| 1 | //! Glob pattern expansion |
| 2 | //! |
| 3 | //! Supports: |
| 4 | //! - Basic patterns: *, ?, [abc], [a-z] |
| 5 | //! - Recursive globbing: ** |
| 6 | //! - Extended glob: !(pattern), ?(pattern), *(pattern), +(pattern), @(pattern) |
| 7 | //! - Dotfile handling |
| 8 | |
| 9 | use std::fs; |
| 10 | use std::path::{Path, PathBuf}; |
| 11 | use globset::{Glob, GlobBuilder, GlobMatcher}; |
| 12 | use regex::Regex; |
| 13 | |
| 14 | #[derive(Debug)] |
| 15 | pub struct GlobOptions { |
| 16 | /// Match dotfiles (files starting with .) |
| 17 | pub match_dotfiles: bool, |
| 18 | /// Allow recursive ** patterns |
| 19 | pub globstar: bool, |
| 20 | /// Return empty list if no matches (vs literal pattern) |
| 21 | pub nullglob: bool, |
| 22 | /// Enable extended glob patterns: !(pat), ?(pat), *(pat), +(pat), @(pat) |
| 23 | pub extglob: bool, |
| 24 | } |
| 25 | |
| 26 | impl Default for GlobOptions { |
| 27 | fn default() -> Self { |
| 28 | Self { |
| 29 | match_dotfiles: false, |
| 30 | globstar: true, |
| 31 | nullglob: false, |
| 32 | extglob: false, |
| 33 | } |
| 34 | } |
| 35 | } |
| 36 | |
| 37 | /// Extended glob pattern representation |
| 38 | #[derive(Debug)] |
| 39 | enum ExtGlobPattern { |
| 40 | /// No extended glob - use standard matching |
| 41 | Standard(String), |
| 42 | /// Negation pattern - requires two-stage filtering |
| 43 | Negation { |
| 44 | base_pattern: String, |
| 45 | exclude_regex: Regex, |
| 46 | }, |
| 47 | } |
| 48 | |
| 49 | /// Expand a glob pattern into matching file paths |
| 50 | pub fn expand_glob(pattern: &str, options: &GlobOptions) -> Result<Vec<String>, String> { |
| 51 | // Check for extended glob patterns first (before metacharacter check) |
| 52 | let is_extglob = options.extglob && ( |
| 53 | pattern.contains("!(") || pattern.contains("?(") || pattern.contains("*(") |
| 54 | || pattern.contains("+(") || pattern.contains("@(") |
| 55 | ); |
| 56 | |
| 57 | // Check if pattern contains glob metacharacters |
| 58 | // Skip this check for extglob patterns as they need processing even without metacharacters |
| 59 | if !is_extglob && !has_glob_chars(pattern) { |
| 60 | // No unescaped glob characters - strip backslash escapes and return as literal |
| 61 | return Ok(vec![strip_backslash_escapes(pattern)]); |
| 62 | } |
| 63 | |
| 64 | // Parse and convert extended glob patterns (if extglob is enabled) |
| 65 | let extglob_result = parse_extglob(pattern, options.extglob)?; |
| 66 | |
| 67 | match extglob_result { |
| 68 | ExtGlobPattern::Standard(base_pattern) => { |
| 69 | // Standard glob - no negation filtering needed |
| 70 | expand_standard_glob(&base_pattern, options) |
| 71 | } |
| 72 | ExtGlobPattern::Negation { base_pattern, exclude_regex } => { |
| 73 | // Two-stage filtering for negation patterns |
| 74 | // Stage 1: Expand base pattern to get candidates |
| 75 | let candidates = expand_standard_glob(&base_pattern, options)?; |
| 76 | |
| 77 | // Stage 2: Filter out matches of the exclude pattern |
| 78 | let filtered: Vec<String> = candidates |
| 79 | .into_iter() |
| 80 | .filter(|path| { |
| 81 | // Extract just the filename for matching |
| 82 | let filename = Path::new(path) |
| 83 | .file_name() |
| 84 | .and_then(|n| n.to_str()) |
| 85 | .unwrap_or(path); |
| 86 | |
| 87 | // Keep if it does NOT match the exclude pattern |
| 88 | !exclude_regex.is_match(filename) |
| 89 | }) |
| 90 | .collect(); |
| 91 | |
| 92 | // Handle no matches |
| 93 | if filtered.is_empty() && !options.nullglob { |
| 94 | Ok(vec![pattern.to_string()]) |
| 95 | } else { |
| 96 | Ok(filtered) |
| 97 | } |
| 98 | } |
| 99 | } |
| 100 | } |
| 101 | |
| 102 | /// Expand a standard (non-negated) glob pattern |
| 103 | fn expand_standard_glob(pattern: &str, options: &GlobOptions) -> Result<Vec<String>, String> { |
| 104 | // Build glob matcher |
| 105 | let glob = GlobBuilder::new(pattern) |
| 106 | .literal_separator(false) // Allow * to match / |
| 107 | .build() |
| 108 | .map_err(|e| format!("Invalid glob pattern: {}", e))?; |
| 109 | |
| 110 | let matcher = glob.compile_matcher(); |
| 111 | |
| 112 | // Get base directory and relative pattern |
| 113 | let (base_dir, rel_pattern) = split_pattern(pattern); |
| 114 | |
| 115 | // Expand the pattern |
| 116 | let mut matches = Vec::new(); |
| 117 | expand_pattern(&base_dir, &rel_pattern, &matcher, options, &mut matches)?; |
| 118 | |
| 119 | // Sort results for consistency |
| 120 | matches.sort(); |
| 121 | |
| 122 | // Handle no matches |
| 123 | if matches.is_empty() { |
| 124 | if options.nullglob { |
| 125 | Ok(vec![]) |
| 126 | } else { |
| 127 | // Return literal pattern if no matches |
| 128 | Ok(vec![pattern.to_string()]) |
| 129 | } |
| 130 | } else { |
| 131 | Ok(matches) |
| 132 | } |
| 133 | } |
| 134 | |
| 135 | /// Check if a pattern contains unescaped glob metacharacters or extended glob patterns |
| 136 | /// Backslash-escaped characters (\*, \?, etc.) are not counted as glob chars |
| 137 | fn has_glob_chars(s: &str) -> bool { |
| 138 | let mut chars = s.chars().peekable(); |
| 139 | while let Some(ch) = chars.next() { |
| 140 | if ch == '\\' { |
| 141 | // Skip the next character (it's escaped) |
| 142 | chars.next(); |
| 143 | continue; |
| 144 | } |
| 145 | if ch == '*' || ch == '?' || ch == '[' || ch == ']' { |
| 146 | return true; |
| 147 | } |
| 148 | } |
| 149 | // Check for extended glob patterns (they can't be easily escaped) |
| 150 | s.contains("!(") || s.contains("?(") || s.contains("*(") |
| 151 | || s.contains("+(") || s.contains("@(") |
| 152 | } |
| 153 | |
| 154 | /// Strip backslash escapes from a string (for returning literal patterns) |
| 155 | /// Only strips backslashes that escape glob metacharacters (* ? [ ]) |
| 156 | /// Preserves other backslashes for escape sequences like \n \t |
| 157 | fn strip_backslash_escapes(s: &str) -> String { |
| 158 | let mut result = String::with_capacity(s.len()); |
| 159 | let mut chars = s.chars().peekable(); |
| 160 | while let Some(ch) = chars.next() { |
| 161 | if ch == '\\' { |
| 162 | if let Some(&next) = chars.peek() { |
| 163 | // Only strip backslash if it's escaping a glob metacharacter |
| 164 | if next == '*' || next == '?' || next == '[' || next == ']' || next == '\\' { |
| 165 | result.push(chars.next().unwrap()); |
| 166 | } else { |
| 167 | // Preserve the backslash for other escape sequences |
| 168 | result.push(ch); |
| 169 | } |
| 170 | } else { |
| 171 | result.push(ch); |
| 172 | } |
| 173 | } else { |
| 174 | result.push(ch); |
| 175 | } |
| 176 | } |
| 177 | result |
| 178 | } |
| 179 | |
| 180 | /// Parse extended glob patterns and convert appropriately |
| 181 | fn parse_extglob(pattern: &str, extglob_enabled: bool) -> Result<ExtGlobPattern, String> { |
| 182 | // If extglob is disabled, treat everything as standard glob |
| 183 | if !extglob_enabled { |
| 184 | return Ok(ExtGlobPattern::Standard(pattern.to_string())); |
| 185 | } |
| 186 | |
| 187 | // Check for extended glob patterns: !(pat), ?(pat), *(pat), +(pat), @(pat) |
| 188 | if !pattern.contains("!(") && !pattern.contains("?(") && |
| 189 | !pattern.contains("*(") && !pattern.contains("+(") && !pattern.contains("@(") { |
| 190 | // No extended glob - return as standard |
| 191 | return Ok(ExtGlobPattern::Standard(pattern.to_string())); |
| 192 | } |
| 193 | |
| 194 | // Convert extglob - check if it contains negation |
| 195 | let (converted_pattern, has_negation, exclude_pattern) = convert_extglob(pattern)?; |
| 196 | |
| 197 | if has_negation { |
| 198 | // Build regex from the exclude pattern |
| 199 | let regex_pattern = glob_to_regex(&exclude_pattern)?; |
| 200 | let exclude_regex = Regex::new(®ex_pattern) |
| 201 | .map_err(|e| format!("Invalid regex pattern: {}", e))?; |
| 202 | |
| 203 | Ok(ExtGlobPattern::Negation { |
| 204 | base_pattern: converted_pattern, |
| 205 | exclude_regex, |
| 206 | }) |
| 207 | } else { |
| 208 | Ok(ExtGlobPattern::Standard(converted_pattern)) |
| 209 | } |
| 210 | } |
| 211 | |
| 212 | /// Convert glob pattern to regex pattern |
| 213 | fn glob_to_regex(pattern: &str) -> Result<String, String> { |
| 214 | let mut regex = String::from("^"); |
| 215 | |
| 216 | for ch in pattern.chars() { |
| 217 | match ch { |
| 218 | '*' => regex.push_str(".*"), |
| 219 | '?' => regex.push('.'), |
| 220 | '.' => regex.push_str("\\."), |
| 221 | '[' => regex.push('['), |
| 222 | ']' => regex.push(']'), |
| 223 | '(' => regex.push_str("\\("), |
| 224 | ')' => regex.push_str("\\)"), |
| 225 | '{' => regex.push_str("\\{"), |
| 226 | '}' => regex.push_str("\\}"), |
| 227 | '+' => regex.push_str("\\+"), |
| 228 | '^' => regex.push_str("\\^"), |
| 229 | '$' => regex.push_str("\\$"), |
| 230 | '|' => regex.push_str("\\|"), |
| 231 | '\\' => regex.push_str("\\\\"), |
| 232 | _ => regex.push(ch), |
| 233 | } |
| 234 | } |
| 235 | |
| 236 | regex.push('$'); |
| 237 | Ok(regex) |
| 238 | } |
| 239 | |
| 240 | /// Convert extended glob patterns to standard glob |
| 241 | /// Returns (converted_pattern, has_negation, exclude_pattern) |
| 242 | fn convert_extglob(pattern: &str) -> Result<(String, bool, String), String> { |
| 243 | let mut result = String::new(); |
| 244 | let mut chars = pattern.chars().peekable(); |
| 245 | let mut has_negation = false; |
| 246 | let mut exclude_pattern = String::new(); |
| 247 | |
| 248 | // For simplicity, we only handle a single !(pattern) at the top level |
| 249 | // Multiple or nested negations are not yet supported |
| 250 | |
| 251 | while let Some(ch) = chars.next() { |
| 252 | match ch { |
| 253 | '!' | '?' | '*' | '+' | '@' => { |
| 254 | if chars.peek() == Some(&'(') { |
| 255 | chars.next(); // consume '(' |
| 256 | |
| 257 | // Find matching ')' |
| 258 | let mut depth = 1; |
| 259 | let mut inner = String::new(); |
| 260 | while let Some(c) = chars.next() { |
| 261 | if c == '(' { |
| 262 | depth += 1; |
| 263 | inner.push(c); |
| 264 | } else if c == ')' { |
| 265 | depth -= 1; |
| 266 | if depth == 0 { |
| 267 | break; |
| 268 | } |
| 269 | inner.push(c); |
| 270 | } else { |
| 271 | inner.push(c); |
| 272 | } |
| 273 | } |
| 274 | |
| 275 | // Convert based on type |
| 276 | match ch { |
| 277 | '!' => { |
| 278 | // !(pattern) - negative match |
| 279 | // Use two-stage filtering: match all (*), then exclude pattern |
| 280 | result.push('*'); |
| 281 | has_negation = true; |
| 282 | exclude_pattern = inner; |
| 283 | } |
| 284 | '?' => { |
| 285 | // ?(pattern) - zero or one |
| 286 | // In glob: not directly supported, approximate with optional match |
| 287 | // For single char patterns, just make it optional |
| 288 | if inner.len() == 1 { |
| 289 | result.push('['); |
| 290 | result.push_str(&inner); |
| 291 | result.push(']'); |
| 292 | result.push('?'); |
| 293 | } else { |
| 294 | // Multi-char: can't express in glob, use pattern or nothing |
| 295 | // This is lossy but better than nothing |
| 296 | result.push('*'); |
| 297 | } |
| 298 | } |
| 299 | '*' => { |
| 300 | // *(pattern) - zero or more |
| 301 | // In glob: approximate with * |
| 302 | result.push('*'); |
| 303 | } |
| 304 | '+' => { |
| 305 | // +(pattern) - one or more |
| 306 | // In glob: approximate with pattern followed by * |
| 307 | result.push_str(&inner); |
| 308 | result.push('*'); |
| 309 | } |
| 310 | '@' => { |
| 311 | // @(pattern) - exactly one |
| 312 | result.push_str(&inner); |
| 313 | } |
| 314 | _ => unreachable!(), |
| 315 | } |
| 316 | } else { |
| 317 | result.push(ch); |
| 318 | } |
| 319 | } |
| 320 | _ => result.push(ch), |
| 321 | } |
| 322 | } |
| 323 | |
| 324 | Ok((result, has_negation, exclude_pattern)) |
| 325 | } |
| 326 | |
| 327 | /// Split pattern into base directory and relative pattern |
| 328 | fn split_pattern(pattern: &str) -> (PathBuf, String) { |
| 329 | let path = Path::new(pattern); |
| 330 | |
| 331 | // Find the first component with glob characters |
| 332 | let mut base = PathBuf::new(); |
| 333 | let mut rel_parts = Vec::new(); |
| 334 | let mut found_glob = false; |
| 335 | |
| 336 | for component in path.components() { |
| 337 | let comp_str = component.as_os_str().to_string_lossy(); |
| 338 | if !found_glob && !has_glob_chars(&comp_str) { |
| 339 | base.push(component); |
| 340 | } else { |
| 341 | found_glob = true; |
| 342 | rel_parts.push(comp_str.to_string()); |
| 343 | } |
| 344 | } |
| 345 | |
| 346 | let rel_pattern = if rel_parts.is_empty() { |
| 347 | pattern.to_string() |
| 348 | } else { |
| 349 | rel_parts.join("/") |
| 350 | }; |
| 351 | |
| 352 | if base.as_os_str().is_empty() { |
| 353 | base = PathBuf::from("."); |
| 354 | } |
| 355 | |
| 356 | (base, rel_pattern) |
| 357 | } |
| 358 | |
| 359 | /// Recursively expand a glob pattern |
| 360 | fn expand_pattern( |
| 361 | dir: &Path, |
| 362 | pattern: &str, |
| 363 | matcher: &GlobMatcher, |
| 364 | options: &GlobOptions, |
| 365 | results: &mut Vec<String>, |
| 366 | ) -> Result<(), String> { |
| 367 | // Handle ** recursive glob |
| 368 | if pattern.starts_with("**/") || pattern == "**" { |
| 369 | if options.globstar { |
| 370 | let rest = if pattern == "**" { |
| 371 | "*" |
| 372 | } else { |
| 373 | &pattern[3..] |
| 374 | }; |
| 375 | |
| 376 | // Recursively search all subdirectories |
| 377 | expand_recursive(dir, rest, matcher, options, results)?; |
| 378 | return Ok(()); |
| 379 | } |
| 380 | } |
| 381 | |
| 382 | // If pattern contains /, we need to handle directory traversal |
| 383 | if pattern.contains('/') { |
| 384 | let parts: Vec<&str> = pattern.split('/').collect(); |
| 385 | if parts.len() > 1 { |
| 386 | let first_part = parts[0]; |
| 387 | let rest = parts[1..].join("/"); |
| 388 | |
| 389 | // Read directory and match first part |
| 390 | let entries = fs::read_dir(dir) |
| 391 | .map_err(|e| format!("Failed to read directory {}: {}", dir.display(), e))?; |
| 392 | |
| 393 | for entry in entries { |
| 394 | let entry = entry.map_err(|e| format!("Failed to read entry: {}", e))?; |
| 395 | let path = entry.path(); |
| 396 | let file_name = entry.file_name(); |
| 397 | let name = file_name.to_string_lossy(); |
| 398 | |
| 399 | // Skip dotfiles unless enabled |
| 400 | if !options.match_dotfiles && name.starts_with('.') { |
| 401 | continue; |
| 402 | } |
| 403 | |
| 404 | // Match first part |
| 405 | if glob_match_simple(first_part, &name) && path.is_dir() { |
| 406 | expand_pattern(&path, &rest, matcher, options, results)?; |
| 407 | } |
| 408 | } |
| 409 | return Ok(()); |
| 410 | } |
| 411 | } |
| 412 | |
| 413 | // Simple case: pattern is just a filename pattern in current directory |
| 414 | let entries = fs::read_dir(dir) |
| 415 | .map_err(|e| format!("Failed to read directory {}: {}", dir.display(), e))?; |
| 416 | |
| 417 | for entry in entries { |
| 418 | let entry = entry.map_err(|e| format!("Failed to read entry: {}", e))?; |
| 419 | let path = entry.path(); |
| 420 | let file_name = entry.file_name(); |
| 421 | let name = file_name.to_string_lossy(); |
| 422 | |
| 423 | // Skip dotfiles unless enabled |
| 424 | if !options.match_dotfiles && name.starts_with('.') { |
| 425 | continue; |
| 426 | } |
| 427 | |
| 428 | // Match against pattern |
| 429 | if glob_match_simple(pattern, &name) { |
| 430 | results.push(path.to_string_lossy().to_string()); |
| 431 | } |
| 432 | } |
| 433 | |
| 434 | Ok(()) |
| 435 | } |
| 436 | |
| 437 | /// Recursively expand with ** pattern |
| 438 | fn expand_recursive( |
| 439 | dir: &Path, |
| 440 | pattern: &str, |
| 441 | matcher: &GlobMatcher, |
| 442 | options: &GlobOptions, |
| 443 | results: &mut Vec<String>, |
| 444 | ) -> Result<(), String> { |
| 445 | // Try to match in current directory |
| 446 | expand_pattern(dir, pattern, matcher, options, results)?; |
| 447 | |
| 448 | // Recurse into subdirectories |
| 449 | let entries = fs::read_dir(dir) |
| 450 | .map_err(|e| format!("Failed to read directory {}: {}", dir.display(), e))?; |
| 451 | |
| 452 | for entry in entries { |
| 453 | let entry = entry.map_err(|e| format!("Failed to read entry: {}", e))?; |
| 454 | let path = entry.path(); |
| 455 | |
| 456 | if path.is_dir() { |
| 457 | let file_name = entry.file_name(); |
| 458 | let name = file_name.to_string_lossy(); |
| 459 | |
| 460 | // Skip dotfiles unless enabled |
| 461 | if !options.match_dotfiles && name.starts_with('.') { |
| 462 | continue; |
| 463 | } |
| 464 | |
| 465 | // Recurse |
| 466 | expand_recursive(&path, pattern, matcher, options, results)?; |
| 467 | } |
| 468 | } |
| 469 | |
| 470 | Ok(()) |
| 471 | } |
| 472 | |
| 473 | /// Simple glob matching for a single path component |
| 474 | fn glob_match_simple(pattern: &str, text: &str) -> bool { |
| 475 | // Handle simple cases |
| 476 | if pattern == "*" { |
| 477 | return true; |
| 478 | } |
| 479 | |
| 480 | if pattern == text { |
| 481 | return true; |
| 482 | } |
| 483 | |
| 484 | // Use globset for more complex patterns |
| 485 | if let Ok(glob) = Glob::new(pattern) { |
| 486 | glob.compile_matcher().is_match(text) |
| 487 | } else { |
| 488 | false |
| 489 | } |
| 490 | } |
| 491 | |
| 492 | #[cfg(test)] |
| 493 | mod tests { |
| 494 | use super::*; |
| 495 | |
| 496 | #[test] |
| 497 | fn test_has_glob_chars() { |
| 498 | assert!(has_glob_chars("*.txt")); |
| 499 | assert!(has_glob_chars("file?.rs")); |
| 500 | assert!(has_glob_chars("[abc]")); |
| 501 | assert!(!has_glob_chars("plain.txt")); |
| 502 | } |
| 503 | |
| 504 | #[test] |
| 505 | fn test_glob_match_simple() { |
| 506 | assert!(glob_match_simple("*", "anything")); |
| 507 | assert!(glob_match_simple("*.txt", "file.txt")); |
| 508 | assert!(glob_match_simple("file", "file")); |
| 509 | assert!(!glob_match_simple("*.txt", "file.rs")); |
| 510 | } |
| 511 | } |
| 512 |