1use std::collections::BTreeMap;
2
3use cpd_core::models::{DetectionToken, Token};
4
5use crate::embedded::{blank_ranges_preserve_newlines, html_only_maps, tokenize_blocks_shifted};
6use crate::line_index::LineIndex;
7use crate::markdown::{offset_detection_tokens, tokens_to_detection};
8use crate::tokenizer::{Mode, TokenMap, TokenizeOptions, tokenize_format_to_detection};
9
10#[derive(Debug, Clone)]
11pub struct Block {
12 pub block_format: String,
13 pub content: String,
14 pub start_offset: usize,
15 pub start_line: u32,
16}
17
18struct SfcBlock {
19 block_format: String,
20 block_start: usize,
21 inner_start: usize,
22 inner_end: usize,
23 block_end: usize,
24}
25
26pub fn tokenize_sfc_maps(
27 source: &str,
28 file_format: &str,
29 options: &TokenizeOptions,
30) -> Vec<TokenMap> {
31 if source.is_empty() {
32 return Vec::new();
33 }
34
35 let blocks = find_sfc_blocks(source, file_format);
36 if blocks.is_empty() {
37 return html_only_maps(source, options);
38 }
39
40 let blank_ranges: Vec<[usize; 2]> = blocks
41 .iter()
42 .filter_map(|b| {
43 if b.inner_start < b.inner_end {
44 Some([b.inner_start, b.inner_end])
45 } else {
46 None
47 }
48 })
49 .collect();
50
51 let sanitized = blank_ranges_preserve_newlines(source, &blank_ranges);
52 let line_index = LineIndex::new(source.as_bytes());
53
54 let mut grouped: BTreeMap<String, Vec<DetectionToken>> = BTreeMap::new();
55
56 let wrapper_tags_only = file_format == "vue";
67
68 let markup_tokens = crate::generic::tokenize_generic(&sanitized, "html");
69 let mut markup_detection = tokens_to_detection(markup_tokens, options);
70 markup_detection.retain(|t| t.range[0] < t.range[1]);
71 if !markup_detection.is_empty() && !wrapper_tags_only {
72 grouped
73 .entry("html".to_string())
74 .or_default()
75 .extend(markup_detection);
76 }
77
78 for block in &blocks {
79 if block.inner_start >= block.inner_end {
80 continue;
81 }
82 let inner = &source[block.inner_start..block.inner_end];
83 let inner_start_loc = line_index.location(block.inner_start);
84
85 let mut inner_tokens = tokenize_format_to_detection(&block.block_format, inner, options);
86 offset_detection_tokens(&mut inner_tokens, block.inner_start, &inner_start_loc);
87
88 grouped
89 .entry(block.block_format.clone())
90 .or_default()
91 .extend(inner_tokens);
92 }
93
94 if let Some(tokens) = grouped.get_mut("html") {
97 tokens.sort_by_key(|token| token.range[0]);
98 }
99
100 grouped
101 .into_iter()
102 .filter(|(_, tokens)| !tokens.is_empty())
103 .map(|(format, tokens)| TokenMap { format, tokens })
104 .collect()
105}
106
107pub fn script_blocks(source: &str, file_format: &str) -> Vec<(String, std::ops::Range<usize>)> {
111 find_sfc_blocks(source, file_format)
112 .into_iter()
113 .filter(|b| b.inner_start < b.inner_end)
114 .filter(|b| {
115 !matches!(
116 b.block_format.as_str(),
117 "html" | "pug" | "css" | "scss" | "less"
118 )
119 })
120 .map(|b| (b.block_format, b.inner_start..b.inner_end))
121 .collect()
122}
123
124fn sfc_tag_names(file_format: &str) -> &[&'static str] {
125 match file_format {
126 "svelte" | "astro" => &["script", "style"],
127 _ => &["template", "script", "style"],
128 }
129}
130
131fn find_sfc_blocks(source: &str, file_format: &str) -> Vec<SfcBlock> {
132 let source_lower = source.to_ascii_lowercase();
133 let mut blocks = Vec::new();
134
135 if file_format == "astro"
136 && let Some(fm) = astro_frontmatter_block(source)
137 {
138 blocks.push(fm);
139 }
140
141 for tag in sfc_tag_names(file_format) {
142 let mut search_from = 0usize;
143 while let Some(block) = find_sfc_tag_block(source, &source_lower, tag, search_from) {
144 search_from = block.block_end;
145 blocks.push(block);
146 }
147 }
148
149 blocks.sort_by_key(|b| b.block_start);
150 let mut deduped = Vec::new();
151 for block in blocks {
152 let nested = deduped.iter().any(|existing: &SfcBlock| {
153 block.block_start >= existing.block_start && block.block_start < existing.block_end
154 });
155 if !nested {
156 deduped.push(block);
157 }
158 }
159 deduped
160}
161
162fn find_tag_bounds(
163 _source: &str,
164 source_lower: &str,
165 tag: &str,
166 from: usize,
167) -> Option<(usize, usize, usize)> {
168 let open_needle = format!("<{}", tag);
169 let close_needle = format!("</{}>", tag);
170
171 let open_start = source_lower[from..].find(&open_needle)? + from;
172 let after_tag_name = open_start + 1 + tag.len();
173 if source_lower
174 .as_bytes()
175 .get(after_tag_name)
176 .is_some_and(|b| b.is_ascii_alphabetic())
177 {
178 return None;
179 }
180 let tag_end = source_lower[open_start..].find('>')? + open_start + 1;
181 let close_start = source_lower[tag_end..].find(&close_needle)? + tag_end;
182
183 Some((open_start, tag_end, close_start))
184}
185
186fn find_sfc_tag_block(
187 source: &str,
188 source_lower: &str,
189 tag: &str,
190 from: usize,
191) -> Option<SfcBlock> {
192 let (open_start, tag_end, close_start) = find_tag_bounds(source, source_lower, tag, from)?;
193 let close_needle = format!("</{}>", tag);
194
195 let attrs = &source[open_start + 1 + tag.len()..tag_end];
196 let inner_start = tag_end;
197 let inner_end = close_start;
198 let block_end = source_lower[close_start..]
199 .find('>')
200 .map(|i| close_start + i + 1)
201 .unwrap_or(close_start + close_needle.len());
202 let block_end = block_end.min(source.len());
203
204 let block_format = detect_sfc_block_format(attrs, tag);
205
206 Some(SfcBlock {
207 block_format,
208 block_start: open_start,
209 inner_start,
210 inner_end: inner_end.max(inner_start),
211 block_end,
212 })
213}
214
215fn detect_block_format(attrs: &str, tag: &str, strict: bool) -> String {
216 let lang = extract_lang_attr_value(attrs);
217 match tag {
218 "script" => match lang.as_deref() {
219 Some("ts" | "typescript") => "typescript".to_string(),
220 Some("js" | "javascript") => "javascript".to_string(),
221 Some(other) if strict => {
222 if crate::formats::get_format_by_extension(other).is_some()
223 || crate::formats::SUPPORTED_FORMATS
224 .iter()
225 .any(|e| e.name == other)
226 {
227 other.to_string()
228 } else {
229 "javascript".to_string()
230 }
231 }
232 _ => "javascript".to_string(),
233 },
234 "style" => match lang.as_deref() {
235 Some("scss" | "sass") => "scss".to_string(),
236 Some("less") => "less".to_string(),
237 _ => "css".to_string(),
238 },
239 "template" => match lang.as_deref() {
240 Some(v) if v == "pug" || v == "jade" => "pug".to_string(),
241 _ => "html".to_string(),
242 },
243 _ => "html".to_string(),
244 }
245}
246
247fn detect_sfc_block_format(attrs: &str, tag: &str) -> String {
248 detect_block_format(attrs, tag, true)
249}
250
251fn detect_display_block_format(attrs: &str, tag: &str) -> String {
252 detect_block_format(attrs, tag, false)
253}
254
255fn astro_frontmatter_block(source: &str) -> Option<SfcBlock> {
256 if !(source.starts_with("---\n") || source.starts_with("---\r\n")) {
257 return None;
258 }
259 let lines = crate::markdown::line_spans(source);
260 let close_idx = lines
261 .iter()
262 .enumerate()
263 .skip(1)
264 .find(|(_, span)| source[span.start..span.end].trim() == "---")
265 .map(|(idx, _)| idx)?;
266 let inner_start = lines.get(1)?.start;
267 let inner_end = source[..lines[close_idx].start]
268 .strip_suffix('\n')
269 .map(|prefix: &str| prefix.len())
270 .unwrap_or(lines[close_idx].start);
271 let block_end = lines[close_idx].next_start.min(source.len());
272 Some(SfcBlock {
273 block_format: "typescript".to_string(),
274 block_start: 0,
275 inner_start,
276 inner_end: inner_end.max(inner_start),
277 block_end,
278 })
279}
280
281fn extract_lang_attr_value(attrs: &str) -> Option<String> {
282 let lower = attrs.to_ascii_lowercase();
283 let lang_pos = lower.find("lang=")?;
284 let rest = &attrs[lang_pos + 5..];
285 let quote = if rest.starts_with('"') {
286 '"'
287 } else if rest.starts_with('\'') {
288 '\''
289 } else {
290 return None;
291 };
292 let value_start = 1;
293 let value_end = rest[value_start..].find(quote)? + value_start;
294 Some(rest[value_start..value_end].to_ascii_lowercase())
295}
296
297pub fn extract_blocks(source: &str, file_format: &str) -> Vec<Block> {
299 let source_lower = source.to_ascii_lowercase();
300 let mut blocks = Vec::new();
301 for tag in sfc_tag_names(file_format) {
302 let mut search_from = 0;
303 while let Some((block, next_from)) =
304 find_display_block(source, &source_lower, tag, search_from)
305 {
306 search_from = next_from;
307 blocks.push(block);
308 }
309 }
310 blocks.sort_by_key(|b: &Block| b.start_offset);
311 blocks
312}
313
314fn find_display_block(
315 source: &str,
316 source_lower: &str,
317 tag: &str,
318 from: usize,
319) -> Option<(Block, usize)> {
320 let (open_start, tag_end, close_start) = find_tag_bounds(source, source_lower, tag, from)?;
321
322 let attrs = &source[open_start + 1 + tag.len()..tag_end];
323 let content = source[tag_end..close_start].to_string();
324 let content_len = content.len();
325 let start_line = source[..tag_end].lines().count() as u32 + 1;
326 let block_format = detect_display_block_format(attrs, tag);
327
328 Some((
329 Block {
330 block_format,
331 content,
332 start_offset: tag_end,
333 start_line,
334 },
335 tag_end + content_len,
336 ))
337}
338
339pub fn tokenize_sfc(source: &str, file_format: &str, mode: Mode) -> Vec<Token> {
340 let blocks = extract_blocks(source, file_format);
341 tokenize_blocks_shifted(
342 blocks
343 .iter()
344 .map(|b| (b.block_format.as_str(), b.content.as_str(), b.start_line)),
345 mode,
346 )
347}
348
349#[cfg(test)]
350mod tests {
351 use super::*;
352
353 const VUE_FILE: &str = r#"<template>
354 <div>Hello</div>
355</template>
356
357<script>
358export default { name: 'Foo' }
359</script>
360
361<style>
362.foo { color: red; }
363</style>
364"#;
365
366 const VUE_TS_FILE: &str = r#"<template>
367 <div>Hello</div>
368</template>
369
370<script lang="ts">
371const x: number = 5;
372</script>
373
374<style lang="scss">
375.foo { color: red; }
376</style>
377"#;
378
379 #[test]
380 fn vue_file_extracts_three_blocks() {
381 let blocks = extract_blocks(VUE_FILE, "vue");
382 assert_eq!(blocks.len(), 3, "must find template, script, style blocks");
383 }
384
385 #[test]
386 fn script_block_default_format_is_javascript() {
387 let blocks = extract_blocks(VUE_FILE, "vue");
388 let script = blocks.iter().find(|b| b.block_format == "javascript");
389 assert!(script.is_some(), "plain <script> must be javascript format");
390 }
391
392 #[test]
393 fn script_lang_ts_produces_typescript_format() {
394 let blocks = extract_blocks(VUE_TS_FILE, "vue");
395 let ts_block = blocks.iter().find(|b| b.block_format == "typescript");
396 assert!(
397 ts_block.is_some(),
398 "<script lang=\"ts\"> must produce typescript format"
399 );
400 }
401
402 #[test]
403 fn unknown_lang_does_not_panic() {
404 let source = "<script lang=\"unknownlang123\">\nconst x = 1;\n</script>\n";
405 let result = std::panic::catch_unwind(|| extract_blocks(source, "vue"));
406 assert!(result.is_ok(), "unknown lang must not panic");
407 }
408
409 #[test]
410 fn no_blocks_returns_empty() {
411 let source = "just plain text no tags";
412 let blocks = extract_blocks(source, "vue");
413 assert!(blocks.is_empty());
414 }
415
416 #[test]
417 fn start_offset_is_after_opening_tag() {
418 let blocks = extract_blocks(VUE_FILE, "vue");
419 for block in &blocks {
420 assert!(block.start_offset > 0);
421 }
422 }
423
424 #[test]
425 fn vue_sfc_maps_produces_multiple_formats() {
426 let options = TokenizeOptions::new(Mode::Mild);
427 let maps = tokenize_sfc_maps(VUE_FILE, "vue", &options);
428 let formats: Vec<&str> = maps.iter().map(|m| m.format.as_str()).collect();
429 assert!(formats.contains(&"javascript"), "must have javascript map");
430 assert!(formats.contains(&"css"), "must have css map");
431 assert!(formats.contains(&"html"), "must have html map");
432 }
433
434 #[test]
435 fn vue_html_map_tokens_remain_in_source_order() {
436 let maps = tokenize_sfc_maps(VUE_FILE, "vue", &TokenizeOptions::new(Mode::Mild));
437 let html = maps.iter().find(|map| map.format == "html").unwrap();
438
439 assert!(
440 html.tokens
441 .windows(2)
442 .all(|pair| pair[0].range[0] <= pair[1].range[0]),
443 "HTML tokens must remain in source order"
444 );
445 }
446
447 #[test]
448 fn vue_html_map_holds_only_the_template_body() {
449 let maps = tokenize_sfc_maps(VUE_FILE, "vue", &TokenizeOptions::new(Mode::Mild));
450 let html = maps.iter().find(|map| map.format == "html").unwrap();
451 let template_start = VUE_FILE.find("<template>").unwrap() + "<template>".len();
452 let template_end = VUE_FILE.find("</template>").unwrap();
453 assert!(
454 html.tokens
455 .iter()
456 .all(|t| t.range[0] >= template_start && t.range[1] <= template_end),
457 "wrapper tags must not enter the html stream: {:?}",
458 html.tokens.iter().map(|t| t.range).collect::<Vec<_>>()
459 );
460 assert_eq!(html.tokens.first().unwrap().start.line, 2);
461 assert_eq!(html.tokens.last().unwrap().end.line, 2);
462 }
463
464 #[test]
465 fn vue_without_template_has_no_html_map() {
466 let source = "<script>\nexport default { render: (h) => h('div', 'x') }\n</script>\n<style>\n.x { margin: 0; }\n</style>\n";
467 let maps = tokenize_sfc_maps(source, "vue", &TokenizeOptions::new(Mode::Mild));
468 assert!(
469 maps.iter().all(|map| map.format != "html"),
470 "wrapper tags alone must not form an html map: {:?}",
471 maps.iter().map(|m| m.format.as_str()).collect::<Vec<_>>()
472 );
473 }
474
475 #[test]
476 fn svelte_keeps_top_level_markup() {
477 let source = "<script>\nlet n = 1;\n</script>\n<h1>Count {n}</h1>\n<button on:click={() => n++}>add</button>\n";
478 let maps = tokenize_sfc_maps(source, "svelte", &TokenizeOptions::new(Mode::Mild));
479 let html = maps.iter().find(|map| map.format == "html").unwrap();
480 assert!(
481 html.tokens.iter().any(|t| t.start.line == 4),
482 "top-level markup is the html map"
483 );
484 }
485
486 #[test]
487 fn vue_ts_maps_produces_typescript() {
488 let options = TokenizeOptions::new(Mode::Mild);
489 let maps = tokenize_sfc_maps(VUE_TS_FILE, "vue", &options);
490 let formats: Vec<&str> = maps.iter().map(|m| m.format.as_str()).collect();
491 assert!(formats.contains(&"typescript"), "must have typescript map");
492 assert!(formats.contains(&"scss"), "must have scss map");
493 }
494
495 #[test]
496 fn empty_sfc_returns_empty() {
497 let options = TokenizeOptions::new(Mode::Mild);
498 let maps = tokenize_sfc_maps("", "vue", &options);
499 assert!(maps.is_empty());
500 }
501
502 #[test]
503 fn svelte_sfc_maps_produces_multiple_formats() {
504 let source = r#"<script>
505 let count = 0;
506</script>
507
508<style>
509 .count { color: blue; }
510</style>
511"#;
512 let options = TokenizeOptions::new(Mode::Mild);
513 let maps = tokenize_sfc_maps(source, "svelte", &options);
514 let formats: Vec<&str> = maps.iter().map(|m| m.format.as_str()).collect();
515 assert!(
516 formats.contains(&"javascript"),
517 "svelte must have javascript map"
518 );
519 assert!(formats.contains(&"css"), "svelte must have css map");
520 assert!(
521 formats.contains(&"html"),
522 "svelte must have html markup map"
523 );
524 }
525
526 #[test]
527 fn svelte_script_containing_style_text_no_panic() {
528 let source = r#"<script>
529 const x = "<style>.red{color:red}</style>";
530</script>
531
532<style>
533 .blue { color: blue; }
534</style>
535"#;
536 let result = std::panic::catch_unwind(|| {
537 let options = TokenizeOptions::new(Mode::Mild);
538 tokenize_sfc_maps(source, "svelte", &options)
539 });
540 assert!(
541 result.is_ok(),
542 "must not panic when <style> text appears inside <script>"
543 );
544 let maps = result.unwrap();
545 let formats: Vec<&str> = maps.iter().map(|m| m.format.as_str()).collect();
546 assert!(
547 formats.contains(&"javascript"),
548 "must have javascript block"
549 );
550 assert!(formats.contains(&"css"), "must have real css block");
551 assert!(formats.contains(&"html"), "must have html markup");
552 }
553}