABLE_STRIKETHROUGH
75 | Options::ENABLE_TASKLISTS
76 | Options::ENABLE_SMART_PUNCTUATION
79 #[derive(Debug, Clone, Copy)]
80 pub enum HeadingOffset {
89 /// When `to_string` is called, this struct will emit the HTML corresponding to
90 /// the rendered version of the contained markdown string.
91 pub struct Markdown<'a> {
93 /// A list of link replacements.
94 pub links: &'a [RenderedLink],
95 /// The current list of used header IDs.
96 pub ids: &'a mut IdMap,
97 /// Whether to allow the use of explicit error codes in doctest lang strings.
98 pub error_codes: ErrorCodes,
99 /// Default edition to use when parsing doctests (to add a `fn main`).
100 pub edition: Edition,
101 pub playground: &'a Option<Playground>,
102 /// Offset at which we render headings.
103 /// E.g. if `heading_offset: HeadingOffset::H2`, then `# something` renders an `<h2>`.
104 pub heading_offset: HeadingOffset,
106 /// A tuple struct like `Markdown` that renders the markdown with a table of contents.
107 pub(crate) struct MarkdownWithToc<'a>(
109 pub(crate) &'a mut IdMap,
110 pub(crate) ErrorCodes,
112 pub(crate) &'a Option<Playground>,
114 /// A tuple struct like `Markdown` that renders the markdown escaping HTML tags.
115 pub(crate) struct MarkdownHtml<'a>(
117 pub(crate) &'a mut IdMap,
118 pub(crate) ErrorCodes,
120 pub(crate) &'a Option<Playground>,
122 /// A tuple struct like `Markdown` that renders only the first paragraph.
123 pub(crate) struct MarkdownSummaryLine<'a>(pub &'a str, pub &'a [RenderedLink]);
125 #[derive(Copy, Clone, PartialEq, Debug)]
126 pub enum ErrorCodes {
132 pub(crate) fn from(b: bool) -> Self {
134 true => ErrorCodes::Yes,
135 false => ErrorCodes::No,
139 pub(crate) fn as_bool(self) -> bool {
141 ErrorCodes::Yes => true,
142 ErrorCodes::No => false,
147 /// Controls whether a line will be hidden or shown in HTML output.
149 /// All lines are used in documentation tests.
156 fn for_html(self) -> Option<Cow<'a, str>> {
158 Line::Shown(l) => Some(l),
159 Line::Hidden(_) => None,
163 fn for_code(self) -> Cow<'a, str> {
166 Line::Hidden(l) => Cow::Borrowed(l),
171 // FIXME: There is a minor inconsistency here. For lines that start with ##, we
172 // have no easy way of removing a potential single space after the hashes, which
173 // is done in the single # case. This inconsistency seems okay, if non-ideal. In
174 // order to fix it we'd have to iterate to find the first non-# character, and
175 // then reallocate to remove it; which would make us return a String.
176 fn map_line(s: &str) -> Line<'_> {
177 let trimmed = s.trim();
178 if trimmed.starts_with("##") {
179 Line::Shown(Cow::Owned(s.replacen("##", "#", 1)))
180 } else if let Some(stripped) = trimmed.strip_prefix("# ") {
182 Line::Hidden(stripped)
183 } else if trimmed == "#" {
184 // We cannot handle '#text' because it could be #[attr].
187 Line::Shown(Cow::Borrowed(s))
191 /// Convert chars from a title for an id.
193 /// "Hello, world!" -> "hello-world"
194 fn slugify(c: char) -> Option<char> {
195 if c.is_alphanumeric() || c == '-' || c == '_' {
196 if c.is_ascii() { Some(c.to_ascii_lowercase()) } else { Some(c) }
197 } else if c.is_whitespace() && c.is_ascii() {
204 #[derive(Clone, Debug)]
205 pub struct Playground {
206 pub crate_name: Option<String>,
210 /// Adds syntax highlighting and playground Run buttons to Rust code blocks.
211 struct CodeBlocks<'p, 'a, I: Iterator<Item = Event<'a>>> {
213 check_error_codes: ErrorCodes,
215 // Information about the playground if a URL has been specified, containing an
216 // optional crate name and the URL.
217 playground: &'p Option<Playground>,
220 impl<'p, 'a, I: Iterator<Item = Event<'a>>> CodeBlocks<'p, 'a, I> {
223 error_codes: ErrorCodes,
225 playground: &'p Option<Playground>,
227 CodeBlocks { inner: iter, check_error_codes: error_codes, edition, playground }
231 impl<'a, I: Iterator<Item = Event<'a>>> Iterator for CodeBlocks<'_, 'a, I> {
232 type Item = Event<'a>;
234 fn next(&mut self) -> Option<Self::Item> {
235 let event = self.inner.next();
240 let Some(Event::Start(Tag::CodeBlock(kind))) = event else {
244 let mut origtext = String::new();
245 for event in &mut self.inner {
247 Event::End(Tag::CodeBlock(..)) => break,
248 Event::Text(ref s) => {
249 origtext.push_str(s);
254 let lines = origtext.lines().filter_map(|l| map_line(l).for_html());
255 let text = lines.intersperse("\n".into()).collect::<String>();
257 let parse_result = match kind {
258 CodeBlockKind::Fenced(ref lang) => {
260 LangString::parse_without_check(lang, self.check_error_codes, false);
261 if !parse_result.rust {
262 return Some(Event::Html(
264 "<div class=\"example-wrap\">\
265 <pre class=\"language-{}\"><code>{}</code></pre>\
275 CodeBlockKind::Indented => Default::default(),
278 compile_fail = parse_result.compile_fail;
279 should_panic = parse_result.should_panic;
280 ignore = parse_result.ignore;
281 edition = parse_result.edition;
283 let explicit_edition = edition.is_some();
284 let edition = edition.unwrap_or(self.edition);
286 let playground_button = self.playground.as_ref().and_then(|playground| {
287 let krate = &playground.crate_name;
288 let url = &playground.url;
294 .map(|l| map_line(l).for_code())
295 .intersperse("\n".into())
296 .collect::<String>();
297 let krate = krate.as_ref().map(|s| &**s);
299 doctest::make_test(&test, krate, false, &Default::default(), edition, None);
300 let channel = if test.contains("#![feature(") { "&version=nightly" } else { "" };
302 // These characters don't need to be escaped in a URI.
303 // FIXME: use a library function for percent encoding.
304 fn dont_escape(c: u8) -> bool {
305 (b'a' <= c && c <= b'z')
306 || (b'A' <= c && c <= b'Z')
307 || (b'0' <= c && c <= b'9')
318 let mut test_escaped = String::new();
319 for b in test.bytes() {
321 test_escaped.push(char::from(b));
323 write!(test_escaped, "%{:02X}", b).unwrap();
327 r#"<a class="test-arrow" target="_blank" href="{}?code={}{}&edition={}">Run</a>"#,
328 url, test_escaped, channel, edition,
332 let tooltip = if ignore != Ignore::None {
333 Some((None, "ignore"))
334 } else if compile_fail {
335 Some((None, "compile_fail"))
336 } else if should_panic {
337 Some((None, "should_panic"))
338 } else if explicit_edition {
339 Some((Some(edition), "edition"))
344 // insert newline to clearly separate it from the
345 // previous block so we can shorten the html output
346 let mut s = Buffer::new();
348 highlight::render_with_highlighting(
352 "rust-example-rendered{}",
353 if let Some((_, class)) = tooltip { format!(" {}", class) } else { String::new() }
355 playground_button.as_deref(),
362 Some(Event::Html(s.into_inner().into()))
366 /// Make headings links with anchor IDs and build up TOC.
367 struct LinkReplacer<'a, I: Iterator<Item = Event<'a>>> {
369 links: &'a [RenderedLink],
370 shortcut_link: Option<&'a RenderedLink>,
373 impl<'a, I: Iterator<Item = Event<'a>>> LinkReplacer<'a, I> {
374 fn new(iter: I, links: &'a [RenderedLink]) -> Self {
375 LinkReplacer { inner: iter, links, shortcut_link: None }
379 impl<'a, I: Iterator<Item = Event<'a>>> Iterator for LinkReplacer<'a, I> {
380 type Item = Event<'a>;
382 fn next(&mut self) -> Option<Self::Item> {
383 let mut event = self.inner.next();
385 // Replace intra-doc links and remove disambiguators from shortcut links (`[fn@f]`).
387 // This is a shortcut link that was resolved by the broken_link_callback: `[fn@f]`
388 // Remove any disambiguator.
389 Some(Event::Start(Tag::Link(
390 // [fn@f] or [fn@f][]
391 LinkType::ShortcutUnknown | LinkType::CollapsedUnknown,
395 debug!("saw start of shortcut link to {} with title {}", dest, title);
396 // If this is a shortcut link, it was resolved by the broken_link_callback.
397 // So the URL will already be updated properly.
398 let link = self.links.iter().find(|&link| *link.href == **dest);
399 // Since this is an external iterator, we can't replace the inner text just yet.
400 // Store that we saw a link so we know to replace it later.
401 if let Some(link) = link {
402 trace!("it matched");
403 assert!(self.shortcut_link.is_none(), "shortcut links cannot be nested");
404 self.shortcut_link = Some(link);
407 // Now that we're done with the shortcut link, don't replace any more text.
408 Some(Event::End(Tag::Link(
409 LinkType::ShortcutUnknown | LinkType::CollapsedUnknown,
413 debug!("saw end of shortcut link to {}", dest);
414 if self.links.iter().any(|link| *link.href == **dest) {
415 assert!(self.shortcut_link.is_some(), "saw closing link without opening tag");
416 self.shortcut_link = None;
419 // Handle backticks in inline code blocks, but only if we're in the middle of a shortcut link.
421 Some(Event::Code(text)) => {
422 trace!("saw code {}", text);
423 if let Some(link) = self.shortcut_link {
424 trace!("original text was {}", link.original_text);
425 // NOTE: this only replaces if the code block is the *entire* text.
426 // If only part of the link has code highlighting, the disambiguator will not be removed.
428 // This is a limitation from `collect_intra_doc_links`: it passes a full link,
429 // and does not distinguish at all between code blocks.
430 // So we could never be sure we weren't replacing too much:
431 // [fn@my_`f`unc] is treated the same as [my_func()] in that pass.
433 // NOTE: &[1..len() - 1] is to strip the backticks
434 if **text == link.original_text[1..link.original_text.len() - 1] {
435 debug!("replacing {} with {}", text, link.new_text);
436 *text = CowStr::Borrowed(&link.new_text);
440 // Replace plain text in links, but only in the middle of a shortcut link.
442 Some(Event::Text(text)) => {
443 trace!("saw text {}", text);
444 if let Some(link) = self.shortcut_link {
445 trace!("original text was {}", link.original_text);
446 // NOTE: same limitations as `Event::Code`
447 if **text == *link.original_text {
448 debug!("replacing {} with {}", text, link.new_text);
449 *text = CowStr::Borrowed(&link.new_text);
453 // If this is a link, but not a shortcut link,
454 // replace the URL, since the broken_link_callback was not called.
455 Some(Event::Start(Tag::Link(_, dest, _))) => {
456 if let Some(link) = self.links.iter().find(|&link| *link.original_text == **dest) {
457 *dest = CowStr::Borrowed(link.href.as_ref());
460 // Anything else couldn't have been a valid Rust path, so no need to replace the text.
464 // Yield the modified event
469 /// Wrap HTML tables into `<div>` to prevent having the doc blocks width being too big.
470 struct TableWrapper<'a, I: Iterator<Item = Event<'a>>> {
472 stored_events: VecDeque<Event<'a>>,
475 impl<'a, I: Iterator<Item = Event<'a>>> TableWrapper<'a, I> {
476 fn new(iter: I) -> Self {
477 Self { inner: iter, stored_events: VecDeque::new() }
481 impl<'a, I: Iterator<Item = Event<'a>>> Iterator for TableWrapper<'a, I> {
482 type Item = Event<'a>;
484 fn next(&mut self) -> Option<Self::Item> {
485 if let Some(first) = self.stored_events.pop_front() {
489 let event = self.inner.next()?;
492 Event::Start(Tag::Table(t)) => {
493 self.stored_events.push_back(Event::Start(Tag::Table(t)));
494 Event::Html(CowStr::Borrowed("<div>"))
496 Event::End(Tag::Table(t)) => {
497 self.stored_events.push_back(Event::Html(CowStr::Borrowed("</div>")));
498 Event::End(Tag::Table(t))
505 type SpannedEvent<'a> = (Event<'a>, Range<usize>);
507 /// Make headings links with anchor IDs and build up TOC.
508 struct HeadingLinks<'a, 'b, 'ids, I> {
510 toc: Option<&'b mut TocBuilder>,
511 buf: VecDeque<SpannedEvent<'a>>,
512 id_map: &'ids mut IdMap,
513 heading_offset: HeadingOffset,
516 impl<'a, 'b, 'ids, I> HeadingLinks<'a, 'b, 'ids, I> {
519 toc: Option<&'b mut TocBuilder>,
520 ids: &'ids mut IdMap,
521 heading_offset: HeadingOffset,
523 HeadingLinks { inner: iter, toc, buf: VecDeque::new(), id_map: ids, heading_offset }
527 impl<'a, 'b, 'ids, I: Iterator<Item = SpannedEvent<'a>>> Iterator
528 for HeadingLinks<'a, 'b, 'ids, I>
530 type Item = SpannedEvent<'a>;
532 fn next(&mut self) -> Option<Self::Item> {
533 if let Some(e) = self.buf.pop_front() {
537 let event = self.inner.next();
538 if let Some((Event::Start(Tag::Heading(level, _, _)), _)) = event {
539 let mut id = String::new();
540 for event in &mut self.inner {
542 Event::End(Tag::Heading(..)) => break,
543 Event::Start(Tag::Link(_, _, _)) | Event::End(Tag::Link(..)) => {}
544 Event::Text(text) | Event::Code(text) => {
545 id.extend(text.chars().filter_map(slugify));
546 self.buf.push_back(event);
548 _ => self.buf.push_back(event),
551 let id = self.id_map.derive(id);
553 if let Some(ref mut builder) = self.toc {
554 let mut html_header = String::new();
555 html::push_html(&mut html_header, self.buf.iter().map(|(ev, _)| ev.clone()));
556 let sec = builder.push(level as u32, html_header, id.clone());
557 self.buf.push_front((Event::Html(format!("{} ", sec).into()), 0..0));
561 std::cmp::min(level as u32 + (self.heading_offset as u32), MAX_HEADER_LEVEL);
562 self.buf.push_back((Event::Html(format!("</a></h{}>", level).into()), 0..0));
564 let start_tags = format!(
565 "<h{level} id=\"{id}\">\
570 return Some((Event::Html(start_tags.into()), 0..0));
576 /// Extracts just the first paragraph.
577 struct SummaryLine<'a, I: Iterator<Item = Event<'a>>> {
583 impl<'a, I: Iterator<Item = Event<'a>>> SummaryLine<'a, I> {
584 fn new(iter: I) -> Self {
585 SummaryLine { inner: iter, started: false, depth: 0 }
589 fn check_if_allowed_tag(t: &Tag<'_>) -> bool {
592 Tag::Paragraph | Tag::Item | Tag::Emphasis | Tag::Strong | Tag::Link(..) | Tag::BlockQuote
596 fn is_forbidden_tag(t: &Tag<'_>) -> bool {
597 matches!(t, Tag::CodeBlock(_) | Tag::Table(_) | Tag::TableHead | Tag::TableRow | Tag::TableCell)
600 impl<'a, I: Iterator<Item = Event<'a>>> Iterator for SummaryLine<'a, I> {
601 type Item = Event<'a>;
603 fn next(&mut self) -> Option<Self::Item> {
604 if self.started && self.depth == 0 {
610 if let Some(event) = self.inner.next() {
611 let mut is_start = true;
612 let is_allowed_tag = match event {
613 Event::Start(ref c) => {
614 if is_forbidden_tag(c) {
618 check_if_allowed_tag(c)
620 Event::End(ref c) => {
621 if is_forbidden_tag(c) {
626 check_if_allowed_tag(c)
630 return if !is_allowed_tag {
632 Some(Event::Start(Tag::Paragraph))
634 Some(Event::End(Tag::Paragraph))
644 /// Moves all footnote definitions to the end and add back links to the
646 struct Footnotes<'a, I> {
648 footnotes: FxHashMap<String, (Vec<Event<'a>>, u16)>,
651 impl<'a, I> Footnotes<'a, I> {
652 fn new(iter: I) -> Self {
653 Footnotes { inner: iter, footnotes: FxHashMap::default() }
656 fn get_entry(&mut self, key: &str) -> &mut (Vec<Event<'a>>, u16) {
657 let new_id = self.footnotes.len() + 1;
658 let key = key.to_owned();
659 self.footnotes.entry(key).or_insert((Vec::new(), new_id as u16))
663 impl<'a, I: Iterator<Item = SpannedEvent<'a>>> Iterator for Footnotes<'a, I> {
664 type Item = SpannedEvent<'a>;
666 fn next(&mut self) -> Option<Self::Item> {
668 match self.inner.next() {
669 Some((Event::FootnoteReference(ref reference), range)) => {
670 let entry = self.get_entry(reference);
671 let reference = format!(
672 "<sup id=\"fnref{0}\"><a href=\"#fn{0}\">{0}</a></sup>",
675 return Some((Event::Html(reference.into()), range));
677 Some((Event::Start(Tag::FootnoteDefinition(def)), _)) => {
678 let mut content = Vec::new();
679 for (event, _) in &mut self.inner {
680 if let Event::End(Tag::FootnoteDefinition(..)) = event {
685 let entry = self.get_entry(&def);
686 (*entry).0 = content;
688 Some(e) => return Some(e),
690 if !self.footnotes.is_empty() {
691 let mut v: Vec<_> = self.footnotes.drain().map(|(_, x)| x).collect();
692 v.sort_by(|a, b| a.1.cmp(&b.1));
693 let mut ret = String::from("<div class=\"footnotes\"><hr><ol>");
694 for (mut content, id) in v {
695 write!(ret, "<li id=\"fn{}\">", id).unwrap();
696 let mut is_paragraph = false;
697 if let Some(&Event::End(Tag::Paragraph)) = content.last() {
701 html::push_html(&mut ret, content.into_iter());
702 write!(ret, " <a href=\"#fnref{}\">↩</a>", id).unwrap();
704 ret.push_str("</p>");
706 ret.push_str("</li>");
708 ret.push_str("</ol></div>");
709 return Some((Event::Html(ret.into()), 0..0));
719 pub(crate) fn find_testable_code<T: doctest::Tester>(
722 error_codes: ErrorCodes,
723 enable_per_target_ignores: bool,
724 extra_info: Option<&ExtraInfo<'_>>,
726 let mut parser = Parser::new(doc).into_offset_iter();
727 let mut prev_offset = 0;
728 let mut nb_lines = 0;
729 let mut register_header = None;
730 while let Some((event, offset)) = parser.next() {
732 Event::Start(Tag::CodeBlock(kind)) => {
733 let block_info = match kind {
734 CodeBlockKind::Fenced(ref lang) => {
741 enable_per_target_ignores,
746 CodeBlockKind::Indented => Default::default(),
748 if !block_info.rust {
752 let mut test_s = String::new();
754 while let Some((Event::Text(s), _)) = parser.next() {
759 .map(|l| map_line(l).for_code())
760 .collect::<Vec<Cow<'_, str>>>()
763 nb_lines += doc[prev_offset..offset.start].lines().count();
764 // If there are characters between the preceding line ending and
765 // this code block, `str::lines` will return an additional line,
766 // which we subtract here.
767 if nb_lines != 0 && !&doc[prev_offset..offset.start].ends_with('\n') {
770 let line = tests.get_line() + nb_lines + 1;
771 tests.add_test(text, block_info, line);
772 prev_offset = offset.start;
774 Event::Start(Tag::Heading(level, _, _)) => {
775 register_header = Some(level as u32);
777 Event::Text(ref s) if register_header.is_some() => {
778 let level = register_header.unwrap();
780 tests.register_header("", level);
782 tests.register_header(s, level);
784 register_header = None;
791 pub(crate) struct ExtraInfo<'tcx> {
802 impl<'tcx> ExtraInfo<'tcx> {
803 pub(crate) fn new(tcx: TyCtxt<'tcx>, hir_id: HirId, sp: Span) -> ExtraInfo<'tcx> {
804 ExtraInfo { id: ExtraInfoId::Hir(hir_id), sp, tcx }
807 pub(crate) fn new_did(tcx: TyCtxt<'tcx>, did: DefId, sp: Span) -> ExtraInfo<'tcx> {
808 ExtraInfo { id: ExtraInfoId::Def(did), sp, tcx }
811 fn error_invalid_codeblock_attr(&self, msg: &str, help: &str) {
812 let hir_id = match self.id {
813 ExtraInfoId::Hir(hir_id) => hir_id,
814 ExtraInfoId::Def(item_did) => {
815 match item_did.as_local() {
816 Some(item_did) => self.tcx.hir().local_def_id_to_hir_id(item_did),
818 // If non-local, no need to check anything.
824 self.tcx.struct_span_lint_hir(
825 crate::lint::INVALID_CODEBLOCK_ATTRIBUTES,
829 let mut diag = lint.build(msg);
837 #[derive(Eq, PartialEq, Clone, Debug)]
838 pub(crate) struct LangString {
840 pub(crate) should_panic: bool,
841 pub(crate) no_run: bool,
842 pub(crate) ignore: Ignore,
843 pub(crate) rust: bool,
844 pub(crate) test_harness: bool,
845 pub(crate) compile_fail: bool,
846 pub(crate) error_codes: Vec<String>,
847 pub(crate) edition: Option<Edition>,
850 #[derive(Eq, PartialEq, Clone, Debug)]
851 pub(crate) enum Ignore {
857 impl Default for LangString {
858 fn default() -> Self {
860 original: String::new(),
863 ignore: Ignore::None,
867 error_codes: Vec::new(),
874 fn parse_without_check(
876 allow_error_code_check: ErrorCodes,
877 enable_per_target_ignores: bool,
879 Self::parse(string, allow_error_code_check, enable_per_target_ignores, None)
882 fn tokens(string: &str) -> impl Iterator<Item = &str> {
883 // Pandoc, which Rust once used for generating documentation,
884 // expects lang strings to be surrounded by `{}` and for each token
885 // to be proceeded by a `.`. Since some of these lang strings are still
886 // loose in the wild, we strip a pair of surrounding `{}` from the lang
887 // string and a leading `.` from each token.
889 let string = string.trim();
891 let first = string.chars().next();
892 let last = string.chars().last();
894 let string = if first == Some('{') && last == Some('}') {
895 &string[1..string.len() - 1]
901 .split(|c| c == ',' || c == ' ' || c == '\t')
903 .map(|token| token.strip_prefix('.').unwrap_or(token))
904 .filter(|token| !token.is_empty())
909 allow_error_code_check: ErrorCodes,
910 enable_per_target_ignores: bool,
911 extra: Option<&ExtraInfo<'_>>,
913 let allow_error_code_check = allow_error_code_check.as_bool();
914 let mut seen_rust_tags = false;
915 let mut seen_other_tags = false;
916 let mut data = LangString::default();
917 let mut ignores = vec![];
919 data.original = string.to_owned();
921 for token in Self::tokens(string) {
924 data.should_panic = true;
925 seen_rust_tags = !seen_other_tags;
929 seen_rust_tags = !seen_other_tags;
932 data.ignore = Ignore::All;
933 seen_rust_tags = !seen_other_tags;
935 x if x.starts_with("ignore-") => {
936 if enable_per_target_ignores {
937 ignores.push(x.trim_start_matches("ignore-").to_owned());
938 seen_rust_tags = !seen_other_tags;
943 seen_rust_tags = true;
946 data.test_harness = true;
947 seen_rust_tags = !seen_other_tags || seen_rust_tags;
950 data.compile_fail = true;
951 seen_rust_tags = !seen_other_tags || seen_rust_tags;
954 x if x.starts_with("edition") => {
955 data.edition = x[7..].parse::<Edition>().ok();
957 x if allow_error_code_check && x.starts_with('E') && x.len() == 5 => {
958 if x[1..].parse::<u32>().is_ok() {
959 data.error_codes.push(x.to_owned());
960 seen_rust_tags = !seen_other_tags || seen_rust_tags;
962 seen_other_tags = true;
965 x if extra.is_some() => {
966 let s = x.to_lowercase();
967 if let Some((flag, help)) = if s == "compile-fail"
968 || s == "compile_fail"
969 || s == "compilefail"
973 "the code block will either not be tested if not marked as a rust one \
974 or won't fail if it compiles successfully",
976 } else if s == "should-panic" || s == "should_panic" || s == "shouldpanic" {
979 "the code block will either not be tested if not marked as a rust one \
980 or won't fail if it doesn't panic when running",
982 } else if s == "no-run" || s == "no_run" || s == "norun" {
985 "the code block will either not be tested if not marked as a rust one \
986 or will be run (which you might not want)",
988 } else if s == "test-harness" || s == "test_harness" || s == "testharness" {
991 "the code block will either not be tested if not marked as a rust one \
992 or the code will be wrapped inside a main function",
997 if let Some(extra) = extra {
998 extra.error_invalid_codeblock_attr(
999 &format!("unknown attribute `{}`. Did you mean `{}`?", x, flag),
1004 seen_other_tags = true;
1006 _ => seen_other_tags = true,
1010 // ignore-foo overrides ignore
1011 if !ignores.is_empty() {
1012 data.ignore = Ignore::Some(ignores);
1015 data.rust &= !seen_other_tags || seen_rust_tags;
1022 pub fn into_string(self) -> String {
1033 // This is actually common enough to special-case
1035 return String::new();
1037 let mut replacer = |broken_link: BrokenLink<'_>| {
1040 .find(|link| link.original_text.as_str() == &*broken_link.reference)
1041 .map(|link| (link.href.as_str().into(), link.new_text.as_str().into()))
1044 let p = Parser::new_with_broken_link_callback(md, main_body_opts(), Some(&mut replacer));
1045 let p = p.into_offset_iter();
1047 let mut s = String::with_capacity(md.len() * 3 / 2);
1049 let p = HeadingLinks::new(p, None, ids, heading_offset);
1050 let p = Footnotes::new(p);
1051 let p = LinkReplacer::new(p.map(|(ev, _)| ev), links);
1052 let p = TableWrapper::new(p);
1053 let p = CodeBlocks::new(p, codes, edition, playground);
1054 html::push_html(&mut s, p);
1060 impl MarkdownWithToc<'_> {
1061 pub(crate) fn into_string(self) -> String {
1062 let MarkdownWithToc(md, ids, codes, edition, playground) = self;
1064 let p = Parser::new_ext(md, main_body_opts()).into_offset_iter();
1066 let mut s = String::with_capacity(md.len() * 3 / 2);
1068 let mut toc = TocBuilder::new();
1071 let p = HeadingLinks::new(p, Some(&mut toc), ids, HeadingOffset::H1);
1072 let p = Footnotes::new(p);
1073 let p = TableWrapper::new(p.map(|(ev, _)| ev));
1074 let p = CodeBlocks::new(p, codes, edition, playground);
1075 html::push_html(&mut s, p);
1078 format!("<nav id=\"TOC\">{}</nav>{}", toc.into_toc().print(), s)
1082 impl MarkdownHtml<'_> {
1083 pub(crate) fn into_string(self) -> String {
1084 let MarkdownHtml(md, ids, codes, edition, playground) = self;
1086 // This is actually common enough to special-case
1088 return String::new();
1090 let p = Parser::new_ext(md, main_body_opts()).into_offset_iter();
1092 // Treat inline HTML as plain text.
1093 let p = p.map(|event| match event.0 {
1094 Event::Html(text) => (Event::Text(text), event.1),
1098 let mut s = String::with_capacity(md.len() * 3 / 2);
1100 let p = HeadingLinks::new(p, None, ids, HeadingOffset::H1);
1101 let p = Footnotes::new(p);
1102 let p = TableWrapper::new(p.map(|(ev, _)| ev));
1103 let p = CodeBlocks::new(p, codes, edition, playground);
1104 html::push_html(&mut s, p);
1110 impl MarkdownSummaryLine<'_> {
1111 pub(crate) fn into_string(self) -> String {
1112 let MarkdownSummaryLine(md, links) = self;
1113 // This is actually common enough to special-case
1115 return String::new();
1118 let mut replacer = |broken_link: BrokenLink<'_>| {
1121 .find(|link| link.original_text.as_str() == &*broken_link.reference)
1122 .map(|link| (link.href.as_str().into(), link.new_text.as_str().into()))
1125 let p = Parser::new_with_broken_link_callback(md, summary_opts(), Some(&mut replacer));
1127 let mut s = String::new();
1129 html::push_html(&mut s, LinkReplacer::new(SummaryLine::new(p), links));
1135 /// Renders a subset of Markdown in the first paragraph of the provided Markdown.
1137 /// - *Italics*, **bold**, and `inline code` styles **are** rendered.
1138 /// - Headings and links are stripped (though the text *is* rendered).
1139 /// - HTML, code blocks, and everything else are ignored.
1141 /// Returns a tuple of the rendered HTML string and whether the output was shortened
1142 /// due to the provided `length_limit`.
1143 fn markdown_summary_with_limit(
1145 link_names: &[RenderedLink],
1146 length_limit: usize,
1147 ) -> (String, bool) {
1149 return (String::new(), false);
1152 let mut replacer = |broken_link: BrokenLink<'_>| {
1155 .find(|link| link.original_text.as_str() == &*broken_link.reference)
1156 .map(|link| (link.href.as_str().into(), link.new_text.as_str().into()))
1159 let p = Parser::new_with_broken_link_callback(md, summary_opts(), Some(&mut replacer));
1160 let mut p = LinkReplacer::new(p, link_names);
1162 let mut buf = HtmlWithLimit::new(length_limit);
1163 let mut stopped_early = false;
1164 p.try_for_each(|event| {
1166 Event::Text(text) => {
1168 text.split_inclusive(char::is_whitespace).try_for_each(|word| buf.push(word));
1170 stopped_early = true;
1174 Event::Code(code) => {
1175 buf.open_tag("code");
1176 let r = buf.push(code);
1178 stopped_early = true;
1184 Event::Start(tag) => match tag {
1185 Tag::Emphasis => buf.open_tag("em"),
1186 Tag::Strong => buf.open_tag("strong"),
1187 Tag::CodeBlock(..) => return ControlFlow::BREAK,
1190 Event::End(tag) => match tag {
1191 Tag::Emphasis | Tag::Strong => buf.close_tag(),
1192 Tag::Paragraph | Tag::Heading(..) => return ControlFlow::BREAK,
1195 Event::HardBreak | Event::SoftBreak => buf.push(" ")?,
1198 ControlFlow::CONTINUE
1201 (buf.finish(), stopped_early)
1204 /// Renders a shortened first paragraph of the given Markdown as a subset of Markdown,
1205 /// making it suitable for contexts like the search index.
1207 /// Will shorten to 59 or 60 characters, including an ellipsis (…) if it was shortened.
1209 /// See [`markdown_summary_with_limit`] for details about what is rendered and what is not.
1210 pub(crate) fn short_markdown_summary(markdown: &str, link_names: &[RenderedLink]) -> String {
1211 let (mut s, was_shortened) = markdown_summary_with_limit(markdown, link_names, 59);
1220 /// Renders the first paragraph of the provided markdown as plain text.
1221 /// Useful for alt-text.
1223 /// - Headings, links, and formatting are stripped.
1224 /// - Inline code is rendered as-is, surrounded by backticks.
1225 /// - HTML and code blocks are ignored.
1226 pub(crate) fn plain_text_summary(md: &str) -> String {
1228 return String::new();
1231 let mut s = String::with_capacity(md.len() * 3 / 2);
1233 for event in Parser::new_ext(md, summary_opts()) {
1235 Event::Text(text) => s.push_str(text),
1236 Event::Code(code) => {
1241 Event::HardBreak | Event::SoftBreak => s.push(' '),
1242 Event::Start(Tag::CodeBlock(..)) => break,
1243 Event::End(Tag::Paragraph) => break,
1244 Event::End(Tag::Heading(..)) => break,
1253 pub(crate) struct MarkdownLink {
1256 pub range: Range<usize>,
1259 pub(crate) fn markdown_links<R>(
1261 filter_map: impl Fn(MarkdownLink) -> Option<R>,
1267 let links = RefCell::new(vec![]);
1269 // FIXME: remove this function once pulldown_cmark can provide spans for link definitions.
1270 let locate = |s: &str, fallback: Range<usize>| unsafe {
1271 let s_start = s.as_ptr();
1272 let s_end = s_start.add(s.len());
1273 let md_start = md.as_ptr();
1274 let md_end = md_start.add(md.len());
1275 if md_start <= s_start && s_end <= md_end {
1276 let start = s_start.offset_from(md_start) as usize;
1277 let end = s_end.offset_from(md_start) as usize;
1284 let span_for_link = |link: &CowStr<'_>, span: Range<usize>| {
1285 // For diagnostics, we want to underline the link's definition but `span` will point at
1286 // where the link is used. This is a problem for reference-style links, where the definition
1287 // is separate from the usage.
1289 // `Borrowed` variant means the string (the link's destination) may come directly from
1290 // the markdown text and we can locate the original link destination.
1291 // NOTE: LinkReplacer also provides `Borrowed` but possibly from other sources,
1292 // so `locate()` can fall back to use `span`.
1293 CowStr::Borrowed(s) => locate(s, span),
1295 // For anything else, we can only use the provided range.
1296 CowStr::Boxed(_) | CowStr::Inlined(_) => span,
1300 let mut push = |link: BrokenLink<'_>| {
1301 let span = span_for_link(&link.reference, link.span);
1302 filter_map(MarkdownLink {
1303 kind: LinkType::ShortcutUnknown,
1304 link: link.reference.to_string(),
1307 .map(|link| links.borrow_mut().push(link));
1310 let p = Parser::new_with_broken_link_callback(md, main_body_opts(), Some(&mut push))
1311 .into_offset_iter();
1313 // There's no need to thread an IdMap through to here because
1314 // the IDs generated aren't going to be emitted anywhere.
1315 let mut ids = IdMap::new();
1316 let iter = Footnotes::new(HeadingLinks::new(p, None, &mut ids, HeadingOffset::H1));
1319 if let Event::Start(Tag::Link(
1320 // `<>` links cannot be intra-doc links so we skip them.
1321 kind @ (LinkType::Inline
1322 | LinkType::Reference
1323 | LinkType::ReferenceUnknown
1324 | LinkType::Collapsed
1325 | LinkType::CollapsedUnknown
1326 | LinkType::Shortcut
1327 | LinkType::ShortcutUnknown),
1332 debug!("found link: {dest}");
1333 let span = span_for_link(&dest, ev.1);
1334 filter_map(MarkdownLink { kind, link: dest.into_string(), range: span })
1335 .map(|link| links.borrow_mut().push(link));
1343 pub(crate) struct RustCodeBlock {
1344 /// The range in the markdown that the code block occupies. Note that this includes the fences
1345 /// for fenced code blocks.
1346 pub(crate) range: Range<usize>,
1347 /// The range in the markdown that the code within the code block occupies.
1348 pub(crate) code: Range<usize>,
1349 pub(crate) is_fenced: bool,
1350 pub(crate) lang_string: LangString,
1353 /// Returns a range of bytes for each code block in the markdown that is tagged as `rust` or
1354 /// untagged (and assumed to be rust).
1355 pub(crate) fn rust_code_blocks(md: &str, extra_info: &ExtraInfo<'_>) -> Vec<RustCodeBlock> {
1356 let mut code_blocks = vec![];
1362 let mut p = Parser::new_ext(md, main_body_opts()).into_offset_iter();
1364 while let Some((event, offset)) = p.next() {
1365 if let Event::Start(Tag::CodeBlock(syntax)) = event {
1366 let (lang_string, code_start, code_end, range, is_fenced) = match syntax {
1367 CodeBlockKind::Fenced(syntax) => {
1368 let syntax = syntax.as_ref();
1369 let lang_string = if syntax.is_empty() {
1372 LangString::parse(&*syntax, ErrorCodes::Yes, false, Some(extra_info))
1374 if !lang_string.rust {
1377 let (code_start, mut code_end) = match p.next() {
1378 Some((Event::Text(_), offset)) => (offset.start, offset.end),
1379 Some((_, sub_offset)) => {
1380 let code = Range { start: sub_offset.start, end: sub_offset.start };
1381 code_blocks.push(RustCodeBlock {
1390 let code = Range { start: offset.end, end: offset.end };
1391 code_blocks.push(RustCodeBlock {
1400 while let Some((Event::Text(_), offset)) = p.next() {
1401 code_end = offset.end;
1403 (lang_string, code_start, code_end, offset, true)
1405 CodeBlockKind::Indented => {
1406 // The ending of the offset goes too far sometime so we reduce it by one in
1408 if offset.end > offset.start && md.get(offset.end..=offset.end) == Some("\n") {
1410 LangString::default(),
1413 Range { start: offset.start, end: offset.end - 1 },
1417 (LangString::default(), offset.start, offset.end, offset, false)
1422 code_blocks.push(RustCodeBlock {
1425 code: Range { start: code_start, end: code_end },
1434 #[derive(Clone, Default, Debug)]
1436 map: FxHashMap<Cow<'static, str>, usize>,
1439 // The map is pre-initialized and cloned each time to avoid reinitializing it repeatedly.
1440 static DEFAULT_ID_MAP: Lazy<FxHashMap<Cow<'static, str>, usize>> = Lazy::new(|| init_id_map());
1442 fn init_id_map() -> FxHashMap<Cow<'static, str>, usize> {
1443 let mut map = FxHashMap::default();
1444 // This is the list of IDs used in Javascript.
1445 map.insert("help".into(), 1);
1446 map.insert("settings".into(), 1);
1447 map.insert("not-displayed".into(), 1);
1448 map.insert("alternative-display".into(), 1);
1449 map.insert("search".into(), 1);
1450 // This is the list of IDs used in HTML generated in Rust (including the ones
1451 // used in tera template files).
1452 map.insert("mainThemeStyle".into(), 1);
1453 map.insert("themeStyle".into(), 1);
1454 map.insert("settings-menu".into(), 1);
1455 map.insert("help-button".into(), 1);
1456 map.insert("main-content".into(), 1);
1457 map.insert("crate-search".into(), 1);
1458 map.insert("render-detail".into(), 1);
1459 map.insert("toggle-all-docs".into(), 1);
1460 map.insert("all-types".into(), 1);
1461 map.insert("default-settings".into(), 1);
1462 map.insert("rustdoc-vars".into(), 1);
1463 map.insert("sidebar-vars".into(), 1);
1464 map.insert("copy-path".into(), 1);
1465 map.insert("TOC".into(), 1);
1466 // This is the list of IDs used by rustdoc sections (but still generated by
1468 map.insert("fields".into(), 1);
1469 map.insert("variants".into(), 1);
1470 map.insert("implementors-list".into(), 1);
1471 map.insert("synthetic-implementors-list".into(), 1);
1472 map.insert("foreign-impls".into(), 1);
1473 map.insert("implementations".into(), 1);
1474 map.insert("trait-implementations".into(), 1);
1475 map.insert("synthetic-implementations".into(), 1);
1476 map.insert("blanket-implementations".into(), 1);
1477 map.insert("required-associated-types".into(), 1);
1478 map.insert("provided-associated-types".into(), 1);
1479 map.insert("provided-associated-consts".into(), 1);
1480 map.insert("required-associated-consts".into(), 1);
1481 map.insert("required-methods".into(), 1);
1482 map.insert("provided-methods".into(), 1);
1483 map.insert("implementors".into(), 1);
1484 map.insert("synthetic-implementors".into(), 1);
1485 map.insert("implementations-list".into(), 1);
1486 map.insert("trait-implementations-list".into(), 1);
1487 map.insert("synthetic-implementations-list".into(), 1);
1488 map.insert("blanket-implementations-list".into(), 1);
1489 map.insert("deref-methods".into(), 1);
1494 pub fn new() -> Self {
1495 IdMap { map: DEFAULT_ID_MAP.clone() }
1498 pub(crate) fn derive<S: AsRef<str> + ToString>(&mut self, candidate: S) -> String {
1499 let id = match self.map.get_mut(candidate.as_ref()) {
1500 None => candidate.to_string(),
1502 let id = format!("{}-{}", candidate.as_ref(), *a);
1508 self.map.insert(id.clone().into(), 1);