1use std::collections::{BTreeMap, BTreeSet, HashMap, HashSet};
2use std::fs::{self, File};
3use std::io::{BufReader, BufWriter, Cursor, Read, Seek, Write};
4use std::path::{Component, Path, PathBuf};
5use std::process::Command;
6use std::sync::{
7 atomic::{AtomicBool, AtomicUsize, Ordering},
8 Arc, Mutex,
9};
10use std::time::{Duration, SystemTime, UNIX_EPOCH};
11
12use globset::{Glob, GlobSet, GlobSetBuilder};
13use ignore::WalkBuilder;
14use rayon::prelude::*;
15use regex::bytes::{Regex, RegexBuilder};
16use regex_syntax::hir::{Hir, HirKind};
17
18use crate::cache_freshness::{self, FileFreshness, FreshnessVerdict};
19use crate::fs_lock;
20
21const DEFAULT_MAX_FILE_SIZE: u64 = 1_048_576;
22const CACHE_MAGIC: u32 = 0x3144_4958; const INDEX_MAGIC: &[u8; 8] = b"AFTIDX01";
24const LOOKUP_MAGIC: &[u8; 8] = b"AFTLKP01";
25const INDEX_VERSION: u32 = 3;
26const PREVIEW_BYTES: usize = 8 * 1024;
27const EOF_SENTINEL: u8 = 0;
28const MAX_ENTRIES: usize = 10_000_000;
29const MIN_FILE_ENTRY_BYTES: usize = 57;
30const LOOKUP_ENTRY_BYTES: usize = 16;
31const POSTING_BYTES: usize = 6;
32static CACHE_LOCK_ACQUIRE_MUTEX: Mutex<()> = Mutex::new(());
33
34pub struct CacheLock {
35 _guard: fs_lock::LockGuard,
36}
37
38impl CacheLock {
39 pub fn acquire(cache_dir: &Path) -> std::io::Result<Self> {
40 fs::create_dir_all(cache_dir)?;
41 let path = cache_dir.join("cache.lock");
42 let _acquire_guard = CACHE_LOCK_ACQUIRE_MUTEX
43 .lock()
44 .map_err(|_| std::io::Error::other("search cache lock acquisition mutex poisoned"))?;
45 fs_lock::try_acquire(&path, Duration::from_secs(2))
46 .map(|guard| Self { _guard: guard })
47 .map_err(|error| match error {
48 fs_lock::AcquireError::Timeout => {
49 std::io::Error::other("timed out acquiring search cache lock")
50 }
51 fs_lock::AcquireError::Io(error) => error,
52 })
53 }
54}
55
56#[derive(Clone, Debug)]
57pub struct SearchIndex {
58 pub postings: HashMap<u32, Vec<Posting>>,
59 pub files: Vec<FileEntry>,
60 pub path_to_id: HashMap<PathBuf, u32>,
61 pub ready: bool,
62 project_root: PathBuf,
63 git_head: Option<String>,
64 max_file_size: u64,
65 pub file_trigrams: HashMap<u32, Vec<u32>>,
66 unindexed_files: HashSet<u32>,
67}
68
69impl SearchIndex {
70 pub fn file_count(&self) -> usize {
72 self.files.len()
73 }
74
75 pub fn trigram_count(&self) -> usize {
77 self.postings.len()
78 }
79
80 pub fn query_trigrams_from_tokens(tokens: &[&str]) -> Vec<u32> {
82 query_trigrams_from_tokens(tokens)
83 }
84
85 pub fn lexical_rank(
87 &self,
88 query_trigrams: &[u32],
89 candidate_filter: Option<&dyn Fn(&Path) -> bool>,
90 max_files: usize,
91 ) -> Vec<(PathBuf, f32)> {
92 if query_trigrams.is_empty() || max_files == 0 {
93 return Vec::new();
94 }
95
96 let mut non_zero: Vec<(u32, usize)> = query_trigrams
97 .iter()
98 .filter_map(|trigram| {
99 let posting_count = self.postings.get(trigram).map_or(0, Vec::len);
100 (posting_count > 0).then_some((*trigram, posting_count))
101 })
102 .collect();
103 if non_zero.is_empty() {
104 return Vec::new();
105 }
106
107 non_zero.sort_unstable_by_key(|(_, posting_count)| *posting_count);
108 let selected_count = non_zero.len().min(3);
109 let candidate_cap = if selected_count == 3 { 200 } else { 500 };
110
111 let mut candidate_ids = BTreeSet::new();
112 for (trigram, _) in non_zero.iter().take(selected_count) {
113 if let Some(postings) = self.postings.get(trigram) {
114 for posting in postings {
115 if self.is_active_file(posting.file_id) {
116 candidate_ids.insert(posting.file_id);
117 }
118 }
119 }
120 }
121
122 let mut ranked = Vec::new();
123 for file_id in candidate_ids.into_iter().take(candidate_cap) {
124 let Some(entry) = self.files.get(file_id as usize) else {
125 continue;
126 };
127 if let Some(filter) = candidate_filter {
128 if !filter(&entry.path) {
129 continue;
130 }
131 }
132 let score = lexical_score(self, query_trigrams, file_id);
133 if score > 0.0 {
134 ranked.push((entry.path.clone(), score));
135 }
136 }
137
138 ranked.sort_by(|a, b| b.1.partial_cmp(&a.1).unwrap_or(std::cmp::Ordering::Equal));
139 ranked.truncate(max_files);
140 ranked
141 }
142}
143
144#[derive(Clone, Debug, PartialEq, Eq)]
145pub struct Posting {
146 pub file_id: u32,
147 pub next_mask: u8,
148 pub loc_mask: u8,
149}
150
151#[derive(Clone, Debug)]
152pub struct FileEntry {
153 pub path: PathBuf,
154 pub size: u64,
155 pub modified: SystemTime,
156 pub content_hash: blake3::Hash,
157}
158
159#[derive(Clone, Debug, PartialEq, Eq)]
160pub struct GrepMatch {
161 pub file: PathBuf,
162 pub line: u32,
163 pub column: u32,
164 pub line_text: String,
165 pub match_text: String,
166}
167
168#[derive(Clone, Debug)]
169pub struct GrepResult {
170 pub matches: Vec<GrepMatch>,
171 pub total_matches: usize,
172 pub files_searched: usize,
173 pub files_with_matches: usize,
174 pub index_status: IndexStatus,
175 pub truncated: bool,
176}
177
178#[derive(Clone, Copy, Debug, PartialEq, Eq)]
179pub enum IndexStatus {
180 Ready,
181 Building,
182 Fallback,
183}
184
185impl IndexStatus {
186 pub fn as_str(&self) -> &'static str {
187 match self {
188 IndexStatus::Ready => "Ready",
189 IndexStatus::Building => "Building",
190 IndexStatus::Fallback => "Fallback",
191 }
192 }
193}
194
195#[derive(Clone, Debug, Default)]
196pub struct RegexQuery {
197 pub and_trigrams: Vec<u32>,
198 pub or_groups: Vec<Vec<u32>>,
199 pub(crate) and_filters: HashMap<u32, PostingFilter>,
200 pub(crate) or_filters: Vec<HashMap<u32, PostingFilter>>,
201}
202
203#[derive(Clone, Copy, Debug, Default)]
204pub(crate) struct PostingFilter {
205 next_mask: u8,
206 loc_mask: u8,
207}
208
209#[derive(Clone, Debug, Default)]
210struct QueryBuild {
211 and_runs: Vec<Vec<u8>>,
212 or_groups: Vec<Vec<Vec<u8>>>,
213}
214
215#[derive(Clone, Debug, Default)]
216pub(crate) struct PathFilters {
217 includes: Option<GlobSet>,
218 excludes: Option<GlobSet>,
219}
220
221#[derive(Clone, Debug)]
222pub(crate) struct SearchScope {
223 pub root: PathBuf,
224 pub use_index: bool,
225}
226
227#[derive(Clone, Debug)]
228struct SharedGrepMatch {
229 file: Arc<PathBuf>,
230 line: u32,
231 column: u32,
232 line_text: String,
233 match_text: String,
234}
235
236#[derive(Clone, Debug)]
237enum SearchMatcher {
238 Literal(LiteralSearch),
239 Regex(Regex),
240}
241
242#[derive(Clone, Debug)]
243enum LiteralSearch {
244 CaseSensitive(Vec<u8>),
245 AsciiCaseInsensitive(Vec<u8>),
246}
247
248impl SearchIndex {
249 pub fn new() -> Self {
250 SearchIndex {
251 postings: HashMap::new(),
252 files: Vec::new(),
253 path_to_id: HashMap::new(),
254 ready: false,
255 project_root: PathBuf::new(),
256 git_head: None,
257 max_file_size: DEFAULT_MAX_FILE_SIZE,
258 file_trigrams: HashMap::new(),
259 unindexed_files: HashSet::new(),
260 }
261 }
262
263 pub fn build(root: &Path) -> Self {
264 Self::build_with_limit(root, DEFAULT_MAX_FILE_SIZE)
265 }
266
267 pub(crate) fn build_with_limit(root: &Path, max_file_size: u64) -> Self {
268 let project_root = fs::canonicalize(root).unwrap_or_else(|_| root.to_path_buf());
269 let mut index = SearchIndex {
270 project_root: project_root.clone(),
271 max_file_size,
272 ..SearchIndex::new()
273 };
274
275 let filters = PathFilters::default();
276 for path in walk_project_files(&project_root, &filters) {
277 index.update_file(&path);
278 }
279
280 index.git_head = current_git_head(&project_root);
281 index.ready = true;
282 index
283 }
284
285 pub fn index_file(&mut self, path: &Path, content: &[u8]) {
286 self.remove_file(path);
287
288 let file_id = match self.allocate_file_id(path, content.len() as u64) {
289 Some(file_id) => file_id,
290 None => return,
291 };
292 if let Some(file) = self.files.get_mut(file_id as usize) {
293 file.content_hash = cache_freshness::hash_bytes(content);
294 }
295
296 let mut trigram_map: BTreeMap<u32, PostingFilter> = BTreeMap::new();
297 for (trigram, next_char, position) in extract_trigrams(content) {
298 let entry = trigram_map.entry(trigram).or_default();
299 entry.next_mask |= mask_for_next_char(next_char);
300 entry.loc_mask |= mask_for_position(position);
301 }
302
303 let mut file_trigrams = Vec::with_capacity(trigram_map.len());
304 for (trigram, filter) in trigram_map {
305 let postings = self.postings.entry(trigram).or_default();
306 postings.push(Posting {
307 file_id,
308 next_mask: filter.next_mask,
309 loc_mask: filter.loc_mask,
310 });
311 if postings.len() > 1
315 && postings[postings.len() - 2].file_id > postings[postings.len() - 1].file_id
316 {
317 postings.sort_unstable_by_key(|p| p.file_id);
318 }
319 file_trigrams.push(trigram);
320 }
321
322 self.file_trigrams.insert(file_id, file_trigrams);
323 self.unindexed_files.remove(&file_id);
324 }
325
326 pub fn remove_file(&mut self, path: &Path) {
327 let Some(file_id) = self.path_to_id.remove(path) else {
328 return;
329 };
330
331 if let Some(trigrams) = self.file_trigrams.remove(&file_id) {
332 for trigram in trigrams {
333 let should_remove = if let Some(postings) = self.postings.get_mut(&trigram) {
334 postings.retain(|posting| posting.file_id != file_id);
335 postings.is_empty()
336 } else {
337 false
338 };
339
340 if should_remove {
341 self.postings.remove(&trigram);
342 }
343 }
344 }
345
346 self.unindexed_files.remove(&file_id);
347 if let Some(file) = self.files.get_mut(file_id as usize) {
348 file.path = PathBuf::new();
349 file.size = 0;
350 file.modified = UNIX_EPOCH;
351 file.content_hash = cache_freshness::zero_hash();
352 }
353 }
354
355 pub fn update_file(&mut self, path: &Path) {
356 self.remove_file(path);
357
358 let metadata = match fs::metadata(path) {
359 Ok(metadata) if metadata.is_file() => metadata,
360 _ => return,
361 };
362
363 if is_binary_path(path, metadata.len()) {
364 self.track_unindexed_file(path, &metadata);
365 return;
366 }
367
368 if metadata.len() > self.max_file_size {
369 self.track_unindexed_file(path, &metadata);
370 return;
371 }
372
373 let content = match fs::read(path) {
374 Ok(content) => content,
375 Err(_) => return,
376 };
377
378 if is_binary_bytes(&content) {
379 self.track_unindexed_file(path, &metadata);
380 return;
381 }
382
383 self.index_file(path, &content);
384 }
385
386 pub fn grep(
387 &self,
388 pattern: &str,
389 case_sensitive: bool,
390 include: &[String],
391 exclude: &[String],
392 search_root: &Path,
393 max_results: usize,
394 ) -> GrepResult {
395 self.search_grep(
396 pattern,
397 case_sensitive,
398 include,
399 exclude,
400 search_root,
401 max_results,
402 )
403 }
404
405 pub fn search_grep(
406 &self,
407 pattern: &str,
408 case_sensitive: bool,
409 include: &[String],
410 exclude: &[String],
411 search_root: &Path,
412 max_results: usize,
413 ) -> GrepResult {
414 let is_literal = !pattern.chars().any(|c| {
417 matches!(
418 c,
419 '.' | '*' | '+' | '?' | '(' | ')' | '[' | ']' | '{' | '}' | '|' | '^' | '$' | '\\'
420 )
421 });
422
423 let literal_search = if is_literal {
424 if case_sensitive {
425 Some(LiteralSearch::CaseSensitive(pattern.as_bytes().to_vec()))
426 } else if pattern.is_ascii() {
427 Some(LiteralSearch::AsciiCaseInsensitive(
428 pattern
429 .as_bytes()
430 .iter()
431 .map(|byte| byte.to_ascii_lowercase())
432 .collect(),
433 ))
434 } else {
435 None
436 }
437 } else {
438 None
439 };
440
441 let regex = if literal_search.is_some() {
443 None
444 } else {
445 let regex_pattern = if is_literal {
446 regex::escape(pattern)
447 } else {
448 pattern.to_string()
449 };
450 let mut builder = RegexBuilder::new(®ex_pattern);
451 builder.case_insensitive(!case_sensitive);
452 builder.multi_line(true);
454 match builder.build() {
455 Ok(r) => Some(r),
456 Err(_) => {
457 return GrepResult {
458 matches: Vec::new(),
459 total_matches: 0,
460 files_searched: 0,
461 files_with_matches: 0,
462 index_status: if self.ready {
463 IndexStatus::Ready
464 } else {
465 IndexStatus::Building
466 },
467 truncated: false,
468 };
469 }
470 }
471 };
472
473 let matcher = if let Some(literal_search) = literal_search {
474 SearchMatcher::Literal(literal_search)
475 } else {
476 SearchMatcher::Regex(
477 regex.expect("regex should exist when literal matcher is unavailable"),
478 )
479 };
480
481 let filters = match build_path_filters(include, exclude) {
482 Ok(filters) => filters,
483 Err(_) => PathFilters::default(),
484 };
485 let search_root = canonicalize_or_normalize(search_root);
486
487 let query = decompose_regex(pattern);
488 let candidate_ids = self.candidates(&query);
489
490 let candidate_files: Vec<&FileEntry> = candidate_ids
491 .into_iter()
492 .filter_map(|file_id| self.files.get(file_id as usize))
493 .filter(|file| !file.path.as_os_str().is_empty())
494 .filter(|file| is_within_search_root(&search_root, &file.path))
495 .filter(|file| filters.matches(&self.project_root, &file.path))
496 .collect();
497
498 let total_matches = AtomicUsize::new(0);
499 let files_searched = AtomicUsize::new(0);
500 let files_with_matches = AtomicUsize::new(0);
501 let truncated = AtomicBool::new(false);
502 let stop_after = max_results.saturating_mul(2);
503
504 let mut matches = if candidate_files.len() > 10 {
505 candidate_files
506 .par_iter()
507 .map(|file| {
508 search_candidate_file(
509 file,
510 &matcher,
511 max_results,
512 stop_after,
513 &total_matches,
514 &files_searched,
515 &files_with_matches,
516 &truncated,
517 )
518 })
519 .reduce(Vec::new, |mut left, mut right| {
520 left.append(&mut right);
521 left
522 })
523 } else {
524 let mut matches = Vec::new();
525 for file in candidate_files {
526 matches.extend(search_candidate_file(
527 file,
528 &matcher,
529 max_results,
530 stop_after,
531 &total_matches,
532 &files_searched,
533 &files_with_matches,
534 &truncated,
535 ));
536
537 if should_stop_search(&truncated, &total_matches, stop_after) {
538 break;
539 }
540 }
541 matches
542 };
543
544 sort_shared_grep_matches_by_cached_mtime_desc(&mut matches, |path| {
545 self.path_to_id
546 .get(path)
547 .and_then(|file_id| self.files.get(*file_id as usize))
548 .map(|file| file.modified)
549 });
550
551 let matches = matches
552 .into_iter()
553 .map(|matched| GrepMatch {
554 file: matched.file.as_ref().clone(),
555 line: matched.line,
556 column: matched.column,
557 line_text: matched.line_text,
558 match_text: matched.match_text,
559 })
560 .collect();
561
562 GrepResult {
563 total_matches: total_matches.load(Ordering::Relaxed),
564 matches,
565 files_searched: files_searched.load(Ordering::Relaxed),
566 files_with_matches: files_with_matches.load(Ordering::Relaxed),
567 index_status: if self.ready {
568 IndexStatus::Ready
569 } else {
570 IndexStatus::Building
571 },
572 truncated: truncated.load(Ordering::Relaxed),
573 }
574 }
575
576 pub fn glob(&self, pattern: &str, search_root: &Path) -> Vec<PathBuf> {
577 let filters = match build_path_filters(&[pattern.to_string()], &[]) {
578 Ok(filters) => filters,
579 Err(_) => return Vec::new(),
580 };
581 let search_root = canonicalize_or_normalize(search_root);
582 let mut entries = self
583 .files
584 .iter()
585 .filter(|file| !file.path.as_os_str().is_empty())
586 .filter(|file| is_within_search_root(&search_root, &file.path))
587 .filter(|file| filters.matches(&self.project_root, &file.path))
588 .map(|file| (file.path.clone(), file.modified))
589 .collect::<Vec<_>>();
590
591 entries.sort_by(|(left_path, left_mtime), (right_path, right_mtime)| {
592 right_mtime
593 .cmp(left_mtime)
594 .then_with(|| left_path.cmp(right_path))
595 });
596
597 entries.into_iter().map(|(path, _)| path).collect()
598 }
599
600 pub fn candidates(&self, query: &RegexQuery) -> Vec<u32> {
601 if query.and_trigrams.is_empty() && query.or_groups.is_empty() {
602 return self.active_file_ids();
603 }
604
605 let mut and_trigrams = query.and_trigrams.clone();
606 and_trigrams.sort_unstable_by_key(|trigram| self.postings.get(trigram).map_or(0, Vec::len));
607
608 let mut current: Option<Vec<u32>> = None;
609
610 for trigram in and_trigrams {
611 let filter = query.and_filters.get(&trigram).copied();
612 let matches = self.postings_for_trigram(trigram, filter);
613 current = Some(match current.take() {
614 Some(existing) => intersect_sorted_ids(&existing, &matches),
615 None => matches,
616 });
617
618 if current.as_ref().is_some_and(|ids| ids.is_empty()) {
619 break;
620 }
621 }
622
623 let mut current = current.unwrap_or_else(|| self.active_file_ids());
624
625 for (index, group) in query.or_groups.iter().enumerate() {
626 let mut group_matches = Vec::new();
627 let filters = query.or_filters.get(index);
628
629 for trigram in group {
630 let filter = filters.and_then(|filters| filters.get(trigram).copied());
631 let matches = self.postings_for_trigram(*trigram, filter);
632 if group_matches.is_empty() {
633 group_matches = matches;
634 } else {
635 group_matches = union_sorted_ids(&group_matches, &matches);
636 }
637 }
638
639 current = intersect_sorted_ids(¤t, &group_matches);
640 if current.is_empty() {
641 break;
642 }
643 }
644
645 let mut unindexed = self
646 .unindexed_files
647 .iter()
648 .copied()
649 .filter(|file_id| self.is_active_file(*file_id))
650 .collect::<Vec<_>>();
651 if !unindexed.is_empty() {
652 unindexed.sort_unstable();
653 current = union_sorted_ids(¤t, &unindexed);
654 }
655
656 current
657 }
658
659 pub fn write_to_disk(&self, cache_dir: &Path, git_head: Option<&str>) {
660 if fs::create_dir_all(cache_dir).is_err() {
661 return;
662 }
663
664 let cache_path = cache_dir.join("cache.bin");
665 let tmp_cache = cache_dir.join(format!(
666 "cache.bin.tmp.{}.{}",
667 std::process::id(),
668 SystemTime::now()
669 .duration_since(UNIX_EPOCH)
670 .unwrap_or(Duration::ZERO)
671 .as_nanos()
672 ));
673
674 let active_ids = self.active_file_ids();
675 let mut id_map = HashMap::new();
676 for (new_id, old_id) in active_ids.iter().enumerate() {
677 let Ok(new_id_u32) = u32::try_from(new_id) else {
678 return;
679 };
680 id_map.insert(*old_id, new_id_u32);
681 }
682
683 let write_result = (|| -> std::io::Result<()> {
684 let mut postings_writer = BufWriter::new(Cursor::new(Vec::new()));
685
686 postings_writer.write_all(INDEX_MAGIC)?;
687 write_u32(&mut postings_writer, INDEX_VERSION)?;
688
689 let head = git_head.unwrap_or_default();
690 let root = self.project_root.to_string_lossy();
691 let head_len = u32::try_from(head.len())
692 .map_err(|_| std::io::Error::other("git head too large to cache"))?;
693 let root_len = u32::try_from(root.len())
694 .map_err(|_| std::io::Error::other("project root too large to cache"))?;
695 let file_count = u32::try_from(active_ids.len())
696 .map_err(|_| std::io::Error::other("too many files to cache"))?;
697
698 write_u32(&mut postings_writer, head_len)?;
699 write_u32(&mut postings_writer, root_len)?;
700 write_u64(&mut postings_writer, self.max_file_size)?;
701 write_u32(&mut postings_writer, file_count)?;
702 postings_writer.write_all(head.as_bytes())?;
703 postings_writer.write_all(root.as_bytes())?;
704
705 for old_id in &active_ids {
706 let Some(file) = self.files.get(*old_id as usize) else {
707 return Err(std::io::Error::other("missing file entry for cache write"));
708 };
709 let path =
710 cache_relative_path(&self.project_root, &file.path).ok_or_else(|| {
711 std::io::Error::other(format!(
712 "refusing to cache path outside project root: {}",
713 file.path.display()
714 ))
715 })?;
716 let path = path.to_string_lossy();
717 let path_len = u32::try_from(path.len())
718 .map_err(|_| std::io::Error::other("cached path too large"))?;
719 let modified = file
720 .modified
721 .duration_since(UNIX_EPOCH)
722 .unwrap_or(Duration::ZERO);
723 let unindexed = if self.unindexed_files.contains(old_id) {
724 1u8
725 } else {
726 0u8
727 };
728
729 postings_writer.write_all(&[unindexed])?;
730 write_u32(&mut postings_writer, path_len)?;
731 write_u64(&mut postings_writer, file.size)?;
732 write_u64(&mut postings_writer, modified.as_secs())?;
733 write_u32(&mut postings_writer, modified.subsec_nanos())?;
734 postings_writer.write_all(file.content_hash.as_bytes())?;
735 postings_writer.write_all(path.as_bytes())?;
736 }
737
738 let mut lookup_entries = Vec::new();
739 let mut postings_blob = Vec::new();
740 let mut sorted_postings: Vec<_> = self.postings.iter().collect();
741 sorted_postings.sort_by_key(|(trigram, _)| **trigram);
742
743 for (trigram, postings) in sorted_postings {
744 let offset = u64::try_from(postings_blob.len())
745 .map_err(|_| std::io::Error::other("postings blob too large"))?;
746 let mut count = 0u32;
747
748 for posting in postings {
749 let Some(mapped_file_id) = id_map.get(&posting.file_id).copied() else {
750 continue;
751 };
752
753 postings_blob.extend_from_slice(&mapped_file_id.to_le_bytes());
754 postings_blob.push(posting.next_mask);
755 postings_blob.push(posting.loc_mask);
756 count = count.saturating_add(1);
757 }
758
759 if count > 0 {
760 lookup_entries.push((*trigram, offset, count));
761 }
762 }
763
764 write_u64(
765 &mut postings_writer,
766 u64::try_from(postings_blob.len())
767 .map_err(|_| std::io::Error::other("postings blob too large"))?,
768 )?;
769 postings_writer.write_all(&postings_blob)?;
770 postings_writer.flush()?;
771 let mut postings_blob_file = postings_writer
772 .into_inner()
773 .map_err(|error| std::io::Error::other(error.to_string()))?
774 .into_inner();
775 let checksum = crc32fast::hash(&postings_blob_file);
776 postings_blob_file.extend_from_slice(&checksum.to_le_bytes());
777
778 let mut lookup_writer = BufWriter::new(Cursor::new(Vec::new()));
779 let entry_count = u32::try_from(lookup_entries.len())
780 .map_err(|_| std::io::Error::other("too many lookup entries to cache"))?;
781
782 lookup_writer.write_all(LOOKUP_MAGIC)?;
783 write_u32(&mut lookup_writer, INDEX_VERSION)?;
784 write_u32(&mut lookup_writer, entry_count)?;
785
786 for (trigram, offset, count) in lookup_entries {
787 write_u32(&mut lookup_writer, trigram)?;
788 write_u64(&mut lookup_writer, offset)?;
789 write_u32(&mut lookup_writer, count)?;
790 }
791
792 lookup_writer.flush()?;
793 let mut lookup_blob_file = lookup_writer
794 .into_inner()
795 .map_err(|error| std::io::Error::other(error.to_string()))?
796 .into_inner();
797 let checksum = crc32fast::hash(&lookup_blob_file);
798 lookup_blob_file.extend_from_slice(&checksum.to_le_bytes());
799
800 let mut cache_writer = BufWriter::new(File::create(&tmp_cache)?);
801 write_u32(&mut cache_writer, CACHE_MAGIC)?;
802 write_u32(&mut cache_writer, INDEX_VERSION)?;
803 write_u64(
804 &mut cache_writer,
805 u64::try_from(postings_blob_file.len())
806 .map_err(|_| std::io::Error::other("postings section too large"))?,
807 )?;
808 cache_writer.write_all(&postings_blob_file)?;
809 cache_writer.write_all(&lookup_blob_file)?;
810 cache_writer.flush()?;
811 cache_writer.get_ref().sync_all()?;
812 drop(cache_writer);
813 fs::rename(&tmp_cache, &cache_path)?;
814
815 Ok(())
816 })();
817
818 if write_result.is_err() {
819 let _ = fs::remove_file(&tmp_cache);
820 }
821 }
822
823 pub fn read_from_disk(cache_dir: &Path, current_canonical_root: &Path) -> Option<Self> {
824 debug_assert!(current_canonical_root.is_absolute());
825 let cache_path = cache_dir.join("cache.bin");
826 let cache_bytes = fs::read(&cache_path).ok()?;
827 if cache_bytes.len() < 16 {
828 return None;
829 }
830 let mut header = Cursor::new(&cache_bytes);
831 if read_u32(&mut header).ok()? != CACHE_MAGIC {
832 return None;
833 }
834 if read_u32(&mut header).ok()? != INDEX_VERSION {
835 return None;
836 }
837 let postings_len_total = usize::try_from(read_u64(&mut header).ok()?).ok()?;
838 let start = usize::try_from(header.position()).ok()?;
839 let postings_end = start.checked_add(postings_len_total)?;
840 if postings_end > cache_bytes.len() {
841 return None;
842 }
843 let postings_bytes = &cache_bytes[start..postings_end];
844 let lookup_bytes = &cache_bytes[postings_end..];
845 let lookup_len_total = lookup_bytes.len();
846 let mut postings_reader = BufReader::new(Cursor::new(postings_bytes));
847 let mut lookup_reader = BufReader::new(Cursor::new(lookup_bytes));
848 if postings_len_total < 4 || lookup_len_total < 4 {
849 return None;
850 }
851 verify_crc32_bytes_slice(postings_bytes).ok()?;
852 verify_crc32_bytes_slice(lookup_bytes).ok()?;
853
854 let mut magic = [0u8; 8];
855 postings_reader.read_exact(&mut magic).ok()?;
856 if &magic != INDEX_MAGIC {
857 return None;
858 }
859 if read_u32(&mut postings_reader).ok()? != INDEX_VERSION {
860 return None;
861 }
862
863 let head_len = read_u32(&mut postings_reader).ok()? as usize;
864 let root_len = read_u32(&mut postings_reader).ok()? as usize;
865 let max_file_size = read_u64(&mut postings_reader).ok()?;
866 let file_count = read_u32(&mut postings_reader).ok()? as usize;
867 if file_count > MAX_ENTRIES {
868 return None;
869 }
870 let postings_body_len = postings_len_total.checked_sub(4)?;
871 let lookup_body_len = lookup_len_total.checked_sub(4)?;
872
873 let remaining_postings = remaining_bytes(&mut postings_reader, postings_body_len)?;
874 let minimum_file_bytes = file_count.checked_mul(MIN_FILE_ENTRY_BYTES)?;
875 if minimum_file_bytes > remaining_postings {
876 return None;
877 }
878
879 if head_len > remaining_bytes(&mut postings_reader, postings_body_len)? {
880 return None;
881 }
882 let mut head_bytes = vec![0u8; head_len];
883 postings_reader.read_exact(&mut head_bytes).ok()?;
884 let git_head = String::from_utf8(head_bytes)
885 .ok()
886 .filter(|head| !head.is_empty());
887
888 if root_len > remaining_bytes(&mut postings_reader, postings_body_len)? {
889 return None;
890 }
891 let mut root_bytes = vec![0u8; root_len];
892 postings_reader.read_exact(&mut root_bytes).ok()?;
893 let _stored_project_root = PathBuf::from(String::from_utf8(root_bytes).ok()?);
894 let project_root = current_canonical_root.to_path_buf();
895
896 let mut files = Vec::with_capacity(file_count);
897 let mut path_to_id = HashMap::new();
898 let mut unindexed_files = HashSet::new();
899
900 for file_id in 0..file_count {
901 let mut unindexed = [0u8; 1];
902 postings_reader.read_exact(&mut unindexed).ok()?;
903 let path_len = read_u32(&mut postings_reader).ok()? as usize;
904 let size = read_u64(&mut postings_reader).ok()?;
905 let secs = read_u64(&mut postings_reader).ok()?;
906 let nanos = read_u32(&mut postings_reader).ok()?;
907 let mut hash_bytes = [0u8; 32];
908 postings_reader.read_exact(&mut hash_bytes).ok()?;
909 let content_hash = blake3::Hash::from_bytes(hash_bytes);
910 if nanos >= 1_000_000_000 {
911 return None;
912 }
913 if path_len > remaining_bytes(&mut postings_reader, postings_body_len)? {
914 return None;
915 }
916 let mut path_bytes = vec![0u8; path_len];
917 postings_reader.read_exact(&mut path_bytes).ok()?;
918 let relative_path = PathBuf::from(String::from_utf8(path_bytes).ok()?);
919 let full_path = cached_path_under_root(&project_root, &relative_path)?;
920 let file_id_u32 = u32::try_from(file_id).ok()?;
921
922 files.push(FileEntry {
923 path: full_path.clone(),
924 size,
925 modified: UNIX_EPOCH + Duration::new(secs, nanos),
926 content_hash,
927 });
928 path_to_id.insert(full_path, file_id_u32);
929 if unindexed[0] == 1 {
930 unindexed_files.insert(file_id_u32);
931 }
932 }
933
934 let postings_len = read_u64(&mut postings_reader).ok()? as usize;
935 let max_postings_bytes = MAX_ENTRIES.checked_mul(POSTING_BYTES)?;
936 if postings_len > max_postings_bytes {
937 return None;
938 }
939 if postings_len > remaining_bytes(&mut postings_reader, postings_body_len)? {
940 return None;
941 }
942 let mut postings_blob = vec![0u8; postings_len];
943 postings_reader.read_exact(&mut postings_blob).ok()?;
944
945 let mut lookup_magic = [0u8; 8];
946 lookup_reader.read_exact(&mut lookup_magic).ok()?;
947 if &lookup_magic != LOOKUP_MAGIC {
948 return None;
949 }
950 if read_u32(&mut lookup_reader).ok()? != INDEX_VERSION {
951 return None;
952 }
953 let entry_count = read_u32(&mut lookup_reader).ok()? as usize;
954 if entry_count > MAX_ENTRIES {
955 return None;
956 }
957 let remaining_lookup = remaining_bytes(&mut lookup_reader, lookup_body_len)?;
958 let minimum_lookup_bytes = entry_count.checked_mul(LOOKUP_ENTRY_BYTES)?;
959 if minimum_lookup_bytes > remaining_lookup {
960 return None;
961 }
962
963 let mut postings = HashMap::new();
964 let mut file_trigrams: HashMap<u32, Vec<u32>> = HashMap::new();
965
966 for _ in 0..entry_count {
967 let trigram = read_u32(&mut lookup_reader).ok()?;
968 let offset = read_u64(&mut lookup_reader).ok()? as usize;
969 let count = read_u32(&mut lookup_reader).ok()? as usize;
970 if count > MAX_ENTRIES {
971 return None;
972 }
973 let bytes_len = count.checked_mul(POSTING_BYTES)?;
974 let end = offset.checked_add(bytes_len)?;
975 if end > postings_blob.len() {
976 return None;
977 }
978
979 let mut trigram_postings = Vec::with_capacity(count);
980 for chunk in postings_blob[offset..end].chunks_exact(6) {
981 let file_id = u32::from_le_bytes([chunk[0], chunk[1], chunk[2], chunk[3]]);
982 let posting = Posting {
983 file_id,
984 next_mask: chunk[4],
985 loc_mask: chunk[5],
986 };
987 trigram_postings.push(posting.clone());
988 file_trigrams.entry(file_id).or_default().push(trigram);
989 }
990 postings.insert(trigram, trigram_postings);
991 }
992
993 Some(SearchIndex {
994 postings,
995 files,
996 path_to_id,
997 ready: false,
998 project_root,
999 git_head,
1000 max_file_size,
1001 file_trigrams,
1002 unindexed_files,
1003 })
1004 }
1005
1006 pub(crate) fn stored_git_head(&self) -> Option<&str> {
1007 self.git_head.as_deref()
1008 }
1009
1010 pub(crate) fn set_ready(&mut self, ready: bool) {
1011 self.ready = ready;
1012 }
1013
1014 pub(crate) fn verify_against_disk(&mut self, current_head: Option<String>) {
1015 self.git_head = current_head;
1016 verify_file_mtimes(self);
1017 self.ready = true;
1018 }
1019
1020 pub(crate) fn rebuild_or_refresh(
1021 root: &Path,
1022 max_file_size: u64,
1023 current_head: Option<String>,
1024 baseline: Option<SearchIndex>,
1025 ) -> Self {
1026 if let Some(mut baseline) = baseline {
1027 baseline.project_root = fs::canonicalize(root).unwrap_or_else(|_| root.to_path_buf());
1028 baseline.max_file_size = max_file_size;
1029
1030 if baseline.git_head == current_head || current_head.is_none() {
1031 baseline.git_head = current_head;
1038 verify_file_mtimes(&mut baseline);
1039 baseline.ready = true;
1040 return baseline;
1041 }
1042
1043 if let (Some(previous), Some(current)) =
1044 (baseline.git_head.clone(), current_head.clone())
1045 {
1046 let project_root = baseline.project_root.clone();
1047 if apply_git_diff_updates(&mut baseline, &project_root, &previous, ¤t) {
1048 baseline.git_head = Some(current);
1049 verify_file_mtimes(&mut baseline);
1050 baseline.ready = true;
1051 return baseline;
1052 }
1053 }
1054 }
1055
1056 SearchIndex::build_with_limit(root, max_file_size)
1057 }
1058
1059 fn allocate_file_id(&mut self, path: &Path, size_hint: u64) -> Option<u32> {
1060 let file_id = u32::try_from(self.files.len()).ok()?;
1061 let metadata = fs::metadata(path).ok();
1062 let size = metadata
1063 .as_ref()
1064 .map_or(size_hint, |metadata| metadata.len());
1065 let modified = metadata
1066 .and_then(|metadata| metadata.modified().ok())
1067 .unwrap_or(UNIX_EPOCH);
1068
1069 self.files.push(FileEntry {
1070 path: path.to_path_buf(),
1071 size,
1072 modified,
1073 content_hash: cache_freshness::zero_hash(),
1074 });
1075 self.path_to_id.insert(path.to_path_buf(), file_id);
1076 Some(file_id)
1077 }
1078
1079 fn track_unindexed_file(&mut self, path: &Path, metadata: &fs::Metadata) {
1080 let Some(file_id) = self.allocate_file_id(path, metadata.len()) else {
1081 return;
1082 };
1083 self.unindexed_files.insert(file_id);
1084 self.file_trigrams.insert(file_id, Vec::new());
1085 }
1086
1087 fn active_file_ids(&self) -> Vec<u32> {
1088 let mut ids: Vec<u32> = self.path_to_id.values().copied().collect();
1089 ids.sort_unstable();
1090 ids
1091 }
1092
1093 fn is_active_file(&self, file_id: u32) -> bool {
1094 self.files
1095 .get(file_id as usize)
1096 .map(|file| !file.path.as_os_str().is_empty())
1097 .unwrap_or(false)
1098 }
1099
1100 fn postings_for_trigram(&self, trigram: u32, filter: Option<PostingFilter>) -> Vec<u32> {
1101 let Some(postings) = self.postings.get(&trigram) else {
1102 return Vec::new();
1103 };
1104
1105 let mut matches = Vec::with_capacity(postings.len());
1106
1107 for posting in postings {
1108 if let Some(filter) = filter {
1109 if filter.next_mask != 0 && posting.next_mask & filter.next_mask == 0 {
1112 continue;
1113 }
1114 }
1119 if self.is_active_file(posting.file_id) {
1120 matches.push(posting.file_id);
1121 }
1122 }
1123
1124 matches
1125 }
1126}
1127
1128fn search_candidate_file(
1129 file: &FileEntry,
1130 matcher: &SearchMatcher,
1131 max_results: usize,
1132 stop_after: usize,
1133 total_matches: &AtomicUsize,
1134 files_searched: &AtomicUsize,
1135 files_with_matches: &AtomicUsize,
1136 truncated: &AtomicBool,
1137) -> Vec<SharedGrepMatch> {
1138 if should_stop_search(truncated, total_matches, stop_after) {
1139 return Vec::new();
1140 }
1141
1142 let content = match read_indexed_file_bytes(&file.path) {
1143 Some(content) => content,
1144 None => return Vec::new(),
1145 };
1146 if is_binary_bytes(&content) {
1153 return Vec::new();
1154 }
1155 files_searched.fetch_add(1, Ordering::Relaxed);
1156
1157 let shared_path = Arc::new(file.path.clone());
1158 let mut matches = Vec::new();
1159 let mut line_starts = None;
1160 let mut seen_lines = HashSet::new();
1161 let mut matched_this_file = false;
1162
1163 match matcher {
1164 SearchMatcher::Literal(LiteralSearch::CaseSensitive(needle)) => {
1165 let finder = memchr::memmem::Finder::new(needle);
1166 let mut start = 0;
1167
1168 while let Some(position) = finder.find(&content[start..]) {
1169 if should_stop_search(truncated, total_matches, stop_after) {
1170 break;
1171 }
1172
1173 let offset = start + position;
1174 start = offset + 1;
1175
1176 let line_starts = line_starts.get_or_insert_with(|| line_starts_bytes(&content));
1177 let (line, column, line_text) = line_details_bytes(&content, line_starts, offset);
1178 if !seen_lines.insert(line) {
1179 continue;
1180 }
1181
1182 matched_this_file = true;
1183 let match_number = total_matches.fetch_add(1, Ordering::Relaxed) + 1;
1184 if match_number > max_results {
1185 truncated.store(true, Ordering::Relaxed);
1186 break;
1187 }
1188
1189 let end = offset + needle.len();
1190 matches.push(SharedGrepMatch {
1191 file: shared_path.clone(),
1192 line,
1193 column,
1194 line_text,
1195 match_text: String::from_utf8_lossy(&content[offset..end]).into_owned(),
1196 });
1197 }
1198 }
1199 SearchMatcher::Literal(LiteralSearch::AsciiCaseInsensitive(needle)) => {
1200 let search_content = content.to_ascii_lowercase();
1201 let finder = memchr::memmem::Finder::new(needle);
1202 let mut start = 0;
1203
1204 while let Some(position) = finder.find(&search_content[start..]) {
1205 if should_stop_search(truncated, total_matches, stop_after) {
1206 break;
1207 }
1208
1209 let offset = start + position;
1210 start = offset + 1;
1211
1212 let line_starts = line_starts.get_or_insert_with(|| line_starts_bytes(&content));
1213 let (line, column, line_text) = line_details_bytes(&content, line_starts, offset);
1214 if !seen_lines.insert(line) {
1215 continue;
1216 }
1217
1218 matched_this_file = true;
1219 let match_number = total_matches.fetch_add(1, Ordering::Relaxed) + 1;
1220 if match_number > max_results {
1221 truncated.store(true, Ordering::Relaxed);
1222 break;
1223 }
1224
1225 let end = offset + needle.len();
1226 matches.push(SharedGrepMatch {
1227 file: shared_path.clone(),
1228 line,
1229 column,
1230 line_text,
1231 match_text: String::from_utf8_lossy(&content[offset..end]).into_owned(),
1232 });
1233 }
1234 }
1235 SearchMatcher::Regex(regex) => {
1236 for matched in regex.find_iter(&content) {
1237 if should_stop_search(truncated, total_matches, stop_after) {
1238 break;
1239 }
1240
1241 let line_starts = line_starts.get_or_insert_with(|| line_starts_bytes(&content));
1242 let (line, column, line_text) =
1243 line_details_bytes(&content, line_starts, matched.start());
1244 if !seen_lines.insert(line) {
1245 continue;
1246 }
1247
1248 matched_this_file = true;
1249 let match_number = total_matches.fetch_add(1, Ordering::Relaxed) + 1;
1250 if match_number > max_results {
1251 truncated.store(true, Ordering::Relaxed);
1252 break;
1253 }
1254
1255 matches.push(SharedGrepMatch {
1256 file: shared_path.clone(),
1257 line,
1258 column,
1259 line_text,
1260 match_text: String::from_utf8_lossy(matched.as_bytes()).into_owned(),
1261 });
1262 }
1263 }
1264 }
1265
1266 if matched_this_file {
1267 files_with_matches.fetch_add(1, Ordering::Relaxed);
1268 }
1269
1270 matches
1271}
1272
1273fn should_stop_search(
1274 truncated: &AtomicBool,
1275 total_matches: &AtomicUsize,
1276 stop_after: usize,
1277) -> bool {
1278 truncated.load(Ordering::Relaxed) && total_matches.load(Ordering::Relaxed) >= stop_after
1279}
1280
1281fn intersect_sorted_ids(left: &[u32], right: &[u32]) -> Vec<u32> {
1282 let mut merged = Vec::with_capacity(left.len().min(right.len()));
1283 let mut left_index = 0;
1284 let mut right_index = 0;
1285
1286 while left_index < left.len() && right_index < right.len() {
1287 match left[left_index].cmp(&right[right_index]) {
1288 std::cmp::Ordering::Less => left_index += 1,
1289 std::cmp::Ordering::Greater => right_index += 1,
1290 std::cmp::Ordering::Equal => {
1291 merged.push(left[left_index]);
1292 left_index += 1;
1293 right_index += 1;
1294 }
1295 }
1296 }
1297
1298 merged
1299}
1300
1301fn union_sorted_ids(left: &[u32], right: &[u32]) -> Vec<u32> {
1302 let mut merged = Vec::with_capacity(left.len() + right.len());
1303 let mut left_index = 0;
1304 let mut right_index = 0;
1305
1306 while left_index < left.len() && right_index < right.len() {
1307 match left[left_index].cmp(&right[right_index]) {
1308 std::cmp::Ordering::Less => {
1309 merged.push(left[left_index]);
1310 left_index += 1;
1311 }
1312 std::cmp::Ordering::Greater => {
1313 merged.push(right[right_index]);
1314 right_index += 1;
1315 }
1316 std::cmp::Ordering::Equal => {
1317 merged.push(left[left_index]);
1318 left_index += 1;
1319 right_index += 1;
1320 }
1321 }
1322 }
1323
1324 merged.extend_from_slice(&left[left_index..]);
1325 merged.extend_from_slice(&right[right_index..]);
1326 merged
1327}
1328
1329pub fn decompose_regex(pattern: &str) -> RegexQuery {
1330 let hir = match regex_syntax::parse(pattern) {
1331 Ok(hir) => hir,
1332 Err(_) => return RegexQuery::default(),
1333 };
1334
1335 let build = build_query(&hir);
1336 build.into_query()
1337}
1338
1339pub fn pack_trigram(a: u8, b: u8, c: u8) -> u32 {
1340 ((a as u32) << 16) | ((b as u32) << 8) | c as u32
1341}
1342
1343pub fn normalize_char(c: u8) -> u8 {
1344 c.to_ascii_lowercase()
1345}
1346
1347pub fn extract_trigrams(content: &[u8]) -> Vec<(u32, u8, usize)> {
1348 if content.len() < 3 {
1349 return Vec::new();
1350 }
1351
1352 let mut trigrams = Vec::with_capacity(content.len().saturating_sub(2));
1353 for start in 0..=content.len() - 3 {
1354 let trigram = pack_trigram(
1355 normalize_char(content[start]),
1356 normalize_char(content[start + 1]),
1357 normalize_char(content[start + 2]),
1358 );
1359 let next_char = content.get(start + 3).copied().unwrap_or(EOF_SENTINEL);
1360 trigrams.push((trigram, next_char, start));
1361 }
1362 trigrams
1363}
1364
1365pub fn query_trigrams_from_tokens(tokens: &[&str]) -> Vec<u32> {
1366 let mut seen = HashSet::new();
1367 let mut out = Vec::new();
1368 for token in tokens {
1369 for (trigram, _, _) in extract_trigrams(token.as_bytes()) {
1370 if seen.insert(trigram) {
1371 out.push(trigram);
1372 }
1373 }
1374 }
1375 out
1376}
1377
1378pub fn lexical_score(index: &SearchIndex, query_trigrams: &[u32], file_id: u32) -> f32 {
1379 if query_trigrams.is_empty() {
1380 return 0.0;
1381 }
1382
1383 let mut hits = 0u32;
1384 for &trigram in query_trigrams {
1385 if let Some(postings) = index.postings.get(&trigram) {
1386 if postings
1387 .binary_search_by(|posting| posting.file_id.cmp(&file_id))
1388 .is_ok()
1389 {
1390 hits += 1;
1391 }
1392 }
1393 }
1394
1395 if hits == 0 {
1396 return 0.0;
1397 }
1398
1399 let file_trigram_count = index
1400 .file_trigrams
1401 .get(&file_id)
1402 .map_or(1, |trigrams| trigrams.len().max(1)) as f32;
1403 (hits as f32) / (1.0 + file_trigram_count.ln())
1404}
1405
1406pub fn resolve_cache_dir(project_root: &Path, storage_dir: Option<&Path>) -> PathBuf {
1407 if let Some(override_dir) = std::env::var_os("AFT_CACHE_DIR") {
1409 return PathBuf::from(override_dir)
1410 .join("index")
1411 .join(project_cache_key(project_root));
1412 }
1413 if let Some(dir) = storage_dir {
1415 return dir.join("index").join(project_cache_key(project_root));
1416 }
1417 let home = std::env::var_os("HOME")
1422 .or_else(|| std::env::var_os("USERPROFILE"))
1423 .map(PathBuf::from)
1424 .unwrap_or_else(std::env::temp_dir);
1425 home.join(".cache")
1426 .join("aft")
1427 .join("index")
1428 .join(project_cache_key(project_root))
1429}
1430
1431pub(crate) fn build_path_filters(
1432 include: &[String],
1433 exclude: &[String],
1434) -> Result<PathFilters, String> {
1435 Ok(PathFilters {
1436 includes: build_globset(include)?,
1437 excludes: build_globset(exclude)?,
1438 })
1439}
1440
1441pub(crate) fn walk_project_files(root: &Path, filters: &PathFilters) -> Vec<PathBuf> {
1442 walk_project_files_from(root, root, filters)
1443}
1444
1445pub(crate) fn walk_project_files_from(
1446 filter_root: &Path,
1447 search_root: &Path,
1448 filters: &PathFilters,
1449) -> Vec<PathBuf> {
1450 let mut builder = WalkBuilder::new(search_root);
1451 builder
1452 .hidden(false)
1453 .git_ignore(true)
1454 .git_global(true)
1455 .git_exclude(true)
1456 .filter_entry(|entry| {
1457 let name = entry.file_name().to_string_lossy();
1458 if entry.file_type().map_or(false, |ft| ft.is_dir()) {
1459 return !matches!(
1460 name.as_ref(),
1461 "node_modules"
1462 | "target"
1463 | "venv"
1464 | ".venv"
1465 | ".git"
1466 | "__pycache__"
1467 | ".tox"
1468 | "dist"
1469 | "build"
1470 );
1471 }
1472 true
1473 });
1474
1475 let mut files = Vec::new();
1476 for entry in builder.build().filter_map(|entry| entry.ok()) {
1477 if !entry
1478 .file_type()
1479 .map_or(false, |file_type| file_type.is_file())
1480 {
1481 continue;
1482 }
1483 let path = entry.into_path();
1484 if filters.matches(filter_root, &path) {
1485 files.push(path);
1486 }
1487 }
1488
1489 sort_paths_by_mtime_desc(&mut files);
1490 files
1491}
1492
1493pub(crate) fn read_searchable_text(path: &Path) -> Option<String> {
1494 let bytes = fs::read(path).ok()?;
1495 if is_binary_bytes(&bytes) {
1496 return None;
1497 }
1498 String::from_utf8(bytes).ok()
1499}
1500
1501fn read_indexed_file_bytes(path: &Path) -> Option<Vec<u8>> {
1502 fs::read(path).ok()
1503}
1504
1505pub(crate) fn relative_to_root(root: &Path, path: &Path) -> PathBuf {
1506 path.strip_prefix(root)
1507 .map(PathBuf::from)
1508 .unwrap_or_else(|_| path.to_path_buf())
1509}
1510
1511pub(crate) fn cache_relative_path(root: &Path, path: &Path) -> Option<PathBuf> {
1512 let normalized_root = normalize_path(root);
1513 let normalized_path = normalize_path(path);
1514 let relative = normalized_path.strip_prefix(&normalized_root).ok()?;
1515 validate_cached_relative_path(relative)
1516}
1517
1518pub(crate) fn cached_path_under_root(root: &Path, relative_path: &Path) -> Option<PathBuf> {
1519 let relative = validate_cached_relative_path(relative_path)?;
1520 let normalized_root = normalize_path(root);
1521 let full_path = normalize_path(&normalized_root.join(relative));
1522 full_path.starts_with(&normalized_root).then_some(full_path)
1523}
1524
1525pub(crate) fn validate_cached_relative_path(path: &Path) -> Option<PathBuf> {
1526 if path.is_absolute() {
1527 return None;
1528 }
1529
1530 let mut normalized = PathBuf::new();
1531 for component in path.components() {
1532 match component {
1533 Component::Normal(part) => normalized.push(part),
1534 Component::CurDir => {}
1535 Component::ParentDir | Component::RootDir | Component::Prefix(_) => return None,
1536 }
1537 }
1538 (!normalized.as_os_str().is_empty()).then_some(normalized)
1539}
1540
1541pub(crate) fn sort_paths_by_mtime_desc(paths: &mut [PathBuf]) {
1554 use std::collections::HashMap;
1555 let mut mtimes: HashMap<PathBuf, Option<SystemTime>> = HashMap::with_capacity(paths.len());
1556 for path in paths.iter() {
1557 mtimes
1558 .entry(path.clone())
1559 .or_insert_with(|| path_modified_time(path));
1560 }
1561 paths.sort_by(|left, right| {
1562 let left_mtime = mtimes.get(left).and_then(|v| *v);
1563 let right_mtime = mtimes.get(right).and_then(|v| *v);
1564 right_mtime.cmp(&left_mtime).then_with(|| left.cmp(right))
1565 });
1566}
1567
1568pub(crate) fn sort_grep_matches_by_mtime_desc(matches: &mut [GrepMatch], project_root: &Path) {
1571 use std::collections::HashMap;
1572 let mut mtimes: HashMap<PathBuf, Option<SystemTime>> = HashMap::new();
1573 for m in matches.iter() {
1574 mtimes.entry(m.file.clone()).or_insert_with(|| {
1575 let resolved = resolve_match_path(project_root, &m.file);
1576 path_modified_time(&resolved)
1577 });
1578 }
1579 matches.sort_by(|left, right| {
1580 let left_mtime = mtimes.get(&left.file).and_then(|v| *v);
1581 let right_mtime = mtimes.get(&right.file).and_then(|v| *v);
1582 right_mtime
1583 .cmp(&left_mtime)
1584 .then_with(|| left.file.cmp(&right.file))
1585 .then_with(|| left.line.cmp(&right.line))
1586 .then_with(|| left.column.cmp(&right.column))
1587 });
1588}
1589
1590fn sort_shared_grep_matches_by_cached_mtime_desc<F>(
1595 matches: &mut [SharedGrepMatch],
1596 modified_for_path: F,
1597) where
1598 F: Fn(&Path) -> Option<SystemTime>,
1599{
1600 use std::collections::HashMap;
1601 let mut mtimes: HashMap<PathBuf, Option<SystemTime>> = HashMap::with_capacity(matches.len());
1602 for m in matches.iter() {
1603 let path = m.file.as_path().to_path_buf();
1604 mtimes
1605 .entry(path.clone())
1606 .or_insert_with(|| modified_for_path(&path));
1607 }
1608 matches.sort_by(|left, right| {
1609 let left_mtime = mtimes.get(left.file.as_path()).and_then(|v| *v);
1610 let right_mtime = mtimes.get(right.file.as_path()).and_then(|v| *v);
1611 right_mtime
1612 .cmp(&left_mtime)
1613 .then_with(|| left.file.as_path().cmp(right.file.as_path()))
1614 .then_with(|| left.line.cmp(&right.line))
1615 .then_with(|| left.column.cmp(&right.column))
1616 });
1617}
1618
1619pub(crate) fn resolve_search_scope(project_root: &Path, path: Option<&str>) -> SearchScope {
1620 let resolved_project_root = canonicalize_or_normalize(project_root);
1621 let root = match path {
1622 Some(path) => {
1623 let path = PathBuf::from(path);
1624 if path.is_absolute() {
1625 canonicalize_or_normalize(&path)
1626 } else {
1627 normalize_path(&resolved_project_root.join(path))
1628 }
1629 }
1630 None => resolved_project_root.clone(),
1631 };
1632
1633 let use_index = is_within_search_root(&resolved_project_root, &root);
1634 SearchScope { root, use_index }
1635}
1636
1637pub(crate) fn is_binary_bytes(content: &[u8]) -> bool {
1638 content_inspector::inspect(content).is_binary()
1639}
1640
1641pub(crate) fn current_git_head(root: &Path) -> Option<String> {
1642 run_git(root, &["rev-parse", "HEAD"])
1643}
1644
1645pub fn project_cache_key(project_root: &Path) -> String {
1646 use sha2::{Digest, Sha256};
1647
1648 let mut hasher = Sha256::new();
1649
1650 if let Some(root_commit) = run_git(project_root, &["rev-list", "--max-parents=0", "HEAD"]) {
1651 hasher.update(root_commit.as_bytes());
1654 } else {
1655 let canonical_root = canonicalize_or_normalize(project_root);
1657 hasher.update(canonical_root.to_string_lossy().as_bytes());
1658 }
1659
1660 let digest = format!("{:x}", hasher.finalize());
1661 digest[..16].to_string()
1662}
1663
1664impl PathFilters {
1665 fn matches(&self, root: &Path, path: &Path) -> bool {
1666 let relative = to_glob_path(&relative_to_root(root, path));
1667 if self
1668 .includes
1669 .as_ref()
1670 .is_some_and(|includes| !includes.is_match(&relative))
1671 {
1672 return false;
1673 }
1674 if self
1675 .excludes
1676 .as_ref()
1677 .is_some_and(|excludes| excludes.is_match(&relative))
1678 {
1679 return false;
1680 }
1681 true
1682 }
1683}
1684
1685fn canonicalize_or_normalize(path: &Path) -> PathBuf {
1686 fs::canonicalize(path).unwrap_or_else(|_| normalize_path(path))
1687}
1688
1689fn resolve_match_path(project_root: &Path, path: &Path) -> PathBuf {
1690 if path.is_absolute() {
1691 path.to_path_buf()
1692 } else {
1693 project_root.join(path)
1694 }
1695}
1696
1697fn path_modified_time(path: &Path) -> Option<SystemTime> {
1698 fs::metadata(path)
1699 .and_then(|metadata| metadata.modified())
1700 .ok()
1701}
1702
1703fn normalize_path(path: &Path) -> PathBuf {
1704 let mut result = PathBuf::new();
1705 for component in path.components() {
1706 match component {
1707 Component::ParentDir => {
1708 if !result.pop() {
1709 result.push(component);
1710 }
1711 }
1712 Component::CurDir => {}
1713 _ => result.push(component),
1714 }
1715 }
1716 result
1717}
1718
1719fn verify_file_mtimes(index: &mut SearchIndex) {
1722 let mut stale_paths = Vec::new();
1724 for entry in &index.files {
1725 if entry.path.as_os_str().is_empty() {
1726 continue; }
1728 let cached = FileFreshness {
1729 mtime: entry.modified,
1730 size: entry.size,
1731 content_hash: entry.content_hash,
1732 };
1733 match cache_freshness::verify_file(&entry.path, &cached) {
1734 FreshnessVerdict::HotFresh | FreshnessVerdict::ContentFresh { .. } => {}
1735 FreshnessVerdict::Stale | FreshnessVerdict::Deleted => {
1736 stale_paths.push(entry.path.clone())
1737 }
1738 }
1739 }
1740
1741 for entry in &mut index.files {
1742 if entry.path.as_os_str().is_empty() {
1743 continue;
1744 }
1745 let cached = FileFreshness {
1746 mtime: entry.modified,
1747 size: entry.size,
1748 content_hash: entry.content_hash,
1749 };
1750 if let FreshnessVerdict::ContentFresh {
1751 new_mtime,
1752 new_size,
1753 } = cache_freshness::verify_file(&entry.path, &cached)
1754 {
1755 entry.modified = new_mtime;
1756 entry.size = new_size;
1757 }
1758 }
1759
1760 for path in &stale_paths {
1762 index.update_file(path);
1763 }
1764
1765 let filters = PathFilters::default();
1767 for path in walk_project_files(&index.project_root, &filters) {
1768 if !index.path_to_id.contains_key(&path) {
1769 index.update_file(&path);
1770 }
1771 }
1772
1773 if !stale_paths.is_empty() {
1774 crate::slog_info!(
1775 "search index: refreshed {} stale file(s) from disk cache",
1776 stale_paths.len()
1777 );
1778 }
1779}
1780
1781fn is_within_search_root(search_root: &Path, path: &Path) -> bool {
1782 normalize_path(path).starts_with(normalize_path(search_root))
1783}
1784
1785impl QueryBuild {
1786 fn into_query(self) -> RegexQuery {
1787 let mut query = RegexQuery::default();
1788
1789 for run in self.and_runs {
1790 add_run_to_and_query(&mut query, &run);
1791 }
1792
1793 for group in self.or_groups {
1794 let mut trigrams = BTreeSet::new();
1795 let mut filters = HashMap::new();
1796 for run in group {
1797 for (trigram, filter) in trigram_filters(&run) {
1798 trigrams.insert(trigram);
1799 merge_filter(filters.entry(trigram).or_default(), filter);
1800 }
1801 }
1802 if !trigrams.is_empty() {
1803 query.or_groups.push(trigrams.into_iter().collect());
1804 query.or_filters.push(filters);
1805 }
1806 }
1807
1808 query
1809 }
1810}
1811
1812fn build_query(hir: &Hir) -> QueryBuild {
1813 match hir.kind() {
1814 HirKind::Literal(literal) => {
1815 if literal.0.len() >= 3 {
1816 QueryBuild {
1817 and_runs: vec![literal.0.to_vec()],
1818 or_groups: Vec::new(),
1819 }
1820 } else {
1821 QueryBuild::default()
1822 }
1823 }
1824 HirKind::Capture(capture) => build_query(&capture.sub),
1825 HirKind::Concat(parts) => {
1826 let mut build = QueryBuild::default();
1827 for part in parts {
1828 let part_build = build_query(part);
1829 build.and_runs.extend(part_build.and_runs);
1830 build.or_groups.extend(part_build.or_groups);
1831 }
1832 build
1833 }
1834 HirKind::Alternation(parts) => {
1835 let mut group = Vec::new();
1836 for part in parts {
1837 let Some(mut choices) = guaranteed_run_choices(part) else {
1838 return QueryBuild::default();
1839 };
1840 group.append(&mut choices);
1841 }
1842 if group.is_empty() {
1843 QueryBuild::default()
1844 } else {
1845 QueryBuild {
1846 and_runs: Vec::new(),
1847 or_groups: vec![group],
1848 }
1849 }
1850 }
1851 HirKind::Repetition(repetition) => {
1852 if repetition.min == 0 {
1853 QueryBuild::default()
1854 } else {
1855 build_query(&repetition.sub)
1856 }
1857 }
1858 HirKind::Empty | HirKind::Class(_) | HirKind::Look(_) => QueryBuild::default(),
1859 }
1860}
1861
1862fn guaranteed_run_choices(hir: &Hir) -> Option<Vec<Vec<u8>>> {
1863 match hir.kind() {
1864 HirKind::Literal(literal) => {
1865 if literal.0.len() >= 3 {
1866 Some(vec![literal.0.to_vec()])
1867 } else {
1868 None
1869 }
1870 }
1871 HirKind::Capture(capture) => guaranteed_run_choices(&capture.sub),
1872 HirKind::Concat(parts) => {
1873 let mut runs = Vec::new();
1874 for part in parts {
1875 if let Some(mut part_runs) = guaranteed_run_choices(part) {
1876 runs.append(&mut part_runs);
1877 }
1878 }
1879 if runs.is_empty() {
1880 None
1881 } else {
1882 Some(runs)
1883 }
1884 }
1885 HirKind::Alternation(parts) => {
1886 let mut runs = Vec::new();
1887 for part in parts {
1888 let Some(mut part_runs) = guaranteed_run_choices(part) else {
1889 return None;
1890 };
1891 runs.append(&mut part_runs);
1892 }
1893 if runs.is_empty() {
1894 None
1895 } else {
1896 Some(runs)
1897 }
1898 }
1899 HirKind::Repetition(repetition) => {
1900 if repetition.min == 0 {
1901 None
1902 } else {
1903 guaranteed_run_choices(&repetition.sub)
1904 }
1905 }
1906 HirKind::Empty | HirKind::Class(_) | HirKind::Look(_) => None,
1907 }
1908}
1909
1910fn add_run_to_and_query(query: &mut RegexQuery, run: &[u8]) {
1911 for (trigram, filter) in trigram_filters(run) {
1912 if !query.and_trigrams.contains(&trigram) {
1913 query.and_trigrams.push(trigram);
1914 }
1915 merge_filter(query.and_filters.entry(trigram).or_default(), filter);
1916 }
1917}
1918
1919fn trigram_filters(run: &[u8]) -> Vec<(u32, PostingFilter)> {
1920 let mut filters: BTreeMap<u32, PostingFilter> = BTreeMap::new();
1921 for (trigram, next_char, position) in extract_trigrams(run) {
1922 let entry: &mut PostingFilter = filters.entry(trigram).or_default();
1923 if next_char != EOF_SENTINEL {
1924 entry.next_mask |= mask_for_next_char(next_char);
1925 }
1926 entry.loc_mask |= mask_for_position(position);
1927 }
1928 filters.into_iter().collect()
1929}
1930
1931fn merge_filter(target: &mut PostingFilter, filter: PostingFilter) {
1932 target.next_mask |= filter.next_mask;
1933 target.loc_mask |= filter.loc_mask;
1934}
1935
1936fn mask_for_next_char(next_char: u8) -> u8 {
1937 let bit = (normalize_char(next_char).wrapping_mul(31) & 7) as u32;
1938 1u8 << bit
1939}
1940
1941fn mask_for_position(position: usize) -> u8 {
1942 1u8 << (position % 8)
1943}
1944
1945fn build_globset(patterns: &[String]) -> Result<Option<GlobSet>, String> {
1946 if patterns.is_empty() {
1947 return Ok(None);
1948 }
1949
1950 let mut builder = GlobSetBuilder::new();
1951 for pattern in patterns {
1952 let glob = Glob::new(pattern).map_err(|error| error.to_string())?;
1953 builder.add(glob);
1954 }
1955 builder.build().map(Some).map_err(|error| error.to_string())
1956}
1957
1958fn read_u32<R: Read>(reader: &mut R) -> std::io::Result<u32> {
1959 let mut buffer = [0u8; 4];
1960 reader.read_exact(&mut buffer)?;
1961 Ok(u32::from_le_bytes(buffer))
1962}
1963
1964fn read_u64<R: Read>(reader: &mut R) -> std::io::Result<u64> {
1965 let mut buffer = [0u8; 8];
1966 reader.read_exact(&mut buffer)?;
1967 Ok(u64::from_le_bytes(buffer))
1968}
1969
1970fn write_u32<W: Write>(writer: &mut W, value: u32) -> std::io::Result<()> {
1971 writer.write_all(&value.to_le_bytes())
1972}
1973
1974fn write_u64<W: Write>(writer: &mut W, value: u64) -> std::io::Result<()> {
1975 writer.write_all(&value.to_le_bytes())
1976}
1977
1978fn verify_crc32_bytes_slice(bytes: &[u8]) -> std::io::Result<()> {
1979 let Some((body, stored)) = bytes.split_last_chunk::<4>() else {
1980 return Err(std::io::Error::other("search index checksum missing"));
1981 };
1982 let expected = u32::from_le_bytes(*stored);
1983 let actual = crc32fast::hash(body);
1984 if actual != expected {
1985 return Err(std::io::Error::other("search index checksum mismatch"));
1986 }
1987 Ok(())
1988}
1989
1990fn remaining_bytes<R: Seek>(reader: &mut R, total_len: usize) -> Option<usize> {
1991 let pos = usize::try_from(reader.stream_position().ok()?).ok()?;
1992 total_len.checked_sub(pos)
1993}
1994
1995fn run_git(root: &Path, args: &[&str]) -> Option<String> {
1996 let output = Command::new("git")
1997 .arg("-C")
1998 .arg(root)
1999 .args(args)
2000 .output()
2001 .ok()?;
2002 if !output.status.success() {
2003 return None;
2004 }
2005 let value = String::from_utf8(output.stdout).ok()?;
2006 let value = value.trim().to_string();
2007 if value.is_empty() {
2008 None
2009 } else {
2010 Some(value)
2011 }
2012}
2013
2014fn apply_git_diff_updates(index: &mut SearchIndex, root: &Path, from: &str, to: &str) -> bool {
2015 let diff_range = format!("{}..{}", from, to);
2016 let output = match Command::new("git")
2017 .arg("-C")
2018 .arg(root)
2019 .args(["diff", "--name-status", "-M", &diff_range])
2020 .output()
2021 {
2022 Ok(output) => output,
2023 Err(_) => return false,
2024 };
2025
2026 if !output.status.success() {
2027 return false;
2028 }
2029
2030 let Ok(diff) = String::from_utf8(output.stdout) else {
2031 return false;
2032 };
2033
2034 for line in diff.lines().map(str::trim).filter(|line| !line.is_empty()) {
2035 let mut fields = line.split('\t');
2036 let Some(status) = fields.next() else {
2037 continue;
2038 };
2039
2040 if status.starts_with('R') {
2041 let Some(old_path) = fields
2042 .next()
2043 .and_then(|path| cached_path_under_root(root, &PathBuf::from(path)))
2044 else {
2045 continue;
2046 };
2047 let Some(new_path) = fields
2048 .next()
2049 .and_then(|path| cached_path_under_root(root, &PathBuf::from(path)))
2050 else {
2051 continue;
2052 };
2053 index.remove_file(&old_path);
2054 index.update_file(&new_path);
2055 continue;
2056 }
2057
2058 let Some(path) = fields
2059 .next()
2060 .and_then(|path| cached_path_under_root(root, &PathBuf::from(path)))
2061 else {
2062 continue;
2063 };
2064 if status.starts_with('D') || !path.exists() {
2065 index.remove_file(&path);
2066 } else {
2067 index.update_file(&path);
2068 }
2069 }
2070
2071 true
2072}
2073
2074fn is_binary_path(path: &Path, size: u64) -> bool {
2075 if size == 0 {
2076 return false;
2077 }
2078
2079 let mut file = match File::open(path) {
2080 Ok(file) => file,
2081 Err(_) => return true,
2082 };
2083
2084 let mut preview = vec![0u8; PREVIEW_BYTES.min(size as usize)];
2085 match file.read(&mut preview) {
2086 Ok(read) => is_binary_bytes(&preview[..read]),
2087 Err(_) => true,
2088 }
2089}
2090
2091fn line_starts_bytes(content: &[u8]) -> Vec<usize> {
2092 let mut starts = vec![0usize];
2093 for (index, byte) in content.iter().copied().enumerate() {
2094 if byte == b'\n' {
2095 starts.push(index + 1);
2096 }
2097 }
2098 starts
2099}
2100
2101fn line_details_bytes(content: &[u8], line_starts: &[usize], offset: usize) -> (u32, u32, String) {
2102 let line_index = match line_starts.binary_search(&offset) {
2103 Ok(index) => index,
2104 Err(index) => index.saturating_sub(1),
2105 };
2106 let line_start = line_starts.get(line_index).copied().unwrap_or(0);
2107 let line_end = content[line_start..]
2108 .iter()
2109 .position(|byte| *byte == b'\n')
2110 .map(|length| line_start + length)
2111 .unwrap_or(content.len());
2112 let mut line_slice = &content[line_start..line_end];
2113 if line_slice.ends_with(b"\r") {
2114 line_slice = &line_slice[..line_slice.len() - 1];
2115 }
2116 let line_text = String::from_utf8_lossy(line_slice).into_owned();
2117 let column = String::from_utf8_lossy(&content[line_start..offset])
2118 .chars()
2119 .count() as u32
2120 + 1;
2121 (line_index as u32 + 1, column, line_text)
2122}
2123
2124fn to_glob_path(path: &Path) -> String {
2125 path.to_string_lossy().replace('\\', "/")
2126}
2127
2128#[cfg(test)]
2129mod tests {
2130 use std::process::Command;
2131
2132 use super::*;
2133
2134 #[test]
2135 fn extract_trigrams_tracks_next_char_and_position() {
2136 let trigrams = extract_trigrams(b"Rust");
2137 assert_eq!(trigrams.len(), 2);
2138 assert_eq!(trigrams[0], (pack_trigram(b'r', b'u', b's'), b't', 0));
2139 assert_eq!(
2140 trigrams[1],
2141 (pack_trigram(b'u', b's', b't'), EOF_SENTINEL, 1)
2142 );
2143 }
2144
2145 #[test]
2146 fn decompose_regex_extracts_literals_and_alternations() {
2147 let query = decompose_regex("abc(def|ghi)xyz");
2148 assert!(query.and_trigrams.contains(&pack_trigram(b'a', b'b', b'c')));
2149 assert!(query.and_trigrams.contains(&pack_trigram(b'x', b'y', b'z')));
2150 assert_eq!(query.or_groups.len(), 1);
2151 assert!(query.or_groups[0].contains(&pack_trigram(b'd', b'e', b'f')));
2152 assert!(query.or_groups[0].contains(&pack_trigram(b'g', b'h', b'i')));
2153 }
2154
2155 #[test]
2156 fn candidates_intersect_posting_lists() {
2157 let mut index = SearchIndex::new();
2158 let dir = tempfile::tempdir().expect("create temp dir");
2159 let alpha = dir.path().join("alpha.txt");
2160 let beta = dir.path().join("beta.txt");
2161 fs::write(&alpha, "abcdef").expect("write alpha");
2162 fs::write(&beta, "abcxyz").expect("write beta");
2163 index.project_root = dir.path().to_path_buf();
2164 index.index_file(&alpha, b"abcdef");
2165 index.index_file(&beta, b"abcxyz");
2166
2167 let query = RegexQuery {
2168 and_trigrams: vec![
2169 pack_trigram(b'a', b'b', b'c'),
2170 pack_trigram(b'd', b'e', b'f'),
2171 ],
2172 ..RegexQuery::default()
2173 };
2174
2175 let candidates = index.candidates(&query);
2176 assert_eq!(candidates.len(), 1);
2177 assert_eq!(index.files[candidates[0] as usize].path, alpha);
2178 }
2179
2180 #[test]
2181 fn candidates_apply_bloom_filters() {
2182 let mut index = SearchIndex::new();
2183 let dir = tempfile::tempdir().expect("create temp dir");
2184 let file = dir.path().join("sample.txt");
2185 fs::write(&file, "abcd efgh").expect("write sample");
2186 index.project_root = dir.path().to_path_buf();
2187 index.index_file(&file, b"abcd efgh");
2188
2189 let trigram = pack_trigram(b'a', b'b', b'c');
2190 let matching_filter = PostingFilter {
2191 next_mask: mask_for_next_char(b'd'),
2192 loc_mask: mask_for_position(0),
2193 };
2194 let non_matching_filter = PostingFilter {
2195 next_mask: mask_for_next_char(b'z'),
2196 loc_mask: mask_for_position(0),
2197 };
2198
2199 assert_eq!(
2200 index
2201 .postings_for_trigram(trigram, Some(matching_filter))
2202 .len(),
2203 1
2204 );
2205 assert!(index
2206 .postings_for_trigram(trigram, Some(non_matching_filter))
2207 .is_empty());
2208 }
2209
2210 #[test]
2211 fn disk_round_trip_preserves_postings_and_files() {
2212 let dir = tempfile::tempdir().expect("create temp dir");
2213 let project = dir.path().join("project");
2214 fs::create_dir_all(&project).expect("create project dir");
2215 let file = project.join("src.txt");
2216 fs::write(&file, "abcdef").expect("write source");
2217
2218 let mut index = SearchIndex::build(&project);
2219 index.git_head = Some("deadbeef".to_string());
2220 let cache_dir = dir.path().join("cache");
2221 index.write_to_disk(&cache_dir, index.git_head.as_deref());
2222
2223 let loaded =
2224 SearchIndex::read_from_disk(&cache_dir, &project).expect("load index from disk");
2225 assert_eq!(loaded.stored_git_head(), Some("deadbeef"));
2226 assert_eq!(loaded.files.len(), 1);
2227 assert_eq!(
2228 relative_to_root(&loaded.project_root, &loaded.files[0].path),
2229 PathBuf::from("src.txt")
2230 );
2231 assert_eq!(loaded.postings.len(), index.postings.len());
2232 assert!(loaded
2233 .postings
2234 .contains_key(&pack_trigram(b'a', b'b', b'c')));
2235 }
2236
2237 #[test]
2238 fn cache_path_helpers_reject_absolute_and_parent_paths() {
2239 let root = PathBuf::from("/tmp/aft-project");
2240
2241 assert_eq!(
2242 cache_relative_path(&root, &root.join("src/lib.rs")),
2243 Some(PathBuf::from("src/lib.rs"))
2244 );
2245 assert!(cache_relative_path(&root, Path::new("/tmp/outside.rs")).is_none());
2246 assert!(cached_path_under_root(&root, Path::new("../outside.rs")).is_none());
2247 assert!(cached_path_under_root(&root, Path::new("/tmp/outside.rs")).is_none());
2248 assert_eq!(
2249 cached_path_under_root(&root, Path::new("src/./lib.rs")),
2250 Some(root.join("src/lib.rs"))
2251 );
2252 }
2253
2254 #[test]
2255 fn refresh_after_head_change_removes_renames_and_detects_local_files() {
2256 let dir = tempfile::tempdir().expect("create temp dir");
2257 let project = dir.path().join("project");
2258 fs::create_dir_all(&project).expect("create project dir");
2259 let canonical_project = fs::canonicalize(&project).expect("canonical project");
2260 fs::write(project.join("old.txt"), "old token\n").expect("write old");
2261 fs::write(project.join("unchanged.txt"), "before\n").expect("write unchanged");
2262
2263 Command::new("git")
2264 .arg("init")
2265 .arg(&project)
2266 .status()
2267 .expect("git init");
2268 for args in [
2269 ["config", "user.email", "aft@example.invalid"],
2270 ["config", "user.name", "AFT Test"],
2271 ] {
2272 Command::new("git")
2273 .arg("-C")
2274 .arg(&project)
2275 .args(args)
2276 .status()
2277 .expect("git config");
2278 }
2279 Command::new("git")
2280 .arg("-C")
2281 .arg(&project)
2282 .args(["add", "."])
2283 .status()
2284 .expect("git add initial");
2285 Command::new("git")
2286 .arg("-C")
2287 .arg(&project)
2288 .args(["commit", "-m", "initial"])
2289 .status()
2290 .expect("git commit initial");
2291 let previous = run_git(&project, &["rev-parse", "HEAD"]).expect("previous head");
2292 let mut baseline = SearchIndex::build(&project);
2293 baseline.git_head = Some(previous.clone());
2294
2295 fs::rename(project.join("old.txt"), project.join("new.txt")).expect("rename file");
2296 Command::new("git")
2297 .arg("-C")
2298 .arg(&project)
2299 .args(["add", "-A"])
2300 .status()
2301 .expect("git add rename");
2302 Command::new("git")
2303 .arg("-C")
2304 .arg(&project)
2305 .args(["commit", "-m", "rename"])
2306 .status()
2307 .expect("git commit rename");
2308 let current = run_git(&project, &["rev-parse", "HEAD"]).expect("current head");
2309
2310 fs::write(project.join("unchanged.txt"), "after local edit\n").expect("local edit");
2311 fs::write(project.join("untracked.txt"), "untracked token\n").expect("untracked");
2312
2313 let refreshed = SearchIndex::rebuild_or_refresh(
2314 &project,
2315 DEFAULT_MAX_FILE_SIZE,
2316 Some(current),
2317 Some(baseline),
2318 );
2319
2320 assert!(!refreshed
2321 .path_to_id
2322 .contains_key(&canonical_project.join("old.txt")));
2323 assert!(refreshed
2324 .path_to_id
2325 .contains_key(&canonical_project.join("new.txt")));
2326 assert!(refreshed
2327 .path_to_id
2328 .contains_key(&canonical_project.join("untracked.txt")));
2329 let matches =
2330 refreshed.search_grep("after local edit", true, &[], &[], &canonical_project, 10);
2331 assert_eq!(matches.matches.len(), 1);
2332 }
2333
2334 #[test]
2335 fn read_from_disk_rejects_corrupt_postings_checksum() {
2336 let dir = tempfile::tempdir().expect("create temp dir");
2337 let project = dir.path().join("project");
2338 fs::create_dir_all(&project).expect("create project dir");
2339 fs::write(project.join("src.txt"), "abcdef").expect("write source");
2340
2341 let index = SearchIndex::build(&project);
2342 let cache_dir = dir.path().join("cache");
2343 index.write_to_disk(&cache_dir, None);
2344
2345 let cache_path = cache_dir.join("cache.bin");
2346 let mut bytes = fs::read(&cache_path).expect("read cache");
2347 let middle = bytes.len() / 2;
2348 bytes[middle] ^= 0xff;
2349 fs::write(&cache_path, bytes).expect("write corrupted cache");
2350
2351 assert!(SearchIndex::read_from_disk(&cache_dir, &project).is_none());
2352 }
2353
2354 #[test]
2355 fn write_to_disk_uses_temp_files_and_cleans_them_up() {
2356 let dir = tempfile::tempdir().expect("create temp dir");
2357 let project = dir.path().join("project");
2358 fs::create_dir_all(&project).expect("create project dir");
2359 fs::write(project.join("src.txt"), "abcdef").expect("write source");
2360
2361 let index = SearchIndex::build(&project);
2362 let cache_dir = dir.path().join("cache");
2363 index.write_to_disk(&cache_dir, None);
2364
2365 assert!(cache_dir.join("cache.bin").is_file());
2366 assert!(fs::read_dir(&cache_dir)
2367 .expect("read cache dir")
2368 .all(|entry| !entry
2369 .expect("cache entry")
2370 .file_name()
2371 .to_string_lossy()
2372 .contains(".tmp.")));
2373 }
2374
2375 #[test]
2376 fn concurrent_search_index_writes_do_not_corrupt() {
2377 let dir = tempfile::tempdir().expect("create temp dir");
2378 let project = dir.path().join("project");
2379 fs::create_dir_all(&project).expect("create project dir");
2380 fs::write(project.join("src.txt"), "abcdef\n").expect("write source");
2381 let cache_dir = dir.path().join("cache");
2382
2383 let a_project = project.clone();
2384 let a_cache = cache_dir.clone();
2385 let a = std::thread::spawn(move || {
2386 let _lock = CacheLock::acquire(&a_cache).expect("acquire cache lock a");
2387 let index = SearchIndex::build(&a_project);
2388 index.write_to_disk(&a_cache, None);
2389 });
2390 let b_project = project.clone();
2391 let b_cache = cache_dir.clone();
2392 let b = std::thread::spawn(move || {
2393 let _lock = CacheLock::acquire(&b_cache).expect("acquire cache lock b");
2394 let index = SearchIndex::build(&b_project);
2395 index.write_to_disk(&b_cache, None);
2396 });
2397 a.join().expect("writer a");
2398 b.join().expect("writer b");
2399
2400 assert!(SearchIndex::read_from_disk(&cache_dir, &project).is_some());
2401 }
2402
2403 #[test]
2404 fn search_index_atomic_rename_survives_partial_write() {
2405 let dir = tempfile::tempdir().expect("create temp dir");
2406 let cache_dir = dir.path().join("cache");
2407 fs::create_dir_all(&cache_dir).expect("create cache dir");
2408 fs::write(cache_dir.join("cache.bin.tmp.1.1"), b"partial").expect("write partial tmp");
2409
2410 assert!(SearchIndex::read_from_disk(&cache_dir, dir.path()).is_none());
2411 }
2412
2413 #[test]
2414 fn project_cache_key_includes_checkout_path() {
2415 let dir = tempfile::tempdir().expect("create temp dir");
2416 let source = dir.path().join("source");
2417 fs::create_dir_all(&source).expect("create source repo dir");
2418 fs::write(source.join("tracked.txt"), "content\n").expect("write tracked file");
2419
2420 assert!(Command::new("git")
2421 .current_dir(&source)
2422 .args(["init"])
2423 .status()
2424 .expect("init git repo")
2425 .success());
2426 assert!(Command::new("git")
2427 .current_dir(&source)
2428 .args(["add", "."])
2429 .status()
2430 .expect("git add")
2431 .success());
2432 assert!(Command::new("git")
2433 .current_dir(&source)
2434 .args([
2435 "-c",
2436 "user.name=AFT Tests",
2437 "-c",
2438 "user.email=aft-tests@example.com",
2439 "commit",
2440 "-m",
2441 "initial",
2442 ])
2443 .status()
2444 .expect("git commit")
2445 .success());
2446
2447 let clone = dir.path().join("clone");
2448 assert!(Command::new("git")
2449 .args(["clone", "--quiet"])
2450 .arg(&source)
2451 .arg(&clone)
2452 .status()
2453 .expect("git clone")
2454 .success());
2455
2456 let source_key = project_cache_key(&source);
2457 let clone_key = project_cache_key(&clone);
2458
2459 assert_eq!(source_key.len(), 16);
2460 assert_eq!(clone_key.len(), 16);
2461 assert_eq!(source_key, clone_key);
2463 }
2464
2465 #[test]
2466 fn git_head_unchanged_picks_up_local_edits() {
2467 let dir = tempfile::tempdir().expect("create temp dir");
2468 let project = dir.path().join("repo");
2469 fs::create_dir_all(&project).expect("create repo dir");
2470 let file = project.join("tracked.txt");
2471 fs::write(&file, "oldtoken\n").expect("write file");
2472 assert!(Command::new("git")
2473 .current_dir(&project)
2474 .arg("init")
2475 .status()
2476 .unwrap()
2477 .success());
2478 assert!(Command::new("git")
2479 .current_dir(&project)
2480 .args(["add", "."])
2481 .status()
2482 .unwrap()
2483 .success());
2484 assert!(Command::new("git")
2485 .current_dir(&project)
2486 .args([
2487 "-c",
2488 "user.name=AFT Tests",
2489 "-c",
2490 "user.email=aft-tests@example.com",
2491 "commit",
2492 "-m",
2493 "initial"
2494 ])
2495 .status()
2496 .unwrap()
2497 .success());
2498 let head = current_git_head(&project);
2499 let mut baseline = SearchIndex::build(&project);
2500 baseline.git_head = head.clone();
2501 fs::write(&file, "newtoken\n").expect("edit tracked file");
2502
2503 let refreshed =
2504 SearchIndex::rebuild_or_refresh(&project, DEFAULT_MAX_FILE_SIZE, head, Some(baseline));
2505 let result = refreshed.search_grep("newtoken", true, &[], &[], &project, 10);
2506
2507 assert_eq!(result.total_matches, 1);
2508 }
2509
2510 #[test]
2511 fn non_git_project_reuses_cache_when_files_unchanged() {
2512 let dir = tempfile::tempdir().expect("create temp dir");
2513 let project = dir.path().join("project");
2514 fs::create_dir_all(&project).expect("create project dir");
2515 fs::write(project.join("file.txt"), "unchangedtoken\n").expect("write file");
2516 let baseline = SearchIndex::build(&project);
2517 let baseline_file_count = baseline.file_count();
2518
2519 let refreshed =
2520 SearchIndex::rebuild_or_refresh(&project, DEFAULT_MAX_FILE_SIZE, None, Some(baseline));
2521
2522 assert_eq!(refreshed.file_count(), baseline_file_count);
2523 assert_eq!(
2524 refreshed
2525 .search_grep("unchangedtoken", true, &[], &[], &project, 10)
2526 .total_matches,
2527 1
2528 );
2529 }
2530
2531 #[test]
2532 fn resolve_search_scope_disables_index_for_external_path() {
2533 let dir = tempfile::tempdir().expect("create temp dir");
2534 let project = dir.path().join("project");
2535 let outside = dir.path().join("outside");
2536 fs::create_dir_all(&project).expect("create project dir");
2537 fs::create_dir_all(&outside).expect("create outside dir");
2538
2539 let scope = resolve_search_scope(&project, outside.to_str());
2540
2541 assert_eq!(
2542 scope.root,
2543 fs::canonicalize(&outside).expect("canonicalize outside")
2544 );
2545 assert!(!scope.use_index);
2546 }
2547
2548 #[test]
2549 fn grep_filters_matches_to_search_root() {
2550 let dir = tempfile::tempdir().expect("create temp dir");
2551 let project = dir.path().join("project");
2552 let src = project.join("src");
2553 let docs = project.join("docs");
2554 fs::create_dir_all(&src).expect("create src dir");
2555 fs::create_dir_all(&docs).expect("create docs dir");
2556 fs::write(src.join("main.rs"), "pub struct SearchIndex;\n").expect("write src file");
2557 fs::write(docs.join("guide.md"), "SearchIndex guide\n").expect("write docs file");
2558
2559 let index = SearchIndex::build(&project);
2560 let result = index.search_grep("SearchIndex", true, &[], &[], &src, 10);
2561
2562 assert_eq!(result.files_searched, 1);
2563 assert_eq!(result.files_with_matches, 1);
2564 assert_eq!(result.matches.len(), 1);
2565 let expected = fs::canonicalize(src.join("main.rs")).expect("canonicalize");
2567 assert_eq!(result.matches[0].file, expected);
2568 }
2569
2570 #[test]
2571 fn grep_deduplicates_multiple_matches_on_same_line() {
2572 let dir = tempfile::tempdir().expect("create temp dir");
2573 let project = dir.path().join("project");
2574 let src = project.join("src");
2575 fs::create_dir_all(&src).expect("create src dir");
2576 fs::write(src.join("main.rs"), "SearchIndex SearchIndex\n").expect("write src file");
2577
2578 let index = SearchIndex::build(&project);
2579 let result = index.search_grep("SearchIndex", true, &[], &[], &src, 10);
2580
2581 assert_eq!(result.total_matches, 1);
2582 assert_eq!(result.matches.len(), 1);
2583 }
2584
2585 #[test]
2586 fn grep_reports_total_matches_before_truncation() {
2587 let dir = tempfile::tempdir().expect("create temp dir");
2588 let project = dir.path().join("project");
2589 let src = project.join("src");
2590 fs::create_dir_all(&src).expect("create src dir");
2591 fs::write(src.join("main.rs"), "SearchIndex\nSearchIndex\n").expect("write src file");
2592
2593 let index = SearchIndex::build(&project);
2594 let result = index.search_grep("SearchIndex", true, &[], &[], &src, 1);
2595
2596 assert_eq!(result.total_matches, 2);
2597 assert_eq!(result.matches.len(), 1);
2598 assert!(result.truncated);
2599 }
2600
2601 #[test]
2602 fn glob_filters_results_to_search_root() {
2603 let dir = tempfile::tempdir().expect("create temp dir");
2604 let project = dir.path().join("project");
2605 let src = project.join("src");
2606 let scripts = project.join("scripts");
2607 fs::create_dir_all(&src).expect("create src dir");
2608 fs::create_dir_all(&scripts).expect("create scripts dir");
2609 fs::write(src.join("main.rs"), "pub fn main() {}\n").expect("write src file");
2610 fs::write(scripts.join("tool.rs"), "pub fn tool() {}\n").expect("write scripts file");
2611
2612 let index = SearchIndex::build(&project);
2613 let files = index.glob("**/*.rs", &src);
2614
2615 assert_eq!(
2616 files,
2617 vec![fs::canonicalize(src.join("main.rs")).expect("canonicalize src file")]
2618 );
2619 }
2620
2621 #[test]
2622 fn glob_includes_hidden_and_binary_files() {
2623 let dir = tempfile::tempdir().expect("create temp dir");
2624 let project = dir.path().join("project");
2625 let hidden_dir = project.join(".hidden");
2626 fs::create_dir_all(&hidden_dir).expect("create hidden dir");
2627 let hidden_file = hidden_dir.join("data.bin");
2628 fs::write(&hidden_file, [0u8, 159, 146, 150]).expect("write binary file");
2629
2630 let index = SearchIndex::build(&project);
2631 let files = index.glob("**/*.bin", &project);
2632
2633 assert_eq!(
2634 files,
2635 vec![fs::canonicalize(hidden_file).expect("canonicalize binary file")]
2636 );
2637 }
2638
2639 #[test]
2640 fn read_from_disk_rejects_invalid_nanos() {
2641 let dir = tempfile::tempdir().expect("create temp dir");
2642 let cache_dir = dir.path().join("cache");
2643 fs::create_dir_all(&cache_dir).expect("create cache dir");
2644
2645 let mut postings = Vec::new();
2646 postings.extend_from_slice(INDEX_MAGIC);
2647 postings.extend_from_slice(&INDEX_VERSION.to_le_bytes());
2648 postings.extend_from_slice(&0u32.to_le_bytes());
2649 postings.extend_from_slice(&1u32.to_le_bytes());
2650 postings.extend_from_slice(&DEFAULT_MAX_FILE_SIZE.to_le_bytes());
2651 postings.extend_from_slice(&1u32.to_le_bytes());
2652 postings.extend_from_slice(b"/");
2653 postings.push(0u8);
2654 postings.extend_from_slice(&1u32.to_le_bytes());
2655 postings.extend_from_slice(&0u64.to_le_bytes());
2656 postings.extend_from_slice(&0u64.to_le_bytes());
2657 postings.extend_from_slice(&1_000_000_000u32.to_le_bytes());
2658 postings.extend_from_slice(b"a");
2659 postings.extend_from_slice(&0u64.to_le_bytes());
2660
2661 let mut lookup = Vec::new();
2662 lookup.extend_from_slice(LOOKUP_MAGIC);
2663 lookup.extend_from_slice(&INDEX_VERSION.to_le_bytes());
2664 lookup.extend_from_slice(&0u32.to_le_bytes());
2665
2666 let postings_checksum = crc32fast::hash(&postings);
2667 postings.extend_from_slice(&postings_checksum.to_le_bytes());
2668 let lookup_checksum = crc32fast::hash(&lookup);
2669 lookup.extend_from_slice(&lookup_checksum.to_le_bytes());
2670 let mut cache = Vec::new();
2671 cache.extend_from_slice(&CACHE_MAGIC.to_le_bytes());
2672 cache.extend_from_slice(&INDEX_VERSION.to_le_bytes());
2673 cache.extend_from_slice(&(postings.len() as u64).to_le_bytes());
2674 cache.extend_from_slice(&postings);
2675 cache.extend_from_slice(&lookup);
2676 fs::write(cache_dir.join("cache.bin"), cache).expect("write cache");
2677
2678 assert!(SearchIndex::read_from_disk(&cache_dir, dir.path()).is_none());
2679 }
2680
2681 #[test]
2696 fn sort_paths_by_mtime_desc_does_not_panic_on_missing_files() {
2697 let dir = tempfile::tempdir().expect("create tempdir");
2701 let mut paths: Vec<PathBuf> = Vec::new();
2702 for i in 0..30 {
2703 let path = if i % 2 == 0 {
2705 let p = dir.path().join(format!("real-{i}.rs"));
2706 fs::write(&p, format!("// {i}\n")).expect("write");
2707 p
2708 } else {
2709 dir.path().join(format!("missing-{i}.rs"))
2710 };
2711 paths.push(path);
2712 }
2713
2714 for _ in 0..50 {
2717 let mut copy = paths.clone();
2718 sort_paths_by_mtime_desc(&mut copy);
2719 assert_eq!(copy.len(), paths.len());
2720 }
2721 }
2722
2723 #[test]
2727 fn sort_grep_matches_by_mtime_desc_does_not_panic_on_missing_files() {
2728 let dir = tempfile::tempdir().expect("create tempdir");
2729 let mut matches: Vec<GrepMatch> = Vec::new();
2730 for i in 0..30 {
2731 let file = if i % 2 == 0 {
2732 let p = dir.path().join(format!("real-{i}.rs"));
2733 fs::write(&p, format!("// {i}\n")).expect("write");
2734 p
2735 } else {
2736 dir.path().join(format!("missing-{i}.rs"))
2737 };
2738 matches.push(GrepMatch {
2739 file,
2740 line: u32::try_from(i).unwrap_or(0),
2741 column: 0,
2742 line_text: format!("match {i}"),
2743 match_text: format!("match {i}"),
2744 });
2745 }
2746
2747 for _ in 0..50 {
2748 let mut copy = matches.clone();
2749 sort_grep_matches_by_mtime_desc(&mut copy, dir.path());
2750 assert_eq!(copy.len(), matches.len());
2751 }
2752 }
2753}