1use crate::markdown_elements::*;
2use async_recursion::async_recursion;
3use collections::FxHashMap;
4use gpui::FontWeight;
5use language::LanguageRegistry;
6use pulldown_cmark::{Alignment, Event, Options, Parser, Tag, TagEnd};
7use std::{ops::Range, path::PathBuf, sync::Arc};
8
9pub async fn parse_markdown(
10 markdown_input: &str,
11 file_location_directory: Option<PathBuf>,
12 language_registry: Option<Arc<LanguageRegistry>>,
13) -> ParsedMarkdown {
14 let mut options = Options::all();
15 options.remove(pulldown_cmark::Options::ENABLE_DEFINITION_LIST);
16
17 let parser = Parser::new_ext(markdown_input, options);
18 let parser = MarkdownParser::new(
19 parser.into_offset_iter().collect(),
20 file_location_directory,
21 language_registry,
22 );
23 let renderer = parser.parse_document().await;
24 ParsedMarkdown {
25 children: renderer.parsed,
26 }
27}
28
29struct MarkdownParser<'a> {
30 tokens: Vec<(Event<'a>, Range<usize>)>,
31 /// The current index in the tokens array
32 cursor: usize,
33 /// The blocks that we have successfully parsed so far
34 parsed: Vec<ParsedMarkdownElement>,
35 file_location_directory: Option<PathBuf>,
36 language_registry: Option<Arc<LanguageRegistry>>,
37}
38
39struct MarkdownListItem {
40 content: Vec<ParsedMarkdownElement>,
41 item_type: ParsedMarkdownListItemType,
42}
43
44impl Default for MarkdownListItem {
45 fn default() -> Self {
46 Self {
47 content: Vec::new(),
48 item_type: ParsedMarkdownListItemType::Unordered,
49 }
50 }
51}
52
53impl<'a> MarkdownParser<'a> {
54 fn new(
55 tokens: Vec<(Event<'a>, Range<usize>)>,
56 file_location_directory: Option<PathBuf>,
57 language_registry: Option<Arc<LanguageRegistry>>,
58 ) -> Self {
59 Self {
60 tokens,
61 file_location_directory,
62 language_registry,
63 cursor: 0,
64 parsed: vec![],
65 }
66 }
67
68 fn eof(&self) -> bool {
69 if self.tokens.is_empty() {
70 return true;
71 }
72 self.cursor >= self.tokens.len() - 1
73 }
74
75 fn peek(&self, steps: usize) -> Option<&(Event, Range<usize>)> {
76 if self.eof() || (steps + self.cursor) >= self.tokens.len() {
77 return self.tokens.last();
78 }
79 return self.tokens.get(self.cursor + steps);
80 }
81
82 fn previous(&self) -> Option<&(Event, Range<usize>)> {
83 if self.cursor == 0 || self.cursor > self.tokens.len() {
84 return None;
85 }
86 return self.tokens.get(self.cursor - 1);
87 }
88
89 fn current(&self) -> Option<&(Event, Range<usize>)> {
90 return self.peek(0);
91 }
92
93 fn current_event(&self) -> Option<&Event> {
94 return self.current().map(|(event, _)| event);
95 }
96
97 fn is_text_like(event: &Event) -> bool {
98 match event {
99 Event::Text(_)
100 // Represent an inline code block
101 | Event::Code(_)
102 | Event::Html(_)
103 | Event::FootnoteReference(_)
104 | Event::Start(Tag::Link { link_type: _, dest_url: _, title: _, id: _ })
105 | Event::Start(Tag::Emphasis)
106 | Event::Start(Tag::Strong)
107 | Event::Start(Tag::Strikethrough)
108 | Event::Start(Tag::Image { link_type: _, dest_url: _, title: _, id: _ }) => {
109 true
110 }
111 _ => false,
112 }
113 }
114
115 async fn parse_document(mut self) -> Self {
116 while !self.eof() {
117 if let Some(block) = self.parse_block().await {
118 self.parsed.extend(block);
119 } else {
120 self.cursor += 1;
121 }
122 }
123 self
124 }
125
126 #[async_recursion]
127 async fn parse_block(&mut self) -> Option<Vec<ParsedMarkdownElement>> {
128 let (current, source_range) = self.current().unwrap();
129 let source_range = source_range.clone();
130 match current {
131 Event::Start(tag) => match tag {
132 Tag::Paragraph => {
133 self.cursor += 1;
134 let text = self.parse_text(false, Some(source_range));
135 Some(vec![ParsedMarkdownElement::Paragraph(text)])
136 }
137 Tag::Heading {
138 level,
139 id: _,
140 classes: _,
141 attrs: _,
142 } => {
143 let level = *level;
144 self.cursor += 1;
145 let heading = self.parse_heading(level);
146 Some(vec![ParsedMarkdownElement::Heading(heading)])
147 }
148 Tag::Table(alignment) => {
149 let alignment = alignment.clone();
150 self.cursor += 1;
151 let table = self.parse_table(alignment);
152 Some(vec![ParsedMarkdownElement::Table(table)])
153 }
154 Tag::List(order) => {
155 let order = *order;
156 self.cursor += 1;
157 let list = self.parse_list(order).await;
158 Some(list)
159 }
160 Tag::BlockQuote(_kind) => {
161 self.cursor += 1;
162 let block_quote = self.parse_block_quote().await;
163 Some(vec![ParsedMarkdownElement::BlockQuote(block_quote)])
164 }
165 Tag::CodeBlock(kind) => {
166 let language = match kind {
167 pulldown_cmark::CodeBlockKind::Indented => None,
168 pulldown_cmark::CodeBlockKind::Fenced(language) => {
169 if language.is_empty() {
170 None
171 } else {
172 Some(language.to_string())
173 }
174 }
175 };
176
177 self.cursor += 1;
178
179 let code_block = self.parse_code_block(language).await;
180 Some(vec![ParsedMarkdownElement::CodeBlock(code_block)])
181 }
182 _ => None,
183 },
184 Event::Rule => {
185 let source_range = source_range.clone();
186 self.cursor += 1;
187 Some(vec![ParsedMarkdownElement::HorizontalRule(source_range)])
188 }
189 _ => None,
190 }
191 }
192
193 fn parse_text(
194 &mut self,
195 should_complete_on_soft_break: bool,
196 source_range: Option<Range<usize>>,
197 ) -> ParsedMarkdownText {
198 let source_range = source_range.unwrap_or_else(|| {
199 self.current()
200 .map(|(_, range)| range.clone())
201 .unwrap_or_default()
202 });
203
204 let mut text = String::new();
205 let mut bold_depth = 0;
206 let mut italic_depth = 0;
207 let mut strikethrough_depth = 0;
208 let mut link: Option<Link> = None;
209 let mut region_ranges: Vec<Range<usize>> = vec![];
210 let mut regions: Vec<ParsedRegion> = vec![];
211 let mut highlights: Vec<(Range<usize>, MarkdownHighlight)> = vec![];
212
213 let mut link_urls: Vec<String> = vec![];
214 let mut link_ranges: Vec<Range<usize>> = vec![];
215
216 loop {
217 if self.eof() {
218 break;
219 }
220
221 let (current, _source_range) = self.current().unwrap();
222 let prev_len = text.len();
223 match current {
224 Event::SoftBreak => {
225 if should_complete_on_soft_break {
226 break;
227 }
228
229 // `Some text\nSome more text` should be treated as a single line.
230 text.push(' ');
231 }
232
233 Event::HardBreak => {
234 text.push('\n');
235 }
236
237 Event::Text(t) => {
238 text.push_str(t.as_ref());
239
240 let mut style = MarkdownHighlightStyle::default();
241
242 if bold_depth > 0 {
243 style.weight = FontWeight::BOLD;
244 }
245
246 if italic_depth > 0 {
247 style.italic = true;
248 }
249
250 if strikethrough_depth > 0 {
251 style.strikethrough = true;
252 }
253
254 let last_run_len = if let Some(link) = link.clone() {
255 region_ranges.push(prev_len..text.len());
256 regions.push(ParsedRegion {
257 code: false,
258 link: Some(link),
259 });
260 style.underline = true;
261 prev_len
262 } else {
263 // Manually scan for links
264 let mut finder = linkify::LinkFinder::new();
265 finder.kinds(&[linkify::LinkKind::Url]);
266 let mut last_link_len = prev_len;
267 for link in finder.links(t) {
268 let start = link.start();
269 let end = link.end();
270 let range = (prev_len + start)..(prev_len + end);
271 link_ranges.push(range.clone());
272 link_urls.push(link.as_str().to_string());
273
274 // If there is a style before we match a link, we have to add this to the highlighted ranges
275 if style != MarkdownHighlightStyle::default()
276 && last_link_len < link.start()
277 {
278 highlights.push((
279 last_link_len..link.start(),
280 MarkdownHighlight::Style(style.clone()),
281 ));
282 }
283
284 highlights.push((
285 range.clone(),
286 MarkdownHighlight::Style(MarkdownHighlightStyle {
287 underline: true,
288 ..style
289 }),
290 ));
291 region_ranges.push(range.clone());
292 regions.push(ParsedRegion {
293 code: false,
294 link: Some(Link::Web {
295 url: link.as_str().to_string(),
296 }),
297 });
298
299 last_link_len = end;
300 }
301 last_link_len
302 };
303
304 if style != MarkdownHighlightStyle::default() && last_run_len < text.len() {
305 let mut new_highlight = true;
306 if let Some((last_range, last_style)) = highlights.last_mut() {
307 if last_range.end == last_run_len
308 && last_style == &MarkdownHighlight::Style(style.clone())
309 {
310 last_range.end = text.len();
311 new_highlight = false;
312 }
313 }
314 if new_highlight {
315 highlights
316 .push((last_run_len..text.len(), MarkdownHighlight::Style(style)));
317 }
318 }
319 }
320
321 // Note: This event means "inline code" and not "code block"
322 Event::Code(t) => {
323 text.push_str(t.as_ref());
324 region_ranges.push(prev_len..text.len());
325
326 if link.is_some() {
327 highlights.push((
328 prev_len..text.len(),
329 MarkdownHighlight::Style(MarkdownHighlightStyle {
330 underline: true,
331 ..Default::default()
332 }),
333 ));
334 }
335
336 regions.push(ParsedRegion {
337 code: true,
338 link: link.clone(),
339 });
340 }
341
342 Event::Start(tag) => match tag {
343 Tag::Emphasis => italic_depth += 1,
344 Tag::Strong => bold_depth += 1,
345 Tag::Strikethrough => strikethrough_depth += 1,
346 Tag::Link {
347 link_type: _,
348 dest_url,
349 title: _,
350 id: _,
351 } => {
352 link = Link::identify(
353 self.file_location_directory.clone(),
354 dest_url.to_string(),
355 );
356 }
357 _ => {
358 break;
359 }
360 },
361
362 Event::End(tag) => match tag {
363 TagEnd::Emphasis => {
364 italic_depth -= 1;
365 }
366 TagEnd::Strong => {
367 bold_depth -= 1;
368 }
369 TagEnd::Strikethrough => {
370 strikethrough_depth -= 1;
371 }
372 TagEnd::Link => {
373 link = None;
374 }
375 TagEnd::Paragraph => {
376 self.cursor += 1;
377 break;
378 }
379 _ => {
380 break;
381 }
382 },
383
384 _ => {
385 break;
386 }
387 }
388
389 self.cursor += 1;
390 }
391
392 ParsedMarkdownText {
393 source_range,
394 contents: text,
395 highlights,
396 regions,
397 region_ranges,
398 }
399 }
400
401 fn parse_heading(&mut self, level: pulldown_cmark::HeadingLevel) -> ParsedMarkdownHeading {
402 let (_event, source_range) = self.previous().unwrap();
403 let source_range = source_range.clone();
404 let text = self.parse_text(true, None);
405
406 // Advance past the heading end tag
407 self.cursor += 1;
408
409 ParsedMarkdownHeading {
410 source_range: source_range.clone(),
411 level: match level {
412 pulldown_cmark::HeadingLevel::H1 => HeadingLevel::H1,
413 pulldown_cmark::HeadingLevel::H2 => HeadingLevel::H2,
414 pulldown_cmark::HeadingLevel::H3 => HeadingLevel::H3,
415 pulldown_cmark::HeadingLevel::H4 => HeadingLevel::H4,
416 pulldown_cmark::HeadingLevel::H5 => HeadingLevel::H5,
417 pulldown_cmark::HeadingLevel::H6 => HeadingLevel::H6,
418 },
419 contents: text,
420 }
421 }
422
423 fn parse_table(&mut self, alignment: Vec<Alignment>) -> ParsedMarkdownTable {
424 let (_event, source_range) = self.previous().unwrap();
425 let source_range = source_range.clone();
426 let mut header = ParsedMarkdownTableRow::new();
427 let mut body = vec![];
428 let mut current_row = vec![];
429 let mut in_header = true;
430 let column_alignments = alignment.iter().map(Self::convert_alignment).collect();
431
432 loop {
433 if self.eof() {
434 break;
435 }
436
437 let (current, source_range) = self.current().unwrap();
438 let source_range = source_range.clone();
439 match current {
440 Event::Start(Tag::TableHead)
441 | Event::Start(Tag::TableRow)
442 | Event::End(TagEnd::TableCell) => {
443 self.cursor += 1;
444 }
445 Event::Start(Tag::TableCell) => {
446 self.cursor += 1;
447 let cell_contents = self.parse_text(false, Some(source_range));
448 current_row.push(cell_contents);
449 }
450 Event::End(TagEnd::TableHead) | Event::End(TagEnd::TableRow) => {
451 self.cursor += 1;
452 let new_row = std::mem::take(&mut current_row);
453 if in_header {
454 header.children = new_row;
455 in_header = false;
456 } else {
457 let row = ParsedMarkdownTableRow::with_children(new_row);
458 body.push(row);
459 }
460 }
461 Event::End(TagEnd::Table) => {
462 self.cursor += 1;
463 break;
464 }
465 _ => {
466 break;
467 }
468 }
469 }
470
471 ParsedMarkdownTable {
472 source_range,
473 header,
474 body,
475 column_alignments,
476 }
477 }
478
479 fn convert_alignment(alignment: &Alignment) -> ParsedMarkdownTableAlignment {
480 match alignment {
481 Alignment::None => ParsedMarkdownTableAlignment::None,
482 Alignment::Left => ParsedMarkdownTableAlignment::Left,
483 Alignment::Center => ParsedMarkdownTableAlignment::Center,
484 Alignment::Right => ParsedMarkdownTableAlignment::Right,
485 }
486 }
487
488 async fn parse_list(&mut self, order: Option<u64>) -> Vec<ParsedMarkdownElement> {
489 let (_, list_source_range) = self.previous().unwrap();
490
491 let mut items = Vec::new();
492 let mut items_stack = vec![MarkdownListItem::default()];
493 let mut depth = 1;
494 let mut order = order;
495 let mut order_stack = Vec::new();
496
497 let mut insertion_indices = FxHashMap::default();
498 let mut source_ranges = FxHashMap::default();
499 let mut start_item_range = list_source_range.clone();
500
501 while !self.eof() {
502 let (current, source_range) = self.current().unwrap();
503 match current {
504 Event::Start(Tag::List(new_order)) => {
505 if items_stack.last().is_some() && !insertion_indices.contains_key(&depth) {
506 insertion_indices.insert(depth, items.len());
507 }
508
509 // We will use the start of the nested list as the end for the current item's range,
510 // because we don't care about the hierarchy of list items
511 if let collections::hash_map::Entry::Vacant(e) = source_ranges.entry(depth) {
512 e.insert(start_item_range.start..source_range.start);
513 }
514
515 order_stack.push(order);
516 order = *new_order;
517 self.cursor += 1;
518 depth += 1;
519 }
520 Event::End(TagEnd::List(_)) => {
521 order = order_stack.pop().flatten();
522 self.cursor += 1;
523 depth -= 1;
524
525 if depth == 0 {
526 break;
527 }
528 }
529 Event::Start(Tag::Item) => {
530 start_item_range = source_range.clone();
531
532 self.cursor += 1;
533 items_stack.push(MarkdownListItem::default());
534
535 let mut task_list = None;
536 // Check for task list marker (`- [ ]` or `- [x]`)
537 if let Some(event) = self.current_event() {
538 // If there is a linebreak in between two list items the task list marker will actually be the first element of the paragraph
539 if event == &Event::Start(Tag::Paragraph) {
540 self.cursor += 1;
541 }
542
543 if let Some((Event::TaskListMarker(checked), range)) = self.current() {
544 task_list = Some((*checked, range.clone()));
545 self.cursor += 1;
546 }
547 }
548
549 if let Some((event, range)) = self.current() {
550 // This is a plain list item.
551 // For example `- some text` or `1. [Docs](./docs.md)`
552 if MarkdownParser::is_text_like(event) {
553 let text = self.parse_text(false, Some(range.clone()));
554 let block = ParsedMarkdownElement::Paragraph(text);
555 if let Some(content) = items_stack.last_mut() {
556 let item_type = if let Some((checked, range)) = task_list {
557 ParsedMarkdownListItemType::Task(checked, range)
558 } else if let Some(order) = order {
559 ParsedMarkdownListItemType::Ordered(order)
560 } else {
561 ParsedMarkdownListItemType::Unordered
562 };
563 content.item_type = item_type;
564 content.content.push(block);
565 }
566 } else {
567 let block = self.parse_block().await;
568 if let Some(block) = block {
569 if let Some(list_item) = items_stack.last_mut() {
570 list_item.content.extend(block);
571 }
572 }
573 }
574 }
575
576 // If there is a linebreak in between two list items the task list marker will actually be the first element of the paragraph
577 if self.current_event() == Some(&Event::End(TagEnd::Paragraph)) {
578 self.cursor += 1;
579 }
580 }
581 Event::End(TagEnd::Item) => {
582 self.cursor += 1;
583
584 if let Some(current) = order {
585 order = Some(current + 1);
586 }
587
588 if let Some(list_item) = items_stack.pop() {
589 let source_range = source_ranges
590 .remove(&depth)
591 .unwrap_or(start_item_range.clone());
592
593 // We need to remove the last character of the source range, because it includes the newline character
594 let source_range = source_range.start..source_range.end - 1;
595 let item = ParsedMarkdownElement::ListItem(ParsedMarkdownListItem {
596 source_range,
597 content: list_item.content,
598 depth,
599 item_type: list_item.item_type,
600 });
601
602 if let Some(index) = insertion_indices.get(&depth) {
603 items.insert(*index, item);
604 insertion_indices.remove(&depth);
605 } else {
606 items.push(item);
607 }
608 }
609 }
610 _ => {
611 if depth == 0 {
612 break;
613 }
614 // This can only happen if a list item starts with more then one paragraph,
615 // or the list item contains blocks that should be rendered after the nested list items
616 let block = self.parse_block().await;
617 if let Some(block) = block {
618 if let Some(list_item) = items_stack.last_mut() {
619 // If we did not insert any nested items yet (in this case insertion index is set), we can append the block to the current list item
620 if !insertion_indices.contains_key(&depth) {
621 list_item.content.extend(block);
622 continue;
623 }
624 }
625
626 // Otherwise we need to insert the block after all the nested items
627 // that have been parsed so far
628 items.extend(block);
629 }
630 }
631 }
632 }
633
634 items
635 }
636
637 #[async_recursion]
638 async fn parse_block_quote(&mut self) -> ParsedMarkdownBlockQuote {
639 let (_event, source_range) = self.previous().unwrap();
640 let source_range = source_range.clone();
641 let mut nested_depth = 1;
642
643 let mut children: Vec<ParsedMarkdownElement> = vec![];
644
645 while !self.eof() {
646 let block = self.parse_block().await;
647
648 if let Some(block) = block {
649 children.extend(block);
650 } else {
651 break;
652 }
653
654 if self.eof() {
655 break;
656 }
657
658 let (current, _source_range) = self.current().unwrap();
659 match current {
660 // This is a nested block quote.
661 // Record that we're in a nested block quote and continue parsing.
662 // We don't need to advance the cursor since the next
663 // call to `parse_block` will handle it.
664 Event::Start(Tag::BlockQuote(_kind)) => {
665 nested_depth += 1;
666 }
667 Event::End(TagEnd::BlockQuote(_kind)) => {
668 nested_depth -= 1;
669 if nested_depth == 0 {
670 self.cursor += 1;
671 break;
672 }
673 }
674 _ => {}
675 };
676 }
677
678 ParsedMarkdownBlockQuote {
679 source_range,
680 children,
681 }
682 }
683
684 async fn parse_code_block(&mut self, language: Option<String>) -> ParsedMarkdownCodeBlock {
685 let (_event, source_range) = self.previous().unwrap();
686 let source_range = source_range.clone();
687 let mut code = String::new();
688
689 while !self.eof() {
690 let (current, _source_range) = self.current().unwrap();
691 match current {
692 Event::Text(text) => {
693 code.push_str(text);
694 self.cursor += 1;
695 }
696 Event::End(TagEnd::CodeBlock) => {
697 self.cursor += 1;
698 break;
699 }
700 _ => {
701 break;
702 }
703 }
704 }
705
706 let highlights = if let Some(language) = &language {
707 if let Some(registry) = &self.language_registry {
708 let rope: language::Rope = code.as_str().into();
709 registry
710 .language_for_name_or_extension(language)
711 .await
712 .map(|l| l.highlight_text(&rope, 0..code.len()))
713 .ok()
714 } else {
715 None
716 }
717 } else {
718 None
719 };
720
721 ParsedMarkdownCodeBlock {
722 source_range,
723 contents: code.trim().to_string().into(),
724 language,
725 highlights,
726 }
727 }
728}
729
730#[cfg(test)]
731mod tests {
732 use super::*;
733
734 use gpui::BackgroundExecutor;
735 use language::{tree_sitter_rust, HighlightId, Language, LanguageConfig, LanguageMatcher};
736 use pretty_assertions::assert_eq;
737 use ParsedMarkdownListItemType::*;
738
739 async fn parse(input: &str) -> ParsedMarkdown {
740 parse_markdown(input, None, None).await
741 }
742
743 #[gpui::test]
744 async fn test_headings() {
745 let parsed = parse("# Heading one\n## Heading two\n### Heading three").await;
746
747 assert_eq!(
748 parsed.children,
749 vec![
750 h1(text("Heading one", 2..13), 0..14),
751 h2(text("Heading two", 17..28), 14..29),
752 h3(text("Heading three", 33..46), 29..46),
753 ]
754 );
755 }
756
757 #[gpui::test]
758 async fn test_newlines_dont_new_paragraphs() {
759 let parsed = parse("Some text **that is bolded**\n and *italicized*").await;
760
761 assert_eq!(
762 parsed.children,
763 vec![p("Some text that is bolded and italicized", 0..46)]
764 );
765 }
766
767 #[gpui::test]
768 async fn test_heading_with_paragraph() {
769 let parsed = parse("# Zed\nThe editor").await;
770
771 assert_eq!(
772 parsed.children,
773 vec![h1(text("Zed", 2..5), 0..6), p("The editor", 6..16),]
774 );
775 }
776
777 #[gpui::test]
778 async fn test_double_newlines_do_new_paragraphs() {
779 let parsed = parse("Some text **that is bolded**\n\n and *italicized*").await;
780
781 assert_eq!(
782 parsed.children,
783 vec![
784 p("Some text that is bolded", 0..29),
785 p("and italicized", 31..47),
786 ]
787 );
788 }
789
790 #[gpui::test]
791 async fn test_bold_italic_text() {
792 let parsed = parse("Some text **that is bolded** and *italicized*").await;
793
794 assert_eq!(
795 parsed.children,
796 vec![p("Some text that is bolded and italicized", 0..45)]
797 );
798 }
799
800 #[gpui::test]
801 async fn test_nested_bold_strikethrough_text() {
802 let parsed = parse("Some **bo~~strikethrough~~ld** text").await;
803
804 assert_eq!(parsed.children.len(), 1);
805 assert_eq!(
806 parsed.children[0],
807 ParsedMarkdownElement::Paragraph(ParsedMarkdownText {
808 source_range: 0..35,
809 contents: "Some bostrikethroughld text".to_string(),
810 highlights: Vec::new(),
811 region_ranges: Vec::new(),
812 regions: Vec::new(),
813 })
814 );
815
816 let paragraph = if let ParsedMarkdownElement::Paragraph(text) = &parsed.children[0] {
817 text
818 } else {
819 panic!("Expected a paragraph");
820 };
821 assert_eq!(
822 paragraph.highlights,
823 vec![
824 (
825 5..7,
826 MarkdownHighlight::Style(MarkdownHighlightStyle {
827 weight: FontWeight::BOLD,
828 ..Default::default()
829 }),
830 ),
831 (
832 7..20,
833 MarkdownHighlight::Style(MarkdownHighlightStyle {
834 weight: FontWeight::BOLD,
835 strikethrough: true,
836 ..Default::default()
837 }),
838 ),
839 (
840 20..22,
841 MarkdownHighlight::Style(MarkdownHighlightStyle {
842 weight: FontWeight::BOLD,
843 ..Default::default()
844 }),
845 ),
846 ]
847 );
848 }
849
850 #[gpui::test]
851 async fn test_raw_links_detection() {
852 let parsed = parse("Checkout this https://zed.dev link").await;
853
854 assert_eq!(
855 parsed.children,
856 vec![p("Checkout this https://zed.dev link", 0..34)]
857 );
858
859 let paragraph = if let ParsedMarkdownElement::Paragraph(text) = &parsed.children[0] {
860 text
861 } else {
862 panic!("Expected a paragraph");
863 };
864 assert_eq!(
865 paragraph.highlights,
866 vec![(
867 14..29,
868 MarkdownHighlight::Style(MarkdownHighlightStyle {
869 underline: true,
870 ..Default::default()
871 }),
872 )]
873 );
874 assert_eq!(
875 paragraph.regions,
876 vec![ParsedRegion {
877 code: false,
878 link: Some(Link::Web {
879 url: "https://zed.dev".to_string()
880 }),
881 }]
882 );
883 assert_eq!(paragraph.region_ranges, vec![14..29]);
884 }
885
886 #[gpui::test]
887 async fn test_header_only_table() {
888 let markdown = "\
889| Header 1 | Header 2 |
890|----------|----------|
891
892Some other content
893";
894
895 let expected_table = table(
896 0..48,
897 row(vec![text("Header 1", 1..11), text("Header 2", 12..22)]),
898 vec![],
899 );
900
901 assert_eq!(
902 parse(markdown).await.children[0],
903 ParsedMarkdownElement::Table(expected_table)
904 );
905 }
906
907 #[gpui::test]
908 async fn test_basic_table() {
909 let markdown = "\
910| Header 1 | Header 2 |
911|----------|----------|
912| Cell 1 | Cell 2 |
913| Cell 3 | Cell 4 |";
914
915 let expected_table = table(
916 0..95,
917 row(vec![text("Header 1", 1..11), text("Header 2", 12..22)]),
918 vec![
919 row(vec![text("Cell 1", 49..59), text("Cell 2", 60..70)]),
920 row(vec![text("Cell 3", 73..83), text("Cell 4", 84..94)]),
921 ],
922 );
923
924 assert_eq!(
925 parse(markdown).await.children[0],
926 ParsedMarkdownElement::Table(expected_table)
927 );
928 }
929
930 #[gpui::test]
931 async fn test_list_basic() {
932 let parsed = parse(
933 "\
934* Item 1
935* Item 2
936* Item 3
937",
938 )
939 .await;
940
941 assert_eq!(
942 parsed.children,
943 vec![
944 list_item(0..8, 1, Unordered, vec![p("Item 1", 2..8)]),
945 list_item(9..17, 1, Unordered, vec![p("Item 2", 11..17)]),
946 list_item(18..26, 1, Unordered, vec![p("Item 3", 20..26)]),
947 ],
948 );
949 }
950
951 #[gpui::test]
952 async fn test_list_with_tasks() {
953 let parsed = parse(
954 "\
955- [ ] TODO
956- [x] Checked
957",
958 )
959 .await;
960
961 assert_eq!(
962 parsed.children,
963 vec![
964 list_item(0..10, 1, Task(false, 2..5), vec![p("TODO", 6..10)]),
965 list_item(11..24, 1, Task(true, 13..16), vec![p("Checked", 17..24)]),
966 ],
967 );
968 }
969
970 #[gpui::test]
971 async fn test_list_with_indented_task() {
972 let parsed = parse(
973 "\
974- [ ] TODO
975 - [x] Checked
976 - Unordered
977 1. Number 1
978 1. Number 2
9791. Number A
980",
981 )
982 .await;
983
984 assert_eq!(
985 parsed.children,
986 vec![
987 list_item(0..12, 1, Task(false, 2..5), vec![p("TODO", 6..10)]),
988 list_item(13..26, 2, Task(true, 15..18), vec![p("Checked", 19..26)]),
989 list_item(29..40, 2, Unordered, vec![p("Unordered", 31..40)]),
990 list_item(43..54, 2, Ordered(1), vec![p("Number 1", 46..54)]),
991 list_item(57..68, 2, Ordered(2), vec![p("Number 2", 60..68)]),
992 list_item(69..80, 1, Ordered(1), vec![p("Number A", 72..80)]),
993 ],
994 );
995 }
996
997 #[gpui::test]
998 async fn test_list_with_linebreak_is_handled_correctly() {
999 let parsed = parse(
1000 "\
1001- [ ] Task 1
1002
1003- [x] Task 2
1004",
1005 )
1006 .await;
1007
1008 assert_eq!(
1009 parsed.children,
1010 vec![
1011 list_item(0..13, 1, Task(false, 2..5), vec![p("Task 1", 6..12)]),
1012 list_item(14..26, 1, Task(true, 16..19), vec![p("Task 2", 20..26)]),
1013 ],
1014 );
1015 }
1016
1017 #[gpui::test]
1018 async fn test_list_nested() {
1019 let parsed = parse(
1020 "\
1021* Item 1
1022* Item 2
1023* Item 3
1024
10251. Hello
10261. Two
1027 1. Three
10282. Four
10293. Five
1030
1031* First
1032 1. Hello
1033 1. Goodbyte
1034 - Inner
1035 - Inner
1036 2. Goodbyte
1037 - Next item empty
1038 -
1039* Last
1040",
1041 )
1042 .await;
1043
1044 assert_eq!(
1045 parsed.children,
1046 vec![
1047 list_item(0..8, 1, Unordered, vec![p("Item 1", 2..8)]),
1048 list_item(9..17, 1, Unordered, vec![p("Item 2", 11..17)]),
1049 list_item(18..27, 1, Unordered, vec![p("Item 3", 20..26)]),
1050 list_item(28..36, 1, Ordered(1), vec![p("Hello", 31..36)]),
1051 list_item(37..46, 1, Ordered(2), vec![p("Two", 40..43),]),
1052 list_item(47..55, 2, Ordered(1), vec![p("Three", 50..55)]),
1053 list_item(56..63, 1, Ordered(3), vec![p("Four", 59..63)]),
1054 list_item(64..72, 1, Ordered(4), vec![p("Five", 67..71)]),
1055 list_item(73..82, 1, Unordered, vec![p("First", 75..80)]),
1056 list_item(83..96, 2, Ordered(1), vec![p("Hello", 86..91)]),
1057 list_item(97..116, 3, Ordered(1), vec![p("Goodbyte", 100..108)]),
1058 list_item(117..124, 4, Unordered, vec![p("Inner", 119..124)]),
1059 list_item(133..140, 4, Unordered, vec![p("Inner", 135..140)]),
1060 list_item(143..159, 2, Ordered(2), vec![p("Goodbyte", 146..154)]),
1061 list_item(160..180, 3, Unordered, vec![p("Next item empty", 165..180)]),
1062 list_item(186..190, 3, Unordered, vec![]),
1063 list_item(191..197, 1, Unordered, vec![p("Last", 193..197)]),
1064 ]
1065 );
1066 }
1067
1068 #[gpui::test]
1069 async fn test_list_with_nested_content() {
1070 let parsed = parse(
1071 "\
1072* This is a list item with two paragraphs.
1073
1074 This is the second paragraph in the list item.
1075",
1076 )
1077 .await;
1078
1079 assert_eq!(
1080 parsed.children,
1081 vec![list_item(
1082 0..96,
1083 1,
1084 Unordered,
1085 vec![
1086 p("This is a list item with two paragraphs.", 4..44),
1087 p("This is the second paragraph in the list item.", 50..97)
1088 ],
1089 ),],
1090 );
1091 }
1092
1093 #[gpui::test]
1094 async fn test_nested_list_with_paragraph_inside() {
1095 let parsed = parse(
1096 "\
10971. a
1098 1. b
1099 1. c
1100
1101 text
1102
1103 1. d
1104",
1105 )
1106 .await;
1107
1108 assert_eq!(
1109 parsed.children,
1110 vec![
1111 list_item(0..7, 1, Ordered(1), vec![p("a", 3..4)],),
1112 list_item(8..20, 2, Ordered(1), vec![p("b", 12..13),],),
1113 list_item(21..27, 3, Ordered(1), vec![p("c", 25..26),],),
1114 p("text", 32..37),
1115 list_item(41..46, 2, Ordered(1), vec![p("d", 45..46),],),
1116 ],
1117 );
1118 }
1119
1120 #[gpui::test]
1121 async fn test_list_with_leading_text() {
1122 let parsed = parse(
1123 "\
1124* `code`
1125* **bold**
1126* [link](https://example.com)
1127",
1128 )
1129 .await;
1130
1131 assert_eq!(
1132 parsed.children,
1133 vec![
1134 list_item(0..8, 1, Unordered, vec![p("code", 2..8)]),
1135 list_item(9..19, 1, Unordered, vec![p("bold", 11..19)]),
1136 list_item(20..49, 1, Unordered, vec![p("link", 22..49)],)
1137 ],
1138 );
1139 }
1140
1141 #[gpui::test]
1142 async fn test_simple_block_quote() {
1143 let parsed = parse("> Simple block quote with **styled text**").await;
1144
1145 assert_eq!(
1146 parsed.children,
1147 vec![block_quote(
1148 vec![p("Simple block quote with styled text", 2..41)],
1149 0..41
1150 )]
1151 );
1152 }
1153
1154 #[gpui::test]
1155 async fn test_simple_block_quote_with_multiple_lines() {
1156 let parsed = parse(
1157 "\
1158> # Heading
1159> More
1160> text
1161>
1162> More text
1163",
1164 )
1165 .await;
1166
1167 assert_eq!(
1168 parsed.children,
1169 vec![block_quote(
1170 vec![
1171 h1(text("Heading", 4..11), 2..12),
1172 p("More text", 14..26),
1173 p("More text", 30..40)
1174 ],
1175 0..40
1176 )]
1177 );
1178 }
1179
1180 #[gpui::test]
1181 async fn test_nested_block_quote() {
1182 let parsed = parse(
1183 "\
1184> A
1185>
1186> > # B
1187>
1188> C
1189
1190More text
1191",
1192 )
1193 .await;
1194
1195 assert_eq!(
1196 parsed.children,
1197 vec![
1198 block_quote(
1199 vec![
1200 p("A", 2..4),
1201 block_quote(vec![h1(text("B", 12..13), 10..14)], 8..14),
1202 p("C", 18..20)
1203 ],
1204 0..20
1205 ),
1206 p("More text", 21..31)
1207 ]
1208 );
1209 }
1210
1211 #[gpui::test]
1212 async fn test_code_block() {
1213 let parsed = parse(
1214 "\
1215```
1216fn main() {
1217 return 0;
1218}
1219```
1220",
1221 )
1222 .await;
1223
1224 assert_eq!(
1225 parsed.children,
1226 vec![code_block(
1227 None,
1228 "fn main() {\n return 0;\n}",
1229 0..35,
1230 None
1231 )]
1232 );
1233 }
1234
1235 #[gpui::test]
1236 async fn test_code_block_with_language(executor: BackgroundExecutor) {
1237 let language_registry = Arc::new(LanguageRegistry::test(executor.clone()));
1238 language_registry.add(rust_lang());
1239
1240 let parsed = parse_markdown(
1241 "\
1242```rust
1243fn main() {
1244 return 0;
1245}
1246```
1247",
1248 None,
1249 Some(language_registry),
1250 )
1251 .await;
1252
1253 assert_eq!(
1254 parsed.children,
1255 vec![code_block(
1256 Some("rust".to_string()),
1257 "fn main() {\n return 0;\n}",
1258 0..39,
1259 Some(vec![])
1260 )]
1261 );
1262 }
1263
1264 fn rust_lang() -> Arc<Language> {
1265 Arc::new(Language::new(
1266 LanguageConfig {
1267 name: "Rust".into(),
1268 matcher: LanguageMatcher {
1269 path_suffixes: vec!["rs".into()],
1270 ..Default::default()
1271 },
1272 collapsed_placeholder: " /* ... */ ".to_string(),
1273 ..Default::default()
1274 },
1275 Some(tree_sitter_rust::LANGUAGE.into()),
1276 ))
1277 }
1278
1279 fn h1(contents: ParsedMarkdownText, source_range: Range<usize>) -> ParsedMarkdownElement {
1280 ParsedMarkdownElement::Heading(ParsedMarkdownHeading {
1281 source_range,
1282 level: HeadingLevel::H1,
1283 contents,
1284 })
1285 }
1286
1287 fn h2(contents: ParsedMarkdownText, source_range: Range<usize>) -> ParsedMarkdownElement {
1288 ParsedMarkdownElement::Heading(ParsedMarkdownHeading {
1289 source_range,
1290 level: HeadingLevel::H2,
1291 contents,
1292 })
1293 }
1294
1295 fn h3(contents: ParsedMarkdownText, source_range: Range<usize>) -> ParsedMarkdownElement {
1296 ParsedMarkdownElement::Heading(ParsedMarkdownHeading {
1297 source_range,
1298 level: HeadingLevel::H3,
1299 contents,
1300 })
1301 }
1302
1303 fn p(contents: &str, source_range: Range<usize>) -> ParsedMarkdownElement {
1304 ParsedMarkdownElement::Paragraph(text(contents, source_range))
1305 }
1306
1307 fn text(contents: &str, source_range: Range<usize>) -> ParsedMarkdownText {
1308 ParsedMarkdownText {
1309 highlights: Vec::new(),
1310 region_ranges: Vec::new(),
1311 regions: Vec::new(),
1312 source_range,
1313 contents: contents.to_string(),
1314 }
1315 }
1316
1317 fn block_quote(
1318 children: Vec<ParsedMarkdownElement>,
1319 source_range: Range<usize>,
1320 ) -> ParsedMarkdownElement {
1321 ParsedMarkdownElement::BlockQuote(ParsedMarkdownBlockQuote {
1322 source_range,
1323 children,
1324 })
1325 }
1326
1327 fn code_block(
1328 language: Option<String>,
1329 code: &str,
1330 source_range: Range<usize>,
1331 highlights: Option<Vec<(Range<usize>, HighlightId)>>,
1332 ) -> ParsedMarkdownElement {
1333 ParsedMarkdownElement::CodeBlock(ParsedMarkdownCodeBlock {
1334 source_range,
1335 language,
1336 contents: code.to_string().into(),
1337 highlights,
1338 })
1339 }
1340
1341 fn list_item(
1342 source_range: Range<usize>,
1343 depth: u16,
1344 item_type: ParsedMarkdownListItemType,
1345 content: Vec<ParsedMarkdownElement>,
1346 ) -> ParsedMarkdownElement {
1347 ParsedMarkdownElement::ListItem(ParsedMarkdownListItem {
1348 source_range,
1349 item_type,
1350 depth,
1351 content,
1352 })
1353 }
1354
1355 fn table(
1356 source_range: Range<usize>,
1357 header: ParsedMarkdownTableRow,
1358 body: Vec<ParsedMarkdownTableRow>,
1359 ) -> ParsedMarkdownTable {
1360 ParsedMarkdownTable {
1361 column_alignments: Vec::new(),
1362 source_range,
1363 header,
1364 body,
1365 }
1366 }
1367
1368 fn row(children: Vec<ParsedMarkdownText>) -> ParsedMarkdownTableRow {
1369 ParsedMarkdownTableRow { children }
1370 }
1371
1372 impl PartialEq for ParsedMarkdownTable {
1373 fn eq(&self, other: &Self) -> bool {
1374 self.source_range == other.source_range
1375 && self.header == other.header
1376 && self.body == other.body
1377 }
1378 }
1379
1380 impl PartialEq for ParsedMarkdownText {
1381 fn eq(&self, other: &Self) -> bool {
1382 self.source_range == other.source_range && self.contents == other.contents
1383 }
1384 }
1385}