1use crate::{AgentTool, ToolCallEventStream};
2use agent_client_protocol as acp;
3use anyhow::{Result, anyhow};
4use futures::StreamExt;
5use gpui::{App, Entity, SharedString, Task};
6use language::{OffsetRangeExt, ParseStatus, Point};
7use project::{
8 Project, WorktreeSettings,
9 search::{SearchQuery, SearchResult},
10};
11use schemars::JsonSchema;
12use serde::{Deserialize, Serialize};
13use settings::Settings;
14use std::{cmp, fmt::Write, sync::Arc};
15use util::RangeExt;
16use util::markdown::MarkdownInlineCode;
17use util::paths::PathMatcher;
18
19/// Searches the contents of files in the project with a regular expression
20///
21/// - Prefer this tool to path search when searching for symbols in the project, because you won't need to guess what path it's in.
22/// - Supports full regex syntax (eg. "log.*Error", "function\\s+\\w+", etc.)
23/// - Pass an `include_pattern` if you know how to narrow your search on the files system
24/// - Never use this tool to search for paths. Only search file contents with this tool.
25/// - Use this tool when you need to find files containing specific patterns
26/// - Results are paginated with 20 matches per page. Use the optional 'offset' parameter to request subsequent pages.
27/// - DO NOT use HTML entities solely to escape characters in the tool parameters.
28#[derive(Debug, Serialize, Deserialize, JsonSchema)]
29pub struct GrepToolInput {
30 /// A regex pattern to search for in the entire project. Note that the regex will be parsed by the Rust `regex` crate.
31 ///
32 /// Do NOT specify a path here! This will only be matched against the code **content**.
33 pub regex: String,
34 /// A glob pattern for the paths of files to include in the search.
35 /// Supports standard glob patterns like "**/*.rs" or "src/**/*.ts".
36 /// If omitted, all files in the project will be searched.
37 pub include_pattern: Option<String>,
38 /// Optional starting position for paginated results (0-based).
39 /// When not provided, starts from the beginning.
40 #[serde(default)]
41 pub offset: u32,
42 /// Whether the regex is case-sensitive. Defaults to false (case-insensitive).
43 #[serde(default)]
44 pub case_sensitive: bool,
45}
46
47impl GrepToolInput {
48 /// Which page of search results this is.
49 pub fn page(&self) -> u32 {
50 1 + (self.offset / RESULTS_PER_PAGE)
51 }
52}
53
54const RESULTS_PER_PAGE: u32 = 20;
55
56pub struct GrepTool {
57 project: Entity<Project>,
58}
59
60impl GrepTool {
61 pub fn new(project: Entity<Project>) -> Self {
62 Self { project }
63 }
64}
65
66impl AgentTool for GrepTool {
67 type Input = GrepToolInput;
68 type Output = String;
69
70 fn name() -> &'static str {
71 "grep"
72 }
73
74 fn kind() -> acp::ToolKind {
75 acp::ToolKind::Search
76 }
77
78 fn initial_title(
79 &self,
80 input: Result<Self::Input, serde_json::Value>,
81 _cx: &mut App,
82 ) -> SharedString {
83 match input {
84 Ok(input) => {
85 let page = input.page();
86 let regex_str = MarkdownInlineCode(&input.regex);
87 let case_info = if input.case_sensitive {
88 " (case-sensitive)"
89 } else {
90 ""
91 };
92
93 if page > 1 {
94 format!("Get page {page} of search results for regex {regex_str}{case_info}")
95 } else {
96 format!("Search files for regex {regex_str}{case_info}")
97 }
98 }
99 Err(_) => "Search with regex".into(),
100 }
101 .into()
102 }
103
104 fn run(
105 self: Arc<Self>,
106 input: Self::Input,
107 _event_stream: ToolCallEventStream,
108 cx: &mut App,
109 ) -> Task<Result<Self::Output>> {
110 const CONTEXT_LINES: u32 = 2;
111 const MAX_ANCESTOR_LINES: u32 = 10;
112
113 let path_style = self.project.read(cx).path_style(cx);
114
115 let include_matcher = match PathMatcher::new(
116 input
117 .include_pattern
118 .as_ref()
119 .into_iter()
120 .collect::<Vec<_>>(),
121 path_style,
122 ) {
123 Ok(matcher) => matcher,
124 Err(error) => {
125 return Task::ready(Err(anyhow!("invalid include glob pattern: {error}")));
126 }
127 };
128
129 // Exclude global file_scan_exclusions and private_files settings
130 let exclude_matcher = {
131 let global_settings = WorktreeSettings::get_global(cx);
132 let exclude_patterns = global_settings
133 .file_scan_exclusions
134 .sources()
135 .iter()
136 .chain(global_settings.private_files.sources().iter());
137
138 match PathMatcher::new(exclude_patterns, path_style) {
139 Ok(matcher) => matcher,
140 Err(error) => {
141 return Task::ready(Err(anyhow!("invalid exclude pattern: {error}")));
142 }
143 }
144 };
145
146 let query = match SearchQuery::regex(
147 &input.regex,
148 false,
149 input.case_sensitive,
150 false,
151 false,
152 include_matcher,
153 exclude_matcher,
154 true, // Always match file include pattern against *full project paths* that start with a project root.
155 None,
156 ) {
157 Ok(query) => query,
158 Err(error) => return Task::ready(Err(error)),
159 };
160
161 let results = self
162 .project
163 .update(cx, |project, cx| project.search(query, cx));
164
165 let project = self.project.downgrade();
166 cx.spawn(async move |cx| {
167 futures::pin_mut!(results);
168
169 let mut output = String::new();
170 let mut skips_remaining = input.offset;
171 let mut matches_found = 0;
172 let mut has_more_matches = false;
173
174 'outer: while let Some(SearchResult::Buffer { buffer, ranges }) = results.next().await {
175 if ranges.is_empty() {
176 continue;
177 }
178
179 let Ok((Some(path), mut parse_status)) = buffer.read_with(cx, |buffer, cx| {
180 (buffer.file().map(|file| file.full_path(cx)), buffer.parse_status())
181 }) else {
182 continue;
183 };
184
185 // Check if this file should be excluded based on its worktree settings
186 if let Ok(Some(project_path)) = project.read_with(cx, |project, cx| {
187 project.find_project_path(&path, cx)
188 })
189 && cx.update(|cx| {
190 let worktree_settings = WorktreeSettings::get(Some((&project_path).into()), cx);
191 worktree_settings.is_path_excluded(&project_path.path)
192 || worktree_settings.is_path_private(&project_path.path)
193 }).unwrap_or(false) {
194 continue;
195 }
196
197 while *parse_status.borrow() != ParseStatus::Idle {
198 parse_status.changed().await?;
199 }
200
201 let snapshot = buffer.read_with(cx, |buffer, _cx| buffer.snapshot())?;
202
203 let mut ranges = ranges
204 .into_iter()
205 .map(|range| {
206 let matched = range.to_point(&snapshot);
207 let matched_end_line_len = snapshot.line_len(matched.end.row);
208 let full_lines = Point::new(matched.start.row, 0)..Point::new(matched.end.row, matched_end_line_len);
209 let symbols = snapshot.symbols_containing(matched.start, None);
210
211 if let Some(ancestor_node) = snapshot.syntax_ancestor(full_lines.clone()) {
212 let full_ancestor_range = ancestor_node.byte_range().to_point(&snapshot);
213 let end_row = full_ancestor_range.end.row.min(full_ancestor_range.start.row + MAX_ANCESTOR_LINES);
214 let end_col = snapshot.line_len(end_row);
215 let capped_ancestor_range = Point::new(full_ancestor_range.start.row, 0)..Point::new(end_row, end_col);
216
217 if capped_ancestor_range.contains_inclusive(&full_lines) {
218 return (capped_ancestor_range, Some(full_ancestor_range), symbols)
219 }
220 }
221
222 let mut matched = matched;
223 matched.start.column = 0;
224 matched.start.row =
225 matched.start.row.saturating_sub(CONTEXT_LINES);
226 matched.end.row = cmp::min(
227 snapshot.max_point().row,
228 matched.end.row + CONTEXT_LINES,
229 );
230 matched.end.column = snapshot.line_len(matched.end.row);
231
232 (matched, None, symbols)
233 })
234 .peekable();
235
236 let mut file_header_written = false;
237
238 while let Some((mut range, ancestor_range, parent_symbols)) = ranges.next(){
239 if skips_remaining > 0 {
240 skips_remaining -= 1;
241 continue;
242 }
243
244 // We'd already found a full page of matches, and we just found one more.
245 if matches_found >= RESULTS_PER_PAGE {
246 has_more_matches = true;
247 break 'outer;
248 }
249
250 while let Some((next_range, _, _)) = ranges.peek() {
251 if range.end.row >= next_range.start.row {
252 range.end = next_range.end;
253 ranges.next();
254 } else {
255 break;
256 }
257 }
258
259 if !file_header_written {
260 writeln!(output, "\n## Matches in {}", path.display())?;
261 file_header_written = true;
262 }
263
264 let end_row = range.end.row;
265 output.push_str("\n### ");
266
267 for symbol in parent_symbols {
268 write!(output, "{} › ", symbol.text)?;
269 }
270
271 if range.start.row == end_row {
272 writeln!(output, "L{}", range.start.row + 1)?;
273 } else {
274 writeln!(output, "L{}-{}", range.start.row + 1, end_row + 1)?;
275 }
276
277 output.push_str("```\n");
278 output.extend(snapshot.text_for_range(range));
279 output.push_str("\n```\n");
280
281 if let Some(ancestor_range) = ancestor_range
282 && end_row < ancestor_range.end.row {
283 let remaining_lines = ancestor_range.end.row - end_row;
284 writeln!(output, "\n{} lines remaining in ancestor node. Read the file to see all.", remaining_lines)?;
285 }
286
287 matches_found += 1;
288 }
289 }
290
291 if matches_found == 0 {
292 Ok("No matches found".into())
293 } else if has_more_matches {
294 Ok(format!(
295 "Showing matches {}-{} (there were more matches found; use offset: {} to see next page):\n{output}",
296 input.offset + 1,
297 input.offset + matches_found,
298 input.offset + RESULTS_PER_PAGE,
299 ))
300 } else {
301 Ok(format!("Found {matches_found} matches:\n{output}"))
302 }
303 })
304 }
305}
306
307#[cfg(test)]
308mod tests {
309 use crate::ToolCallEventStream;
310
311 use super::*;
312 use gpui::{TestAppContext, UpdateGlobal};
313 use language::{Language, LanguageConfig, LanguageMatcher};
314 use project::{FakeFs, Project};
315 use serde_json::json;
316 use settings::SettingsStore;
317 use unindent::Unindent;
318 use util::path;
319
320 #[gpui::test]
321 async fn test_grep_tool_with_include_pattern(cx: &mut TestAppContext) {
322 init_test(cx);
323 cx.executor().allow_parking();
324
325 let fs = FakeFs::new(cx.executor());
326 fs.insert_tree(
327 path!("/root"),
328 serde_json::json!({
329 "src": {
330 "main.rs": "fn main() {\n println!(\"Hello, world!\");\n}",
331 "utils": {
332 "helper.rs": "fn helper() {\n println!(\"I'm a helper!\");\n}",
333 },
334 },
335 "tests": {
336 "test_main.rs": "fn test_main() {\n assert!(true);\n}",
337 }
338 }),
339 )
340 .await;
341
342 let project = Project::test(fs.clone(), [path!("/root").as_ref()], cx).await;
343
344 // Test with include pattern for Rust files inside the root of the project
345 let input = GrepToolInput {
346 regex: "println".to_string(),
347 include_pattern: Some("root/**/*.rs".to_string()),
348 offset: 0,
349 case_sensitive: false,
350 };
351
352 let result = run_grep_tool(input, project.clone(), cx).await;
353 assert!(result.contains("main.rs"), "Should find matches in main.rs");
354 assert!(
355 result.contains("helper.rs"),
356 "Should find matches in helper.rs"
357 );
358 assert!(
359 !result.contains("test_main.rs"),
360 "Should not include test_main.rs even though it's a .rs file (because it doesn't have the pattern)"
361 );
362
363 // Test with include pattern for src directory only
364 let input = GrepToolInput {
365 regex: "fn".to_string(),
366 include_pattern: Some("root/**/src/**".to_string()),
367 offset: 0,
368 case_sensitive: false,
369 };
370
371 let result = run_grep_tool(input, project.clone(), cx).await;
372 assert!(
373 result.contains("main.rs"),
374 "Should find matches in src/main.rs"
375 );
376 assert!(
377 result.contains("helper.rs"),
378 "Should find matches in src/utils/helper.rs"
379 );
380 assert!(
381 !result.contains("test_main.rs"),
382 "Should not include test_main.rs as it's not in src directory"
383 );
384
385 // Test with empty include pattern (should default to all files)
386 let input = GrepToolInput {
387 regex: "fn".to_string(),
388 include_pattern: None,
389 offset: 0,
390 case_sensitive: false,
391 };
392
393 let result = run_grep_tool(input, project.clone(), cx).await;
394 assert!(result.contains("main.rs"), "Should find matches in main.rs");
395 assert!(
396 result.contains("helper.rs"),
397 "Should find matches in helper.rs"
398 );
399 assert!(
400 result.contains("test_main.rs"),
401 "Should include test_main.rs"
402 );
403 }
404
405 #[gpui::test]
406 async fn test_grep_tool_with_case_sensitivity(cx: &mut TestAppContext) {
407 init_test(cx);
408 cx.executor().allow_parking();
409
410 let fs = FakeFs::new(cx.executor());
411 fs.insert_tree(
412 path!("/root"),
413 serde_json::json!({
414 "case_test.txt": "This file has UPPERCASE and lowercase text.\nUPPERCASE patterns should match only with case_sensitive: true",
415 }),
416 )
417 .await;
418
419 let project = Project::test(fs.clone(), [path!("/root").as_ref()], cx).await;
420
421 // Test case-insensitive search (default)
422 let input = GrepToolInput {
423 regex: "uppercase".to_string(),
424 include_pattern: Some("**/*.txt".to_string()),
425 offset: 0,
426 case_sensitive: false,
427 };
428
429 let result = run_grep_tool(input, project.clone(), cx).await;
430 assert!(
431 result.contains("UPPERCASE"),
432 "Case-insensitive search should match uppercase"
433 );
434
435 // Test case-sensitive search
436 let input = GrepToolInput {
437 regex: "uppercase".to_string(),
438 include_pattern: Some("**/*.txt".to_string()),
439 offset: 0,
440 case_sensitive: true,
441 };
442
443 let result = run_grep_tool(input, project.clone(), cx).await;
444 assert!(
445 !result.contains("UPPERCASE"),
446 "Case-sensitive search should not match uppercase"
447 );
448
449 // Test case-sensitive search
450 let input = GrepToolInput {
451 regex: "LOWERCASE".to_string(),
452 include_pattern: Some("**/*.txt".to_string()),
453 offset: 0,
454 case_sensitive: true,
455 };
456
457 let result = run_grep_tool(input, project.clone(), cx).await;
458
459 assert!(
460 !result.contains("lowercase"),
461 "Case-sensitive search should match lowercase"
462 );
463
464 // Test case-sensitive search for lowercase pattern
465 let input = GrepToolInput {
466 regex: "lowercase".to_string(),
467 include_pattern: Some("**/*.txt".to_string()),
468 offset: 0,
469 case_sensitive: true,
470 };
471
472 let result = run_grep_tool(input, project.clone(), cx).await;
473 assert!(
474 result.contains("lowercase"),
475 "Case-sensitive search should match lowercase text"
476 );
477 }
478
479 /// Helper function to set up a syntax test environment
480 async fn setup_syntax_test(cx: &mut TestAppContext) -> Entity<Project> {
481 use unindent::Unindent;
482 init_test(cx);
483 cx.executor().allow_parking();
484
485 let fs = FakeFs::new(cx.executor());
486
487 // Create test file with syntax structures
488 fs.insert_tree(
489 path!("/root"),
490 serde_json::json!({
491 "test_syntax.rs": r#"
492 fn top_level_function() {
493 println!("This is at the top level");
494 }
495
496 mod feature_module {
497 pub mod nested_module {
498 pub fn nested_function(
499 first_arg: String,
500 second_arg: i32,
501 ) {
502 println!("Function in nested module");
503 println!("{first_arg}");
504 println!("{second_arg}");
505 }
506 }
507 }
508
509 struct MyStruct {
510 field1: String,
511 field2: i32,
512 }
513
514 impl MyStruct {
515 fn method_with_block() {
516 let condition = true;
517 if condition {
518 println!("Inside if block");
519 }
520 }
521
522 fn long_function() {
523 println!("Line 1");
524 println!("Line 2");
525 println!("Line 3");
526 println!("Line 4");
527 println!("Line 5");
528 println!("Line 6");
529 println!("Line 7");
530 println!("Line 8");
531 println!("Line 9");
532 println!("Line 10");
533 println!("Line 11");
534 println!("Line 12");
535 }
536 }
537
538 trait Processor {
539 fn process(&self, input: &str) -> String;
540 }
541
542 impl Processor for MyStruct {
543 fn process(&self, input: &str) -> String {
544 format!("Processed: {}", input)
545 }
546 }
547 "#.unindent().trim(),
548 }),
549 )
550 .await;
551
552 let project = Project::test(fs.clone(), [path!("/root").as_ref()], cx).await;
553
554 project.update(cx, |project, _cx| {
555 project.languages().add(rust_lang().into())
556 });
557
558 project
559 }
560
561 #[gpui::test]
562 async fn test_grep_top_level_function(cx: &mut TestAppContext) {
563 let project = setup_syntax_test(cx).await;
564
565 // Test: Line at the top level of the file
566 let input = GrepToolInput {
567 regex: "This is at the top level".to_string(),
568 include_pattern: Some("**/*.rs".to_string()),
569 offset: 0,
570 case_sensitive: false,
571 };
572
573 let result = run_grep_tool(input, project.clone(), cx).await;
574 let expected = r#"
575 Found 1 matches:
576
577 ## Matches in root/test_syntax.rs
578
579 ### fn top_level_function › L1-3
580 ```
581 fn top_level_function() {
582 println!("This is at the top level");
583 }
584 ```
585 "#
586 .unindent();
587 assert_eq!(result, expected);
588 }
589
590 #[gpui::test]
591 async fn test_grep_function_body(cx: &mut TestAppContext) {
592 let project = setup_syntax_test(cx).await;
593
594 // Test: Line inside a function body
595 let input = GrepToolInput {
596 regex: "Function in nested module".to_string(),
597 include_pattern: Some("**/*.rs".to_string()),
598 offset: 0,
599 case_sensitive: false,
600 };
601
602 let result = run_grep_tool(input, project.clone(), cx).await;
603 let expected = r#"
604 Found 1 matches:
605
606 ## Matches in root/test_syntax.rs
607
608 ### mod feature_module › pub mod nested_module › pub fn nested_function › L10-14
609 ```
610 ) {
611 println!("Function in nested module");
612 println!("{first_arg}");
613 println!("{second_arg}");
614 }
615 ```
616 "#
617 .unindent();
618 assert_eq!(result, expected);
619 }
620
621 #[gpui::test]
622 async fn test_grep_function_args_and_body(cx: &mut TestAppContext) {
623 let project = setup_syntax_test(cx).await;
624
625 // Test: Line with a function argument
626 let input = GrepToolInput {
627 regex: "second_arg".to_string(),
628 include_pattern: Some("**/*.rs".to_string()),
629 offset: 0,
630 case_sensitive: false,
631 };
632
633 let result = run_grep_tool(input, project.clone(), cx).await;
634 let expected = r#"
635 Found 1 matches:
636
637 ## Matches in root/test_syntax.rs
638
639 ### mod feature_module › pub mod nested_module › pub fn nested_function › L7-14
640 ```
641 pub fn nested_function(
642 first_arg: String,
643 second_arg: i32,
644 ) {
645 println!("Function in nested module");
646 println!("{first_arg}");
647 println!("{second_arg}");
648 }
649 ```
650 "#
651 .unindent();
652 assert_eq!(result, expected);
653 }
654
655 #[gpui::test]
656 async fn test_grep_if_block(cx: &mut TestAppContext) {
657 use unindent::Unindent;
658 let project = setup_syntax_test(cx).await;
659
660 // Test: Line inside an if block
661 let input = GrepToolInput {
662 regex: "Inside if block".to_string(),
663 include_pattern: Some("**/*.rs".to_string()),
664 offset: 0,
665 case_sensitive: false,
666 };
667
668 let result = run_grep_tool(input, project.clone(), cx).await;
669 let expected = r#"
670 Found 1 matches:
671
672 ## Matches in root/test_syntax.rs
673
674 ### impl MyStruct › fn method_with_block › L26-28
675 ```
676 if condition {
677 println!("Inside if block");
678 }
679 ```
680 "#
681 .unindent();
682 assert_eq!(result, expected);
683 }
684
685 #[gpui::test]
686 async fn test_grep_long_function_top(cx: &mut TestAppContext) {
687 use unindent::Unindent;
688 let project = setup_syntax_test(cx).await;
689
690 // Test: Line in the middle of a long function - should show message about remaining lines
691 let input = GrepToolInput {
692 regex: "Line 5".to_string(),
693 include_pattern: Some("**/*.rs".to_string()),
694 offset: 0,
695 case_sensitive: false,
696 };
697
698 let result = run_grep_tool(input, project.clone(), cx).await;
699 let expected = r#"
700 Found 1 matches:
701
702 ## Matches in root/test_syntax.rs
703
704 ### impl MyStruct › fn long_function › L31-41
705 ```
706 fn long_function() {
707 println!("Line 1");
708 println!("Line 2");
709 println!("Line 3");
710 println!("Line 4");
711 println!("Line 5");
712 println!("Line 6");
713 println!("Line 7");
714 println!("Line 8");
715 println!("Line 9");
716 println!("Line 10");
717 ```
718
719 3 lines remaining in ancestor node. Read the file to see all.
720 "#
721 .unindent();
722 assert_eq!(result, expected);
723 }
724
725 #[gpui::test]
726 async fn test_grep_long_function_bottom(cx: &mut TestAppContext) {
727 use unindent::Unindent;
728 let project = setup_syntax_test(cx).await;
729
730 // Test: Line in the long function
731 let input = GrepToolInput {
732 regex: "Line 12".to_string(),
733 include_pattern: Some("**/*.rs".to_string()),
734 offset: 0,
735 case_sensitive: false,
736 };
737
738 let result = run_grep_tool(input, project.clone(), cx).await;
739 let expected = r#"
740 Found 1 matches:
741
742 ## Matches in root/test_syntax.rs
743
744 ### impl MyStruct › fn long_function › L41-45
745 ```
746 println!("Line 10");
747 println!("Line 11");
748 println!("Line 12");
749 }
750 }
751 ```
752 "#
753 .unindent();
754 assert_eq!(result, expected);
755 }
756
757 async fn run_grep_tool(
758 input: GrepToolInput,
759 project: Entity<Project>,
760 cx: &mut TestAppContext,
761 ) -> String {
762 let tool = Arc::new(GrepTool { project });
763 let task = cx.update(|cx| tool.run(input, ToolCallEventStream::test().0, cx));
764
765 match task.await {
766 Ok(result) => {
767 if cfg!(windows) {
768 result.replace("root\\", "root/")
769 } else {
770 result
771 }
772 }
773 Err(e) => panic!("Failed to run grep tool: {}", e),
774 }
775 }
776
777 fn init_test(cx: &mut TestAppContext) {
778 cx.update(|cx| {
779 let settings_store = SettingsStore::test(cx);
780 cx.set_global(settings_store);
781 });
782 }
783
784 fn rust_lang() -> Language {
785 Language::new(
786 LanguageConfig {
787 name: "Rust".into(),
788 matcher: LanguageMatcher {
789 path_suffixes: vec!["rs".to_string()],
790 ..Default::default()
791 },
792 ..Default::default()
793 },
794 Some(tree_sitter_rust::LANGUAGE.into()),
795 )
796 .with_outline_query(include_str!("../../../languages/src/rust/outline.scm"))
797 .unwrap()
798 }
799
800 #[gpui::test]
801 async fn test_grep_security_boundaries(cx: &mut TestAppContext) {
802 init_test(cx);
803
804 let fs = FakeFs::new(cx.executor());
805
806 fs.insert_tree(
807 path!("/"),
808 json!({
809 "project_root": {
810 "allowed_file.rs": "fn main() { println!(\"This file is in the project\"); }",
811 ".mysecrets": "SECRET_KEY=abc123\nfn secret() { /* private */ }",
812 ".secretdir": {
813 "config": "fn special_configuration() { /* excluded */ }"
814 },
815 ".mymetadata": "fn custom_metadata() { /* excluded */ }",
816 "subdir": {
817 "normal_file.rs": "fn normal_file_content() { /* Normal */ }",
818 "special.privatekey": "fn private_key_content() { /* private */ }",
819 "data.mysensitive": "fn sensitive_data() { /* private */ }"
820 }
821 },
822 "outside_project": {
823 "sensitive_file.rs": "fn outside_function() { /* This file is outside the project */ }"
824 }
825 }),
826 )
827 .await;
828
829 cx.update(|cx| {
830 use gpui::UpdateGlobal;
831 use settings::SettingsStore;
832 SettingsStore::update_global(cx, |store, cx| {
833 store.update_user_settings(cx, |settings| {
834 settings.project.worktree.file_scan_exclusions = Some(vec![
835 "**/.secretdir".to_string(),
836 "**/.mymetadata".to_string(),
837 ]);
838 settings.project.worktree.private_files = Some(
839 vec![
840 "**/.mysecrets".to_string(),
841 "**/*.privatekey".to_string(),
842 "**/*.mysensitive".to_string(),
843 ]
844 .into(),
845 );
846 });
847 });
848 });
849
850 let project = Project::test(fs.clone(), [path!("/project_root").as_ref()], cx).await;
851
852 // Searching for files outside the project worktree should return no results
853 let result = run_grep_tool(
854 GrepToolInput {
855 regex: "outside_function".to_string(),
856 include_pattern: None,
857 offset: 0,
858 case_sensitive: false,
859 },
860 project.clone(),
861 cx,
862 )
863 .await;
864 let paths = extract_paths_from_results(&result);
865 assert!(
866 paths.is_empty(),
867 "grep_tool should not find files outside the project worktree"
868 );
869
870 // Searching within the project should succeed
871 let result = run_grep_tool(
872 GrepToolInput {
873 regex: "main".to_string(),
874 include_pattern: None,
875 offset: 0,
876 case_sensitive: false,
877 },
878 project.clone(),
879 cx,
880 )
881 .await;
882 let paths = extract_paths_from_results(&result);
883 assert!(
884 paths.iter().any(|p| p.contains("allowed_file.rs")),
885 "grep_tool should be able to search files inside worktrees"
886 );
887
888 // Searching files that match file_scan_exclusions should return no results
889 let result = run_grep_tool(
890 GrepToolInput {
891 regex: "special_configuration".to_string(),
892 include_pattern: None,
893 offset: 0,
894 case_sensitive: false,
895 },
896 project.clone(),
897 cx,
898 )
899 .await;
900 let paths = extract_paths_from_results(&result);
901 assert!(
902 paths.is_empty(),
903 "grep_tool should not search files in .secretdir (file_scan_exclusions)"
904 );
905
906 let result = run_grep_tool(
907 GrepToolInput {
908 regex: "custom_metadata".to_string(),
909 include_pattern: None,
910 offset: 0,
911 case_sensitive: false,
912 },
913 project.clone(),
914 cx,
915 )
916 .await;
917 let paths = extract_paths_from_results(&result);
918 assert!(
919 paths.is_empty(),
920 "grep_tool should not search .mymetadata files (file_scan_exclusions)"
921 );
922
923 // Searching private files should return no results
924 let result = run_grep_tool(
925 GrepToolInput {
926 regex: "SECRET_KEY".to_string(),
927 include_pattern: None,
928 offset: 0,
929 case_sensitive: false,
930 },
931 project.clone(),
932 cx,
933 )
934 .await;
935 let paths = extract_paths_from_results(&result);
936 assert!(
937 paths.is_empty(),
938 "grep_tool should not search .mysecrets (private_files)"
939 );
940
941 let result = run_grep_tool(
942 GrepToolInput {
943 regex: "private_key_content".to_string(),
944 include_pattern: None,
945 offset: 0,
946 case_sensitive: false,
947 },
948 project.clone(),
949 cx,
950 )
951 .await;
952 let paths = extract_paths_from_results(&result);
953
954 assert!(
955 paths.is_empty(),
956 "grep_tool should not search .privatekey files (private_files)"
957 );
958
959 let result = run_grep_tool(
960 GrepToolInput {
961 regex: "sensitive_data".to_string(),
962 include_pattern: None,
963 offset: 0,
964 case_sensitive: false,
965 },
966 project.clone(),
967 cx,
968 )
969 .await;
970 let paths = extract_paths_from_results(&result);
971 assert!(
972 paths.is_empty(),
973 "grep_tool should not search .mysensitive files (private_files)"
974 );
975
976 // Searching a normal file should still work, even with private_files configured
977 let result = run_grep_tool(
978 GrepToolInput {
979 regex: "normal_file_content".to_string(),
980 include_pattern: None,
981 offset: 0,
982 case_sensitive: false,
983 },
984 project.clone(),
985 cx,
986 )
987 .await;
988 let paths = extract_paths_from_results(&result);
989 assert!(
990 paths.iter().any(|p| p.contains("normal_file.rs")),
991 "Should be able to search normal files"
992 );
993
994 // Path traversal attempts with .. in include_pattern should not escape project
995 let result = run_grep_tool(
996 GrepToolInput {
997 regex: "outside_function".to_string(),
998 include_pattern: Some("../outside_project/**/*.rs".to_string()),
999 offset: 0,
1000 case_sensitive: false,
1001 },
1002 project.clone(),
1003 cx,
1004 )
1005 .await;
1006 let paths = extract_paths_from_results(&result);
1007 assert!(
1008 paths.is_empty(),
1009 "grep_tool should not allow escaping project boundaries with relative paths"
1010 );
1011 }
1012
1013 #[gpui::test]
1014 async fn test_grep_with_multiple_worktree_settings(cx: &mut TestAppContext) {
1015 init_test(cx);
1016
1017 let fs = FakeFs::new(cx.executor());
1018
1019 // Create first worktree with its own private files
1020 fs.insert_tree(
1021 path!("/worktree1"),
1022 json!({
1023 ".zed": {
1024 "settings.json": r#"{
1025 "file_scan_exclusions": ["**/fixture.*"],
1026 "private_files": ["**/secret.rs"]
1027 }"#
1028 },
1029 "src": {
1030 "main.rs": "fn main() { let secret_key = \"hidden\"; }",
1031 "secret.rs": "const API_KEY: &str = \"secret_value\";",
1032 "utils.rs": "pub fn get_config() -> String { \"config\".to_string() }"
1033 },
1034 "tests": {
1035 "test.rs": "fn test_secret() { assert!(true); }",
1036 "fixture.sql": "SELECT * FROM secret_table;"
1037 }
1038 }),
1039 )
1040 .await;
1041
1042 // Create second worktree with different private files
1043 fs.insert_tree(
1044 path!("/worktree2"),
1045 json!({
1046 ".zed": {
1047 "settings.json": r#"{
1048 "file_scan_exclusions": ["**/internal.*"],
1049 "private_files": ["**/private.js", "**/data.json"]
1050 }"#
1051 },
1052 "lib": {
1053 "public.js": "export function getSecret() { return 'public'; }",
1054 "private.js": "const SECRET_KEY = \"private_value\";",
1055 "data.json": "{\"secret_data\": \"hidden\"}"
1056 },
1057 "docs": {
1058 "README.md": "# Documentation with secret info",
1059 "internal.md": "Internal secret documentation"
1060 }
1061 }),
1062 )
1063 .await;
1064
1065 // Set global settings
1066 cx.update(|cx| {
1067 SettingsStore::update_global(cx, |store, cx| {
1068 store.update_user_settings(cx, |settings| {
1069 settings.project.worktree.file_scan_exclusions =
1070 Some(vec!["**/.git".to_string(), "**/node_modules".to_string()]);
1071 settings.project.worktree.private_files =
1072 Some(vec!["**/.env".to_string()].into());
1073 });
1074 });
1075 });
1076
1077 let project = Project::test(
1078 fs.clone(),
1079 [path!("/worktree1").as_ref(), path!("/worktree2").as_ref()],
1080 cx,
1081 )
1082 .await;
1083
1084 // Wait for worktrees to be fully scanned
1085 cx.executor().run_until_parked();
1086
1087 // Search for "secret" - should exclude files based on worktree-specific settings
1088 let result = run_grep_tool(
1089 GrepToolInput {
1090 regex: "secret".to_string(),
1091 include_pattern: None,
1092 offset: 0,
1093 case_sensitive: false,
1094 },
1095 project.clone(),
1096 cx,
1097 )
1098 .await;
1099 let paths = extract_paths_from_results(&result);
1100
1101 // Should find matches in non-private files
1102 assert!(
1103 paths.iter().any(|p| p.contains("main.rs")),
1104 "Should find 'secret' in worktree1/src/main.rs"
1105 );
1106 assert!(
1107 paths.iter().any(|p| p.contains("test.rs")),
1108 "Should find 'secret' in worktree1/tests/test.rs"
1109 );
1110 assert!(
1111 paths.iter().any(|p| p.contains("public.js")),
1112 "Should find 'secret' in worktree2/lib/public.js"
1113 );
1114 assert!(
1115 paths.iter().any(|p| p.contains("README.md")),
1116 "Should find 'secret' in worktree2/docs/README.md"
1117 );
1118
1119 // Should NOT find matches in private/excluded files based on worktree settings
1120 assert!(
1121 !paths.iter().any(|p| p.contains("secret.rs")),
1122 "Should not search in worktree1/src/secret.rs (local private_files)"
1123 );
1124 assert!(
1125 !paths.iter().any(|p| p.contains("fixture.sql")),
1126 "Should not search in worktree1/tests/fixture.sql (local file_scan_exclusions)"
1127 );
1128 assert!(
1129 !paths.iter().any(|p| p.contains("private.js")),
1130 "Should not search in worktree2/lib/private.js (local private_files)"
1131 );
1132 assert!(
1133 !paths.iter().any(|p| p.contains("data.json")),
1134 "Should not search in worktree2/lib/data.json (local private_files)"
1135 );
1136 assert!(
1137 !paths.iter().any(|p| p.contains("internal.md")),
1138 "Should not search in worktree2/docs/internal.md (local file_scan_exclusions)"
1139 );
1140
1141 // Test with `include_pattern` specific to one worktree
1142 let result = run_grep_tool(
1143 GrepToolInput {
1144 regex: "secret".to_string(),
1145 include_pattern: Some("worktree1/**/*.rs".to_string()),
1146 offset: 0,
1147 case_sensitive: false,
1148 },
1149 project.clone(),
1150 cx,
1151 )
1152 .await;
1153
1154 let paths = extract_paths_from_results(&result);
1155
1156 // Should only find matches in worktree1 *.rs files (excluding private ones)
1157 assert!(
1158 paths.iter().any(|p| p.contains("main.rs")),
1159 "Should find match in worktree1/src/main.rs"
1160 );
1161 assert!(
1162 paths.iter().any(|p| p.contains("test.rs")),
1163 "Should find match in worktree1/tests/test.rs"
1164 );
1165 assert!(
1166 !paths.iter().any(|p| p.contains("secret.rs")),
1167 "Should not find match in excluded worktree1/src/secret.rs"
1168 );
1169 assert!(
1170 paths.iter().all(|p| !p.contains("worktree2")),
1171 "Should not find any matches in worktree2"
1172 );
1173 }
1174
1175 // Helper function to extract file paths from grep results
1176 fn extract_paths_from_results(results: &str) -> Vec<String> {
1177 results
1178 .lines()
1179 .filter(|line| line.starts_with("## Matches in "))
1180 .map(|line| {
1181 line.strip_prefix("## Matches in ")
1182 .unwrap()
1183 .trim()
1184 .to_string()
1185 })
1186 .collect()
1187 }
1188}