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, vec};
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 { .. })
105 | Event::Start(Tag::Emphasis)
106 | Event::Start(Tag::Strong)
107 | Event::Start(Tag::Strikethrough)
108 | Event::Start(Tag::Image { .. }) => {
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 { level, .. } => {
138 let level = *level;
139 self.cursor += 1;
140 let heading = self.parse_heading(level);
141 Some(vec![ParsedMarkdownElement::Heading(heading)])
142 }
143 Tag::Table(alignment) => {
144 let alignment = alignment.clone();
145 self.cursor += 1;
146 let table = self.parse_table(alignment);
147 Some(vec![ParsedMarkdownElement::Table(table)])
148 }
149 Tag::List(order) => {
150 let order = *order;
151 self.cursor += 1;
152 let list = self.parse_list(order).await;
153 Some(list)
154 }
155 Tag::BlockQuote(_kind) => {
156 self.cursor += 1;
157 let block_quote = self.parse_block_quote().await;
158 Some(vec![ParsedMarkdownElement::BlockQuote(block_quote)])
159 }
160 Tag::CodeBlock(kind) => {
161 let language = match kind {
162 pulldown_cmark::CodeBlockKind::Indented => None,
163 pulldown_cmark::CodeBlockKind::Fenced(language) => {
164 if language.is_empty() {
165 None
166 } else {
167 Some(language.to_string())
168 }
169 }
170 };
171
172 self.cursor += 1;
173
174 let code_block = self.parse_code_block(language).await;
175 Some(vec![ParsedMarkdownElement::CodeBlock(code_block)])
176 }
177 _ => None,
178 },
179 Event::Rule => {
180 let source_range = source_range.clone();
181 self.cursor += 1;
182 Some(vec![ParsedMarkdownElement::HorizontalRule(source_range)])
183 }
184 _ => None,
185 }
186 }
187
188 fn parse_text(
189 &mut self,
190 should_complete_on_soft_break: bool,
191 source_range: Option<Range<usize>>,
192 ) -> MarkdownParagraph {
193 let source_range = source_range.unwrap_or_else(|| {
194 self.current()
195 .map(|(_, range)| range.clone())
196 .unwrap_or_default()
197 });
198
199 let mut markdown_text_like = Vec::new();
200 let mut text = String::new();
201 let mut bold_depth = 0;
202 let mut italic_depth = 0;
203 let mut strikethrough_depth = 0;
204 let mut link: Option<Link> = None;
205 let mut image: Option<Image> = None;
206 let mut region_ranges: Vec<Range<usize>> = vec![];
207 let mut regions: Vec<ParsedRegion> = vec![];
208 let mut highlights: Vec<(Range<usize>, MarkdownHighlight)> = vec![];
209 let mut link_urls: Vec<String> = vec![];
210 let mut link_ranges: Vec<Range<usize>> = vec![];
211
212 loop {
213 if self.eof() {
214 break;
215 }
216
217 let (current, _) = self.current().unwrap();
218 let prev_len = text.len();
219 match current {
220 Event::SoftBreak => {
221 if should_complete_on_soft_break {
222 break;
223 }
224 text.push(' ');
225 }
226
227 Event::HardBreak => {
228 text.push('\n');
229 }
230
231 // We want to ignore any inline HTML tags in the text but keep
232 // the text between them
233 Event::InlineHtml(_) => {}
234
235 Event::Text(t) => {
236 text.push_str(t.as_ref());
237 let mut style = MarkdownHighlightStyle::default();
238
239 if bold_depth > 0 {
240 style.weight = FontWeight::BOLD;
241 }
242
243 if italic_depth > 0 {
244 style.italic = true;
245 }
246
247 if strikethrough_depth > 0 {
248 style.strikethrough = true;
249 }
250
251 let last_run_len = if let Some(link) = link.clone() {
252 region_ranges.push(prev_len..text.len());
253 regions.push(ParsedRegion {
254 code: false,
255 link: Some(link),
256 });
257 style.underline = true;
258 prev_len
259 } else {
260 // Manually scan for links
261 let mut finder = linkify::LinkFinder::new();
262 finder.kinds(&[linkify::LinkKind::Url]);
263 let mut last_link_len = prev_len;
264 for link in finder.links(t) {
265 let start = link.start();
266 let end = link.end();
267 let range = (prev_len + start)..(prev_len + end);
268 link_ranges.push(range.clone());
269 link_urls.push(link.as_str().to_string());
270
271 // If there is a style before we match a link, we have to add this to the highlighted ranges
272 if style != MarkdownHighlightStyle::default()
273 && last_link_len < link.start()
274 {
275 highlights.push((
276 last_link_len..link.start(),
277 MarkdownHighlight::Style(style.clone()),
278 ));
279 }
280
281 highlights.push((
282 range.clone(),
283 MarkdownHighlight::Style(MarkdownHighlightStyle {
284 underline: true,
285 ..style
286 }),
287 ));
288 region_ranges.push(range.clone());
289 regions.push(ParsedRegion {
290 code: false,
291 link: Some(Link::Web {
292 url: link.as_str().to_string(),
293 }),
294 });
295 last_link_len = end;
296 }
297 last_link_len
298 };
299
300 if style != MarkdownHighlightStyle::default() && last_run_len < text.len() {
301 let mut new_highlight = true;
302 if let Some((last_range, last_style)) = highlights.last_mut() {
303 if last_range.end == last_run_len
304 && last_style == &MarkdownHighlight::Style(style.clone())
305 {
306 last_range.end = text.len();
307 new_highlight = false;
308 }
309 }
310 if new_highlight {
311 highlights.push((
312 last_run_len..text.len(),
313 MarkdownHighlight::Style(style.clone()),
314 ));
315 }
316 }
317 }
318 Event::Code(t) => {
319 text.push_str(t.as_ref());
320 region_ranges.push(prev_len..text.len());
321
322 if link.is_some() {
323 highlights.push((
324 prev_len..text.len(),
325 MarkdownHighlight::Style(MarkdownHighlightStyle {
326 underline: true,
327 ..Default::default()
328 }),
329 ));
330 }
331 regions.push(ParsedRegion {
332 code: true,
333 link: link.clone(),
334 });
335 }
336 Event::Start(tag) => match tag {
337 Tag::Emphasis => italic_depth += 1,
338 Tag::Strong => bold_depth += 1,
339 Tag::Strikethrough => strikethrough_depth += 1,
340 Tag::Link { dest_url, .. } => {
341 link = Link::identify(
342 self.file_location_directory.clone(),
343 dest_url.to_string(),
344 );
345 }
346 Tag::Image { dest_url, .. } => {
347 if !text.is_empty() {
348 let parsed_regions = MarkdownParagraphChunk::Text(ParsedMarkdownText {
349 source_range: source_range.clone(),
350 contents: text.clone(),
351 highlights: highlights.clone(),
352 region_ranges: region_ranges.clone(),
353 regions: regions.clone(),
354 });
355 text = String::new();
356 highlights = vec![];
357 region_ranges = vec![];
358 regions = vec![];
359 markdown_text_like.push(parsed_regions);
360 }
361 image = Image::identify(
362 dest_url.to_string(),
363 source_range.clone(),
364 self.file_location_directory.clone(),
365 );
366 }
367 _ => {
368 break;
369 }
370 },
371
372 Event::End(tag) => match tag {
373 TagEnd::Emphasis => italic_depth -= 1,
374 TagEnd::Strong => bold_depth -= 1,
375 TagEnd::Strikethrough => strikethrough_depth -= 1,
376 TagEnd::Link => {
377 link = None;
378 }
379 TagEnd::Image => {
380 if let Some(mut image) = image.take() {
381 if !text.is_empty() {
382 image.alt_text = Some(std::mem::take(&mut text).into());
383 }
384 markdown_text_like.push(MarkdownParagraphChunk::Image(image));
385 }
386 }
387 TagEnd::Paragraph => {
388 self.cursor += 1;
389 break;
390 }
391 _ => {
392 break;
393 }
394 },
395 _ => {
396 break;
397 }
398 }
399
400 self.cursor += 1;
401 }
402 if !text.is_empty() {
403 markdown_text_like.push(MarkdownParagraphChunk::Text(ParsedMarkdownText {
404 source_range: source_range.clone(),
405 contents: text,
406 highlights,
407 regions,
408 region_ranges,
409 }));
410 }
411 markdown_text_like
412 }
413
414 fn parse_heading(&mut self, level: pulldown_cmark::HeadingLevel) -> ParsedMarkdownHeading {
415 let (_event, source_range) = self.previous().unwrap();
416 let source_range = source_range.clone();
417 let text = self.parse_text(true, None);
418
419 // Advance past the heading end tag
420 self.cursor += 1;
421
422 ParsedMarkdownHeading {
423 source_range: source_range.clone(),
424 level: match level {
425 pulldown_cmark::HeadingLevel::H1 => HeadingLevel::H1,
426 pulldown_cmark::HeadingLevel::H2 => HeadingLevel::H2,
427 pulldown_cmark::HeadingLevel::H3 => HeadingLevel::H3,
428 pulldown_cmark::HeadingLevel::H4 => HeadingLevel::H4,
429 pulldown_cmark::HeadingLevel::H5 => HeadingLevel::H5,
430 pulldown_cmark::HeadingLevel::H6 => HeadingLevel::H6,
431 },
432 contents: text,
433 }
434 }
435
436 fn parse_table(&mut self, alignment: Vec<Alignment>) -> ParsedMarkdownTable {
437 let (_event, source_range) = self.previous().unwrap();
438 let source_range = source_range.clone();
439 let mut header = ParsedMarkdownTableRow::new();
440 let mut body = vec![];
441 let mut current_row = vec![];
442 let mut in_header = true;
443 let column_alignments = alignment.iter().map(Self::convert_alignment).collect();
444
445 loop {
446 if self.eof() {
447 break;
448 }
449
450 let (current, source_range) = self.current().unwrap();
451 let source_range = source_range.clone();
452 match current {
453 Event::Start(Tag::TableHead)
454 | Event::Start(Tag::TableRow)
455 | Event::End(TagEnd::TableCell) => {
456 self.cursor += 1;
457 }
458 Event::Start(Tag::TableCell) => {
459 self.cursor += 1;
460 let cell_contents = self.parse_text(false, Some(source_range));
461 current_row.push(cell_contents);
462 }
463 Event::End(TagEnd::TableHead) | Event::End(TagEnd::TableRow) => {
464 self.cursor += 1;
465 let new_row = std::mem::take(&mut current_row);
466 if in_header {
467 header.children = new_row;
468 in_header = false;
469 } else {
470 let row = ParsedMarkdownTableRow::with_children(new_row);
471 body.push(row);
472 }
473 }
474 Event::End(TagEnd::Table) => {
475 self.cursor += 1;
476 break;
477 }
478 _ => {
479 break;
480 }
481 }
482 }
483
484 ParsedMarkdownTable {
485 source_range,
486 header,
487 body,
488 column_alignments,
489 }
490 }
491
492 fn convert_alignment(alignment: &Alignment) -> ParsedMarkdownTableAlignment {
493 match alignment {
494 Alignment::None => ParsedMarkdownTableAlignment::None,
495 Alignment::Left => ParsedMarkdownTableAlignment::Left,
496 Alignment::Center => ParsedMarkdownTableAlignment::Center,
497 Alignment::Right => ParsedMarkdownTableAlignment::Right,
498 }
499 }
500
501 async fn parse_list(&mut self, order: Option<u64>) -> Vec<ParsedMarkdownElement> {
502 let (_, list_source_range) = self.previous().unwrap();
503
504 let mut items = Vec::new();
505 let mut items_stack = vec![MarkdownListItem::default()];
506 let mut depth = 1;
507 let mut order = order;
508 let mut order_stack = Vec::new();
509
510 let mut insertion_indices = FxHashMap::default();
511 let mut source_ranges = FxHashMap::default();
512 let mut start_item_range = list_source_range.clone();
513
514 while !self.eof() {
515 let (current, source_range) = self.current().unwrap();
516 match current {
517 Event::Start(Tag::List(new_order)) => {
518 if items_stack.last().is_some() && !insertion_indices.contains_key(&depth) {
519 insertion_indices.insert(depth, items.len());
520 }
521
522 // We will use the start of the nested list as the end for the current item's range,
523 // because we don't care about the hierarchy of list items
524 if let collections::hash_map::Entry::Vacant(e) = source_ranges.entry(depth) {
525 e.insert(start_item_range.start..source_range.start);
526 }
527
528 order_stack.push(order);
529 order = *new_order;
530 self.cursor += 1;
531 depth += 1;
532 }
533 Event::End(TagEnd::List(_)) => {
534 order = order_stack.pop().flatten();
535 self.cursor += 1;
536 depth -= 1;
537
538 if depth == 0 {
539 break;
540 }
541 }
542 Event::Start(Tag::Item) => {
543 start_item_range = source_range.clone();
544
545 self.cursor += 1;
546 items_stack.push(MarkdownListItem::default());
547
548 let mut task_list = None;
549 // Check for task list marker (`- [ ]` or `- [x]`)
550 if let Some(event) = self.current_event() {
551 // If there is a linebreak in between two list items the task list marker will actually be the first element of the paragraph
552 if event == &Event::Start(Tag::Paragraph) {
553 self.cursor += 1;
554 }
555
556 if let Some((Event::TaskListMarker(checked), range)) = self.current() {
557 task_list = Some((*checked, range.clone()));
558 self.cursor += 1;
559 }
560 }
561
562 if let Some((event, range)) = self.current() {
563 // This is a plain list item.
564 // For example `- some text` or `1. [Docs](./docs.md)`
565 if MarkdownParser::is_text_like(event) {
566 let text = self.parse_text(false, Some(range.clone()));
567 let block = ParsedMarkdownElement::Paragraph(text);
568 if let Some(content) = items_stack.last_mut() {
569 let item_type = if let Some((checked, range)) = task_list {
570 ParsedMarkdownListItemType::Task(checked, range)
571 } else if let Some(order) = order {
572 ParsedMarkdownListItemType::Ordered(order)
573 } else {
574 ParsedMarkdownListItemType::Unordered
575 };
576 content.item_type = item_type;
577 content.content.push(block);
578 }
579 } else {
580 let block = self.parse_block().await;
581 if let Some(block) = block {
582 if let Some(list_item) = items_stack.last_mut() {
583 list_item.content.extend(block);
584 }
585 }
586 }
587 }
588
589 // If there is a linebreak in between two list items the task list marker will actually be the first element of the paragraph
590 if self.current_event() == Some(&Event::End(TagEnd::Paragraph)) {
591 self.cursor += 1;
592 }
593 }
594 Event::End(TagEnd::Item) => {
595 self.cursor += 1;
596
597 if let Some(current) = order {
598 order = Some(current + 1);
599 }
600
601 if let Some(list_item) = items_stack.pop() {
602 let source_range = source_ranges
603 .remove(&depth)
604 .unwrap_or(start_item_range.clone());
605
606 // We need to remove the last character of the source range, because it includes the newline character
607 let source_range = source_range.start..source_range.end - 1;
608 let item = ParsedMarkdownElement::ListItem(ParsedMarkdownListItem {
609 source_range,
610 content: list_item.content,
611 depth,
612 item_type: list_item.item_type,
613 });
614
615 if let Some(index) = insertion_indices.get(&depth) {
616 items.insert(*index, item);
617 insertion_indices.remove(&depth);
618 } else {
619 items.push(item);
620 }
621 }
622 }
623 _ => {
624 if depth == 0 {
625 break;
626 }
627 // This can only happen if a list item starts with more then one paragraph,
628 // or the list item contains blocks that should be rendered after the nested list items
629 let block = self.parse_block().await;
630 if let Some(block) = block {
631 if let Some(list_item) = items_stack.last_mut() {
632 // 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
633 if !insertion_indices.contains_key(&depth) {
634 list_item.content.extend(block);
635 continue;
636 }
637 }
638
639 // Otherwise we need to insert the block after all the nested items
640 // that have been parsed so far
641 items.extend(block);
642 } else {
643 self.cursor += 1;
644 }
645 }
646 }
647 }
648
649 items
650 }
651
652 #[async_recursion]
653 async fn parse_block_quote(&mut self) -> ParsedMarkdownBlockQuote {
654 let (_event, source_range) = self.previous().unwrap();
655 let source_range = source_range.clone();
656 let mut nested_depth = 1;
657
658 let mut children: Vec<ParsedMarkdownElement> = vec![];
659
660 while !self.eof() {
661 let block = self.parse_block().await;
662
663 if let Some(block) = block {
664 children.extend(block);
665 } else {
666 break;
667 }
668
669 if self.eof() {
670 break;
671 }
672
673 let (current, _source_range) = self.current().unwrap();
674 match current {
675 // This is a nested block quote.
676 // Record that we're in a nested block quote and continue parsing.
677 // We don't need to advance the cursor since the next
678 // call to `parse_block` will handle it.
679 Event::Start(Tag::BlockQuote(_kind)) => {
680 nested_depth += 1;
681 }
682 Event::End(TagEnd::BlockQuote(_kind)) => {
683 nested_depth -= 1;
684 if nested_depth == 0 {
685 self.cursor += 1;
686 break;
687 }
688 }
689 _ => {}
690 };
691 }
692
693 ParsedMarkdownBlockQuote {
694 source_range,
695 children,
696 }
697 }
698
699 async fn parse_code_block(&mut self, language: Option<String>) -> ParsedMarkdownCodeBlock {
700 let (_event, source_range) = self.previous().unwrap();
701 let source_range = source_range.clone();
702 let mut code = String::new();
703
704 while !self.eof() {
705 let (current, _source_range) = self.current().unwrap();
706 match current {
707 Event::Text(text) => {
708 code.push_str(text);
709 self.cursor += 1;
710 }
711 Event::End(TagEnd::CodeBlock) => {
712 self.cursor += 1;
713 break;
714 }
715 _ => {
716 break;
717 }
718 }
719 }
720 let highlights = if let Some(language) = &language {
721 if let Some(registry) = &self.language_registry {
722 let rope: language::Rope = code.as_str().into();
723 registry
724 .language_for_name_or_extension(language)
725 .await
726 .map(|l| l.highlight_text(&rope, 0..code.len()))
727 .ok()
728 } else {
729 None
730 }
731 } else {
732 None
733 };
734
735 ParsedMarkdownCodeBlock {
736 source_range,
737 contents: code.trim().to_string().into(),
738 language,
739 highlights,
740 }
741 }
742}
743
744#[cfg(test)]
745mod tests {
746 use core::panic;
747
748 use super::*;
749
750 use gpui::BackgroundExecutor;
751 use language::{
752 tree_sitter_rust, HighlightId, Language, LanguageConfig, LanguageMatcher, LanguageRegistry,
753 };
754 use pretty_assertions::assert_eq;
755 use ParsedMarkdownListItemType::*;
756
757 async fn parse(input: &str) -> ParsedMarkdown {
758 parse_markdown(input, None, None).await
759 }
760
761 #[gpui::test]
762 async fn test_headings() {
763 let parsed = parse("# Heading one\n## Heading two\n### Heading three").await;
764
765 assert_eq!(
766 parsed.children,
767 vec![
768 h1(text("Heading one", 2..13), 0..14),
769 h2(text("Heading two", 17..28), 14..29),
770 h3(text("Heading three", 33..46), 29..46),
771 ]
772 );
773 }
774
775 #[gpui::test]
776 async fn test_newlines_dont_new_paragraphs() {
777 let parsed = parse("Some text **that is bolded**\n and *italicized*").await;
778
779 assert_eq!(
780 parsed.children,
781 vec![p("Some text that is bolded and italicized", 0..46)]
782 );
783 }
784
785 #[gpui::test]
786 async fn test_heading_with_paragraph() {
787 let parsed = parse("# Zed\nThe editor").await;
788
789 assert_eq!(
790 parsed.children,
791 vec![h1(text("Zed", 2..5), 0..6), p("The editor", 6..16),]
792 );
793 }
794
795 #[gpui::test]
796 async fn test_double_newlines_do_new_paragraphs() {
797 let parsed = parse("Some text **that is bolded**\n\n and *italicized*").await;
798
799 assert_eq!(
800 parsed.children,
801 vec![
802 p("Some text that is bolded", 0..29),
803 p("and italicized", 31..47),
804 ]
805 );
806 }
807
808 #[gpui::test]
809 async fn test_bold_italic_text() {
810 let parsed = parse("Some text **that is bolded** and *italicized*").await;
811
812 assert_eq!(
813 parsed.children,
814 vec![p("Some text that is bolded and italicized", 0..45)]
815 );
816 }
817
818 #[gpui::test]
819 async fn test_nested_bold_strikethrough_text() {
820 let parsed = parse("Some **bo~~strikethrough~~ld** text").await;
821
822 assert_eq!(parsed.children.len(), 1);
823 assert_eq!(
824 parsed.children[0],
825 ParsedMarkdownElement::Paragraph(vec![MarkdownParagraphChunk::Text(
826 ParsedMarkdownText {
827 source_range: 0..35,
828 contents: "Some bostrikethroughld text".to_string(),
829 highlights: Vec::new(),
830 region_ranges: Vec::new(),
831 regions: Vec::new(),
832 }
833 )])
834 );
835
836 let new_text = if let ParsedMarkdownElement::Paragraph(text) = &parsed.children[0] {
837 text
838 } else {
839 panic!("Expected a paragraph");
840 };
841
842 let paragraph = if let MarkdownParagraphChunk::Text(text) = &new_text[0] {
843 text
844 } else {
845 panic!("Expected a text");
846 };
847
848 assert_eq!(
849 paragraph.highlights,
850 vec![
851 (
852 5..7,
853 MarkdownHighlight::Style(MarkdownHighlightStyle {
854 weight: FontWeight::BOLD,
855 ..Default::default()
856 }),
857 ),
858 (
859 7..20,
860 MarkdownHighlight::Style(MarkdownHighlightStyle {
861 weight: FontWeight::BOLD,
862 strikethrough: true,
863 ..Default::default()
864 }),
865 ),
866 (
867 20..22,
868 MarkdownHighlight::Style(MarkdownHighlightStyle {
869 weight: FontWeight::BOLD,
870 ..Default::default()
871 }),
872 ),
873 ]
874 );
875 }
876
877 #[gpui::test]
878 async fn test_text_with_inline_html() {
879 let parsed = parse("This is a paragraph with an inline HTML <sometag>tag</sometag>.").await;
880
881 assert_eq!(
882 parsed.children,
883 vec![p("This is a paragraph with an inline HTML tag.", 0..63),],
884 );
885 }
886
887 #[gpui::test]
888 async fn test_raw_links_detection() {
889 let parsed = parse("Checkout this https://zed.dev link").await;
890
891 assert_eq!(
892 parsed.children,
893 vec![p("Checkout this https://zed.dev link", 0..34)]
894 );
895 }
896
897 #[gpui::test]
898 async fn test_empty_image() {
899 let parsed = parse("![]()").await;
900
901 let paragraph = if let ParsedMarkdownElement::Paragraph(text) = &parsed.children[0] {
902 text
903 } else {
904 panic!("Expected a paragraph");
905 };
906 assert_eq!(paragraph.len(), 0);
907 }
908
909 #[gpui::test]
910 async fn test_image_links_detection() {
911 let parsed = parse("").await;
912
913 let paragraph = if let ParsedMarkdownElement::Paragraph(text) = &parsed.children[0] {
914 text
915 } else {
916 panic!("Expected a paragraph");
917 };
918 assert_eq!(
919 paragraph[0],
920 MarkdownParagraphChunk::Image(Image {
921 source_range: 0..111,
922 link: Link::Web {
923 url: "https://blog.logrocket.com/wp-content/uploads/2024/04/exploring-zed-open-source-code-editor-rust-2.png".to_string(),
924 },
925 alt_text: Some("test".into()),
926 },)
927 );
928 }
929
930 #[gpui::test]
931 async fn test_image_without_alt_text() {
932 let parsed = parse("").await;
933
934 let paragraph = if let ParsedMarkdownElement::Paragraph(text) = &parsed.children[0] {
935 text
936 } else {
937 panic!("Expected a paragraph");
938 };
939 assert_eq!(
940 paragraph[0],
941 MarkdownParagraphChunk::Image(Image {
942 source_range: 0..31,
943 link: Link::Web {
944 url: "http://example.com/foo.png".to_string(),
945 },
946 alt_text: None,
947 },)
948 );
949 }
950
951 #[gpui::test]
952 async fn test_image_with_alt_text_containing_formatting() {
953 let parsed = parse("").await;
954
955 let ParsedMarkdownElement::Paragraph(chunks) = &parsed.children[0] else {
956 panic!("Expected a paragraph");
957 };
958 assert_eq!(
959 chunks,
960 &[MarkdownParagraphChunk::Image(Image {
961 source_range: 0..44,
962 link: Link::Web {
963 url: "http://example.com/foo.png".to_string(),
964 },
965 alt_text: Some("foo bar baz".into()),
966 }),],
967 );
968 }
969
970 #[gpui::test]
971 async fn test_images_with_text_in_between() {
972 let parsed = parse(
973 "\nLorem Ipsum\n",
974 )
975 .await;
976
977 let chunks = if let ParsedMarkdownElement::Paragraph(text) = &parsed.children[0] {
978 text
979 } else {
980 panic!("Expected a paragraph");
981 };
982 assert_eq!(
983 chunks,
984 &vec![
985 MarkdownParagraphChunk::Image(Image {
986 source_range: 0..81,
987 link: Link::Web {
988 url: "http://example.com/foo.png".to_string(),
989 },
990 alt_text: Some("foo".into()),
991 }),
992 MarkdownParagraphChunk::Text(ParsedMarkdownText {
993 source_range: 0..81,
994 contents: " Lorem Ipsum ".to_string(),
995 highlights: Vec::new(),
996 region_ranges: Vec::new(),
997 regions: Vec::new(),
998 }),
999 MarkdownParagraphChunk::Image(Image {
1000 source_range: 0..81,
1001 link: Link::Web {
1002 url: "http://example.com/bar.png".to_string(),
1003 },
1004 alt_text: Some("bar".into()),
1005 })
1006 ]
1007 );
1008 }
1009
1010 #[gpui::test]
1011 async fn test_header_only_table() {
1012 let markdown = "\
1013| Header 1 | Header 2 |
1014|----------|----------|
1015
1016Some other content
1017";
1018
1019 let expected_table = table(
1020 0..48,
1021 row(vec![text("Header 1", 1..11), text("Header 2", 12..22)]),
1022 vec![],
1023 );
1024
1025 assert_eq!(
1026 parse(markdown).await.children[0],
1027 ParsedMarkdownElement::Table(expected_table)
1028 );
1029 }
1030
1031 #[gpui::test]
1032 async fn test_basic_table() {
1033 let markdown = "\
1034| Header 1 | Header 2 |
1035|----------|----------|
1036| Cell 1 | Cell 2 |
1037| Cell 3 | Cell 4 |";
1038
1039 let expected_table = table(
1040 0..95,
1041 row(vec![text("Header 1", 1..11), text("Header 2", 12..22)]),
1042 vec![
1043 row(vec![text("Cell 1", 49..59), text("Cell 2", 60..70)]),
1044 row(vec![text("Cell 3", 73..83), text("Cell 4", 84..94)]),
1045 ],
1046 );
1047
1048 assert_eq!(
1049 parse(markdown).await.children[0],
1050 ParsedMarkdownElement::Table(expected_table)
1051 );
1052 }
1053
1054 #[gpui::test]
1055 async fn test_list_basic() {
1056 let parsed = parse(
1057 "\
1058* Item 1
1059* Item 2
1060* Item 3
1061",
1062 )
1063 .await;
1064
1065 assert_eq!(
1066 parsed.children,
1067 vec![
1068 list_item(0..8, 1, Unordered, vec![p("Item 1", 2..8)]),
1069 list_item(9..17, 1, Unordered, vec![p("Item 2", 11..17)]),
1070 list_item(18..26, 1, Unordered, vec![p("Item 3", 20..26)]),
1071 ],
1072 );
1073 }
1074
1075 #[gpui::test]
1076 async fn test_list_with_tasks() {
1077 let parsed = parse(
1078 "\
1079- [ ] TODO
1080- [x] Checked
1081",
1082 )
1083 .await;
1084
1085 assert_eq!(
1086 parsed.children,
1087 vec![
1088 list_item(0..10, 1, Task(false, 2..5), vec![p("TODO", 6..10)]),
1089 list_item(11..24, 1, Task(true, 13..16), vec![p("Checked", 17..24)]),
1090 ],
1091 );
1092 }
1093
1094 #[gpui::test]
1095 async fn test_list_with_indented_task() {
1096 let parsed = parse(
1097 "\
1098- [ ] TODO
1099 - [x] Checked
1100 - Unordered
1101 1. Number 1
1102 1. Number 2
11031. Number A
1104",
1105 )
1106 .await;
1107
1108 assert_eq!(
1109 parsed.children,
1110 vec![
1111 list_item(0..12, 1, Task(false, 2..5), vec![p("TODO", 6..10)]),
1112 list_item(13..26, 2, Task(true, 15..18), vec![p("Checked", 19..26)]),
1113 list_item(29..40, 2, Unordered, vec![p("Unordered", 31..40)]),
1114 list_item(43..54, 2, Ordered(1), vec![p("Number 1", 46..54)]),
1115 list_item(57..68, 2, Ordered(2), vec![p("Number 2", 60..68)]),
1116 list_item(69..80, 1, Ordered(1), vec![p("Number A", 72..80)]),
1117 ],
1118 );
1119 }
1120
1121 #[gpui::test]
1122 async fn test_list_with_linebreak_is_handled_correctly() {
1123 let parsed = parse(
1124 "\
1125- [ ] Task 1
1126
1127- [x] Task 2
1128",
1129 )
1130 .await;
1131
1132 assert_eq!(
1133 parsed.children,
1134 vec![
1135 list_item(0..13, 1, Task(false, 2..5), vec![p("Task 1", 6..12)]),
1136 list_item(14..26, 1, Task(true, 16..19), vec![p("Task 2", 20..26)]),
1137 ],
1138 );
1139 }
1140
1141 #[gpui::test]
1142 async fn test_list_nested() {
1143 let parsed = parse(
1144 "\
1145* Item 1
1146* Item 2
1147* Item 3
1148
11491. Hello
11501. Two
1151 1. Three
11522. Four
11533. Five
1154
1155* First
1156 1. Hello
1157 1. Goodbyte
1158 - Inner
1159 - Inner
1160 2. Goodbyte
1161 - Next item empty
1162 -
1163* Last
1164",
1165 )
1166 .await;
1167
1168 assert_eq!(
1169 parsed.children,
1170 vec![
1171 list_item(0..8, 1, Unordered, vec![p("Item 1", 2..8)]),
1172 list_item(9..17, 1, Unordered, vec![p("Item 2", 11..17)]),
1173 list_item(18..27, 1, Unordered, vec![p("Item 3", 20..26)]),
1174 list_item(28..36, 1, Ordered(1), vec![p("Hello", 31..36)]),
1175 list_item(37..46, 1, Ordered(2), vec![p("Two", 40..43),]),
1176 list_item(47..55, 2, Ordered(1), vec![p("Three", 50..55)]),
1177 list_item(56..63, 1, Ordered(3), vec![p("Four", 59..63)]),
1178 list_item(64..72, 1, Ordered(4), vec![p("Five", 67..71)]),
1179 list_item(73..82, 1, Unordered, vec![p("First", 75..80)]),
1180 list_item(83..96, 2, Ordered(1), vec![p("Hello", 86..91)]),
1181 list_item(97..116, 3, Ordered(1), vec![p("Goodbyte", 100..108)]),
1182 list_item(117..124, 4, Unordered, vec![p("Inner", 119..124)]),
1183 list_item(133..140, 4, Unordered, vec![p("Inner", 135..140)]),
1184 list_item(143..159, 2, Ordered(2), vec![p("Goodbyte", 146..154)]),
1185 list_item(160..180, 3, Unordered, vec![p("Next item empty", 165..180)]),
1186 list_item(186..190, 3, Unordered, vec![]),
1187 list_item(191..197, 1, Unordered, vec![p("Last", 193..197)]),
1188 ]
1189 );
1190 }
1191
1192 #[gpui::test]
1193 async fn test_list_with_nested_content() {
1194 let parsed = parse(
1195 "\
1196* This is a list item with two paragraphs.
1197
1198 This is the second paragraph in the list item.
1199",
1200 )
1201 .await;
1202
1203 assert_eq!(
1204 parsed.children,
1205 vec![list_item(
1206 0..96,
1207 1,
1208 Unordered,
1209 vec![
1210 p("This is a list item with two paragraphs.", 4..44),
1211 p("This is the second paragraph in the list item.", 50..97)
1212 ],
1213 ),],
1214 );
1215 }
1216
1217 #[gpui::test]
1218 async fn test_list_item_with_inline_html() {
1219 let parsed = parse(
1220 "\
1221* This is a list item with an inline HTML <sometag>tag</sometag>.
1222",
1223 )
1224 .await;
1225
1226 assert_eq!(
1227 parsed.children,
1228 vec![list_item(
1229 0..67,
1230 1,
1231 Unordered,
1232 vec![p("This is a list item with an inline HTML tag.", 4..44),],
1233 ),],
1234 );
1235 }
1236
1237 #[gpui::test]
1238 async fn test_nested_list_with_paragraph_inside() {
1239 let parsed = parse(
1240 "\
12411. a
1242 1. b
1243 1. c
1244
1245 text
1246
1247 1. d
1248",
1249 )
1250 .await;
1251
1252 assert_eq!(
1253 parsed.children,
1254 vec![
1255 list_item(0..7, 1, Ordered(1), vec![p("a", 3..4)],),
1256 list_item(8..20, 2, Ordered(1), vec![p("b", 12..13),],),
1257 list_item(21..27, 3, Ordered(1), vec![p("c", 25..26),],),
1258 p("text", 32..37),
1259 list_item(41..46, 2, Ordered(1), vec![p("d", 45..46),],),
1260 ],
1261 );
1262 }
1263
1264 #[gpui::test]
1265 async fn test_list_with_leading_text() {
1266 let parsed = parse(
1267 "\
1268* `code`
1269* **bold**
1270* [link](https://example.com)
1271",
1272 )
1273 .await;
1274
1275 assert_eq!(
1276 parsed.children,
1277 vec![
1278 list_item(0..8, 1, Unordered, vec![p("code", 2..8)]),
1279 list_item(9..19, 1, Unordered, vec![p("bold", 11..19)]),
1280 list_item(20..49, 1, Unordered, vec![p("link", 22..49)],),
1281 ],
1282 );
1283 }
1284
1285 #[gpui::test]
1286 async fn test_simple_block_quote() {
1287 let parsed = parse("> Simple block quote with **styled text**").await;
1288
1289 assert_eq!(
1290 parsed.children,
1291 vec![block_quote(
1292 vec![p("Simple block quote with styled text", 2..41)],
1293 0..41
1294 )]
1295 );
1296 }
1297
1298 #[gpui::test]
1299 async fn test_simple_block_quote_with_multiple_lines() {
1300 let parsed = parse(
1301 "\
1302> # Heading
1303> More
1304> text
1305>
1306> More text
1307",
1308 )
1309 .await;
1310
1311 assert_eq!(
1312 parsed.children,
1313 vec![block_quote(
1314 vec![
1315 h1(text("Heading", 4..11), 2..12),
1316 p("More text", 14..26),
1317 p("More text", 30..40)
1318 ],
1319 0..40
1320 )]
1321 );
1322 }
1323
1324 #[gpui::test]
1325 async fn test_nested_block_quote() {
1326 let parsed = parse(
1327 "\
1328> A
1329>
1330> > # B
1331>
1332> C
1333
1334More text
1335",
1336 )
1337 .await;
1338
1339 assert_eq!(
1340 parsed.children,
1341 vec![
1342 block_quote(
1343 vec![
1344 p("A", 2..4),
1345 block_quote(vec![h1(text("B", 12..13), 10..14)], 8..14),
1346 p("C", 18..20)
1347 ],
1348 0..20
1349 ),
1350 p("More text", 21..31)
1351 ]
1352 );
1353 }
1354
1355 #[gpui::test]
1356 async fn test_code_block() {
1357 let parsed = parse(
1358 "\
1359```
1360fn main() {
1361 return 0;
1362}
1363```
1364",
1365 )
1366 .await;
1367
1368 assert_eq!(
1369 parsed.children,
1370 vec![code_block(
1371 None,
1372 "fn main() {\n return 0;\n}",
1373 0..35,
1374 None
1375 )]
1376 );
1377 }
1378
1379 #[gpui::test]
1380 async fn test_code_block_with_language(executor: BackgroundExecutor) {
1381 let language_registry = Arc::new(LanguageRegistry::test(executor.clone()));
1382 language_registry.add(rust_lang());
1383
1384 let parsed = parse_markdown(
1385 "\
1386```rust
1387fn main() {
1388 return 0;
1389}
1390```
1391",
1392 None,
1393 Some(language_registry),
1394 )
1395 .await;
1396
1397 assert_eq!(
1398 parsed.children,
1399 vec![code_block(
1400 Some("rust".to_string()),
1401 "fn main() {\n return 0;\n}",
1402 0..39,
1403 Some(vec![])
1404 )]
1405 );
1406 }
1407
1408 fn rust_lang() -> Arc<Language> {
1409 Arc::new(Language::new(
1410 LanguageConfig {
1411 name: "Rust".into(),
1412 matcher: LanguageMatcher {
1413 path_suffixes: vec!["rs".into()],
1414 ..Default::default()
1415 },
1416 collapsed_placeholder: " /* ... */ ".to_string(),
1417 ..Default::default()
1418 },
1419 Some(tree_sitter_rust::LANGUAGE.into()),
1420 ))
1421 }
1422
1423 fn h1(contents: MarkdownParagraph, source_range: Range<usize>) -> ParsedMarkdownElement {
1424 ParsedMarkdownElement::Heading(ParsedMarkdownHeading {
1425 source_range,
1426 level: HeadingLevel::H1,
1427 contents,
1428 })
1429 }
1430
1431 fn h2(contents: MarkdownParagraph, source_range: Range<usize>) -> ParsedMarkdownElement {
1432 ParsedMarkdownElement::Heading(ParsedMarkdownHeading {
1433 source_range,
1434 level: HeadingLevel::H2,
1435 contents,
1436 })
1437 }
1438
1439 fn h3(contents: MarkdownParagraph, source_range: Range<usize>) -> ParsedMarkdownElement {
1440 ParsedMarkdownElement::Heading(ParsedMarkdownHeading {
1441 source_range,
1442 level: HeadingLevel::H3,
1443 contents,
1444 })
1445 }
1446
1447 fn p(contents: &str, source_range: Range<usize>) -> ParsedMarkdownElement {
1448 ParsedMarkdownElement::Paragraph(text(contents, source_range))
1449 }
1450
1451 fn text(contents: &str, source_range: Range<usize>) -> MarkdownParagraph {
1452 vec![MarkdownParagraphChunk::Text(ParsedMarkdownText {
1453 highlights: Vec::new(),
1454 region_ranges: Vec::new(),
1455 regions: Vec::new(),
1456 source_range,
1457 contents: contents.to_string(),
1458 })]
1459 }
1460
1461 fn block_quote(
1462 children: Vec<ParsedMarkdownElement>,
1463 source_range: Range<usize>,
1464 ) -> ParsedMarkdownElement {
1465 ParsedMarkdownElement::BlockQuote(ParsedMarkdownBlockQuote {
1466 source_range,
1467 children,
1468 })
1469 }
1470
1471 fn code_block(
1472 language: Option<String>,
1473 code: &str,
1474 source_range: Range<usize>,
1475 highlights: Option<Vec<(Range<usize>, HighlightId)>>,
1476 ) -> ParsedMarkdownElement {
1477 ParsedMarkdownElement::CodeBlock(ParsedMarkdownCodeBlock {
1478 source_range,
1479 language,
1480 contents: code.to_string().into(),
1481 highlights,
1482 })
1483 }
1484
1485 fn list_item(
1486 source_range: Range<usize>,
1487 depth: u16,
1488 item_type: ParsedMarkdownListItemType,
1489 content: Vec<ParsedMarkdownElement>,
1490 ) -> ParsedMarkdownElement {
1491 ParsedMarkdownElement::ListItem(ParsedMarkdownListItem {
1492 source_range,
1493 item_type,
1494 depth,
1495 content,
1496 })
1497 }
1498
1499 fn table(
1500 source_range: Range<usize>,
1501 header: ParsedMarkdownTableRow,
1502 body: Vec<ParsedMarkdownTableRow>,
1503 ) -> ParsedMarkdownTable {
1504 ParsedMarkdownTable {
1505 column_alignments: Vec::new(),
1506 source_range,
1507 header,
1508 body,
1509 }
1510 }
1511
1512 fn row(children: Vec<MarkdownParagraph>) -> ParsedMarkdownTableRow {
1513 ParsedMarkdownTableRow { children }
1514 }
1515
1516 impl PartialEq for ParsedMarkdownTable {
1517 fn eq(&self, other: &Self) -> bool {
1518 self.source_range == other.source_range
1519 && self.header == other.header
1520 && self.body == other.body
1521 }
1522 }
1523
1524 impl PartialEq for ParsedMarkdownText {
1525 fn eq(&self, other: &Self) -> bool {
1526 self.source_range == other.source_range && self.contents == other.contents
1527 }
1528 }
1529}