| 1 | ! ============================================================================== |
| 2 | ! Module: pipeline_helpers |
| 3 | ! ============================================================================== |
| 4 | ! Shared helpers for command expansion used by both executor and ast_executor. |
| 5 | ! Extracted from executor.f90 and parser.f90 to break the ast_executor → executor |
| 6 | ! dependency for pipeline execution. |
| 7 | ! |
| 8 | ! Functions: |
| 9 | ! - expand_tokens: Variable expansion, IFS field splitting, quote handling |
| 10 | ! - expand_command_globs: Glob pattern expansion for command tokens |
| 11 | ! - process_command_escapes: Post-glob backslash escape processing |
| 12 | ! - has_escaped_spaces: Check for backslash-escaped spaces |
| 13 | ! - interpret_ifs_escapes: Convert IFS escape sequences (\t, \n, \\) |
| 14 | module pipeline_helpers |
| 15 | use shell_types |
| 16 | use iso_fortran_env, only: error_unit |
| 17 | implicit none |
| 18 | private |
| 19 | |
| 20 | public :: expand_tokens |
| 21 | public :: expand_command_globs |
| 22 | public :: process_command_escapes |
| 23 | public :: has_escaped_spaces |
| 24 | public :: interpret_ifs_escapes |
| 25 | |
| 26 | contains |
| 27 | |
| 28 | function has_escaped_spaces(str) result(has_escaped) |
| 29 | character(len=*), intent(in) :: str |
| 30 | logical :: has_escaped |
| 31 | integer :: i, len_str |
| 32 | character(len=1) :: backslash |
| 33 | |
| 34 | has_escaped = .false. |
| 35 | len_str = len_trim(str) |
| 36 | backslash = char(92) ! ASCII code for backslash |
| 37 | |
| 38 | do i = 1, len_str - 1 |
| 39 | if (str(i:i) == backslash .and. str(i+1:i+1) == ' ') then |
| 40 | has_escaped = .true. |
| 41 | return |
| 42 | end if |
| 43 | end do |
| 44 | end function |
| 45 | |
| 46 | ! Interpret escape sequences in IFS string (\t -> tab, \n -> newline) |
| 47 | subroutine interpret_ifs_escapes(input, output, output_len) |
| 48 | character(len=*), intent(in) :: input |
| 49 | character(len=*), intent(out) :: output |
| 50 | integer, intent(out) :: output_len |
| 51 | integer :: i, j, input_len |
| 52 | character(len=1) :: backslash |
| 53 | |
| 54 | backslash = char(92) ! ASCII code for backslash |
| 55 | input_len = len(input) ! Use len(), not len_trim() - input might be all spaces (IFS=" ") |
| 56 | j = 1 |
| 57 | i = 1 |
| 58 | output = '' |
| 59 | |
| 60 | do while (i <= input_len) |
| 61 | if (input(i:i) == backslash .and. i < input_len) then |
| 62 | ! Check for escape sequences |
| 63 | if (input(i+1:i+1) == 't') then |
| 64 | ! \t -> tab |
| 65 | output(j:j) = char(9) |
| 66 | j = j + 1 |
| 67 | i = i + 2 |
| 68 | else if (input(i+1:i+1) == 'n') then |
| 69 | ! \n -> newline |
| 70 | output(j:j) = char(10) |
| 71 | j = j + 1 |
| 72 | i = i + 2 |
| 73 | else if (input(i+1:i+1) == backslash) then |
| 74 | ! \\ -> backslash |
| 75 | output(j:j) = backslash |
| 76 | j = j + 1 |
| 77 | i = i + 2 |
| 78 | else |
| 79 | ! Unknown escape, keep backslash and next char |
| 80 | output(j:j) = input(i:i) |
| 81 | j = j + 1 |
| 82 | i = i + 1 |
| 83 | end if |
| 84 | else |
| 85 | ! Regular character |
| 86 | output(j:j) = input(i:i) |
| 87 | j = j + 1 |
| 88 | i = i + 1 |
| 89 | end if |
| 90 | end do |
| 91 | output_len = j - 1 ! Return actual length of output |
| 92 | end subroutine |
| 93 | |
| 94 | subroutine expand_tokens(cmd, shell) |
| 95 | use expansion, only: field_split |
| 96 | type(command_t), intent(inout) :: cmd |
| 97 | type(shell_state_t), intent(inout) :: shell |
| 98 | integer :: i, j, total_tokens, temp_cap |
| 99 | character(len=:), allocatable :: expanded |
| 100 | type(string_t), allocatable :: temp_tokens(:) |
| 101 | integer, allocatable :: temp_token_lengths(:) ! Track actual lengths of expanded tokens |
| 102 | logical, allocatable :: temp_token_quoted(:) ! Track if original token was quoted |
| 103 | logical :: is_format_string |
| 104 | ! Heap-allocated to avoid static storage in recursive context |
| 105 | character(len=:), allocatable :: split_words(:) |
| 106 | integer :: split_cap, split_char_len |
| 107 | character(len=256) :: ifs_to_use |
| 108 | integer :: word_count, k, ifs_check_i, ifs_len_to_use |
| 109 | logical :: should_split, has_quotes, has_equals, has_escaped, has_ifs_char, ifs_explicitly_set |
| 110 | logical :: is_double_bracket_cmd |
| 111 | logical :: was_originally_quoted |
| 112 | |
| 113 | split_cap = 256 |
| 114 | split_char_len = MAX_TOKEN_LEN |
| 115 | allocate(character(len=split_char_len) :: split_words(split_cap)) |
| 116 | |
| 117 | ! Allocate temporary storage for expanded tokens |
| 118 | temp_cap = max(cmd%num_tokens * 10, 256) |
| 119 | allocate(temp_tokens(temp_cap)) |
| 120 | allocate(temp_token_lengths(temp_cap)) |
| 121 | allocate(temp_token_quoted(temp_cap)) |
| 122 | temp_token_lengths = 0 |
| 123 | temp_token_quoted = .false. |
| 124 | total_tokens = 0 |
| 125 | |
| 126 | ! Check if this is a [[ ]] command - word splitting is NOT performed in [[ ]] |
| 127 | ! per POSIX and bash behavior |
| 128 | is_double_bracket_cmd = .false. |
| 129 | if (cmd%num_tokens > 0) then |
| 130 | if (trim(cmd%tokens(1)) == '[[') then |
| 131 | is_double_bracket_cmd = .true. |
| 132 | end if |
| 133 | end if |
| 134 | |
| 135 | ! Determine IFS characters to use |
| 136 | ! Interpret escape sequences in IFS (\t -> tab, \n -> newline) |
| 137 | ! POSIX: When IFS is set to empty, field splitting is disabled |
| 138 | ! Check if IFS was explicitly set by user (in global variables or local scope) |
| 139 | ifs_explicitly_set = .false. |
| 140 | ! Check global variables array |
| 141 | do ifs_check_i = 1, shell%num_variables |
| 142 | if (trim(shell%variables(ifs_check_i)%name) == 'IFS') then |
| 143 | ifs_explicitly_set = .true. |
| 144 | exit |
| 145 | end if |
| 146 | end do |
| 147 | ! Also check local variable scope (for local IFS=...) |
| 148 | if (.not. ifs_explicitly_set .and. shell%function_depth > 0) then |
| 149 | block |
| 150 | integer :: lv_depth, lv_i |
| 151 | do lv_depth = shell%function_depth, 1, -1 |
| 152 | if (lv_depth <= size(shell%local_var_counts)) then |
| 153 | do lv_i = 1, shell%local_var_counts(lv_depth) |
| 154 | if (trim(shell%local_vars(lv_depth, lv_i)%name) == 'IFS') then |
| 155 | ifs_explicitly_set = .true. |
| 156 | exit |
| 157 | end if |
| 158 | end do |
| 159 | end if |
| 160 | if (ifs_explicitly_set) exit |
| 161 | end do |
| 162 | end block |
| 163 | end if |
| 164 | |
| 165 | if (ifs_explicitly_set) then |
| 166 | ! IFS is explicitly set - use its value (even if empty) |
| 167 | ! Use ifs_len to preserve trailing whitespace (e.g., IFS=" ") |
| 168 | if (shell%ifs_len > 0) then |
| 169 | call interpret_ifs_escapes(shell%ifs(1:shell%ifs_len), ifs_to_use, ifs_len_to_use) |
| 170 | else |
| 171 | ifs_to_use = '' ! Empty IFS disables field splitting |
| 172 | ifs_len_to_use = 0 |
| 173 | end if |
| 174 | else |
| 175 | ! IFS not set - use default |
| 176 | ifs_to_use = ' '//char(9)//char(10) ! space, tab, newline (default IFS) |
| 177 | ifs_len_to_use = 3 |
| 178 | end if |
| 179 | |
| 180 | do i = 1, cmd%num_tokens |
| 181 | ! Track if this token was originally quoted (for preserving trailing whitespace) |
| 182 | was_originally_quoted = .false. |
| 183 | if (allocated(cmd%token_quoted) .and. i <= size(cmd%token_quoted)) then |
| 184 | was_originally_quoted = cmd%token_quoted(i) |
| 185 | else if (allocated(cmd%token_quote_type) .and. i <= size(cmd%token_quote_type)) then |
| 186 | was_originally_quoted = (cmd%token_quote_type(i) == QUOTE_SINGLE .or. & |
| 187 | cmd%token_quote_type(i) == QUOTE_DOUBLE) |
| 188 | end if |
| 189 | |
| 190 | ! POSIX: Special handling for "$@" - expands to separate quoted arguments |
| 191 | ! When a double-quoted token is exactly $@ or just contains $@ as the whole expansion, |
| 192 | ! we need to add each positional parameter as a separate token |
| 193 | if (allocated(cmd%token_quote_type) .and. & |
| 194 | i <= size(cmd%token_quote_type) .and. & |
| 195 | cmd%token_quote_type(i) == QUOTE_DOUBLE .and. & |
| 196 | trim(cmd%tokens(i)) == '$@') then |
| 197 | ! "$@" - add each positional parameter as a separate quoted token |
| 198 | do j = 1, shell%num_positional |
| 199 | total_tokens = total_tokens + 1 |
| 200 | if (total_tokens > temp_cap) call grow_temp_arrays() |
| 201 | temp_tokens(total_tokens)%str = trim(shell%positional_params(j)%str) |
| 202 | temp_token_lengths(total_tokens) = len_trim(shell%positional_params(j)%str) |
| 203 | temp_token_quoted(total_tokens) = .true. ! Positional params from "$@" are quoted |
| 204 | end do |
| 205 | cycle ! Skip normal token processing |
| 206 | end if |
| 207 | |
| 208 | ! Check if this token was single-quoted (no expansion) |
| 209 | if (allocated(cmd%token_quote_type) .and. & |
| 210 | i <= size(cmd%token_quote_type) .and. & |
| 211 | cmd%token_quote_type(i) == QUOTE_SINGLE) then |
| 212 | ! Single quotes - no expansion, use literal value but strip sentinels/quotes |
| 213 | ! Lexer uses char(2) for start sentinel and char(3) for end sentinel |
| 214 | ! Old parser path uses actual quote characters |
| 215 | block |
| 216 | character(len=:), allocatable :: stripped |
| 217 | integer :: strip_j, strip_k, strip_len, start_pos, end_pos |
| 218 | ! Use token_lengths to preserve trailing spaces if available |
| 219 | if (allocated(cmd%token_lengths) .and. i <= size(cmd%token_lengths) .and. & |
| 220 | cmd%token_lengths(i) > 0) then |
| 221 | strip_len = cmd%token_lengths(i) |
| 222 | else |
| 223 | strip_len = len_trim(cmd%tokens(i)) |
| 224 | end if |
| 225 | |
| 226 | ! Determine start and end positions, skipping outer quotes if present |
| 227 | start_pos = 1 |
| 228 | end_pos = strip_len |
| 229 | if (strip_len >= 2) then |
| 230 | ! Check for actual quote characters at start and end (old parser path) |
| 231 | if (cmd%tokens(i)(1:1) == "'" .and. cmd%tokens(i)(strip_len:strip_len) == "'") then |
| 232 | start_pos = 2 |
| 233 | end_pos = strip_len - 1 |
| 234 | end if |
| 235 | end if |
| 236 | |
| 237 | allocate(character(len=end_pos - start_pos + 1) :: stripped) |
| 238 | strip_k = 1 |
| 239 | do strip_j = start_pos, end_pos |
| 240 | ! Skip single-quote sentinels (for new lexer path) |
| 241 | if (cmd%tokens(i)(strip_j:strip_j) /= char(2) .and. & |
| 242 | cmd%tokens(i)(strip_j:strip_j) /= char(3)) then |
| 243 | stripped(strip_k:strip_k) = cmd%tokens(i)(strip_j:strip_j) |
| 244 | strip_k = strip_k + 1 |
| 245 | end if |
| 246 | end do |
| 247 | if (strip_k > 1) then |
| 248 | expanded = stripped(1:strip_k-1) |
| 249 | else |
| 250 | expanded = '' |
| 251 | end if |
| 252 | end block |
| 253 | else |
| 254 | ! No quotes or double quotes - perform expansion |
| 255 | block |
| 256 | use parser, only: expand_variables |
| 257 | use expansion, only: expand_braces_to_words |
| 258 | logical :: is_double_quoted_token |
| 259 | is_double_quoted_token = .false. |
| 260 | if (allocated(cmd%token_quote_type) .and. & |
| 261 | i <= size(cmd%token_quote_type)) then |
| 262 | is_double_quoted_token = (cmd%token_quote_type(i) == QUOTE_DOUBLE) |
| 263 | end if |
| 264 | |
| 265 | if (is_double_quoted_token) then |
| 266 | ! Double-quoted: no brace expansion, preserve trailing whitespace |
| 267 | if (allocated(cmd%token_lengths) .and. i <= size(cmd%token_lengths)) then |
| 268 | if (cmd%token_lengths(i) > 0) then |
| 269 | call expand_variables(cmd%tokens(i)(1:cmd%token_lengths(i)), expanded, shell, was_quoted_in=.true.) |
| 270 | else |
| 271 | expanded = '' |
| 272 | end if |
| 273 | else |
| 274 | call expand_variables(cmd%tokens(i), expanded, shell, was_quoted_in=.true.) |
| 275 | end if |
| 276 | else if (index(cmd%tokens(i), '{') > 0 .and. index(cmd%tokens(i), '}') > 0 .and. & |
| 277 | index(cmd%tokens(i), '${') == 0) then |
| 278 | ! Unquoted token with braces (not ${...}): expand braces into |
| 279 | ! separate words first, then variable-expand each word individually. |
| 280 | ! This bypasses MAX_TOKEN_LEN limits and matches bash/zsh behavior. |
| 281 | block |
| 282 | character(len=MAX_TOKEN_LEN), allocatable :: brace_words(:) |
| 283 | character(len=:), allocatable :: var_expanded |
| 284 | integer :: bw_count, bw_i |
| 285 | ! Fast-path variables for range detection |
| 286 | integer :: br_open, br_close, br_dots, br_dots2, br_step |
| 287 | integer :: br_start, br_end, br_cur, br_count |
| 288 | integer :: br_sc, br_ec, br_cc |
| 289 | character(len=:), allocatable :: br_prefix, br_suffix |
| 290 | character(len=16) :: br_num |
| 291 | integer :: br_nlen, br_plen, br_slen |
| 292 | logical :: br_fast |
| 293 | |
| 294 | br_fast = .false. |
| 295 | ! Try fast path: simple {N..M} or {N..M..S} or {a..z} with optional prefix/suffix |
| 296 | ! Must have exactly one { and one } with no nesting or $ |
| 297 | br_open = index(trim(cmd%tokens(i)), '{') |
| 298 | br_close = index(trim(cmd%tokens(i)), '}', back=.true.) |
| 299 | if (br_open > 0 .and. br_close > br_open .and. & |
| 300 | index(cmd%tokens(i)(br_open+1:br_close-1), '{') == 0 .and. & |
| 301 | index(cmd%tokens(i)(br_open+1:br_close-1), ',') == 0) then |
| 302 | br_dots = index(cmd%tokens(i)(br_open+1:br_close-1), '..') |
| 303 | if (br_dots > 0) then |
| 304 | br_dots = br_open + br_dots ! absolute position |
| 305 | ! Extract prefix and suffix |
| 306 | if (br_open > 1) then |
| 307 | br_prefix = cmd%tokens(i)(1:br_open-1) |
| 308 | else |
| 309 | br_prefix = '' |
| 310 | end if |
| 311 | if (br_close < len_trim(cmd%tokens(i))) then |
| 312 | br_suffix = cmd%tokens(i)(br_close+1:len_trim(cmd%tokens(i))) |
| 313 | else |
| 314 | br_suffix = '' |
| 315 | end if |
| 316 | br_plen = len(br_prefix) |
| 317 | br_slen = len(br_suffix) |
| 318 | |
| 319 | ! Check for step: second .. after the first |
| 320 | br_dots2 = index(cmd%tokens(i)(br_dots+2:br_close-1), '..') |
| 321 | br_step = 1 |
| 322 | if (br_dots2 > 0) then |
| 323 | br_dots2 = br_dots + 1 + br_dots2 ! absolute position |
| 324 | read(cmd%tokens(i)(br_dots2+2:br_close-1), *, iostat=bw_i) br_step |
| 325 | if (bw_i /= 0) br_step = 1 |
| 326 | else |
| 327 | br_dots2 = 0 |
| 328 | end if |
| 329 | |
| 330 | ! Try numeric range |
| 331 | if (br_dots2 > 0) then |
| 332 | read(cmd%tokens(i)(br_open+1:br_dots-1), *, iostat=bw_i) br_start |
| 333 | else |
| 334 | read(cmd%tokens(i)(br_open+1:br_dots-1), *, iostat=bw_i) br_start |
| 335 | end if |
| 336 | if (bw_i == 0) then |
| 337 | if (br_dots2 > 0) then |
| 338 | read(cmd%tokens(i)(br_dots+2:br_dots2-1), *, iostat=bw_i) br_end |
| 339 | else |
| 340 | read(cmd%tokens(i)(br_dots+2:br_close-1), *, iostat=bw_i) br_end |
| 341 | end if |
| 342 | end if |
| 343 | |
| 344 | if (bw_i == 0) then |
| 345 | ! Numeric range — generate tokens directly |
| 346 | br_fast = .true. |
| 347 | if (br_start <= br_end) then |
| 348 | br_count = (br_end - br_start) / br_step + 1 |
| 349 | else |
| 350 | br_count = (br_start - br_end) / br_step + 1 |
| 351 | end if |
| 352 | ! Pre-grow to avoid repeated doubling |
| 353 | if (total_tokens + br_count > temp_cap) & |
| 354 | call grow_temp_arrays(total_tokens + br_count) |
| 355 | br_cur = br_start |
| 356 | if (br_start <= br_end) then |
| 357 | do while (br_cur <= br_end) |
| 358 | write(br_num, '(I0)') br_cur |
| 359 | br_nlen = len_trim(br_num) |
| 360 | total_tokens = total_tokens + 1 |
| 361 | if (br_plen > 0 .and. br_slen > 0) then |
| 362 | temp_tokens(total_tokens)%str = br_prefix // br_num(1:br_nlen) // br_suffix |
| 363 | else if (br_plen > 0) then |
| 364 | temp_tokens(total_tokens)%str = br_prefix // br_num(1:br_nlen) |
| 365 | else if (br_slen > 0) then |
| 366 | temp_tokens(total_tokens)%str = br_num(1:br_nlen) // br_suffix |
| 367 | else |
| 368 | temp_tokens(total_tokens)%str = br_num(1:br_nlen) |
| 369 | end if |
| 370 | temp_token_lengths(total_tokens) = br_plen + br_nlen + br_slen |
| 371 | temp_token_quoted(total_tokens) = .false. |
| 372 | br_cur = br_cur + br_step |
| 373 | end do |
| 374 | else |
| 375 | do while (br_cur >= br_end) |
| 376 | write(br_num, '(I0)') br_cur |
| 377 | br_nlen = len_trim(br_num) |
| 378 | total_tokens = total_tokens + 1 |
| 379 | if (br_plen > 0 .and. br_slen > 0) then |
| 380 | temp_tokens(total_tokens)%str = br_prefix // br_num(1:br_nlen) // br_suffix |
| 381 | else if (br_plen > 0) then |
| 382 | temp_tokens(total_tokens)%str = br_prefix // br_num(1:br_nlen) |
| 383 | else if (br_slen > 0) then |
| 384 | temp_tokens(total_tokens)%str = br_num(1:br_nlen) // br_suffix |
| 385 | else |
| 386 | temp_tokens(total_tokens)%str = br_num(1:br_nlen) |
| 387 | end if |
| 388 | temp_token_lengths(total_tokens) = br_plen + br_nlen + br_slen |
| 389 | temp_token_quoted(total_tokens) = .false. |
| 390 | br_cur = br_cur - br_step |
| 391 | end do |
| 392 | end if |
| 393 | cycle |
| 394 | else if (len_trim(cmd%tokens(i)(br_open+1:br_dots-1)) == 1 .and. & |
| 395 | ((br_dots2 > 0 .and. len_trim(cmd%tokens(i)(br_dots+2:br_dots2-1)) == 1) .or. & |
| 396 | (br_dots2 == 0 .and. len_trim(cmd%tokens(i)(br_dots+2:br_close-1)) == 1))) then |
| 397 | ! Alpha range — generate tokens directly |
| 398 | br_fast = .true. |
| 399 | br_sc = ichar(cmd%tokens(i)(br_open+1:br_open+1)) |
| 400 | if (br_dots2 > 0) then |
| 401 | br_ec = ichar(cmd%tokens(i)(br_dots+2:br_dots+2)) |
| 402 | else |
| 403 | br_ec = ichar(cmd%tokens(i)(br_dots+2:br_dots+2)) |
| 404 | end if |
| 405 | if (br_sc <= br_ec) then |
| 406 | br_count = (br_ec - br_sc) / br_step + 1 |
| 407 | else |
| 408 | br_count = (br_sc - br_ec) / br_step + 1 |
| 409 | end if |
| 410 | if (total_tokens + br_count > temp_cap) & |
| 411 | call grow_temp_arrays(total_tokens + br_count) |
| 412 | br_cc = br_sc |
| 413 | if (br_sc <= br_ec) then |
| 414 | do while (br_cc <= br_ec) |
| 415 | total_tokens = total_tokens + 1 |
| 416 | if (br_plen > 0 .and. br_slen > 0) then |
| 417 | temp_tokens(total_tokens)%str = br_prefix // char(br_cc) // br_suffix |
| 418 | else if (br_plen > 0) then |
| 419 | temp_tokens(total_tokens)%str = br_prefix // char(br_cc) |
| 420 | else if (br_slen > 0) then |
| 421 | temp_tokens(total_tokens)%str = char(br_cc) // br_suffix |
| 422 | else |
| 423 | temp_tokens(total_tokens)%str = char(br_cc) |
| 424 | end if |
| 425 | temp_token_lengths(total_tokens) = br_plen + 1 + br_slen |
| 426 | temp_token_quoted(total_tokens) = .false. |
| 427 | br_cc = br_cc + br_step |
| 428 | end do |
| 429 | else |
| 430 | do while (br_cc >= br_ec) |
| 431 | total_tokens = total_tokens + 1 |
| 432 | if (br_plen > 0 .and. br_slen > 0) then |
| 433 | temp_tokens(total_tokens)%str = br_prefix // char(br_cc) // br_suffix |
| 434 | else if (br_plen > 0) then |
| 435 | temp_tokens(total_tokens)%str = br_prefix // char(br_cc) |
| 436 | else if (br_slen > 0) then |
| 437 | temp_tokens(total_tokens)%str = char(br_cc) // br_suffix |
| 438 | else |
| 439 | temp_tokens(total_tokens)%str = char(br_cc) |
| 440 | end if |
| 441 | temp_token_lengths(total_tokens) = br_plen + 1 + br_slen |
| 442 | temp_token_quoted(total_tokens) = .false. |
| 443 | br_cc = br_cc - br_step |
| 444 | end do |
| 445 | end if |
| 446 | cycle |
| 447 | end if |
| 448 | end if |
| 449 | end if |
| 450 | |
| 451 | ! General brace expansion fallback (comma lists, nested braces) |
| 452 | if (.not. br_fast) then |
| 453 | call expand_braces_to_words(trim(cmd%tokens(i)), brace_words, bw_count) |
| 454 | |
| 455 | if (bw_count > 1) then |
| 456 | ! Multiple words from brace expansion — add each as a separate token |
| 457 | do bw_i = 1, bw_count |
| 458 | ! Expand variables in each brace word individually |
| 459 | if (index(brace_words(bw_i), '$') > 0 .or. & |
| 460 | index(brace_words(bw_i), '`') > 0) then |
| 461 | call expand_variables(trim(brace_words(bw_i)), var_expanded, shell, was_quoted_in=.false.) |
| 462 | total_tokens = total_tokens + 1 |
| 463 | if (total_tokens > temp_cap) call grow_temp_arrays() |
| 464 | if (allocated(var_expanded)) then |
| 465 | temp_tokens(total_tokens)%str = var_expanded |
| 466 | temp_token_lengths(total_tokens) = len(var_expanded) |
| 467 | else |
| 468 | temp_tokens(total_tokens)%str = brace_words(bw_i) |
| 469 | temp_token_lengths(total_tokens) = len_trim(brace_words(bw_i)) |
| 470 | end if |
| 471 | else |
| 472 | total_tokens = total_tokens + 1 |
| 473 | if (total_tokens > temp_cap) call grow_temp_arrays() |
| 474 | temp_tokens(total_tokens)%str = brace_words(bw_i) |
| 475 | temp_token_lengths(total_tokens) = len_trim(brace_words(bw_i)) |
| 476 | end if |
| 477 | temp_token_quoted(total_tokens) = .false. |
| 478 | end do |
| 479 | deallocate(brace_words) |
| 480 | cycle ! All words already added as separate tokens |
| 481 | else |
| 482 | ! Single word — fall through to normal variable expansion |
| 483 | expanded = trim(brace_words(1)) |
| 484 | deallocate(brace_words) |
| 485 | if (index(expanded, '$') > 0 .or. index(expanded, '`') > 0) then |
| 486 | call expand_variables(expanded, var_expanded, shell, was_quoted_in=.false.) |
| 487 | if (allocated(var_expanded)) expanded = var_expanded |
| 488 | end if |
| 489 | end if |
| 490 | end if |
| 491 | end block |
| 492 | else |
| 493 | ! No braces — standard variable expansion |
| 494 | call expand_variables(cmd%tokens(i), expanded, shell, was_quoted_in=.false.) |
| 495 | end if |
| 496 | end block |
| 497 | end if |
| 498 | |
| 499 | ! Determine if we should split this token on IFS characters |
| 500 | ! Only split if: |
| 501 | ! 1. Contains IFS characters |
| 502 | ! 2. NOT quoted (doesn't contain quote characters) |
| 503 | ! 3. NOT an assignment (doesn't contain =, like alias ll='...' or var=value) |
| 504 | ! 4. NOT escaped (doesn't contain escaped IFS chars) |
| 505 | should_split = .false. |
| 506 | |
| 507 | ! Check if expanded string contains any IFS character |
| 508 | has_ifs_char = .false. |
| 509 | if (ifs_len_to_use > 0) then |
| 510 | do k = 1, len(expanded) |
| 511 | ! Only check against actual IFS chars (first ifs_len_to_use chars of ifs_to_use) |
| 512 | if (index(ifs_to_use(1:ifs_len_to_use), expanded(k:k)) > 0) then |
| 513 | has_ifs_char = .true. |
| 514 | exit |
| 515 | end if |
| 516 | end do |
| 517 | end if |
| 518 | ! If ifs_len_to_use == 0 (empty IFS), has_ifs_char stays false, disabling field splitting |
| 519 | |
| 520 | if (has_ifs_char) then |
| 521 | ! Check if ORIGINAL token was quoted (using metadata, not looking for quotes in string) |
| 522 | if (allocated(cmd%token_quoted) .and. i <= size(cmd%token_quoted)) then |
| 523 | has_quotes = cmd%token_quoted(i) |
| 524 | else |
| 525 | ! Fallback: Check if ORIGINAL token had quotes (not expanded, since expand_variables strips them) |
| 526 | has_quotes = (index(cmd%tokens(i), '"') > 0 .or. index(cmd%tokens(i), "'") > 0) |
| 527 | end if |
| 528 | ! Check if it's an assignment (contains =) |
| 529 | has_equals = (index(expanded, '=') > 0) |
| 530 | ! Check if spaces are escaped with backslash in ORIGINAL token |
| 531 | has_escaped = has_escaped_spaces(cmd%tokens(i)) |
| 532 | ! PARSER FIX: Check if token starts with % (printf format string) |
| 533 | is_format_string = (len_trim(expanded) > 0 .and. expanded(1:1) == '%') |
| 534 | |
| 535 | ! Only split if no quotes, no equals sign, no escaped spaces, not a format string, |
| 536 | ! and NOT inside a [[ ]] expression (word splitting is disabled in [[ ]]) |
| 537 | should_split = (.not. has_quotes .and. .not. has_equals .and. .not. has_escaped & |
| 538 | .and. .not. is_format_string .and. .not. is_double_bracket_cmd) |
| 539 | end if |
| 540 | |
| 541 | if (should_split) then |
| 542 | ! Split the expanded string using IFS characters |
| 543 | word_count = 0 |
| 544 | ! Grow split_words if expanded string needs more capacity or longer char length |
| 545 | if (allocated(expanded)) then |
| 546 | if (len(expanded) / 2 + 1 > split_cap .or. len(expanded) > split_char_len) then |
| 547 | deallocate(split_words) |
| 548 | split_cap = max(split_cap, len(expanded) / 2 + 1) |
| 549 | split_char_len = max(split_char_len, len(expanded)) |
| 550 | allocate(character(len=split_char_len) :: split_words(split_cap)) |
| 551 | end if |
| 552 | end if |
| 553 | ! Pass ifs_to_use with exact length - use substring to avoid trailing blanks |
| 554 | if (ifs_len_to_use > 0) then |
| 555 | call field_split(expanded, ifs_to_use(1:ifs_len_to_use), split_words, word_count) |
| 556 | else |
| 557 | ! Empty IFS - no splitting should happen (but we shouldn't reach here) |
| 558 | split_words(1) = expanded |
| 559 | word_count = 1 |
| 560 | end if |
| 561 | |
| 562 | ! Add all split words as separate tokens |
| 563 | do j = 1, word_count |
| 564 | total_tokens = total_tokens + 1 |
| 565 | if (total_tokens > temp_cap) call grow_temp_arrays() |
| 566 | temp_tokens(total_tokens)%str = split_words(j) |
| 567 | temp_token_lengths(total_tokens) = len_trim(split_words(j)) |
| 568 | ! Split tokens from unquoted expansion are not quoted |
| 569 | temp_token_quoted(total_tokens) = .false. |
| 570 | end do |
| 571 | else |
| 572 | ! No IFS chars or shouldn't split, just add as single token |
| 573 | ! POSIX: Skip empty tokens from unquoted variable expansion |
| 574 | ! Only keep empty strings if the original token was quoted |
| 575 | if (len_trim(expanded) == 0) then |
| 576 | ! Check if original token was quoted |
| 577 | if (allocated(cmd%token_quoted) .and. i <= size(cmd%token_quoted)) then |
| 578 | if (.not. cmd%token_quoted(i)) then |
| 579 | cycle ! Skip empty unquoted token |
| 580 | end if |
| 581 | else |
| 582 | ! No metadata - check token for quotes (fallback) |
| 583 | if (index(cmd%tokens(i), '"') == 0 .and. index(cmd%tokens(i), "'") == 0) then |
| 584 | cycle ! Skip empty unquoted token |
| 585 | end if |
| 586 | end if |
| 587 | end if |
| 588 | total_tokens = total_tokens + 1 |
| 589 | if (total_tokens > temp_cap) call grow_temp_arrays() |
| 590 | temp_tokens(total_tokens)%str = expanded |
| 591 | ! Track actual length of expanded token (use len for allocatable to get real length) |
| 592 | if (allocated(expanded)) then |
| 593 | temp_token_lengths(total_tokens) = len(expanded) |
| 594 | else |
| 595 | temp_token_lengths(total_tokens) = 0 |
| 596 | end if |
| 597 | ! Preserve quoted status for trailing whitespace preservation |
| 598 | temp_token_quoted(total_tokens) = was_originally_quoted |
| 599 | end if |
| 600 | end do |
| 601 | |
| 602 | ! Copy string_t temp_tokens to cmd%tokens uniform-length character array. |
| 603 | ! Find max token length to allocate cmd%tokens with sufficient character width. |
| 604 | block |
| 605 | integer :: max_tok_len, tt |
| 606 | max_tok_len = MAX_TOKEN_LEN |
| 607 | do tt = 1, total_tokens |
| 608 | if (allocated(temp_tokens(tt)%str)) then |
| 609 | max_tok_len = max(max_tok_len, len(temp_tokens(tt)%str)) |
| 610 | end if |
| 611 | end do |
| 612 | if (allocated(cmd%tokens)) deallocate(cmd%tokens) |
| 613 | allocate(character(len=max_tok_len) :: cmd%tokens(total_tokens)) |
| 614 | do tt = 1, total_tokens |
| 615 | if (allocated(temp_tokens(tt)%str)) then |
| 616 | cmd%tokens(tt) = temp_tokens(tt)%str |
| 617 | else |
| 618 | cmd%tokens(tt) = '' |
| 619 | end if |
| 620 | end do |
| 621 | deallocate(temp_tokens) |
| 622 | end block |
| 623 | cmd%num_tokens = total_tokens |
| 624 | |
| 625 | ! Update token_lengths and token_quoted — move to avoid extra allocation |
| 626 | if (allocated(cmd%token_lengths)) deallocate(cmd%token_lengths) |
| 627 | call move_alloc(temp_token_lengths, cmd%token_lengths) |
| 628 | |
| 629 | if (allocated(cmd%token_quoted)) deallocate(cmd%token_quoted) |
| 630 | call move_alloc(temp_token_quoted, cmd%token_quoted) |
| 631 | |
| 632 | ! Resize token_quote_type and token_escaped to match expanded token count |
| 633 | if (allocated(cmd%token_quote_type)) then |
| 634 | block |
| 635 | integer, allocatable :: old_qt(:) |
| 636 | integer :: old_qt_size |
| 637 | old_qt_size = size(cmd%token_quote_type) |
| 638 | if (old_qt_size < total_tokens) then |
| 639 | allocate(old_qt(old_qt_size)) |
| 640 | old_qt = cmd%token_quote_type |
| 641 | deallocate(cmd%token_quote_type) |
| 642 | allocate(cmd%token_quote_type(total_tokens)) |
| 643 | cmd%token_quote_type = 0 |
| 644 | cmd%token_quote_type(1:old_qt_size) = old_qt(1:old_qt_size) |
| 645 | deallocate(old_qt) |
| 646 | end if |
| 647 | end block |
| 648 | end if |
| 649 | if (allocated(cmd%token_escaped)) then |
| 650 | block |
| 651 | logical, allocatable :: old_esc(:) |
| 652 | integer :: old_esc_size |
| 653 | old_esc_size = size(cmd%token_escaped) |
| 654 | if (old_esc_size < total_tokens) then |
| 655 | allocate(old_esc(old_esc_size)) |
| 656 | old_esc = cmd%token_escaped |
| 657 | deallocate(cmd%token_escaped) |
| 658 | allocate(cmd%token_escaped(total_tokens)) |
| 659 | cmd%token_escaped = .false. |
| 660 | cmd%token_escaped(1:old_esc_size) = old_esc(1:old_esc_size) |
| 661 | deallocate(old_esc) |
| 662 | end if |
| 663 | end block |
| 664 | end if |
| 665 | |
| 666 | ! temp_tokens, temp_token_lengths, temp_token_quoted moved above via move_alloc |
| 667 | |
| 668 | contains |
| 669 | |
| 670 | subroutine grow_temp_arrays(min_cap) |
| 671 | integer, intent(in), optional :: min_cap |
| 672 | type(string_t), allocatable :: new_tokens(:) |
| 673 | integer, allocatable :: new_lengths(:) |
| 674 | logical, allocatable :: new_quoted(:) |
| 675 | integer :: new_cap, gk |
| 676 | |
| 677 | new_cap = temp_cap * 2 |
| 678 | if (present(min_cap)) then |
| 679 | do while (new_cap < min_cap) |
| 680 | new_cap = new_cap * 2 |
| 681 | end do |
| 682 | end if |
| 683 | allocate(new_tokens(new_cap)) |
| 684 | allocate(new_lengths(new_cap)) |
| 685 | allocate(new_quoted(new_cap)) |
| 686 | new_lengths = 0 |
| 687 | new_quoted = .false. |
| 688 | do gk = 1, temp_cap |
| 689 | if (allocated(temp_tokens(gk)%str)) then |
| 690 | new_tokens(gk)%str = temp_tokens(gk)%str |
| 691 | else |
| 692 | new_tokens(gk)%str = '' |
| 693 | end if |
| 694 | end do |
| 695 | new_lengths(1:temp_cap) = temp_token_lengths(1:temp_cap) |
| 696 | new_quoted(1:temp_cap) = temp_token_quoted(1:temp_cap) |
| 697 | call move_alloc(new_tokens, temp_tokens) |
| 698 | call move_alloc(new_lengths, temp_token_lengths) |
| 699 | call move_alloc(new_quoted, temp_token_quoted) |
| 700 | temp_cap = new_cap |
| 701 | end subroutine grow_temp_arrays |
| 702 | |
| 703 | end subroutine |
| 704 | |
| 705 | subroutine expand_command_globs(cmd, shell) |
| 706 | use glob, only: expand_glob_patterns |
| 707 | type(command_t), intent(inout) :: cmd |
| 708 | type(shell_state_t), intent(in) :: shell |
| 709 | |
| 710 | character(len=MAX_TOKEN_LEN), allocatable :: expanded_tokens(:) |
| 711 | character(len=:), allocatable :: original_tokens(:) |
| 712 | integer :: expanded_count, i, tok_char_len |
| 713 | logical :: has_expandable |
| 714 | |
| 715 | if (.not. allocated(cmd%tokens) .or. cmd%num_tokens == 0) return |
| 716 | |
| 717 | ! Skip glob expansion if noglob option is enabled (set -f) |
| 718 | if (shell%option_noglob) return |
| 719 | |
| 720 | ! Preserve the character width from expand_tokens (may be > MAX_TOKEN_LEN) |
| 721 | tok_char_len = len(cmd%tokens(1)) |
| 722 | |
| 723 | ! Save original tokens with matching character width |
| 724 | allocate(character(len=tok_char_len) :: original_tokens(cmd%num_tokens)) |
| 725 | do i = 1, cmd%num_tokens |
| 726 | original_tokens(i) = cmd%tokens(i) |
| 727 | end do |
| 728 | |
| 729 | ! Don't glob expand tokens that were quoted, escaped, or have backslashes |
| 730 | ! Check metadata if available, otherwise fall back to checking for backslash |
| 731 | has_expandable = .false. |
| 732 | do i = 1, cmd%num_tokens |
| 733 | ! Skip if token was quoted (prevents glob expansion per POSIX) |
| 734 | if (allocated(cmd%token_quoted)) then |
| 735 | if (i <= size(cmd%token_quoted) .and. cmd%token_quoted(i)) then |
| 736 | cycle ! Skip this token - it was quoted |
| 737 | end if |
| 738 | end if |
| 739 | |
| 740 | ! Skip if token was escaped (metadata available) or has backslash (fallback) |
| 741 | if (allocated(cmd%token_escaped)) then |
| 742 | ! Use metadata if available |
| 743 | if (i <= size(cmd%token_escaped) .and. cmd%token_escaped(i)) then |
| 744 | cycle ! Skip this token - it was escaped |
| 745 | end if |
| 746 | else if (index(cmd%tokens(i), '\') > 0) then |
| 747 | ! Fallback: check for backslash in token |
| 748 | cycle ! Skip this token - it has a backslash |
| 749 | end if |
| 750 | |
| 751 | ! Check if token has glob characters |
| 752 | if (index(cmd%tokens(i), '*') > 0 .or. & |
| 753 | index(cmd%tokens(i), '?') > 0 .or. & |
| 754 | index(cmd%tokens(i), '[') > 0) then |
| 755 | has_expandable = .true. |
| 756 | exit |
| 757 | end if |
| 758 | end do |
| 759 | |
| 760 | if (.not. has_expandable) then |
| 761 | ! No tokens need glob expansion |
| 762 | if (allocated(original_tokens)) deallocate(original_tokens) |
| 763 | return |
| 764 | end if |
| 765 | |
| 766 | ! Expand glob patterns (pass token_quoted to prevent glob expansion on quoted tokens) |
| 767 | if (allocated(cmd%token_quoted)) then |
| 768 | call expand_glob_patterns(original_tokens, cmd%num_tokens, expanded_tokens, expanded_count, cmd%token_quoted) |
| 769 | else |
| 770 | call expand_glob_patterns(original_tokens, cmd%num_tokens, expanded_tokens, expanded_count) |
| 771 | end if |
| 772 | |
| 773 | ! Replace command tokens with expanded ones |
| 774 | if (allocated(cmd%tokens)) deallocate(cmd%tokens) |
| 775 | |
| 776 | if (expanded_count > 0) then |
| 777 | allocate(character(len=max(tok_char_len, MAX_TOKEN_LEN)) :: cmd%tokens(expanded_count)) |
| 778 | do i = 1, expanded_count |
| 779 | cmd%tokens(i) = expanded_tokens(i) |
| 780 | end do |
| 781 | cmd%num_tokens = expanded_count |
| 782 | |
| 783 | ! Update token_lengths to match new tokens (use trimmed length) |
| 784 | if (allocated(cmd%token_lengths)) deallocate(cmd%token_lengths) |
| 785 | allocate(cmd%token_lengths(expanded_count)) |
| 786 | do i = 1, expanded_count |
| 787 | cmd%token_lengths(i) = len_trim(expanded_tokens(i)) |
| 788 | end do |
| 789 | |
| 790 | ! Reset token_quoted and token_escaped for expanded tokens |
| 791 | ! (glob-expanded filenames are not quoted) |
| 792 | if (allocated(cmd%token_quoted)) deallocate(cmd%token_quoted) |
| 793 | allocate(cmd%token_quoted(expanded_count)) |
| 794 | cmd%token_quoted = .false. |
| 795 | |
| 796 | if (allocated(cmd%token_escaped)) deallocate(cmd%token_escaped) |
| 797 | allocate(cmd%token_escaped(expanded_count)) |
| 798 | cmd%token_escaped = .false. |
| 799 | else |
| 800 | ! No expansion occurred - restore original |
| 801 | allocate(character(len=tok_char_len) :: cmd%tokens(cmd%num_tokens)) |
| 802 | do i = 1, cmd%num_tokens |
| 803 | cmd%tokens(i) = original_tokens(i) |
| 804 | end do |
| 805 | end if |
| 806 | |
| 807 | ! Cleanup |
| 808 | if (allocated(expanded_tokens)) deallocate(expanded_tokens) |
| 809 | if (allocated(original_tokens)) deallocate(original_tokens) |
| 810 | end subroutine |
| 811 | |
| 812 | subroutine process_command_escapes(cmd) |
| 813 | type(command_t), intent(inout) :: cmd |
| 814 | integer :: i, j, k, token_len |
| 815 | logical :: in_quotes |
| 816 | character(len=1) :: quote_char, backslash |
| 817 | integer :: result_len |
| 818 | |
| 819 | backslash = char(92) ! ASCII for backslash |
| 820 | |
| 821 | ! Use the character length of the token array for result buffer |
| 822 | result_len = len(cmd%tokens(1)) |
| 823 | |
| 824 | do i = 1, cmd%num_tokens |
| 825 | token_len = len_trim(cmd%tokens(i)) |
| 826 | block |
| 827 | character(len=result_len) :: result |
| 828 | result = '' |
| 829 | k = 0 ! Count of characters written to result |
| 830 | j = 1 |
| 831 | in_quotes = .false. |
| 832 | quote_char = ' ' |
| 833 | |
| 834 | do while (j <= token_len) |
| 835 | ! Track quote state |
| 836 | if (.not. in_quotes .and. (cmd%tokens(i)(j:j) == '"' .or. cmd%tokens(i)(j:j) == "'")) then |
| 837 | in_quotes = .true. |
| 838 | quote_char = cmd%tokens(i)(j:j) |
| 839 | k = k + 1 |
| 840 | result(k:k) = cmd%tokens(i)(j:j) |
| 841 | j = j + 1 |
| 842 | else if (in_quotes .and. cmd%tokens(i)(j:j) == quote_char) then |
| 843 | in_quotes = .false. |
| 844 | k = k + 1 |
| 845 | result(k:k) = cmd%tokens(i)(j:j) |
| 846 | j = j + 1 |
| 847 | else if (.not. in_quotes .and. cmd%tokens(i)(j:j) == backslash .and. j < token_len) then |
| 848 | ! Check what character follows the backslash |
| 849 | ! Only process structural escapes (space, glob characters) |
| 850 | if (cmd%tokens(i)(j+1:j+1) == ' ' .or. & |
| 851 | cmd%tokens(i)(j+1:j+1) == '*' .or. & |
| 852 | cmd%tokens(i)(j+1:j+1) == '?' .or. & |
| 853 | cmd%tokens(i)(j+1:j+1) == '[') then |
| 854 | ! Structural escape - skip backslash, keep next char |
| 855 | j = j + 1 |
| 856 | k = k + 1 |
| 857 | result(k:k) = cmd%tokens(i)(j:j) |
| 858 | j = j + 1 |
| 859 | else |
| 860 | ! Non-structural escape (like \n, \t) - keep both backslash and next char |
| 861 | k = k + 1 |
| 862 | result(k:k) = backslash |
| 863 | j = j + 1 |
| 864 | if (j <= token_len) then |
| 865 | k = k + 1 |
| 866 | result(k:k) = cmd%tokens(i)(j:j) |
| 867 | j = j + 1 |
| 868 | end if |
| 869 | end if |
| 870 | else |
| 871 | ! Regular character |
| 872 | k = k + 1 |
| 873 | result(k:k) = cmd%tokens(i)(j:j) |
| 874 | j = j + 1 |
| 875 | end if |
| 876 | end do |
| 877 | |
| 878 | ! Only copy the actual content (k characters) |
| 879 | if (k > 0) then |
| 880 | cmd%tokens(i) = result(1:k) |
| 881 | else |
| 882 | cmd%tokens(i) = '' |
| 883 | end if |
| 884 | end block |
| 885 | end do |
| 886 | end subroutine |
| 887 | |
| 888 | end module pipeline_helpers |
| 889 |