1package tools
2
3import (
4 "bufio"
5 "context"
6 _ "embed"
7 "encoding/json"
8 "fmt"
9 "io"
10 "os"
11 "os/exec"
12 "path/filepath"
13 "regexp"
14 "sort"
15 "strconv"
16 "strings"
17 "sync"
18 "time"
19
20 "github.com/charmbracelet/crush/internal/fsext"
21)
22
23// regexCache provides thread-safe caching of compiled regex patterns
24type regexCache struct {
25 cache map[string]*regexp.Regexp
26 mu sync.RWMutex
27}
28
29// newRegexCache creates a new regex cache
30func newRegexCache() *regexCache {
31 return ®exCache{
32 cache: make(map[string]*regexp.Regexp),
33 }
34}
35
36// get retrieves a compiled regex from cache or compiles and caches it
37func (rc *regexCache) get(pattern string) (*regexp.Regexp, error) {
38 // Try to get from cache first (read lock)
39 rc.mu.RLock()
40 if regex, exists := rc.cache[pattern]; exists {
41 rc.mu.RUnlock()
42 return regex, nil
43 }
44 rc.mu.RUnlock()
45
46 // Compile the regex (write lock)
47 rc.mu.Lock()
48 defer rc.mu.Unlock()
49
50 // Double-check in case another goroutine compiled it while we waited
51 if regex, exists := rc.cache[pattern]; exists {
52 return regex, nil
53 }
54
55 // Compile and cache the regex
56 regex, err := regexp.Compile(pattern)
57 if err != nil {
58 return nil, err
59 }
60
61 rc.cache[pattern] = regex
62 return regex, nil
63}
64
65// Global regex cache instances
66var (
67 searchRegexCache = newRegexCache()
68 globRegexCache = newRegexCache()
69 // Pre-compiled regex for glob conversion (used frequently)
70 globBraceRegex = regexp.MustCompile(`\{([^}]+)\}`)
71)
72
73type GrepParams struct {
74 Pattern string `json:"pattern"`
75 Path string `json:"path"`
76 Include string `json:"include"`
77 LiteralText bool `json:"literal_text"`
78}
79
80type grepMatch struct {
81 path string
82 modTime time.Time
83 lineNum int
84 lineText string
85}
86
87type GrepResponseMetadata struct {
88 NumberOfMatches int `json:"number_of_matches"`
89 Truncated bool `json:"truncated"`
90}
91
92type grepTool struct {
93 workingDir string
94}
95
96const (
97 GrepToolName = "grep"
98 maxGrepContentWidth = 500
99)
100
101//go:embed grep.md
102var grepDescription []byte
103
104func NewGrepTool(workingDir string) BaseTool {
105 return &grepTool{
106 workingDir: workingDir,
107 }
108}
109
110func (g *grepTool) Name() string {
111 return GrepToolName
112}
113
114func (g *grepTool) Info() ToolInfo {
115 return ToolInfo{
116 Name: GrepToolName,
117 Description: string(grepDescription),
118 Parameters: map[string]any{
119 "pattern": map[string]any{
120 "type": "string",
121 "description": "The regex pattern to search for in file contents",
122 },
123 "path": map[string]any{
124 "type": "string",
125 "description": "The directory to search in. Defaults to the current working directory.",
126 },
127 "include": map[string]any{
128 "type": "string",
129 "description": "File pattern to include in the search (e.g. \"*.js\", \"*.{ts,tsx}\")",
130 },
131 "literal_text": map[string]any{
132 "type": "boolean",
133 "description": "If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.",
134 },
135 },
136 Required: []string{"pattern"},
137 }
138}
139
140// escapeRegexPattern escapes special regex characters so they're treated as literal characters
141func escapeRegexPattern(pattern string) string {
142 specialChars := []string{"\\", ".", "+", "*", "?", "(", ")", "[", "]", "{", "}", "^", "$", "|"}
143 escaped := pattern
144
145 for _, char := range specialChars {
146 escaped = strings.ReplaceAll(escaped, char, "\\"+char)
147 }
148
149 return escaped
150}
151
152func (g *grepTool) Run(ctx context.Context, call ToolCall) (ToolResponse, error) {
153 var params GrepParams
154 if err := json.Unmarshal([]byte(call.Input), ¶ms); err != nil {
155 return NewTextErrorResponse(fmt.Sprintf("error parsing parameters: %s", err)), nil
156 }
157
158 if params.Pattern == "" {
159 return NewTextErrorResponse("pattern is required"), nil
160 }
161
162 // If literal_text is true, escape the pattern
163 searchPattern := params.Pattern
164 if params.LiteralText {
165 searchPattern = escapeRegexPattern(params.Pattern)
166 }
167
168 searchPath := params.Path
169 if searchPath == "" {
170 searchPath = g.workingDir
171 }
172
173 matches, truncated, err := searchFiles(ctx, searchPattern, searchPath, params.Include, 100)
174 if err != nil {
175 return ToolResponse{}, fmt.Errorf("error searching files: %w", err)
176 }
177
178 var output strings.Builder
179 if len(matches) == 0 {
180 output.WriteString("No files found")
181 } else {
182 fmt.Fprintf(&output, "Found %d matches\n", len(matches))
183
184 currentFile := ""
185 for _, match := range matches {
186 if currentFile != match.path {
187 if currentFile != "" {
188 output.WriteString("\n")
189 }
190 currentFile = match.path
191 fmt.Fprintf(&output, "%s:\n", match.path)
192 }
193 if match.lineNum > 0 {
194 lineText := match.lineText
195 if len(lineText) > maxGrepContentWidth {
196 lineText = lineText[:maxGrepContentWidth] + "..."
197 }
198 fmt.Fprintf(&output, " Line %d: %s\n", match.lineNum, lineText)
199 } else {
200 fmt.Fprintf(&output, " %s\n", match.path)
201 }
202 }
203
204 if truncated {
205 output.WriteString("\n(Results are truncated. Consider using a more specific path or pattern.)")
206 }
207 }
208
209 return WithResponseMetadata(
210 NewTextResponse(output.String()),
211 GrepResponseMetadata{
212 NumberOfMatches: len(matches),
213 Truncated: truncated,
214 },
215 ), nil
216}
217
218func searchFiles(ctx context.Context, pattern, rootPath, include string, limit int) ([]grepMatch, bool, error) {
219 matches, err := searchWithRipgrep(ctx, getRgSearchCmd, pattern, rootPath, include)
220 if err != nil {
221 matches, err = searchFilesWithRegex(pattern, rootPath, include)
222 if err != nil {
223 return nil, false, err
224 }
225 }
226
227 sort.Slice(matches, func(i, j int) bool {
228 return matches[i].modTime.After(matches[j].modTime)
229 })
230
231 truncated := len(matches) > limit
232 if truncated {
233 matches = matches[:limit]
234 }
235
236 return matches, truncated, nil
237}
238
239// NOTE(tauraamui): ideally I would want to not pass in the search specific args here but will leave for now
240func searchWithRipgrep(ctx context.Context, rgSearchCmd resolveRgSearchCmd, pattern, path, include string) ([]grepMatch, error) {
241 cmd := rgSearchCmd(ctx, pattern, path, include)
242 if cmd == nil {
243 return nil, fmt.Errorf("ripgrep not found in $PATH")
244 }
245
246 // Only add ignore files if they exist
247 for _, ignoreFile := range []string{".gitignore", ".crushignore"} {
248 ignorePath := filepath.Join(path, ignoreFile)
249 if _, err := os.Stat(ignorePath); err == nil {
250 cmd.AddArgs("--ignore-file", ignorePath)
251 }
252 }
253
254 output, err := cmd.Output()
255 if err != nil {
256 if exitErr, ok := err.(*exec.ExitError); ok && exitErr.ExitCode() == 1 {
257 return []grepMatch{}, nil
258 }
259 return nil, err
260 }
261
262 lines := strings.Split(strings.TrimSpace(string(output)), "\n")
263 matches := make([]grepMatch, 0, len(lines))
264
265 for _, line := range lines {
266 if line == "" {
267 continue
268 }
269
270 // Parse ripgrep output using null separation
271 filePath, lineNumStr, lineText, ok := parseRipgrepLine(line)
272 if !ok {
273 continue
274 }
275
276 lineNum, err := strconv.Atoi(lineNumStr)
277 if err != nil {
278 continue
279 }
280
281 fileInfo, err := os.Stat(filePath)
282 if err != nil {
283 continue // Skip files we can't access
284 }
285
286 matches = append(matches, grepMatch{
287 path: filePath,
288 modTime: fileInfo.ModTime(),
289 lineNum: lineNum,
290 lineText: lineText,
291 })
292 }
293
294 return matches, nil
295}
296
297// parseRipgrepLine parses ripgrep output with null separation to handle Windows paths
298func parseRipgrepLine(line string) (filePath, lineNum, lineText string, ok bool) {
299 // Split on null byte first to separate filename from rest
300 parts := strings.SplitN(line, "\x00", 2)
301 if len(parts) != 2 {
302 return "", "", "", false
303 }
304
305 filePath = parts[0]
306 remainder := parts[1]
307
308 // Now split the remainder on first colon: "linenum:content"
309 colonIndex := strings.Index(remainder, ":")
310 if colonIndex == -1 {
311 return "", "", "", false
312 }
313
314 lineNumStr := remainder[:colonIndex]
315 lineText = remainder[colonIndex+1:]
316
317 if _, err := strconv.Atoi(lineNumStr); err != nil {
318 return "", "", "", false
319 }
320
321 return filePath, lineNumStr, lineText, true
322}
323
324func searchFilesWithRegex(pattern, rootPath, include string) ([]grepMatch, error) {
325 matches := []grepMatch{}
326
327 // Use cached regex compilation
328 regex, err := searchRegexCache.get(pattern)
329 if err != nil {
330 return nil, fmt.Errorf("invalid regex pattern: %w", err)
331 }
332
333 var includePattern *regexp.Regexp
334 if include != "" {
335 regexPattern := globToRegex(include)
336 includePattern, err = globRegexCache.get(regexPattern)
337 if err != nil {
338 return nil, fmt.Errorf("invalid include pattern: %w", err)
339 }
340 }
341
342 // Create walker with gitignore and crushignore support
343 walker := fsext.NewFastGlobWalker(rootPath)
344
345 err = filepath.Walk(rootPath, func(path string, info os.FileInfo, err error) error {
346 if err != nil {
347 return nil // Skip errors
348 }
349
350 if info.IsDir() {
351 // Check if directory should be skipped
352 if walker.ShouldSkip(path) {
353 return filepath.SkipDir
354 }
355 return nil // Continue into directory
356 }
357
358 // Use walker's shouldSkip method for files
359 if walker.ShouldSkip(path) {
360 return nil
361 }
362
363 // Skip hidden files (starting with a dot) to match ripgrep's default behavior
364 base := filepath.Base(path)
365 if base != "." && strings.HasPrefix(base, ".") {
366 return nil
367 }
368
369 if includePattern != nil && !includePattern.MatchString(path) {
370 return nil
371 }
372
373 match, lineNum, lineText, err := fileContainsPattern(path, regex)
374 if err != nil {
375 return nil // Skip files we can't read
376 }
377
378 if match {
379 matches = append(matches, grepMatch{
380 path: path,
381 modTime: info.ModTime(),
382 lineNum: lineNum,
383 lineText: lineText,
384 })
385
386 if len(matches) >= 200 {
387 return filepath.SkipAll
388 }
389 }
390
391 return nil
392 })
393 if err != nil {
394 return nil, err
395 }
396
397 return matches, nil
398}
399
400func fileContainsPattern(filePath string, pattern *regexp.Regexp) (bool, int, string, error) {
401 // Quick binary file detection
402 if isBinaryFile(filePath) {
403 return false, 0, "", nil
404 }
405
406 file, err := os.Open(filePath)
407 if err != nil {
408 return false, 0, "", err
409 }
410 defer file.Close()
411
412 scanner := bufio.NewScanner(file)
413 lineNum := 0
414 for scanner.Scan() {
415 lineNum++
416 line := scanner.Text()
417 if pattern.MatchString(line) {
418 return true, lineNum, line, nil
419 }
420 }
421
422 return false, 0, "", scanner.Err()
423}
424
425var binaryExts = map[string]struct{}{
426 ".exe": {}, ".dll": {}, ".so": {}, ".dylib": {},
427 ".bin": {}, ".obj": {}, ".o": {}, ".a": {},
428 ".zip": {}, ".tar": {}, ".gz": {}, ".bz2": {},
429 ".jpg": {}, ".jpeg": {}, ".png": {}, ".gif": {},
430 ".pdf": {}, ".doc": {}, ".docx": {}, ".xls": {},
431 ".mp3": {}, ".mp4": {}, ".avi": {}, ".mov": {},
432}
433
434// isBinaryFile performs a quick check to determine if a file is binary
435func isBinaryFile(filePath string) bool {
436 // Check file extension first (fastest)
437 ext := strings.ToLower(filepath.Ext(filePath))
438 if _, isBinary := binaryExts[ext]; isBinary {
439 return true
440 }
441
442 // Quick content check for files without clear extensions
443 file, err := os.Open(filePath)
444 if err != nil {
445 return false // If we can't open it, let the caller handle the error
446 }
447 defer file.Close()
448
449 // Read first 512 bytes to check for null bytes
450 buffer := make([]byte, 512)
451 n, err := file.Read(buffer)
452 if err != nil && err != io.EOF {
453 return false
454 }
455
456 // Check for null bytes (common in binary files)
457 for i := range n {
458 if buffer[i] == 0 {
459 return true
460 }
461 }
462
463 return false
464}
465
466func globToRegex(glob string) string {
467 regexPattern := strings.ReplaceAll(glob, ".", "\\.")
468 regexPattern = strings.ReplaceAll(regexPattern, "*", ".*")
469 regexPattern = strings.ReplaceAll(regexPattern, "?", ".")
470
471 // Use pre-compiled regex instead of compiling each time
472 regexPattern = globBraceRegex.ReplaceAllStringFunc(regexPattern, func(match string) string {
473 inner := match[1 : len(match)-1]
474 return "(" + strings.ReplaceAll(inner, ",", "|") + ")"
475 })
476
477 return regexPattern
478}