1use regex::Regex;
2use std::sync::LazyLock;
3
4static CAMEL_CASE_RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"[a-z][A-Z]").unwrap());
5static SNAKE_CASE_RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"[a-z]_[a-z]").unwrap());
6static PASCAL_CASE_RE: LazyLock<Regex> =
7 LazyLock::new(|| Regex::new(r"^[A-Z][a-z]+[A-Z]").unwrap());
8static ACRONYM_PASCAL_RE: LazyLock<Regex> =
9 LazyLock::new(|| Regex::new(r"\b[A-Z]{2,}[A-Z][a-z]").unwrap());
10static DOT_PATH_RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"[a-zA-Z]\.[a-zA-Z]").unwrap());
11static FILE_PATH_RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"[/\\].*\.\w{1,5}$").unwrap());
12static HEX_CODE_RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"0x[A-Fa-f0-9]+").unwrap());
13static ERROR_PREFIX_RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"\bERR_\w+").unwrap());
14static NUMERIC_ERROR_RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"\bE\d{4,}").unwrap());
15static TYPESCRIPT_ERROR_RE: LazyLock<Regex> =
16 LazyLock::new(|| Regex::new(r"\bTS\d{4,}\b").unwrap());
17static HTTP_STATUS_RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"\b[1-5]\d{2}\b").unwrap());
18static IDENTIFIER_TOKEN_RE: LazyLock<Regex> = LazyLock::new(|| {
19 Regex::new(r"\b[A-Za-z_$][A-Za-z0-9_$]*(?:\.[A-Za-z_$][A-Za-z0-9_$]*)*\b").unwrap()
20});
21
22static WINDOWS_ABS_PATH_RE: LazyLock<Regex> =
23 LazyLock::new(|| Regex::new(r"^[A-Za-z]:[\\/][A-Za-z0-9_.\-+?\\/' ]+$").unwrap());
24static WINDOWS_REL_PATH_RE: LazyLock<Regex> =
25 LazyLock::new(|| Regex::new(r"^[A-Za-z0-9_.\-+?' ]+(\\[A-Za-z0-9_.\-+?' ]+)+$").unwrap());
26static POSIX_ABS_PATH_RE: LazyLock<Regex> =
27 LazyLock::new(|| Regex::new(r"^/[A-Za-z0-9_.\-+?/' ]+$").unwrap());
28static POSIX_REL_PATH_RE: LazyLock<Regex> =
29 LazyLock::new(|| Regex::new(r"^[A-Za-z0-9_.\-+?' ]+(/[A-Za-z0-9_.\-+?' ]+)+$").unwrap());
30static UNC_PATH_RE: LazyLock<Regex> =
31 LazyLock::new(|| Regex::new(r"^\\\\[A-Za-z0-9_.\-+?\\']+$").unwrap());
32static FILENAME_EXEMPTION_RE: LazyLock<Regex> =
33 LazyLock::new(|| Regex::new(r"^[A-Za-z_][A-Za-z0-9_.\-+'? ]*\.[A-Za-z0-9]{1,8}$").unwrap());
34static BRACE_QUANTIFIER_RE: LazyLock<Regex> =
35 LazyLock::new(|| Regex::new(r"\{\d+(?:,\d*)?\}").unwrap());
36static NAMED_CAPTURE_RE: LazyLock<Regex> = LazyLock::new(|| Regex::new(r"\(\?P<[^>]+>").unwrap());
37static CHAR_RANGE_RE: LazyLock<Regex> =
38 LazyLock::new(|| Regex::new(r"[A-Za-z0-9]-[A-Za-z0-9]").unwrap());
39
40const QUESTION_WORDS: &[&str] = &[
41 "how", "what", "where", "why", "when", "which", "who", "does",
42];
43
44#[derive(Debug, Clone, Copy, PartialEq, Eq)]
45pub enum QueryKind {
46 Identifier,
47 Mixed,
48 ErrorCode,
49 Path,
50 Regex,
51 NaturalLanguage,
52}
53
54#[derive(Debug, Clone, Copy, PartialEq)]
55pub struct ShapeWeights {
56 pub semantic: f32,
57 pub lexical: f32,
58 pub should_use_lexical: bool,
59}
60
61#[derive(Debug, Clone, Copy, PartialEq)]
62pub struct QueryShape {
63 pub kind: QueryKind,
64 pub weights: ShapeWeights,
65}
66
67pub fn classify(query: &str) -> QueryShape {
68 let trimmed = query.trim();
69 if trimmed.is_empty() {
70 return shape(QueryKind::NaturalLanguage);
71 }
72
73 if pre_tier_exempt(trimmed).is_some() {
74 return shape(QueryKind::Path);
75 }
76
77 if looks_like_regex(trimmed) {
78 return shape(QueryKind::Regex);
79 }
80
81 let words: Vec<&str> = trimmed.split_whitespace().collect();
82 let word_count = words.len();
83 let first_word_lower = words[0].to_ascii_lowercase();
84
85 if FILE_PATH_RE.is_match(trimmed) {
86 return shape(QueryKind::Path);
87 }
88
89 let has_question_word = QUESTION_WORDS.contains(&first_word_lower.as_str());
90 let is_long_phrase = word_count > 2;
91 let is_two_word_concept = is_two_word_lowercase_concept(&words);
92 let has_natural_language_signals = has_question_word || is_long_phrase || is_two_word_concept;
93 let has_error_code = contains_error_code(trimmed, word_count);
94
95 if has_error_code && has_natural_language_signals {
96 return shape(QueryKind::Mixed);
97 }
98
99 if has_error_code {
100 return shape(QueryKind::ErrorCode);
101 }
102
103 let has_code_identifier = CAMEL_CASE_RE.is_match(trimmed)
104 || SNAKE_CASE_RE.is_match(trimmed)
105 || PASCAL_CASE_RE.is_match(trimmed)
106 || ACRONYM_PASCAL_RE.is_match(trimmed)
107 || DOT_PATH_RE.is_match(trimmed);
108
109 if has_code_identifier && has_natural_language_signals {
110 return shape(QueryKind::Mixed);
111 }
112
113 if has_code_identifier || (word_count <= 2 && !has_natural_language_signals) {
114 return shape(QueryKind::Identifier);
115 }
116
117 shape(QueryKind::NaturalLanguage)
118}
119
120pub fn extract_tokens(query: &str, shape: &QueryShape) -> Vec<String> {
121 match shape.kind {
122 QueryKind::NaturalLanguage | QueryKind::Regex => Vec::new(),
123 QueryKind::Path => extract_path_tokens(query),
124 QueryKind::ErrorCode => extract_error_code_tokens(query),
125 QueryKind::Identifier => extract_identifier_tokens(query, false),
126 QueryKind::Mixed => extract_identifier_tokens(query, true),
127 }
128}
129
130pub fn pre_tier_exempt(query: &str) -> Option<&'static str> {
131 if let Some(kind) = check_url_exemption(query) {
132 return Some(kind);
133 }
134 check_path_exemption(query)
135}
136
137pub fn looks_like_regex(query: &str) -> bool {
138 crate::pattern_compile::detect_unsupported_features(query).is_some()
139 || tier_a_regex_signal(query)
140 || tier_b_character_class(query)
141 || tier_c_adjacent_meta(query)
142}
143
144fn check_url_exemption(query: &str) -> Option<&'static str> {
145 let parsed = url::Url::parse(query).ok()?;
146 if !matches!(parsed.scheme(), "http" | "https" | "file" | "ftp" | "ssh") {
147 return None;
148 }
149 if has_regex_meta_sequences(query) || has_obvious_regex_chars(query) {
150 return None;
151 }
152 Some("url")
153}
154
155fn check_path_exemption(query: &str) -> Option<&'static str> {
156 let kind = if WINDOWS_ABS_PATH_RE.is_match(query) {
157 "windows_abs"
158 } else if WINDOWS_REL_PATH_RE.is_match(query) {
159 "windows_rel"
160 } else if POSIX_ABS_PATH_RE.is_match(query) {
161 "posix_abs"
162 } else if POSIX_REL_PATH_RE.is_match(query) {
163 "posix_rel"
164 } else if UNC_PATH_RE.is_match(query) {
165 "unc"
166 } else if FILENAME_EXEMPTION_RE.is_match(query) {
167 "filename"
168 } else {
169 return None;
170 };
171 if has_path_regex_meta_sequences(query) || has_obvious_regex_chars(query) {
172 return None;
173 }
174 Some(kind)
175}
176
177fn contains_error_code(query: &str, word_count: usize) -> bool {
178 HEX_CODE_RE.is_match(query)
179 || ERROR_PREFIX_RE.is_match(query)
180 || NUMERIC_ERROR_RE.is_match(query)
181 || TYPESCRIPT_ERROR_RE.is_match(query)
182 || has_http_status(query, word_count)
183}
184
185fn has_http_status(query: &str, word_count: usize) -> bool {
186 HTTP_STATUS_RE.is_match(query)
187 && (word_count <= 3 || query.to_ascii_lowercase().contains("http"))
188}
189
190fn is_two_word_lowercase_concept(words: &[&str]) -> bool {
191 words.len() == 2
192 && words
193 .iter()
194 .all(|word| is_dictionary_style_lowercase_word(word))
195}
196
197fn is_dictionary_style_lowercase_word(word: &str) -> bool {
198 word.len() >= 3 && word.bytes().all(|byte| byte.is_ascii_lowercase())
199}
200
201fn has_regex_meta_sequences(query: &str) -> bool {
202 query.contains(".+")
203 || query.contains(".*")
204 || query.contains(".?")
205 || query.contains(r"\n")
206 || query.contains(r"\t")
207 || query.contains(r"\r")
208 || query.contains(r"\b")
209 || query.contains(r"\B")
210 || query.contains(r"\w")
211 || query.contains(r"\W")
212 || query.contains(r"\d")
213 || query.contains(r"\D")
214 || query.contains(r"\s")
215 || query.contains(r"\S")
216 || query.contains(r"\p{")
217 || query.contains(r"\x")
218 || query.contains(r"\u{")
219 || has_escaped_regex_metachar(query)
220}
221
222fn has_path_regex_meta_sequences(query: &str) -> bool {
223 query.contains(".+")
224 || query.contains(".*")
225 || query.contains(".?")
226 || query.contains(r"\p{")
227 || query.contains(r"\x")
228 || query.contains(r"\u{")
229 || has_path_context_regex_escape(query)
230 || has_escaped_regex_metachar(query)
231}
232
233fn has_path_context_regex_escape(query: &str) -> bool {
234 let chars = query.char_indices().collect::<Vec<_>>();
235 for index in 0..chars.len().saturating_sub(1) {
236 if chars[index].1 != '\\' {
237 continue;
238 }
239 let escaped = chars[index + 1].1;
240 if matches!(escaped, 'b' | 'B' | 'w' | 'W' | 'd' | 'D' | 's' | 'S')
241 && path_escape_looks_like_regex(&chars, index + 1)
242 {
243 return true;
244 }
245 }
246 false
247}
248
249fn path_escape_looks_like_regex(chars: &[(usize, char)], escaped_index: usize) -> bool {
250 let Some((_, next)) = chars.get(escaped_index + 1) else {
251 return true;
252 };
253
254 matches!(
255 *next,
256 '*' | '+' | '?' | '{' | '(' | '[' | '|' | '^' | '$' | '\\' | '/'
257 )
258}
259
260fn has_escaped_regex_metachar(query: &str) -> bool {
261 let mut escaped = false;
262 for ch in query.chars() {
263 if escaped {
264 if is_escaped_metachar(ch) {
265 return true;
266 }
267 escaped = false;
268 continue;
269 }
270 escaped = ch == '\\';
271 }
272 false
273}
274
275fn has_obvious_regex_chars(query: &str) -> bool {
276 query.contains('*')
277 || query.contains('[')
278 || query.contains(']')
279 || query.contains('(')
280 || query.contains(')')
281 || query.contains('|')
282 || query.contains('{')
283 || query.contains('}')
284}
285
286fn tier_a_regex_signal(query: &str) -> bool {
287 query.contains("(?:")
288 || NAMED_CAPTURE_RE.is_match(query)
289 || ["(?i)", "(?m)", "(?s)", "(?x)"]
290 .iter()
291 .any(|signal| query.contains(signal))
292 || [
293 r"\b", r"\B", r"\w", r"\W", r"\d", r"\D", r"\s", r"\S", r"\p{", r"\x", r"\u{", r"\n",
294 r"\t", r"\r",
295 ]
296 .iter()
297 .any(|signal| query.contains(signal))
298 || has_brace_quantifier(query)
299 || has_anchored_identifier(query)
300 || has_contextual_escaped_metachar(query)
301}
302
303fn has_brace_quantifier(query: &str) -> bool {
304 for matched in BRACE_QUANTIFIER_RE.find_iter(query) {
305 if matched.start() > 0
306 && query[..matched.start()]
307 .chars()
308 .last()
309 .is_some_and(|ch| !ch.is_whitespace())
310 {
311 return true;
312 }
313 }
314 false
315}
316
317fn has_anchored_identifier(query: &str) -> bool {
318 let trimmed = query.trim();
319 if let Some(rest) = trimmed.strip_prefix('^') {
320 if leading_identifier_len(rest) >= 3 {
321 return true;
322 }
323 }
324 if let Some(rest) = trimmed.strip_suffix('$') {
325 if trailing_identifier_len(rest) >= 3 {
326 return true;
327 }
328 }
329 false
330}
331
332fn leading_identifier_len(text: &str) -> usize {
333 text.chars()
334 .take_while(|ch| ch.is_ascii_alphanumeric() || *ch == '_')
335 .count()
336}
337
338fn trailing_identifier_len(text: &str) -> usize {
339 text.chars()
340 .rev()
341 .take_while(|ch| ch.is_ascii_alphanumeric() || *ch == '_')
342 .count()
343}
344
345fn has_contextual_escaped_metachar(query: &str) -> bool {
346 let chars: Vec<char> = query.chars().collect();
347 let mut index = 0usize;
348 while index + 1 < chars.len() {
349 if chars[index] == '\\' && is_escaped_metachar(chars[index + 1]) {
350 let literal_after = chars[index + 2..]
351 .iter()
352 .filter(|ch| ch.is_ascii_alphanumeric() || **ch == '_')
353 .count();
354 if literal_after >= 2 {
355 return true;
356 }
357 index += 2;
358 } else {
359 index += 1;
360 }
361 }
362 false
363}
364
365fn is_escaped_metachar(ch: char) -> bool {
366 matches!(
367 ch,
368 '.' | '*' | '+' | '?' | '(' | ')' | '[' | ']' | '{' | '}' | '|' | '^' | '$'
369 )
370}
371
372fn tier_b_character_class(query: &str) -> bool {
373 for content in bracket_contents(query) {
374 if content.starts_with('^')
375 || CHAR_RANGE_RE.is_match(&content)
376 || [r"\w", r"\d", r"\s", r"\W", r"\D", r"\S"]
377 .iter()
378 .any(|signal| content.contains(signal))
379 || multi_char_non_identifier_class(&content)
380 {
381 return true;
382 }
383 }
384 false
385}
386
387fn bracket_contents(query: &str) -> Vec<String> {
388 let mut contents = Vec::new();
389 let mut escaped = false;
390 let mut start = None;
391 for (index, ch) in query.char_indices() {
392 if escaped {
393 escaped = false;
394 continue;
395 }
396 if ch == '\\' {
397 escaped = true;
398 continue;
399 }
400 match ch {
401 '[' if start.is_none() => start = Some(index + ch.len_utf8()),
402 ']' => {
403 if let Some(open) = start.take() {
404 contents.push(query[open..index].to_string());
405 }
406 }
407 _ => {}
408 }
409 }
410 contents
411}
412
413fn multi_char_non_identifier_class(content: &str) -> bool {
414 let char_count = content.chars().count();
415 char_count >= 2
416 && !content.chars().any(|ch| {
417 ch.is_ascii_alphanumeric() || ch == '_' || ch == '"' || ch == '\'' || ch == ';'
418 })
419}
420
421fn tier_c_adjacent_meta(query: &str) -> bool {
422 has_dot_quantifier(query)
423 || has_literal_atom_quantifier(query)
424 || has_regex_pipe(query)
425 || escaped_paren_count(query) >= 2
426}
427
428fn has_dot_quantifier(query: &str) -> bool {
429 [".*", ".+", ".?"]
430 .iter()
431 .any(|signal| query.contains(signal) && query.trim().len() > signal.len())
432}
433
434fn has_literal_atom_quantifier(query: &str) -> bool {
435 let chars = query.char_indices().collect::<Vec<_>>();
436 for (index, (byte_index, ch)) in chars.iter().copied().enumerate() {
437 if !is_bare_quantifier(ch) || is_escaped_at(query, byte_index) {
438 continue;
439 }
440 if chars
441 .get(index + 1)
442 .is_some_and(|(_, next)| is_bare_quantifier(*next))
443 {
444 continue;
445 }
446 if ch == '?'
447 && (sentence_final_question_mark_in_phrase(query, byte_index)
448 || question_mark_is_code_shape(&chars, index))
449 {
450 continue;
451 }
452 if previous_is_literal_atom(&chars, index) {
453 return true;
454 }
455 }
456 false
457}
458
459fn sentence_final_question_mark_in_phrase(query: &str, byte_index: usize) -> bool {
460 query[byte_index + '?'.len_utf8()..].trim().is_empty()
461 && query[..byte_index].split_whitespace().count() > 1
462}
463
464fn question_mark_is_code_shape(chars: &[(usize, char)], question_index: usize) -> bool {
465 question_mark_is_optional_chain(chars, question_index)
466 || question_mark_after_empty_call(chars, question_index)
467 || question_mark_after_index_expression(chars, question_index)
468 || question_mark_is_typescript_optional(chars, question_index)
469}
470
471fn question_mark_is_optional_chain(chars: &[(usize, char)], question_index: usize) -> bool {
472 chars
473 .get(question_index + 1)
474 .is_some_and(|(_, next)| *next == '.')
475 && question_index
476 .checked_sub(1)
477 .and_then(|previous_index| chars.get(previous_index))
478 .is_some_and(|(_, previous)| is_code_expression_tail(*previous))
479}
480
481fn question_mark_after_empty_call(chars: &[(usize, char)], question_index: usize) -> bool {
482 let Some(call_open_index) = question_index.checked_sub(2) else {
483 return false;
484 };
485 chars
486 .get(question_index - 1)
487 .is_some_and(|(_, previous)| *previous == ')')
488 && chars
489 .get(call_open_index)
490 .is_some_and(|(_, open)| *open == '(')
491 && call_open_index
492 .checked_sub(1)
493 .and_then(|callee_index| chars.get(callee_index))
494 .is_some_and(|(_, callee_tail)| is_code_expression_tail(*callee_tail))
495}
496
497fn question_mark_after_index_expression(chars: &[(usize, char)], question_index: usize) -> bool {
498 if chars
499 .get(question_index.checked_sub(1).unwrap_or(usize::MAX))
500 .is_none_or(|(_, previous)| *previous != ']')
501 {
502 return false;
503 }
504
505 let mut depth = 0usize;
506 for index in (0..question_index).rev() {
507 match chars[index].1 {
508 ']' => depth += 1,
509 '[' => {
510 depth = depth.saturating_sub(1);
511 if depth == 0 {
512 return index
513 .checked_sub(1)
514 .and_then(|target_index| chars.get(target_index))
515 .is_some_and(|(_, target_tail)| is_code_expression_tail(*target_tail));
516 }
517 }
518 _ => {}
519 }
520 }
521 false
522}
523
524fn question_mark_is_typescript_optional(chars: &[(usize, char)], question_index: usize) -> bool {
525 let previous_is_identifier = question_index
526 .checked_sub(1)
527 .and_then(|previous_index| chars.get(previous_index))
528 .is_some_and(|(_, previous)| is_identifier_tail(*previous));
529 if !previous_is_identifier {
530 return false;
531 }
532 if chars
533 .get(question_index + 1)
534 .is_none_or(|(_, next)| *next != ':')
535 {
536 return false;
537 }
538
539 chars
540 .get(question_index + 2)
541 .is_none_or(|(_, after_colon)| {
542 after_colon.is_whitespace()
543 || after_colon.is_ascii_alphabetic()
544 || matches!(*after_colon, '_' | '{' | '[' | '(' | '"' | '\'')
545 })
546}
547
548fn is_code_expression_tail(ch: char) -> bool {
549 is_identifier_tail(ch) || matches!(ch, ')' | ']')
550}
551
552fn is_identifier_tail(ch: char) -> bool {
553 ch.is_ascii_alphanumeric() || matches!(ch, '_' | '$')
554}
555
556fn previous_is_literal_atom(chars: &[(usize, char)], quantifier_index: usize) -> bool {
557 let Some((_, previous)) = quantifier_index
558 .checked_sub(1)
559 .and_then(|previous_index| chars.get(previous_index))
560 else {
561 return false;
562 };
563
564 previous.is_ascii_alphanumeric() || *previous == '_' || *previous == ')' || *previous == ']'
565}
566
567fn is_bare_quantifier(ch: char) -> bool {
568 matches!(ch, '*' | '+' | '?')
569}
570
571fn is_escaped_at(query: &str, byte_index: usize) -> bool {
572 let backslash_count = query[..byte_index]
573 .chars()
574 .rev()
575 .take_while(|ch| *ch == '\\')
576 .count();
577 backslash_count % 2 == 1
578}
579
580fn has_regex_pipe(query: &str) -> bool {
581 for (index, ch) in query.char_indices() {
582 if ch != '|' {
583 continue;
584 }
585 let left = trailing_identifier_len(&query[..index]);
586 let right = leading_identifier_len(&query[index + ch.len_utf8()..]);
587 if left >= 3 && right >= 3 {
588 return true;
589 }
590 }
591 false
592}
593
594fn escaped_paren_count(query: &str) -> usize {
595 let mut count = 0usize;
596 let mut escaped = false;
597 for ch in query.chars() {
598 if escaped {
599 if ch == '(' || ch == ')' {
600 count += 1;
601 }
602 escaped = false;
603 continue;
604 }
605 escaped = ch == '\\';
606 }
607 count
608}
609
610fn extract_path_tokens(query: &str) -> Vec<String> {
611 let mut tokens = Vec::new();
612 for segment in query
613 .split(['/', '\\'])
614 .filter(|segment| !segment.is_empty())
615 {
616 if segment.contains('.') {
617 if let Some(stem) = segment.rsplit_once('.').map(|(stem, _)| stem) {
618 push_unique(&mut tokens, stem);
619 }
620 }
621 push_unique(&mut tokens, segment);
622 }
623 tokens
624}
625
626fn extract_error_code_tokens(query: &str) -> Vec<String> {
627 let mut tokens = Vec::new();
628 for regex in [
629 &*HEX_CODE_RE,
630 &*ERROR_PREFIX_RE,
631 &*NUMERIC_ERROR_RE,
632 &*TYPESCRIPT_ERROR_RE,
633 &*HTTP_STATUS_RE,
634 ] {
635 for mat in regex.find_iter(query) {
636 push_unique(&mut tokens, mat.as_str());
637 }
638 }
639 if tokens.is_empty() && !query.trim().is_empty() {
640 push_unique(&mut tokens, query.trim());
641 }
642 tokens
643}
644
645fn extract_identifier_tokens(query: &str, require_code_shape: bool) -> Vec<String> {
646 let mut tokens = Vec::new();
647 for mat in IDENTIFIER_TOKEN_RE.find_iter(query) {
648 let token = mat.as_str();
649 if require_code_shape && !is_code_identifier_token(token) {
650 continue;
651 }
652 push_unique(&mut tokens, token);
653 }
654 tokens
655}
656
657fn is_code_identifier_token(token: &str) -> bool {
658 CAMEL_CASE_RE.is_match(token)
659 || SNAKE_CASE_RE.is_match(token)
660 || PASCAL_CASE_RE.is_match(token)
661 || ACRONYM_PASCAL_RE.is_match(token)
662 || DOT_PATH_RE.is_match(token)
663 || ERROR_PREFIX_RE.is_match(token)
664 || NUMERIC_ERROR_RE.is_match(token)
665 || TYPESCRIPT_ERROR_RE.is_match(token)
666}
667
668fn push_unique(tokens: &mut Vec<String>, token: &str) {
669 if !token.is_empty() && !tokens.iter().any(|existing| existing == token) {
670 tokens.push(token.to_string());
671 }
672}
673
674fn shape(kind: QueryKind) -> QueryShape {
675 QueryShape {
676 kind,
677 weights: weights_for(kind),
678 }
679}
680
681fn weights_for(kind: QueryKind) -> ShapeWeights {
682 match kind {
683 QueryKind::Identifier => ShapeWeights {
684 semantic: 0.2,
685 lexical: 0.8,
686 should_use_lexical: true,
687 },
688 QueryKind::Path | QueryKind::ErrorCode => ShapeWeights {
689 semantic: 0.1,
690 lexical: 0.9,
691 should_use_lexical: true,
692 },
693 QueryKind::Regex => ShapeWeights {
694 semantic: 0.0,
695 lexical: 1.0,
696 should_use_lexical: false,
697 },
698 QueryKind::NaturalLanguage => ShapeWeights {
699 semantic: 0.6,
700 lexical: 0.4,
701 should_use_lexical: false,
702 },
703 QueryKind::Mixed => ShapeWeights {
704 semantic: 0.4,
705 lexical: 0.6,
706 should_use_lexical: true,
707 },
708 }
709}
710
711#[cfg(test)]
712mod tests {
713 use super::*;
714
715 fn kind(query: &str) -> QueryKind {
716 classify(query).kind
717 }
718
719 #[test]
720 fn url_exemptions_allow_common_literal_url_punctuation() {
721 for query in [
722 "https://api.io/path",
723 "https://api.io/foo?q=test",
724 "https://api.io/foo+bar",
725 "https://api.io/foo@bar",
726 "https://api.io/foo#anchor",
727 ] {
728 assert_eq!(pre_tier_exempt(query), Some("url"), "{query}");
729 assert_ne!(kind(query), QueryKind::Regex, "{query}");
730 }
731 }
732
733 #[test]
734 fn url_exemptions_reject_regex_sequences() {
735 for query in [
736 "https://.*",
737 "https://api.io/.+",
738 "file://[^ ]+",
739 "file:///tmp/.+",
740 r"https://api.io/users/\w+",
741 ] {
742 assert_eq!(kind(query), QueryKind::Regex, "{query}");
743 }
744 }
745
746 #[test]
747 fn path_and_filename_exemptions_allow_literal_punctuation() {
748 for (query, expected) in [
749 (r"C:\new\test", "windows_abs"),
750 (r"src\bin\main.rs", "windows_rel"),
751 (r"src\tab\main.ts", "windows_rel"),
752 (r"packages\opencode-plugin\src", "windows_rel"),
753 ("/usr/local/bin", "posix_abs"),
754 ("/Users/John Doe/Documents", "posix_abs"),
755 ("/home/user/.gitignore", "posix_abs"),
756 ("v1/release/notes.md", "posix_rel"),
757 ("/home/user/jeff's-folder", "posix_abs"),
758 ("C++/parser/main.cpp", "posix_rel"),
759 ("foo+bar/baz.ts", "posix_rel"),
760 ("is_valid?.ts", "filename"),
761 ("Cargo.lock", "filename"),
762 ("tsconfig.json", "filename"),
763 ] {
764 assert_eq!(pre_tier_exempt(query), Some(expected), "{query}");
765 assert_eq!(kind(query), QueryKind::Path, "{query}");
766 }
767 assert_eq!(pre_tier_exempt("foo?"), None);
768 }
769
770 #[test]
771 fn path_exemptions_reject_regex_sequences() {
772 for query in [
773 "src/.*",
774 "src/.+",
775 r"C:\bin\foo*.exe",
776 r"C:\Users\\w+",
777 r"src\w+\main.ts",
778 ] {
779 assert_eq!(kind(query), QueryKind::Regex, "{query}");
780 }
781 }
782
783 #[test]
784 fn tier_a_and_c_regex_signals_route_to_regex() {
785 for query in [
786 "^export",
787 "foo$",
788 "^main$",
789 r"foo\.bar",
790 r"\(method\)",
791 r"\bTODO\b",
792 ".*foo",
793 "foo|bar",
794 "(?:foo)",
795 "(?P<n>foo)",
796 "(?i)Todo",
797 r"\p{Lu}",
798 r"\xFF",
799 r"\u{1F600}",
800 "a{3}",
801 r"\n",
806 r"\t",
807 r"\r",
808 r"\tindent",
809 ] {
810 assert_eq!(kind(query), QueryKind::Regex, "{query}");
811 }
812 }
813
814 #[test]
815 fn character_classes_route_only_when_they_look_like_classes() {
816 for query in ["[a-z]+", "[^abc]", r"[\w]+"] {
817 assert_eq!(kind(query), QueryKind::Regex, "{query}");
818 }
819 for query in [
820 "arr[0]",
821 "obj[key]",
822 "config[\"key\"]",
823 "#[derive]",
824 "Vec<[u8; 32]>",
825 ] {
826 assert_ne!(kind(query), QueryKind::Regex, "{query}");
827 }
828 }
829
830 #[test]
831 fn unsupported_regex_syntax_still_routes_to_regex_for_compile_error() {
832 for query in [
833 "(?=foo)",
834 "(?!foo)",
835 "(?<=foo)",
836 "(?<!foo)",
837 "(?P=name)",
838 r"\1",
839 "foo*+",
840 "(?>foo)",
841 ] {
842 assert_eq!(kind(query), QueryKind::Regex, "{query}");
843 }
844 }
845
846 #[test]
847 fn two_word_lowercase_concepts_route_to_natural_language() {
848 for query in ["retry logic", "auth flow", "cache invalidation"] {
849 assert_eq!(kind(query), QueryKind::NaturalLanguage, "{query}");
850 }
851 }
852
853 #[test]
854 fn identifierish_short_queries_stay_identifier() {
855 for query in ["useState hook", "parseConfig", "parse_config option"] {
856 assert_eq!(kind(query), QueryKind::Identifier, "{query}");
857 }
858 }
859
860 #[test]
861 fn question_mark_code_shapes_do_not_route_to_regex() {
862 for query in ["foo()?", "optional?.length", "user?.name", "arr[0]?"] {
863 assert_ne!(kind(query), QueryKind::Regex, "{query}");
864 }
865 }
866
867 #[test]
868 fn question_mark_regex_quantifiers_still_route_to_regex() {
869 for query in ["colou?r", "https?"] {
870 assert_eq!(kind(query), QueryKind::Regex, "{query}");
871 }
872 }
873
874 #[test]
875 fn weak_regex_like_punctuation_does_not_route_to_regex() {
876 for query in [
877 "^id",
878 "id$",
879 "^",
880 "$",
881 "$HOME",
882 r"\.",
883 "array.length",
884 "foo()",
885 "map.get(key)",
886 "a|b",
887 ] {
888 assert_ne!(kind(query), QueryKind::Regex, "{query}");
889 }
890 }
891}