1use aho_corasick::{AhoCorasick, AhoCorasickBuilder};
2use anyhow::Result;
3use client::proto;
4use fancy_regex::{Captures, Regex, RegexBuilder};
5use gpui::Entity;
6use language::{Buffer, BufferSnapshot, CharKind};
7use smol::future::yield_now;
8use std::{
9 borrow::Cow,
10 io::{BufRead, BufReader, Read},
11 ops::Range,
12 path::Path,
13 sync::{Arc, LazyLock},
14};
15use text::Anchor;
16use util::paths::PathMatcher;
17
18pub enum SearchResult {
19 Buffer {
20 buffer: Entity<Buffer>,
21 ranges: Vec<Range<Anchor>>,
22 },
23 LimitReached,
24}
25
26#[derive(Clone, Copy, PartialEq)]
27pub enum SearchInputKind {
28 Query,
29 Include,
30 Exclude,
31}
32
33#[derive(Clone, Debug)]
34pub struct SearchInputs {
35 query: Arc<str>,
36 files_to_include: PathMatcher,
37 files_to_exclude: PathMatcher,
38 buffers: Option<Vec<Entity<Buffer>>>,
39}
40
41impl SearchInputs {
42 pub fn as_str(&self) -> &str {
43 self.query.as_ref()
44 }
45 pub fn files_to_include(&self) -> &PathMatcher {
46 &self.files_to_include
47 }
48 pub fn files_to_exclude(&self) -> &PathMatcher {
49 &self.files_to_exclude
50 }
51 pub fn buffers(&self) -> &Option<Vec<Entity<Buffer>>> {
52 &self.buffers
53 }
54}
55#[derive(Clone, Debug)]
56pub enum SearchQuery {
57 Text {
58 search: AhoCorasick,
59 replacement: Option<String>,
60 whole_word: bool,
61 case_sensitive: bool,
62 include_ignored: bool,
63 inner: SearchInputs,
64 },
65
66 Regex {
67 regex: Regex,
68 replacement: Option<String>,
69 multiline: bool,
70 whole_word: bool,
71 case_sensitive: bool,
72 include_ignored: bool,
73 inner: SearchInputs,
74 },
75}
76
77static WORD_MATCH_TEST: LazyLock<Regex> = LazyLock::new(|| {
78 RegexBuilder::new(r"\B")
79 .build()
80 .expect("Failed to create WORD_MATCH_TEST")
81});
82
83impl SearchQuery {
84 pub fn text(
85 query: impl ToString,
86 whole_word: bool,
87 case_sensitive: bool,
88 include_ignored: bool,
89 files_to_include: PathMatcher,
90 files_to_exclude: PathMatcher,
91 buffers: Option<Vec<Entity<Buffer>>>,
92 ) -> Result<Self> {
93 let query = query.to_string();
94 let search = AhoCorasickBuilder::new()
95 .ascii_case_insensitive(!case_sensitive)
96 .build([&query])?;
97 let inner = SearchInputs {
98 query: query.into(),
99 files_to_exclude,
100 files_to_include,
101 buffers,
102 };
103 Ok(Self::Text {
104 search,
105 replacement: None,
106 whole_word,
107 case_sensitive,
108 include_ignored,
109 inner,
110 })
111 }
112
113 pub fn regex(
114 query: impl ToString,
115 whole_word: bool,
116 case_sensitive: bool,
117 include_ignored: bool,
118 files_to_include: PathMatcher,
119 files_to_exclude: PathMatcher,
120 buffers: Option<Vec<Entity<Buffer>>>,
121 ) -> Result<Self> {
122 let mut query = query.to_string();
123 let initial_query = Arc::from(query.as_str());
124 if whole_word {
125 let mut word_query = String::new();
126 if let Some(first) = query.get(0..1) {
127 if WORD_MATCH_TEST.is_match(first).is_ok_and(|x| !x) {
128 word_query.push_str("\\b");
129 }
130 }
131 word_query.push_str(&query);
132 if let Some(last) = query.get(query.len() - 1..) {
133 if WORD_MATCH_TEST.is_match(last).is_ok_and(|x| !x) {
134 word_query.push_str("\\b");
135 }
136 }
137 query = word_query
138 }
139
140 let multiline = query.contains('\n') || query.contains("\\n");
141 let regex = RegexBuilder::new(&query)
142 .case_insensitive(!case_sensitive)
143 .build()?;
144 let inner = SearchInputs {
145 query: initial_query,
146 files_to_exclude,
147 files_to_include,
148 buffers,
149 };
150 Ok(Self::Regex {
151 regex,
152 replacement: None,
153 multiline,
154 whole_word,
155 case_sensitive,
156 include_ignored,
157 inner,
158 })
159 }
160
161 pub fn from_proto(message: proto::SearchQuery) -> Result<Self> {
162 if message.regex {
163 Self::regex(
164 message.query,
165 message.whole_word,
166 message.case_sensitive,
167 message.include_ignored,
168 deserialize_path_matches(&message.files_to_include)?,
169 deserialize_path_matches(&message.files_to_exclude)?,
170 None, // search opened only don't need search remote
171 )
172 } else {
173 Self::text(
174 message.query,
175 message.whole_word,
176 message.case_sensitive,
177 message.include_ignored,
178 deserialize_path_matches(&message.files_to_include)?,
179 deserialize_path_matches(&message.files_to_exclude)?,
180 None, // search opened only don't need search remote
181 )
182 }
183 }
184
185 pub fn with_replacement(mut self, new_replacement: String) -> Self {
186 match self {
187 Self::Text {
188 ref mut replacement,
189 ..
190 }
191 | Self::Regex {
192 ref mut replacement,
193 ..
194 } => {
195 *replacement = Some(new_replacement);
196 self
197 }
198 }
199 }
200
201 pub fn to_proto(&self) -> proto::SearchQuery {
202 proto::SearchQuery {
203 query: self.as_str().to_string(),
204 regex: self.is_regex(),
205 whole_word: self.whole_word(),
206 case_sensitive: self.case_sensitive(),
207 include_ignored: self.include_ignored(),
208 files_to_include: self.files_to_include().sources().join(","),
209 files_to_exclude: self.files_to_exclude().sources().join(","),
210 }
211 }
212
213 pub(crate) fn detect(
214 &self,
215 mut reader: BufReader<Box<dyn Read + Send + Sync>>,
216 ) -> Result<bool> {
217 if self.as_str().is_empty() {
218 return Ok(false);
219 }
220
221 match self {
222 Self::Text { search, .. } => {
223 let mat = search.stream_find_iter(reader).next();
224 match mat {
225 Some(Ok(_)) => Ok(true),
226 Some(Err(err)) => Err(err.into()),
227 None => Ok(false),
228 }
229 }
230 Self::Regex {
231 regex, multiline, ..
232 } => {
233 if *multiline {
234 let mut text = String::new();
235 if let Err(err) = reader.read_to_string(&mut text) {
236 Err(err.into())
237 } else {
238 Ok(regex.find(&text)?.is_some())
239 }
240 } else {
241 for line in reader.lines() {
242 let line = line?;
243 if regex.find(&line)?.is_some() {
244 return Ok(true);
245 }
246 }
247 Ok(false)
248 }
249 }
250 }
251 }
252 /// Returns the replacement text for this `SearchQuery`.
253 pub fn replacement(&self) -> Option<&str> {
254 match self {
255 SearchQuery::Text { replacement, .. } | SearchQuery::Regex { replacement, .. } => {
256 replacement.as_deref()
257 }
258 }
259 }
260 /// Replaces search hits if replacement is set. `text` is assumed to be a string that matches this `SearchQuery` exactly, without any leftovers on either side.
261 pub fn replacement_for<'a>(&self, text: &'a str) -> Option<Cow<'a, str>> {
262 match self {
263 SearchQuery::Text { replacement, .. } => replacement.clone().map(Cow::from),
264 SearchQuery::Regex {
265 regex, replacement, ..
266 } => {
267 if let Some(replacement) = replacement {
268 static TEXT_REPLACEMENT_SPECIAL_CHARACTERS_REGEX: LazyLock<Regex> =
269 LazyLock::new(|| Regex::new(r"\\\\|\\n|\\t").unwrap());
270 let replacement = TEXT_REPLACEMENT_SPECIAL_CHARACTERS_REGEX.replace_all(
271 replacement,
272 |c: &Captures| match c.get(0).unwrap().as_str() {
273 r"\\" => "\\",
274 r"\n" => "\n",
275 r"\t" => "\t",
276 x => unreachable!("Unexpected escape sequence: {}", x),
277 },
278 );
279 Some(regex.replace(text, replacement))
280 } else {
281 None
282 }
283 }
284 }
285 }
286
287 pub async fn search(
288 &self,
289 buffer: &BufferSnapshot,
290 subrange: Option<Range<usize>>,
291 ) -> Vec<Range<usize>> {
292 const YIELD_INTERVAL: usize = 20000;
293
294 if self.as_str().is_empty() {
295 return Default::default();
296 }
297
298 let range_offset = subrange.as_ref().map(|r| r.start).unwrap_or(0);
299 let rope = if let Some(range) = subrange {
300 buffer.as_rope().slice(range)
301 } else {
302 buffer.as_rope().clone()
303 };
304
305 let mut matches = Vec::new();
306 match self {
307 Self::Text {
308 search, whole_word, ..
309 } => {
310 for (ix, mat) in search
311 .stream_find_iter(rope.bytes_in_range(0..rope.len()))
312 .enumerate()
313 {
314 if (ix + 1) % YIELD_INTERVAL == 0 {
315 yield_now().await;
316 }
317
318 let mat = mat.unwrap();
319 if *whole_word {
320 let classifier = buffer.char_classifier_at(range_offset + mat.start());
321
322 let prev_kind = rope
323 .reversed_chars_at(mat.start())
324 .next()
325 .map(|c| classifier.kind(c));
326 let start_kind =
327 classifier.kind(rope.chars_at(mat.start()).next().unwrap());
328 let end_kind =
329 classifier.kind(rope.reversed_chars_at(mat.end()).next().unwrap());
330 let next_kind = rope.chars_at(mat.end()).next().map(|c| classifier.kind(c));
331 if (Some(start_kind) == prev_kind && start_kind == CharKind::Word)
332 || (Some(end_kind) == next_kind && end_kind == CharKind::Word)
333 {
334 continue;
335 }
336 }
337 matches.push(mat.start()..mat.end())
338 }
339 }
340
341 Self::Regex {
342 regex, multiline, ..
343 } => {
344 if *multiline {
345 let text = rope.to_string();
346 for (ix, mat) in regex.find_iter(&text).enumerate() {
347 if (ix + 1) % YIELD_INTERVAL == 0 {
348 yield_now().await;
349 }
350
351 if let Ok(mat) = mat {
352 matches.push(mat.start()..mat.end());
353 }
354 }
355 } else {
356 let mut line = String::new();
357 let mut line_offset = 0;
358 for (chunk_ix, chunk) in rope.chunks().chain(["\n"]).enumerate() {
359 if (chunk_ix + 1) % YIELD_INTERVAL == 0 {
360 yield_now().await;
361 }
362
363 for (newline_ix, text) in chunk.split('\n').enumerate() {
364 if newline_ix > 0 {
365 for mat in regex.find_iter(&line).flatten() {
366 let start = line_offset + mat.start();
367 let end = line_offset + mat.end();
368 matches.push(start..end);
369 }
370
371 line_offset += line.len() + 1;
372 line.clear();
373 }
374 line.push_str(text);
375 }
376 }
377 }
378 }
379 }
380
381 matches
382 }
383
384 pub fn is_empty(&self) -> bool {
385 self.as_str().is_empty()
386 }
387
388 pub fn as_str(&self) -> &str {
389 self.as_inner().as_str()
390 }
391
392 pub fn whole_word(&self) -> bool {
393 match self {
394 Self::Text { whole_word, .. } => *whole_word,
395 Self::Regex { whole_word, .. } => *whole_word,
396 }
397 }
398
399 pub fn case_sensitive(&self) -> bool {
400 match self {
401 Self::Text { case_sensitive, .. } => *case_sensitive,
402 Self::Regex { case_sensitive, .. } => *case_sensitive,
403 }
404 }
405
406 pub fn include_ignored(&self) -> bool {
407 match self {
408 Self::Text {
409 include_ignored, ..
410 } => *include_ignored,
411 Self::Regex {
412 include_ignored, ..
413 } => *include_ignored,
414 }
415 }
416
417 pub fn is_regex(&self) -> bool {
418 matches!(self, Self::Regex { .. })
419 }
420
421 pub fn files_to_include(&self) -> &PathMatcher {
422 self.as_inner().files_to_include()
423 }
424
425 pub fn files_to_exclude(&self) -> &PathMatcher {
426 self.as_inner().files_to_exclude()
427 }
428
429 pub fn buffers(&self) -> Option<&Vec<Entity<Buffer>>> {
430 self.as_inner().buffers.as_ref()
431 }
432
433 pub fn is_opened_only(&self) -> bool {
434 self.as_inner().buffers.is_some()
435 }
436
437 pub fn filters_path(&self) -> bool {
438 !(self.files_to_exclude().sources().is_empty()
439 && self.files_to_include().sources().is_empty())
440 }
441
442 pub fn file_matches(&self, file_path: &Path) -> bool {
443 let mut path = file_path.to_path_buf();
444 loop {
445 if self.files_to_exclude().is_match(&path) {
446 return false;
447 } else if self.files_to_include().sources().is_empty()
448 || self.files_to_include().is_match(&path)
449 {
450 return true;
451 } else if !path.pop() {
452 return false;
453 }
454 }
455 }
456 pub fn as_inner(&self) -> &SearchInputs {
457 match self {
458 Self::Regex { inner, .. } | Self::Text { inner, .. } => inner,
459 }
460 }
461}
462
463pub fn deserialize_path_matches(glob_set: &str) -> anyhow::Result<PathMatcher> {
464 let globs = glob_set
465 .split(',')
466 .map(str::trim)
467 .filter(|&glob_str| (!glob_str.is_empty()))
468 .map(|glob_str| glob_str.to_owned())
469 .collect::<Vec<_>>();
470 Ok(PathMatcher::new(&globs)?)
471}
472
473#[cfg(test)]
474mod tests {
475 use super::*;
476
477 #[test]
478 fn path_matcher_creation_for_valid_paths() {
479 for valid_path in [
480 "file",
481 "Cargo.toml",
482 ".DS_Store",
483 "~/dir/another_dir/",
484 "./dir/file",
485 "dir/[a-z].txt",
486 "../dir/filé",
487 ] {
488 let path_matcher = PathMatcher::new(&[valid_path.to_owned()]).unwrap_or_else(|e| {
489 panic!("Valid path {valid_path} should be accepted, but got: {e}")
490 });
491 assert!(
492 path_matcher.is_match(valid_path),
493 "Path matcher for valid path {valid_path} should match itself"
494 )
495 }
496 }
497
498 #[test]
499 fn path_matcher_creation_for_globs() {
500 for invalid_glob in ["dir/[].txt", "dir/[a-z.txt", "dir/{file"] {
501 match PathMatcher::new(&[invalid_glob.to_owned()]) {
502 Ok(_) => panic!("Invalid glob {invalid_glob} should not be accepted"),
503 Err(_expected) => {}
504 }
505 }
506
507 for valid_glob in [
508 "dir/?ile",
509 "dir/*.txt",
510 "dir/**/file",
511 "dir/[a-z].txt",
512 "{dir,file}",
513 ] {
514 match PathMatcher::new(&[valid_glob.to_owned()]) {
515 Ok(_expected) => {}
516 Err(e) => panic!("Valid glob should be accepted, but got: {e}"),
517 }
518 }
519 }
520}