lexi.c revision 1.198
1 /* $NetBSD: lexi.c,v 1.198 2023/05/18 04:23:03 rillig Exp $ */ 2 3 /*- 4 * SPDX-License-Identifier: BSD-4-Clause 5 * 6 * Copyright (c) 1985 Sun Microsystems, Inc. 7 * Copyright (c) 1980, 1993 8 * The Regents of the University of California. All rights reserved. 9 * All rights reserved. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. All advertising materials mentioning features or use of this software 20 * must display the following acknowledgement: 21 * This product includes software developed by the University of 22 * California, Berkeley and its contributors. 23 * 4. Neither the name of the University nor the names of its contributors 24 * may be used to endorse or promote products derived from this software 25 * without specific prior written permission. 26 * 27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 37 * SUCH DAMAGE. 38 */ 39 40 #include <sys/cdefs.h> 41 __RCSID("$NetBSD: lexi.c,v 1.198 2023/05/18 04:23:03 rillig Exp $"); 42 43 #include <stdlib.h> 44 #include <string.h> 45 46 #include "indent.h" 47 48 /* In lexi_alnum, this constant marks a type, independent of parentheses. */ 49 #define lsym_type lsym_type_outside_parentheses 50 #define lsym_type_modifier lsym_storage_class 51 52 /* must be sorted alphabetically, is used in binary search */ 53 static const struct keyword { 54 const char name[12]; 55 lexer_symbol lsym; 56 } keywords[] = { 57 {"_Bool", lsym_type}, 58 {"_Complex", lsym_type}, 59 {"_Imaginary", lsym_type}, 60 {"auto", lsym_storage_class}, 61 {"bool", lsym_type}, 62 {"break", lsym_word}, 63 {"case", lsym_case_label}, 64 {"char", lsym_type}, 65 {"complex", lsym_type}, 66 {"const", lsym_type_modifier}, 67 {"continue", lsym_word}, 68 {"default", lsym_case_label}, 69 {"do", lsym_do}, 70 {"double", lsym_type}, 71 {"else", lsym_else}, 72 {"enum", lsym_tag}, 73 {"extern", lsym_storage_class}, 74 {"float", lsym_type}, 75 {"for", lsym_for}, 76 {"goto", lsym_word}, 77 {"if", lsym_if}, 78 {"imaginary", lsym_type}, 79 {"inline", lsym_word}, 80 {"int", lsym_type}, 81 {"long", lsym_type}, 82 {"offsetof", lsym_offsetof}, 83 {"register", lsym_storage_class}, 84 {"restrict", lsym_word}, 85 {"return", lsym_return}, 86 {"short", lsym_type}, 87 {"signed", lsym_type}, 88 {"sizeof", lsym_sizeof}, 89 {"static", lsym_storage_class}, 90 {"struct", lsym_tag}, 91 {"switch", lsym_switch}, 92 {"typedef", lsym_typedef}, 93 {"union", lsym_tag}, 94 {"unsigned", lsym_type}, 95 {"void", lsym_type}, 96 {"volatile", lsym_type_modifier}, 97 {"while", lsym_while} 98 }; 99 100 static struct { 101 const char **items; 102 unsigned int len; 103 unsigned int cap; 104 } typenames; 105 106 /*- 107 * The transition table below was rewritten by hand from lx's output, given 108 * the following definitions. lx is Katherine Flavel's lexer generator. 109 * 110 * O = /[0-7]/; D = /[0-9]/; NZ = /[1-9]/; 111 * H = /[a-f0-9]/i; B = /[0-1]/; HP = /0x/i; 112 * BP = /0b/i; E = /e[+\-]?/i D+; P = /p[+\-]?/i D+; 113 * FS = /[fl]/i; IS = /u/i /(l|L|ll|LL)/? | /(l|L|ll|LL)/ /u/i?; 114 * 115 * D+ E FS? -> $float; 116 * D* "." D+ E? FS? -> $float; 117 * D+ "." E? FS? -> $float; HP H+ IS? -> $int; 118 * HP H+ P FS? -> $float; NZ D* IS? -> $int; 119 * HP H* "." H+ P FS? -> $float; "0" O* IS? -> $int; 120 * HP H+ "." P FS -> $float; BP B+ IS? -> $int; 121 */ 122 /* INDENT OFF */ 123 static const unsigned char lex_number_state[][26] = { 124 /* examples: 125 00 126 s 0xx 127 t 00xaa 128 a 11 101100xxa.. 129 r 11ee0001101lbuuxx.a.pp 130 t.01.e+008bLuxll0Ll.aa.p+0 131 states: ABCDEFGHIJKLMNOPQRSTUVWXYZ */ 132 [0] = "uuiifuufiuuiiuiiiiiuiuuuuu", /* (other) */ 133 [1] = "CEIDEHHHIJQ U Q VUVVZZZ", /* 0 */ 134 [2] = "DEIDEHHHIJQ U Q VUVVZZZ", /* 1 */ 135 [3] = "DEIDEHHHIJ U VUVVZZZ", /* 2 3 4 5 6 7 */ 136 [4] = "DEJDEHHHJJ U VUVVZZZ", /* 8 9 */ 137 [5] = " U VUVV ", /* A a C c D d */ 138 [6] = " K U VUVV ", /* B b */ 139 [7] = " FFF FF U VUVV ", /* E e */ 140 [8] = " f f U VUVV f", /* F f */ 141 [9] = " LLf fL PR Li L f", /* L */ 142 [10] = " OOf fO S P O i O f", /* l */ 143 [11] = " FFX ", /* P p */ 144 [12] = " MM M i iiM M ", /* U u */ 145 [13] = " N ", /* X x */ 146 [14] = " G Y ", /* + - */ 147 [15] = "B EE EE T W ", /* . */ 148 /* ABCDEFGHIJKLMNOPQRSTUVWXYZ */ 149 }; 150 /* INDENT ON */ 151 152 static const unsigned char lex_number_row[] = { 153 ['0'] = 1, 154 ['1'] = 2, 155 ['2'] = 3, ['3'] = 3, ['4'] = 3, ['5'] = 3, ['6'] = 3, ['7'] = 3, 156 ['8'] = 4, ['9'] = 4, 157 ['A'] = 5, ['a'] = 5, ['C'] = 5, ['c'] = 5, ['D'] = 5, ['d'] = 5, 158 ['B'] = 6, ['b'] = 6, 159 ['E'] = 7, ['e'] = 7, 160 ['F'] = 8, ['f'] = 8, 161 ['L'] = 9, 162 ['l'] = 10, 163 ['P'] = 11, ['p'] = 11, 164 ['U'] = 12, ['u'] = 12, 165 ['X'] = 13, ['x'] = 13, 166 ['+'] = 14, ['-'] = 14, 167 ['.'] = 15, 168 }; 169 170 static void 171 token_add_char(char ch) 172 { 173 buf_add_char(&token, ch); 174 } 175 176 177 static void 178 lex_number(void) 179 { 180 for (unsigned char s = 'A'; s != 'f' && s != 'i' && s != 'u';) { 181 unsigned char ch = (unsigned char)inp.st[0]; 182 if (ch == '\\' && inp.st[1] == '\n') { 183 inp.st++; 184 inp_skip(); 185 line_no++; 186 continue; 187 } 188 if (ch >= array_length(lex_number_row) || lex_number_row[ch] == 0) 189 break; 190 191 unsigned char row = lex_number_row[ch]; 192 if (lex_number_state[row][s - 'A'] == ' ') { 193 /*- 194 * lex_number_state[0][s - 'A'] now indicates the type: 195 * f = floating, i = integer, u = unknown 196 */ 197 return; 198 } 199 200 s = lex_number_state[row][s - 'A']; 201 token_add_char(inp_next()); 202 } 203 } 204 205 static bool 206 is_identifier_start(char ch) 207 { 208 return ch_isalpha(ch) || ch == '_' || ch == '$'; 209 } 210 211 static bool 212 is_identifier_part(char ch) 213 { 214 return ch_isalnum(ch) || ch == '_' || ch == '$'; 215 } 216 217 static void 218 lex_word(void) 219 { 220 for (;;) { 221 if (is_identifier_part(inp.st[0])) 222 token_add_char(*inp.st++); 223 else if (inp.st[0] == '\\' && inp.st[1] == '\n') { 224 inp.st++; 225 inp_skip(); 226 line_no++; 227 } else 228 return; 229 } 230 } 231 232 static void 233 lex_char_or_string(void) 234 { 235 for (char delim = token.mem[token.len - 1];;) { 236 if (inp.st[0] == '\n') { 237 diag(1, "Unterminated literal"); 238 return; 239 } 240 241 token_add_char(*inp.st++); 242 if (token.mem[token.len - 1] == delim) 243 return; 244 245 if (token.mem[token.len - 1] == '\\') { 246 if (inp.st[0] == '\n') 247 ++line_no; 248 token_add_char(inp_next()); 249 } 250 } 251 } 252 253 /* Guess whether the current token is a declared type. */ 254 static bool 255 probably_typename(void) 256 { 257 if (ps.prev_token == lsym_storage_class) 258 return true; 259 if (ps.block_init) 260 return false; 261 if (ps.in_stmt_or_decl) /* XXX: this condition looks incorrect */ 262 return false; 263 if (inp.st[0] == '*' && inp.st[1] != '=') 264 goto maybe; 265 /* XXX: is_identifier_start */ 266 if (ch_isalpha(inp.st[0])) 267 goto maybe; 268 return false; 269 maybe: 270 return ps.prev_token == lsym_semicolon || 271 ps.prev_token == lsym_lbrace || 272 ps.prev_token == lsym_rbrace; 273 } 274 275 static int 276 bsearch_typenames(const char *key) 277 { 278 const char **arr = typenames.items; 279 int lo = 0; 280 int hi = (int)typenames.len - 1; 281 282 while (lo <= hi) { 283 int mid = (int)((unsigned)(lo + hi) >> 1); 284 int cmp = strcmp(arr[mid], key); 285 if (cmp < 0) 286 lo = mid + 1; 287 else if (cmp > 0) 288 hi = mid - 1; 289 else 290 return mid; 291 } 292 return -(lo + 1); 293 } 294 295 static bool 296 is_typename(void) 297 { 298 if (opt.auto_typedefs && 299 token.len >= 2 && memcmp(token.mem + token.len - 2, "_t", 2) == 0) 300 return true; 301 302 return bsearch_typenames(token.st) >= 0; 303 } 304 305 static int 306 cmp_keyword_by_name(const void *key, const void *elem) 307 { 308 return strcmp(key, ((const struct keyword *)elem)->name); 309 } 310 311 /* 312 * Looking at something like 'function_name(...)' in a line, guess whether 313 * this starts a function definition or a declaration. 314 */ 315 static bool 316 probably_looking_at_definition(void) 317 { 318 int paren_level = 0; 319 for (const char *p = inp.st; *p != '\n'; p++) { 320 if (*p == '(') 321 paren_level++; 322 if (*p == ')' && --paren_level == 0) { 323 p++; 324 325 while (*p != '\n' && (ch_isspace(*p) || is_identifier_part(*p))) 326 p++; /* '__dead' or '__unused' */ 327 328 if (*p == '\n') /* func(...) */ 329 break; 330 if (*p == ';') /* func(...); */ 331 return false; 332 if (*p == ',') /* double abs(), pi; */ 333 return false; 334 if (*p == '(') /* func(...) __attribute__((...)) */ 335 paren_level++; /* func(...) __printflike(...) 336 */ 337 else 338 break; /* func(...) { ... */ 339 } 340 } 341 342 /* To further reduce the cases where indent wrongly treats an 343 * incomplete function declaration as a function definition, thus 344 * adding a newline before the function name, it may be worth looking 345 * for parameter names, as these are often omitted in function 346 * declarations and only included in function definitions. Or just 347 * increase the lookahead to more than just the current line of input, 348 * until the next '{'. */ 349 return true; 350 } 351 352 /* Read an alphanumeric token into 'token', or return lsym_eof. */ 353 static lexer_symbol 354 lexi_alnum(void) 355 { 356 if (ch_isdigit(inp.st[0]) || 357 (inp.st[0] == '.' && ch_isdigit(inp.st[1]))) { 358 lex_number(); 359 } else if (is_identifier_start(inp.st[0])) { 360 lex_word(); 361 362 if (token.len == 1 && token.st[0] == 'L' && 363 (inp.st[0] == '"' || inp.st[0] == '\'')) { 364 token_add_char(*inp.st++); 365 lex_char_or_string(); 366 ps.next_unary = false; 367 return lsym_word; 368 } 369 } else 370 return lsym_eof; /* just as a placeholder */ 371 372 while (ch_isblank(inp.st[0])) 373 inp.st++; 374 375 ps.next_unary = ps.prev_token == lsym_tag 376 || ps.prev_token == lsym_typedef; 377 378 if (ps.prev_token == lsym_tag && ps.nparen == 0) 379 return lsym_type_outside_parentheses; 380 381 token_add_char('\0'); 382 token.len--; 383 const struct keyword *kw = bsearch(token.st, keywords, 384 array_length(keywords), sizeof(keywords[0]), cmp_keyword_by_name); 385 bool is_type = false; 386 if (kw == NULL) { 387 if (is_typename()) { 388 is_type = true; 389 ps.next_unary = true; 390 if (ps.in_enum == in_enum_enum) 391 ps.in_enum = in_enum_type; 392 goto found_typename; 393 } 394 395 } else { /* we have a keyword */ 396 is_type = kw->lsym == lsym_type; 397 ps.next_unary = true; 398 if (kw->lsym != lsym_tag && kw->lsym != lsym_type) 399 return kw->lsym; 400 401 found_typename: 402 if (ps.nparen > 0) { 403 /* inside parentheses: cast, param list, offsetof or 404 * sizeof */ 405 if (ps.paren[ps.nparen - 1].cast == cast_unknown) 406 ps.paren[ps.nparen - 1].cast = cast_maybe; 407 } 408 if (ps.prev_token != lsym_period && ps.prev_token != lsym_unary_op) { 409 if (kw != NULL && kw->lsym == lsym_tag) { 410 if (token.st[0] == 'e' /* enum */) 411 ps.in_enum = in_enum_enum; 412 return lsym_tag; 413 } 414 if (ps.nparen == 0) 415 return lsym_type_outside_parentheses; 416 } 417 } 418 419 if (inp.st[0] == '(' && ps.tos <= 1 && ps.ind_level == 0 && 420 !ps.in_func_def_params && !ps.block_init) { 421 422 if (ps.nparen == 0 && probably_looking_at_definition()) { 423 ps.is_function_definition = true; 424 if (ps.in_decl) 425 ps.in_func_def_params = true; 426 return lsym_funcname; 427 } 428 429 } else if (ps.nparen == 0 && probably_typename()) { 430 ps.next_unary = true; 431 return lsym_type_outside_parentheses; 432 } 433 434 return is_type ? lsym_type_in_parentheses : lsym_word; 435 } 436 437 static bool 438 is_asterisk_unary(void) 439 { 440 if (ps.next_unary || ps.in_func_def_params) 441 return true; 442 if (ps.prev_token == lsym_word || 443 ps.prev_token == lsym_rparen_or_rbracket) 444 return false; 445 return ps.in_decl && ps.nparen > 0; 446 } 447 448 static void 449 lex_asterisk_unary(void) 450 { 451 while (inp.st[0] == '*' || ch_isspace(inp.st[0])) { 452 if (inp.st[0] == '*') 453 token_add_char('*'); 454 inp_skip(); 455 } 456 457 if (ps.in_decl) { 458 for (const char *tp = inp.st; *tp != '\n';) { 459 if (ch_isspace(*tp)) 460 tp++; 461 else if (is_identifier_start(*tp)) { 462 tp++; 463 while (is_identifier_part(*tp)) 464 tp++; 465 } else { 466 if (*tp == '(') 467 ps.is_function_definition = true; 468 break; 469 } 470 } 471 } 472 } 473 474 static void 475 skip_blank(const char **pp) 476 { 477 while (ch_isblank(**pp)) 478 (*pp)++; 479 } 480 481 static bool 482 skip_string(const char **pp, const char *s) 483 { 484 size_t len = strlen(s); 485 if (strncmp(*pp, s, len) == 0) { 486 *pp += len; 487 return true; 488 } 489 return false; 490 } 491 492 static void 493 lex_indent_comment(void) 494 { 495 const char *p = inp.mem; 496 497 skip_blank(&p); 498 if (!skip_string(&p, "/*")) 499 return; 500 skip_blank(&p); 501 if (!skip_string(&p, "INDENT")) 502 return; 503 504 enum indent_enabled enabled; 505 skip_blank(&p); 506 if (*p == '*' || skip_string(&p, "ON")) 507 enabled = indent_last_off_line; 508 else if (skip_string(&p, "OFF")) 509 enabled = indent_off; 510 else 511 return; 512 513 skip_blank(&p); 514 if (!skip_string(&p, "*/\n")) 515 return; 516 517 if (lab.len > 0 || code.len > 0 || com.len > 0) 518 output_line(); 519 520 indent_enabled = enabled; 521 } 522 523 /* Reads the next token, placing it in the global variable "token". */ 524 lexer_symbol 525 lexi(void) 526 { 527 token.len = 0; 528 ps.curr_col_1 = ps.next_col_1; 529 ps.next_col_1 = false; 530 531 for (;;) { 532 if (ch_isblank(inp.st[0])) { 533 ps.curr_col_1 = false; 534 inp.st++; 535 } else if (inp.st[0] == '\\' && inp.st[1] == '\n') { 536 inp.st++; 537 inp_skip(); 538 line_no++; 539 } else 540 break; 541 } 542 543 lexer_symbol alnum_lsym = lexi_alnum(); 544 if (alnum_lsym != lsym_eof) { 545 debug_parser_state(alnum_lsym); 546 return alnum_lsym; 547 } 548 549 /* Scan a non-alphanumeric token */ 550 551 token_add_char(inp_next()); 552 553 lexer_symbol lsym; 554 bool next_unary; 555 556 switch (token.mem[token.len - 1]) { 557 558 /* INDENT OFF */ 559 case '(': 560 case '[': lsym = lsym_lparen_or_lbracket; next_unary = true; break; 561 case ')': 562 case ']': lsym = lsym_rparen_or_rbracket; next_unary = false; break; 563 case '?': lsym = lsym_question; next_unary = true; break; 564 case ':': lsym = lsym_colon; next_unary = true; break; 565 case ';': lsym = lsym_semicolon; next_unary = true; break; 566 case '{': lsym = lsym_lbrace; next_unary = true; break; 567 case '}': lsym = lsym_rbrace; next_unary = true; break; 568 case ',': lsym = lsym_comma; next_unary = true; break; 569 case '.': lsym = lsym_period; next_unary = false; break; 570 /* INDENT ON */ 571 572 case '\n': 573 /* if data has been exhausted, the '\n' is a dummy. */ 574 lsym = had_eof ? lsym_eof : lsym_newline; 575 next_unary = ps.next_unary; 576 ps.next_col_1 = true; 577 break; 578 579 case '#': 580 lsym = lsym_preprocessing; 581 next_unary = ps.next_unary; 582 break; 583 584 case '\'': 585 case '"': 586 lex_char_or_string(); 587 lsym = lsym_word; 588 next_unary = false; 589 break; 590 591 case '-': 592 case '+': 593 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 594 next_unary = true; 595 596 if (inp.st[0] == token.mem[token.len - 1]) { /* '++' or '--' */ 597 token_add_char(*inp.st++); 598 if (ps.prev_token == lsym_word || 599 ps.prev_token == lsym_rparen_or_rbracket) { 600 lsym = ps.next_unary ? lsym_unary_op : lsym_postfix_op; 601 next_unary = false; 602 } 603 604 } else if (inp.st[0] == '=') { /* '+=' or '-=' */ 605 token_add_char(*inp.st++); 606 607 } else if (inp.st[0] == '>') { /* '->' */ 608 token_add_char(*inp.st++); 609 lsym = lsym_unary_op; 610 next_unary = false; 611 ps.want_blank = false; 612 } 613 break; 614 615 case '=': 616 if (ps.init_or_struct) 617 ps.block_init = true; 618 if (inp.st[0] == '=') 619 token_add_char(*inp.st++); 620 lsym = lsym_binary_op; 621 next_unary = true; 622 break; 623 624 case '>': 625 case '<': 626 case '!': /* ops like <, <<, <=, !=, etc */ 627 if (inp.st[0] == '>' || inp.st[0] == '<' || inp.st[0] == '=') 628 token_add_char(*inp.st++); 629 if (inp.st[0] == '=') 630 token_add_char(*inp.st++); 631 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 632 next_unary = true; 633 break; 634 635 case '*': 636 if (is_asterisk_unary()) { 637 lex_asterisk_unary(); 638 lsym = lsym_unary_op; 639 next_unary = true; 640 } else { 641 if (inp.st[0] == '=') 642 token_add_char(*inp.st++); 643 lsym = lsym_binary_op; 644 next_unary = true; 645 } 646 break; 647 648 default: 649 if (token.mem[token.len - 1] == '/' 650 && (inp.st[0] == '*' || inp.st[0] == '/')) { 651 enum indent_enabled prev = indent_enabled; 652 lex_indent_comment(); 653 if (prev == indent_on && indent_enabled == indent_off) 654 clear_indent_off_text(); 655 token_add_char(*inp.st++); 656 lsym = lsym_comment; 657 next_unary = ps.next_unary; 658 break; 659 } 660 661 /* handle '||', '&&', etc., and also things as in 'int *****i' 662 */ 663 while (inp.st[0] == token.mem[token.len - 1] || inp.st[0] == '=') 664 token_add_char(*inp.st++); 665 666 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 667 next_unary = true; 668 } 669 670 if (ps.in_enum == in_enum_enum || ps.in_enum == in_enum_type) 671 ps.in_enum = lsym == lsym_lbrace ? in_enum_brace : in_enum_no; 672 if (lsym == lsym_rbrace) 673 ps.in_enum = in_enum_no; 674 675 ps.next_unary = next_unary; 676 677 debug_parser_state(lsym); 678 return lsym; 679 } 680 681 void 682 register_typename(const char *name) 683 { 684 if (typenames.len >= typenames.cap) { 685 typenames.cap = 16 + 2 * typenames.cap; 686 typenames.items = nonnull(realloc(typenames.items, 687 sizeof(typenames.items[0]) * typenames.cap)); 688 } 689 690 int pos = bsearch_typenames(name); 691 if (pos >= 0) 692 return; /* already in the list */ 693 694 pos = -(pos + 1); 695 memmove(typenames.items + pos + 1, typenames.items + pos, 696 sizeof(typenames.items[0]) * (typenames.len++ - (unsigned)pos)); 697 typenames.items[pos] = nonnull(strdup(name)); 698 } 699
Indexes created Mon Sep 29 15:09:53 GMT 2025