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::{IdMap, Markdown, ErrorCodes};
13 //! let s = "My *markdown* _text_";
14 //! let mut id_map = IdMap::new();
15 //! let md = Markdown(s, &[], &mut id_map, ErrorCodes::Yes, Edition::Edition2015, &None);
16 //! let html = md.into_string();
17 //! // ... something using html
20 use rustc_data_structures::fx::FxHashMap;
21 use rustc_hir::def_id::DefId;
23 use rustc_middle::ty::TyCtxt;
24 use rustc_span::edition::Edition;
28 use std::cell::RefCell;
29 use std::collections::VecDeque;
30 use std::default::Default;
32 use std::ops::{ControlFlow, Range};
35 use crate::clean::RenderedLink;
37 use crate::html::escape::Escape;
38 use crate::html::format::Buffer;
39 use crate::html::highlight;
40 use crate::html::length_limit::HtmlWithLimit;
41 use crate::html::toc::TocBuilder;
44 html, BrokenLink, CodeBlockKind, CowStr, Event, LinkType, Options, Parser, Tag,
50 /// Options for rendering Markdown in the main body of documentation.
51 pub(crate) fn opts() -> Options {
52 Options::ENABLE_TABLES
53 | Options::ENABLE_FOOTNOTES
54 | Options::ENABLE_STRIKETHROUGH
55 | Options::ENABLE_TASKLISTS
56 | Options::ENABLE_SMART_PUNCTUATION
59 /// A subset of [`opts()`] used for rendering summaries.
60 pub(crate) fn summary_opts() -> Options {
61 Options::ENABLE_STRIKETHROUGH | Options::ENABLE_SMART_PUNCTUATION | Options::ENABLE_TABLES
64 /// When `to_string` is called, this struct will emit the HTML corresponding to
65 /// the rendered version of the contained markdown string.
66 pub struct Markdown<'a>(
68 /// A list of link replacements.
69 pub &'a [RenderedLink],
70 /// The current list of used header IDs.
72 /// Whether to allow the use of explicit error codes in doctest lang strings.
74 /// Default edition to use when parsing doctests (to add a `fn main`).
76 pub &'a Option<Playground>,
78 /// A tuple struct like `Markdown` that renders the markdown with a table of contents.
79 crate struct MarkdownWithToc<'a>(
84 crate &'a Option<Playground>,
86 /// A tuple struct like `Markdown` that renders the markdown escaping HTML tags.
87 crate struct MarkdownHtml<'a>(
92 crate &'a Option<Playground>,
94 /// A tuple struct like `Markdown` that renders only the first paragraph.
95 crate struct MarkdownSummaryLine<'a>(pub &'a str, pub &'a [RenderedLink]);
97 #[derive(Copy, Clone, PartialEq, Debug)]
104 crate fn from(b: bool) -> Self {
106 true => ErrorCodes::Yes,
107 false => ErrorCodes::No,
111 crate fn as_bool(self) -> bool {
113 ErrorCodes::Yes => true,
114 ErrorCodes::No => false,
119 /// Controls whether a line will be hidden or shown in HTML output.
121 /// All lines are used in documentation tests.
128 fn for_html(self) -> Option<Cow<'a, str>> {
130 Line::Shown(l) => Some(l),
131 Line::Hidden(_) => None,
135 fn for_code(self) -> Cow<'a, str> {
138 Line::Hidden(l) => Cow::Borrowed(l),
143 // FIXME: There is a minor inconsistency here. For lines that start with ##, we
144 // have no easy way of removing a potential single space after the hashes, which
145 // is done in the single # case. This inconsistency seems okay, if non-ideal. In
146 // order to fix it we'd have to iterate to find the first non-# character, and
147 // then reallocate to remove it; which would make us return a String.
148 fn map_line(s: &str) -> Line<'_> {
149 let trimmed = s.trim();
150 if trimmed.starts_with("##") {
151 Line::Shown(Cow::Owned(s.replacen("##", "#", 1)))
152 } else if let Some(stripped) = trimmed.strip_prefix("# ") {
154 Line::Hidden(&stripped)
155 } else if trimmed == "#" {
156 // We cannot handle '#text' because it could be #[attr].
159 Line::Shown(Cow::Borrowed(s))
163 /// Convert chars from a title for an id.
165 /// "Hello, world!" -> "hello-world"
166 fn slugify(c: char) -> Option<char> {
167 if c.is_alphanumeric() || c == '-' || c == '_' {
168 if c.is_ascii() { Some(c.to_ascii_lowercase()) } else { Some(c) }
169 } else if c.is_whitespace() && c.is_ascii() {
176 #[derive(Clone, Debug)]
177 pub struct Playground {
178 pub crate_name: Option<String>,
182 /// Adds syntax highlighting and playground Run buttons to Rust code blocks.
183 struct CodeBlocks<'p, 'a, I: Iterator<Item = Event<'a>>> {
185 check_error_codes: ErrorCodes,
187 // Information about the playground if a URL has been specified, containing an
188 // optional crate name and the URL.
189 playground: &'p Option<Playground>,
192 impl<'p, 'a, I: Iterator<Item = Event<'a>>> CodeBlocks<'p, 'a, I> {
195 error_codes: ErrorCodes,
197 playground: &'p Option<Playground>,
199 CodeBlocks { inner: iter, check_error_codes: error_codes, edition, playground }
203 impl<'a, I: Iterator<Item = Event<'a>>> Iterator for CodeBlocks<'_, 'a, I> {
204 type Item = Event<'a>;
206 fn next(&mut self) -> Option<Self::Item> {
207 let event = self.inner.next();
212 let kind = if let Some(Event::Start(Tag::CodeBlock(kind))) = event {
218 let mut origtext = String::new();
219 for event in &mut self.inner {
221 Event::End(Tag::CodeBlock(..)) => break,
222 Event::Text(ref s) => {
223 origtext.push_str(s);
228 let lines = origtext.lines().filter_map(|l| map_line(l).for_html());
229 let text = lines.collect::<Vec<Cow<'_, str>>>().join("\n");
231 let parse_result = match kind {
232 CodeBlockKind::Fenced(ref lang) => {
234 LangString::parse_without_check(&lang, self.check_error_codes, false);
235 if !parse_result.rust {
236 return Some(Event::Html(
238 "<div class=\"example-wrap\">\
239 <pre class=\"language-{}\"><code>{}</code></pre>\
249 CodeBlockKind::Indented => Default::default(),
252 compile_fail = parse_result.compile_fail;
253 should_panic = parse_result.should_panic;
254 ignore = parse_result.ignore;
255 edition = parse_result.edition;
257 let explicit_edition = edition.is_some();
258 let edition = edition.unwrap_or(self.edition);
260 let playground_button = self.playground.as_ref().and_then(|playground| {
261 let krate = &playground.crate_name;
262 let url = &playground.url;
268 .map(|l| map_line(l).for_code())
269 .collect::<Vec<Cow<'_, str>>>()
271 let krate = krate.as_ref().map(|s| &**s);
273 doctest::make_test(&test, krate, false, &Default::default(), edition, None);
274 let channel = if test.contains("#![feature(") { "&version=nightly" } else { "" };
276 let edition_string = format!("&edition={}", edition);
278 // These characters don't need to be escaped in a URI.
279 // FIXME: use a library function for percent encoding.
280 fn dont_escape(c: u8) -> bool {
281 (b'a' <= c && c <= b'z')
282 || (b'A' <= c && c <= b'Z')
283 || (b'0' <= c && c <= b'9')
294 let mut test_escaped = String::new();
295 for b in test.bytes() {
297 test_escaped.push(char::from(b));
299 write!(test_escaped, "%{:02X}", b).unwrap();
303 r#"<a class="test-arrow" target="_blank" href="{}?code={}{}{}">Run</a>"#,
304 url, test_escaped, channel, edition_string
308 let tooltip = if ignore != Ignore::None {
309 Some((None, "ignore"))
310 } else if compile_fail {
311 Some((None, "compile_fail"))
312 } else if should_panic {
313 Some((None, "should_panic"))
314 } else if explicit_edition {
315 Some((Some(edition), "edition"))
320 // insert newline to clearly separate it from the
321 // previous block so we can shorten the html output
322 let mut s = Buffer::new();
324 highlight::render_with_highlighting(
328 "rust-example-rendered{}",
329 if let Some((_, class)) = tooltip { format!(" {}", class) } else { String::new() }
331 playground_button.as_deref(),
337 Some(Event::Html(s.into_inner().into()))
341 /// Make headings links with anchor IDs and build up TOC.
342 struct LinkReplacer<'a, I: Iterator<Item = Event<'a>>> {
344 links: &'a [RenderedLink],
345 shortcut_link: Option<&'a RenderedLink>,
348 impl<'a, I: Iterator<Item = Event<'a>>> LinkReplacer<'a, I> {
349 fn new(iter: I, links: &'a [RenderedLink]) -> Self {
350 LinkReplacer { inner: iter, links, shortcut_link: None }
354 impl<'a, I: Iterator<Item = Event<'a>>> Iterator for LinkReplacer<'a, I> {
355 type Item = Event<'a>;
357 fn next(&mut self) -> Option<Self::Item> {
358 let mut event = self.inner.next();
360 // Replace intra-doc links and remove disambiguators from shortcut links (`[fn@f]`).
362 // This is a shortcut link that was resolved by the broken_link_callback: `[fn@f]`
363 // Remove any disambiguator.
364 Some(Event::Start(Tag::Link(
365 // [fn@f] or [fn@f][]
366 LinkType::ShortcutUnknown | LinkType::CollapsedUnknown,
370 debug!("saw start of shortcut link to {} with title {}", dest, title);
371 // If this is a shortcut link, it was resolved by the broken_link_callback.
372 // So the URL will already be updated properly.
373 let link = self.links.iter().find(|&link| *link.href == **dest);
374 // Since this is an external iterator, we can't replace the inner text just yet.
375 // Store that we saw a link so we know to replace it later.
376 if let Some(link) = link {
377 trace!("it matched");
378 assert!(self.shortcut_link.is_none(), "shortcut links cannot be nested");
379 self.shortcut_link = Some(link);
382 // Now that we're done with the shortcut link, don't replace any more text.
383 Some(Event::End(Tag::Link(
384 LinkType::ShortcutUnknown | LinkType::CollapsedUnknown,
388 debug!("saw end of shortcut link to {}", dest);
389 if self.links.iter().any(|link| *link.href == **dest) {
390 assert!(self.shortcut_link.is_some(), "saw closing link without opening tag");
391 self.shortcut_link = None;
394 // Handle backticks in inline code blocks, but only if we're in the middle of a shortcut link.
396 Some(Event::Code(text)) => {
397 trace!("saw code {}", text);
398 if let Some(link) = self.shortcut_link {
399 trace!("original text was {}", link.original_text);
400 // NOTE: this only replaces if the code block is the *entire* text.
401 // If only part of the link has code highlighting, the disambiguator will not be removed.
403 // This is a limitation from `collect_intra_doc_links`: it passes a full link,
404 // and does not distinguish at all between code blocks.
405 // So we could never be sure we weren't replacing too much:
406 // [fn@my_`f`unc] is treated the same as [my_func()] in that pass.
408 // NOTE: &[1..len() - 1] is to strip the backticks
409 if **text == link.original_text[1..link.original_text.len() - 1] {
410 debug!("replacing {} with {}", text, link.new_text);
411 *text = CowStr::Borrowed(&link.new_text);
415 // Replace plain text in links, but only in the middle of a shortcut link.
417 Some(Event::Text(text)) => {
418 trace!("saw text {}", text);
419 if let Some(link) = self.shortcut_link {
420 trace!("original text was {}", link.original_text);
421 // NOTE: same limitations as `Event::Code`
422 if **text == *link.original_text {
423 debug!("replacing {} with {}", text, link.new_text);
424 *text = CowStr::Borrowed(&link.new_text);
428 // If this is a link, but not a shortcut link,
429 // replace the URL, since the broken_link_callback was not called.
430 Some(Event::Start(Tag::Link(_, dest, _))) => {
431 if let Some(link) = self.links.iter().find(|&link| *link.original_text == **dest) {
432 *dest = CowStr::Borrowed(link.href.as_ref());
435 // Anything else couldn't have been a valid Rust path, so no need to replace the text.
439 // Yield the modified event
444 type SpannedEvent<'a> = (Event<'a>, Range<usize>);
446 /// Make headings links with anchor IDs and build up TOC.
447 struct HeadingLinks<'a, 'b, 'ids, I> {
449 toc: Option<&'b mut TocBuilder>,
450 buf: VecDeque<SpannedEvent<'a>>,
451 id_map: &'ids mut IdMap,
454 impl<'a, 'b, 'ids, I> HeadingLinks<'a, 'b, 'ids, I> {
455 fn new(iter: I, toc: Option<&'b mut TocBuilder>, ids: &'ids mut IdMap) -> Self {
456 HeadingLinks { inner: iter, toc, buf: VecDeque::new(), id_map: ids }
460 impl<'a, 'b, 'ids, I: Iterator<Item = SpannedEvent<'a>>> Iterator
461 for HeadingLinks<'a, 'b, 'ids, I>
463 type Item = SpannedEvent<'a>;
465 fn next(&mut self) -> Option<Self::Item> {
466 if let Some(e) = self.buf.pop_front() {
470 let event = self.inner.next();
471 if let Some((Event::Start(Tag::Heading(level)), _)) = event {
472 let mut id = String::new();
473 for event in &mut self.inner {
475 Event::End(Tag::Heading(..)) => break,
476 Event::Start(Tag::Link(_, _, _)) | Event::End(Tag::Link(..)) => {}
477 Event::Text(text) | Event::Code(text) => {
478 id.extend(text.chars().filter_map(slugify));
479 self.buf.push_back(event);
481 _ => self.buf.push_back(event),
484 let id = self.id_map.derive(id);
486 if let Some(ref mut builder) = self.toc {
487 let mut html_header = String::new();
488 html::push_html(&mut html_header, self.buf.iter().map(|(ev, _)| ev.clone()));
489 let sec = builder.push(level as u32, html_header, id.clone());
490 self.buf.push_front((Event::Html(format!("{} ", sec).into()), 0..0));
493 self.buf.push_back((Event::Html(format!("</a></h{}>", level).into()), 0..0));
495 let start_tags = format!(
496 "<h{level} id=\"{id}\" class=\"section-header\">\
501 return Some((Event::Html(start_tags.into()), 0..0));
507 /// Extracts just the first paragraph.
508 struct SummaryLine<'a, I: Iterator<Item = Event<'a>>> {
514 impl<'a, I: Iterator<Item = Event<'a>>> SummaryLine<'a, I> {
515 fn new(iter: I) -> Self {
516 SummaryLine { inner: iter, started: false, depth: 0 }
520 fn check_if_allowed_tag(t: &Tag<'_>) -> bool {
523 Tag::Paragraph | Tag::Item | Tag::Emphasis | Tag::Strong | Tag::Link(..) | Tag::BlockQuote
527 fn is_forbidden_tag(t: &Tag<'_>) -> bool {
528 matches!(t, Tag::CodeBlock(_) | Tag::Table(_) | Tag::TableHead | Tag::TableRow | Tag::TableCell)
531 impl<'a, I: Iterator<Item = Event<'a>>> Iterator for SummaryLine<'a, I> {
532 type Item = Event<'a>;
534 fn next(&mut self) -> Option<Self::Item> {
535 if self.started && self.depth == 0 {
541 if let Some(event) = self.inner.next() {
542 let mut is_start = true;
543 let is_allowed_tag = match event {
544 Event::Start(ref c) => {
545 if is_forbidden_tag(c) {
549 check_if_allowed_tag(c)
551 Event::End(ref c) => {
552 if is_forbidden_tag(c) {
557 check_if_allowed_tag(c)
561 return if !is_allowed_tag {
563 Some(Event::Start(Tag::Paragraph))
565 Some(Event::End(Tag::Paragraph))
575 /// Moves all footnote definitions to the end and add back links to the
577 struct Footnotes<'a, I> {
579 footnotes: FxHashMap<String, (Vec<Event<'a>>, u16)>,
582 impl<'a, I> Footnotes<'a, I> {
583 fn new(iter: I) -> Self {
584 Footnotes { inner: iter, footnotes: FxHashMap::default() }
587 fn get_entry(&mut self, key: &str) -> &mut (Vec<Event<'a>>, u16) {
588 let new_id = self.footnotes.keys().count() + 1;
589 let key = key.to_owned();
590 self.footnotes.entry(key).or_insert((Vec::new(), new_id as u16))
594 impl<'a, I: Iterator<Item = SpannedEvent<'a>>> Iterator for Footnotes<'a, I> {
595 type Item = SpannedEvent<'a>;
597 fn next(&mut self) -> Option<Self::Item> {
599 match self.inner.next() {
600 Some((Event::FootnoteReference(ref reference), range)) => {
601 let entry = self.get_entry(&reference);
602 let reference = format!(
603 "<sup id=\"fnref{0}\"><a href=\"#fn{0}\">{0}</a></sup>",
606 return Some((Event::Html(reference.into()), range));
608 Some((Event::Start(Tag::FootnoteDefinition(def)), _)) => {
609 let mut content = Vec::new();
610 for (event, _) in &mut self.inner {
611 if let Event::End(Tag::FootnoteDefinition(..)) = event {
616 let entry = self.get_entry(&def);
617 (*entry).0 = content;
619 Some(e) => return Some(e),
621 if !self.footnotes.is_empty() {
622 let mut v: Vec<_> = self.footnotes.drain().map(|(_, x)| x).collect();
623 v.sort_by(|a, b| a.1.cmp(&b.1));
624 let mut ret = String::from("<div class=\"footnotes\"><hr><ol>");
625 for (mut content, id) in v {
626 write!(ret, "<li id=\"fn{}\">", id).unwrap();
627 let mut is_paragraph = false;
628 if let Some(&Event::End(Tag::Paragraph)) = content.last() {
632 html::push_html(&mut ret, content.into_iter());
633 write!(ret, " <a href=\"#fnref{}\">↩</a>", id).unwrap();
635 ret.push_str("</p>");
637 ret.push_str("</li>");
639 ret.push_str("</ol></div>");
640 return Some((Event::Html(ret.into()), 0..0));
650 crate fn find_testable_code<T: doctest::Tester>(
653 error_codes: ErrorCodes,
654 enable_per_target_ignores: bool,
655 extra_info: Option<&ExtraInfo<'_>>,
657 let mut parser = Parser::new(doc).into_offset_iter();
658 let mut prev_offset = 0;
659 let mut nb_lines = 0;
660 let mut register_header = None;
661 while let Some((event, offset)) = parser.next() {
663 Event::Start(Tag::CodeBlock(kind)) => {
664 let block_info = match kind {
665 CodeBlockKind::Fenced(ref lang) => {
672 enable_per_target_ignores,
677 CodeBlockKind::Indented => Default::default(),
679 if !block_info.rust {
683 let mut test_s = String::new();
685 while let Some((Event::Text(s), _)) = parser.next() {
690 .map(|l| map_line(l).for_code())
691 .collect::<Vec<Cow<'_, str>>>()
694 nb_lines += doc[prev_offset..offset.start].lines().count();
695 let line = tests.get_line() + nb_lines + 1;
696 tests.add_test(text, block_info, line);
697 prev_offset = offset.start;
699 Event::Start(Tag::Heading(level)) => {
700 register_header = Some(level as u32);
702 Event::Text(ref s) if register_header.is_some() => {
703 let level = register_header.unwrap();
705 tests.register_header("", level);
707 tests.register_header(s, level);
709 register_header = None;
716 crate struct ExtraInfo<'tcx> {
727 impl<'tcx> ExtraInfo<'tcx> {
728 crate fn new(tcx: TyCtxt<'tcx>, hir_id: HirId, sp: Span) -> ExtraInfo<'tcx> {
729 ExtraInfo { id: ExtraInfoId::Hir(hir_id), sp, tcx }
732 crate fn new_did(tcx: TyCtxt<'tcx>, did: DefId, sp: Span) -> ExtraInfo<'tcx> {
733 ExtraInfo { id: ExtraInfoId::Def(did), sp, tcx }
736 fn error_invalid_codeblock_attr(&self, msg: &str, help: &str) {
737 let hir_id = match self.id {
738 ExtraInfoId::Hir(hir_id) => hir_id,
739 ExtraInfoId::Def(item_did) => {
740 match item_did.as_local() {
741 Some(item_did) => self.tcx.hir().local_def_id_to_hir_id(item_did),
743 // If non-local, no need to check anything.
749 self.tcx.struct_span_lint_hir(
750 crate::lint::INVALID_CODEBLOCK_ATTRIBUTES,
754 let mut diag = lint.build(msg);
762 #[derive(Eq, PartialEq, Clone, Debug)]
763 crate struct LangString {
765 crate should_panic: bool,
767 crate ignore: Ignore,
769 crate test_harness: bool,
770 crate compile_fail: bool,
771 crate error_codes: Vec<String>,
772 crate allow_fail: bool,
773 crate edition: Option<Edition>,
776 #[derive(Eq, PartialEq, Clone, Debug)]
783 impl Default for LangString {
784 fn default() -> Self {
786 original: String::new(),
789 ignore: Ignore::None,
793 error_codes: Vec::new(),
801 fn parse_without_check(
803 allow_error_code_check: ErrorCodes,
804 enable_per_target_ignores: bool,
806 Self::parse(string, allow_error_code_check, enable_per_target_ignores, None)
809 fn tokens(string: &str) -> impl Iterator<Item = &str> {
810 // Pandoc, which Rust once used for generating documentation,
811 // expects lang strings to be surrounded by `{}` and for each token
812 // to be proceeded by a `.`. Since some of these lang strings are still
813 // loose in the wild, we strip a pair of surrounding `{}` from the lang
814 // string and a leading `.` from each token.
816 let string = string.trim();
818 let first = string.chars().next();
819 let last = string.chars().last();
821 let string = if first == Some('{') && last == Some('}') {
822 &string[1..string.len() - 1]
828 .split(|c| c == ',' || c == ' ' || c == '\t')
830 .map(|token| if token.chars().next() == Some('.') { &token[1..] } else { token })
831 .filter(|token| !token.is_empty())
836 allow_error_code_check: ErrorCodes,
837 enable_per_target_ignores: bool,
838 extra: Option<&ExtraInfo<'_>>,
840 let allow_error_code_check = allow_error_code_check.as_bool();
841 let mut seen_rust_tags = false;
842 let mut seen_other_tags = false;
843 let mut data = LangString::default();
844 let mut ignores = vec![];
846 data.original = string.to_owned();
848 let tokens = Self::tokens(string).collect::<Vec<&str>>();
850 for token in tokens {
853 data.should_panic = true;
854 seen_rust_tags = !seen_other_tags;
858 seen_rust_tags = !seen_other_tags;
861 data.ignore = Ignore::All;
862 seen_rust_tags = !seen_other_tags;
864 x if x.starts_with("ignore-") => {
865 if enable_per_target_ignores {
866 ignores.push(x.trim_start_matches("ignore-").to_owned());
867 seen_rust_tags = !seen_other_tags;
871 data.allow_fail = true;
872 seen_rust_tags = !seen_other_tags;
876 seen_rust_tags = true;
879 data.test_harness = true;
880 seen_rust_tags = !seen_other_tags || seen_rust_tags;
883 data.compile_fail = true;
884 seen_rust_tags = !seen_other_tags || seen_rust_tags;
887 x if x.starts_with("edition") => {
888 data.edition = x[7..].parse::<Edition>().ok();
890 x if allow_error_code_check && x.starts_with('E') && x.len() == 5 => {
891 if x[1..].parse::<u32>().is_ok() {
892 data.error_codes.push(x.to_owned());
893 seen_rust_tags = !seen_other_tags || seen_rust_tags;
895 seen_other_tags = true;
898 x if extra.is_some() => {
899 let s = x.to_lowercase();
900 match if s == "compile-fail" || s == "compile_fail" || s == "compilefail" {
903 "the code block will either not be tested if not marked as a rust one \
904 or won't fail if it compiles successfully",
906 } else if s == "should-panic" || s == "should_panic" || s == "shouldpanic" {
909 "the code block will either not be tested if not marked as a rust one \
910 or won't fail if it doesn't panic when running",
912 } else if s == "no-run" || s == "no_run" || s == "norun" {
915 "the code block will either not be tested if not marked as a rust one \
916 or will be run (which you might not want)",
918 } else if s == "allow-fail" || s == "allow_fail" || s == "allowfail" {
921 "the code block will either not be tested if not marked as a rust one \
922 or will be run (which you might not want)",
924 } else if s == "test-harness" || s == "test_harness" || s == "testharness" {
927 "the code block will either not be tested if not marked as a rust one \
928 or the code will be wrapped inside a main function",
933 Some((flag, help)) => {
934 if let Some(ref extra) = extra {
935 extra.error_invalid_codeblock_attr(
936 &format!("unknown attribute `{}`. Did you mean `{}`?", x, flag),
943 seen_other_tags = true;
945 _ => seen_other_tags = true,
949 // ignore-foo overrides ignore
950 if !ignores.is_empty() {
951 data.ignore = Ignore::Some(ignores);
954 data.rust &= !seen_other_tags || seen_rust_tags;
961 pub fn into_string(self) -> String {
962 let Markdown(md, links, mut ids, codes, edition, playground) = self;
964 // This is actually common enough to special-case
966 return String::new();
968 let mut replacer = |broken_link: BrokenLink<'_>| {
970 links.iter().find(|link| &*link.original_text == broken_link.reference)
972 Some((link.href.as_str().into(), link.new_text.as_str().into()))
978 let p = Parser::new_with_broken_link_callback(md, opts(), Some(&mut replacer));
979 let p = p.into_offset_iter();
981 let mut s = String::with_capacity(md.len() * 3 / 2);
983 let p = HeadingLinks::new(p, None, &mut ids);
984 let p = Footnotes::new(p);
985 let p = LinkReplacer::new(p.map(|(ev, _)| ev), links);
986 let p = CodeBlocks::new(p, codes, edition, playground);
987 html::push_html(&mut s, p);
993 impl MarkdownWithToc<'_> {
994 crate fn into_string(self) -> String {
995 let MarkdownWithToc(md, mut ids, codes, edition, playground) = self;
997 let p = Parser::new_ext(md, opts()).into_offset_iter();
999 let mut s = String::with_capacity(md.len() * 3 / 2);
1001 let mut toc = TocBuilder::new();
1004 let p = HeadingLinks::new(p, Some(&mut toc), &mut ids);
1005 let p = Footnotes::new(p);
1006 let p = CodeBlocks::new(p.map(|(ev, _)| ev), codes, edition, playground);
1007 html::push_html(&mut s, p);
1010 format!("<nav id=\"TOC\">{}</nav>{}", toc.into_toc().print(), s)
1014 impl MarkdownHtml<'_> {
1015 crate fn into_string(self) -> String {
1016 let MarkdownHtml(md, mut ids, codes, edition, playground) = self;
1018 // This is actually common enough to special-case
1020 return String::new();
1022 let p = Parser::new_ext(md, opts()).into_offset_iter();
1024 // Treat inline HTML as plain text.
1025 let p = p.map(|event| match event.0 {
1026 Event::Html(text) => (Event::Text(text), event.1),
1030 let mut s = String::with_capacity(md.len() * 3 / 2);
1032 let p = HeadingLinks::new(p, None, &mut ids);
1033 let p = Footnotes::new(p);
1034 let p = CodeBlocks::new(p.map(|(ev, _)| ev), codes, edition, playground);
1035 html::push_html(&mut s, p);
1041 impl MarkdownSummaryLine<'_> {
1042 crate fn into_string(self) -> String {
1043 let MarkdownSummaryLine(md, links) = self;
1044 // This is actually common enough to special-case
1046 return String::new();
1049 let mut replacer = |broken_link: BrokenLink<'_>| {
1051 links.iter().find(|link| &*link.original_text == broken_link.reference)
1053 Some((link.href.as_str().into(), link.new_text.as_str().into()))
1059 let p = Parser::new_with_broken_link_callback(md, summary_opts(), Some(&mut replacer));
1061 let mut s = String::new();
1063 html::push_html(&mut s, LinkReplacer::new(SummaryLine::new(p), links));
1069 /// Renders a subset of Markdown in the first paragraph of the provided Markdown.
1071 /// - *Italics*, **bold**, and `inline code` styles **are** rendered.
1072 /// - Headings and links are stripped (though the text *is* rendered).
1073 /// - HTML, code blocks, and everything else are ignored.
1075 /// Returns a tuple of the rendered HTML string and whether the output was shortened
1076 /// due to the provided `length_limit`.
1077 fn markdown_summary_with_limit(
1079 link_names: &[RenderedLink],
1080 length_limit: usize,
1081 ) -> (String, bool) {
1083 return (String::new(), false);
1086 let mut replacer = |broken_link: BrokenLink<'_>| {
1088 link_names.iter().find(|link| &*link.original_text == broken_link.reference)
1090 Some((link.href.as_str().into(), link.new_text.as_str().into()))
1096 let p = Parser::new_with_broken_link_callback(md, opts(), Some(&mut replacer));
1097 let mut p = LinkReplacer::new(p, link_names);
1099 let mut buf = HtmlWithLimit::new(length_limit);
1100 let mut stopped_early = false;
1101 p.try_for_each(|event| {
1103 Event::Text(text) => {
1105 text.split_inclusive(char::is_whitespace).try_for_each(|word| buf.push(word));
1107 stopped_early = true;
1111 Event::Code(code) => {
1112 buf.open_tag("code");
1113 let r = buf.push(code);
1115 stopped_early = true;
1121 Event::Start(tag) => match tag {
1122 Tag::Emphasis => buf.open_tag("em"),
1123 Tag::Strong => buf.open_tag("strong"),
1124 Tag::CodeBlock(..) => return ControlFlow::BREAK,
1127 Event::End(tag) => match tag {
1128 Tag::Emphasis | Tag::Strong => buf.close_tag(),
1129 Tag::Paragraph | Tag::Heading(..) => return ControlFlow::BREAK,
1132 Event::HardBreak | Event::SoftBreak => buf.push(" ")?,
1135 ControlFlow::CONTINUE
1138 (buf.finish(), stopped_early)
1141 /// Renders a shortened first paragraph of the given Markdown as a subset of Markdown,
1142 /// making it suitable for contexts like the search index.
1144 /// Will shorten to 59 or 60 characters, including an ellipsis (…) if it was shortened.
1146 /// See [`markdown_summary_with_limit`] for details about what is rendered and what is not.
1147 crate fn short_markdown_summary(markdown: &str, link_names: &[RenderedLink]) -> String {
1148 let (mut s, was_shortened) = markdown_summary_with_limit(markdown, link_names, 59);
1157 /// Renders the first paragraph of the provided markdown as plain text.
1158 /// Useful for alt-text.
1160 /// - Headings, links, and formatting are stripped.
1161 /// - Inline code is rendered as-is, surrounded by backticks.
1162 /// - HTML and code blocks are ignored.
1163 crate fn plain_text_summary(md: &str) -> String {
1165 return String::new();
1168 let mut s = String::with_capacity(md.len() * 3 / 2);
1170 for event in Parser::new_ext(md, summary_opts()) {
1172 Event::Text(text) => s.push_str(text),
1173 Event::Code(code) => {
1178 Event::HardBreak | Event::SoftBreak => s.push(' '),
1179 Event::Start(Tag::CodeBlock(..)) => break,
1180 Event::End(Tag::Paragraph) => break,
1181 Event::End(Tag::Heading(..)) => break,
1190 crate struct MarkdownLink {
1193 pub range: Range<usize>,
1196 crate fn markdown_links(md: &str) -> Vec<MarkdownLink> {
1201 let links = RefCell::new(vec![]);
1203 // FIXME: remove this function once pulldown_cmark can provide spans for link definitions.
1204 let locate = |s: &str, fallback: Range<usize>| unsafe {
1205 let s_start = s.as_ptr();
1206 let s_end = s_start.add(s.len());
1207 let md_start = md.as_ptr();
1208 let md_end = md_start.add(md.len());
1209 if md_start <= s_start && s_end <= md_end {
1210 let start = s_start.offset_from(md_start) as usize;
1211 let end = s_end.offset_from(md_start) as usize;
1218 let span_for_link = |link: &CowStr<'_>, span: Range<usize>| {
1219 // For diagnostics, we want to underline the link's definition but `span` will point at
1220 // where the link is used. This is a problem for reference-style links, where the definition
1221 // is separate from the usage.
1223 // `Borrowed` variant means the string (the link's destination) may come directly from
1224 // the markdown text and we can locate the original link destination.
1225 // NOTE: LinkReplacer also provides `Borrowed` but possibly from other sources,
1226 // so `locate()` can fall back to use `span`.
1227 CowStr::Borrowed(s) => locate(s, span),
1229 // For anything else, we can only use the provided range.
1230 CowStr::Boxed(_) | CowStr::Inlined(_) => span,
1234 let mut push = |link: BrokenLink<'_>| {
1235 let span = span_for_link(&CowStr::Borrowed(link.reference), link.span);
1236 links.borrow_mut().push(MarkdownLink {
1237 kind: LinkType::ShortcutUnknown,
1238 link: link.reference.to_owned(),
1243 let p = Parser::new_with_broken_link_callback(md, opts(), Some(&mut push)).into_offset_iter();
1245 // There's no need to thread an IdMap through to here because
1246 // the IDs generated aren't going to be emitted anywhere.
1247 let mut ids = IdMap::new();
1248 let iter = Footnotes::new(HeadingLinks::new(p, None, &mut ids));
1251 if let Event::Start(Tag::Link(kind, dest, _)) = ev.0 {
1252 debug!("found link: {}", dest);
1253 let span = span_for_link(&dest, ev.1);
1254 links.borrow_mut().push(MarkdownLink { kind, link: dest.into_string(), range: span });
1262 crate struct RustCodeBlock {
1263 /// The range in the markdown that the code block occupies. Note that this includes the fences
1264 /// for fenced code blocks.
1265 crate range: Range<usize>,
1266 /// The range in the markdown that the code within the code block occupies.
1267 crate code: Range<usize>,
1268 crate is_fenced: bool,
1269 crate syntax: Option<String>,
1270 crate is_ignore: bool,
1273 /// Returns a range of bytes for each code block in the markdown that is tagged as `rust` or
1274 /// untagged (and assumed to be rust).
1275 crate fn rust_code_blocks(md: &str, extra_info: &ExtraInfo<'_>) -> Vec<RustCodeBlock> {
1276 let mut code_blocks = vec![];
1282 let mut p = Parser::new_ext(md, opts()).into_offset_iter();
1284 while let Some((event, offset)) = p.next() {
1285 if let Event::Start(Tag::CodeBlock(syntax)) = event {
1286 let (syntax, code_start, code_end, range, is_fenced, is_ignore) = match syntax {
1287 CodeBlockKind::Fenced(syntax) => {
1288 let syntax = syntax.as_ref();
1289 let lang_string = if syntax.is_empty() {
1292 LangString::parse(&*syntax, ErrorCodes::Yes, false, Some(extra_info))
1294 if !lang_string.rust {
1297 let is_ignore = lang_string.ignore != Ignore::None;
1298 let syntax = if syntax.is_empty() { None } else { Some(syntax.to_owned()) };
1299 let (code_start, mut code_end) = match p.next() {
1300 Some((Event::Text(_), offset)) => (offset.start, offset.end),
1301 Some((_, sub_offset)) => {
1302 let code = Range { start: sub_offset.start, end: sub_offset.start };
1303 code_blocks.push(RustCodeBlock {
1313 let code = Range { start: offset.end, end: offset.end };
1314 code_blocks.push(RustCodeBlock {
1324 while let Some((Event::Text(_), offset)) = p.next() {
1325 code_end = offset.end;
1327 (syntax, code_start, code_end, offset, true, is_ignore)
1329 CodeBlockKind::Indented => {
1330 // The ending of the offset goes too far sometime so we reduce it by one in
1332 if offset.end > offset.start && md.get(offset.end..=offset.end) == Some(&"\n") {
1337 Range { start: offset.start, end: offset.end - 1 },
1342 (None, offset.start, offset.end, offset, false, false)
1347 code_blocks.push(RustCodeBlock {
1350 code: Range { start: code_start, end: code_end },
1360 #[derive(Clone, Default, Debug)]
1362 map: FxHashMap<String, usize>,
1365 fn init_id_map() -> FxHashMap<String, usize> {
1366 let mut map = FxHashMap::default();
1367 // This is the list of IDs used in Javascript.
1368 map.insert("help".to_owned(), 1);
1369 // This is the list of IDs used in HTML generated in Rust (including the ones
1370 // used in tera template files).
1371 map.insert("mainThemeStyle".to_owned(), 1);
1372 map.insert("themeStyle".to_owned(), 1);
1373 map.insert("theme-picker".to_owned(), 1);
1374 map.insert("theme-choices".to_owned(), 1);
1375 map.insert("settings-menu".to_owned(), 1);
1376 map.insert("help-button".to_owned(), 1);
1377 map.insert("main".to_owned(), 1);
1378 map.insert("search".to_owned(), 1);
1379 map.insert("crate-search".to_owned(), 1);
1380 map.insert("render-detail".to_owned(), 1);
1381 map.insert("toggle-all-docs".to_owned(), 1);
1382 map.insert("all-types".to_owned(), 1);
1383 map.insert("default-settings".to_owned(), 1);
1384 map.insert("rustdoc-vars".to_owned(), 1);
1385 map.insert("sidebar-vars".to_owned(), 1);
1386 map.insert("copy-path".to_owned(), 1);
1387 map.insert("TOC".to_owned(), 1);
1388 // This is the list of IDs used by rustdoc sections (but still generated by
1390 map.insert("fields".to_owned(), 1);
1391 map.insert("variants".to_owned(), 1);
1392 map.insert("implementors-list".to_owned(), 1);
1393 map.insert("synthetic-implementors-list".to_owned(), 1);
1394 map.insert("foreign-impls".to_owned(), 1);
1395 map.insert("implementations".to_owned(), 1);
1396 map.insert("trait-implementations".to_owned(), 1);
1397 map.insert("synthetic-implementations".to_owned(), 1);
1398 map.insert("blanket-implementations".to_owned(), 1);
1399 map.insert("associated-types".to_owned(), 1);
1400 map.insert("associated-const".to_owned(), 1);
1401 map.insert("required-methods".to_owned(), 1);
1402 map.insert("provided-methods".to_owned(), 1);
1403 map.insert("implementors".to_owned(), 1);
1404 map.insert("synthetic-implementors".to_owned(), 1);
1405 map.insert("trait-implementations-list".to_owned(), 1);
1406 map.insert("synthetic-implementations-list".to_owned(), 1);
1407 map.insert("blanket-implementations-list".to_owned(), 1);
1408 map.insert("deref-methods".to_owned(), 1);
1413 pub fn new() -> Self {
1414 IdMap { map: init_id_map() }
1417 crate fn derive<S: AsRef<str> + ToString>(&mut self, candidate: S) -> String {
1418 let id = match self.map.get_mut(candidate.as_ref()) {
1419 None => candidate.to_string(),
1421 let id = format!("{}-{}", candidate.as_ref(), *a);
1427 self.map.insert(id.clone(), 1);