lexi.c revision 1.211
1 /* $NetBSD: lexi.c,v 1.211 2023/06/04 14:20:00 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 1.211 2023/06/04 14:20:00 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 51 /* must be sorted alphabetically, is used in binary search */ 52 static const struct keyword { 53 const char name[12]; 54 lexer_symbol lsym; 55 } keywords[] = { 56 {"_Bool", lsym_type}, 57 {"_Complex", lsym_type}, 58 {"_Imaginary", lsym_type}, 59 {"auto", lsym_modifier}, 60 {"bool", lsym_type}, 61 {"break", lsym_word}, 62 {"case", lsym_case}, 63 {"char", lsym_type}, 64 {"complex", lsym_type}, 65 {"const", lsym_modifier}, 66 {"continue", lsym_word}, 67 {"default", lsym_default}, 68 {"do", lsym_do}, 69 {"double", lsym_type}, 70 {"else", lsym_else}, 71 {"enum", lsym_tag}, 72 {"extern", lsym_modifier}, 73 {"float", lsym_type}, 74 {"for", lsym_for}, 75 {"goto", lsym_word}, 76 {"if", lsym_if}, 77 {"imaginary", lsym_type}, 78 {"inline", lsym_modifier}, 79 {"int", lsym_type}, 80 {"long", lsym_type}, 81 {"offsetof", lsym_offsetof}, 82 {"register", lsym_modifier}, 83 {"restrict", lsym_word}, 84 {"return", lsym_return}, 85 {"short", lsym_type}, 86 {"signed", lsym_type}, 87 {"sizeof", lsym_sizeof}, 88 {"static", lsym_modifier}, 89 {"struct", lsym_tag}, 90 {"switch", lsym_switch}, 91 {"typedef", lsym_typedef}, 92 {"union", lsym_tag}, 93 {"unsigned", lsym_type}, 94 {"void", lsym_type}, 95 {"volatile", lsym_modifier}, 96 {"while", lsym_while} 97 }; 98 99 static struct { 100 const char **items; 101 unsigned int len; 102 unsigned int cap; 103 } typenames; 104 105 /*- 106 * The transition table below was rewritten by hand from lx's output, given 107 * the following definitions. lx is Katherine Flavel's lexer generator. 108 * 109 * O = /[0-7]/; D = /[0-9]/; NZ = /[1-9]/; 110 * H = /[a-f0-9]/i; B = /[0-1]/; HP = /0x/i; 111 * BP = /0b/i; E = /e[+\-]?/i D+; P = /p[+\-]?/i D+; 112 * FS = /[fl]/i; IS = /u/i /(l|L|ll|LL)/? | /(l|L|ll|LL)/ /u/i?; 113 * 114 * D+ E FS? -> $float; 115 * D* "." D+ E? FS? -> $float; 116 * D+ "." E? FS? -> $float; HP H+ IS? -> $int; 117 * HP H+ P FS? -> $float; NZ D* IS? -> $int; 118 * HP H* "." H+ P FS? -> $float; "0" O* IS? -> $int; 119 * HP H+ "." P FS -> $float; BP B+ IS? -> $int; 120 */ 121 /* INDENT OFF */ 122 static const unsigned char lex_number_state[][26] = { 123 /* examples: 124 00 125 s 0xx 126 t 00xaa 127 a 11 101100xxa.. 128 r 11ee0001101lbuuxx.a.pp 129 t.01.e+008bLuxll0Ll.aa.p+0 130 states: ABCDEFGHIJKLMNOPQRSTUVWXYZ */ 131 [0] = "uuiifuufiuuiiuiiiiiuiuuuuu", /* (other) */ 132 [1] = "CEIDEHHHIJQ U Q VUVVZZZ", /* 0 */ 133 [2] = "DEIDEHHHIJQ U Q VUVVZZZ", /* 1 */ 134 [3] = "DEIDEHHHIJ U VUVVZZZ", /* 2 3 4 5 6 7 */ 135 [4] = "DEJDEHHHJJ U VUVVZZZ", /* 8 9 */ 136 [5] = " U VUVV ", /* A a C c D d */ 137 [6] = " K U VUVV ", /* B b */ 138 [7] = " FFF FF U VUVV ", /* E e */ 139 [8] = " f f U VUVV f", /* F f */ 140 [9] = " LLf fL PR Li L f", /* L */ 141 [10] = " OOf fO S P O i O f", /* l */ 142 [11] = " FFX ", /* P p */ 143 [12] = " MM M i iiM M ", /* U u */ 144 [13] = " N ", /* X x */ 145 [14] = " G Y ", /* + - */ 146 [15] = "B EE EE T W ", /* . */ 147 /* ABCDEFGHIJKLMNOPQRSTUVWXYZ */ 148 }; 149 /* INDENT ON */ 150 151 static const unsigned char lex_number_row[] = { 152 ['0'] = 1, 153 ['1'] = 2, 154 ['2'] = 3, ['3'] = 3, ['4'] = 3, ['5'] = 3, ['6'] = 3, ['7'] = 3, 155 ['8'] = 4, ['9'] = 4, 156 ['A'] = 5, ['a'] = 5, ['C'] = 5, ['c'] = 5, ['D'] = 5, ['d'] = 5, 157 ['B'] = 6, ['b'] = 6, 158 ['E'] = 7, ['e'] = 7, 159 ['F'] = 8, ['f'] = 8, 160 ['L'] = 9, 161 ['l'] = 10, 162 ['P'] = 11, ['p'] = 11, 163 ['U'] = 12, ['u'] = 12, 164 ['X'] = 13, ['x'] = 13, 165 ['+'] = 14, ['-'] = 14, 166 ['.'] = 15, 167 }; 168 169 static void 170 token_add_char(char ch) 171 { 172 buf_add_char(&token, ch); 173 } 174 175 static void 176 lex_number(void) 177 { 178 for (unsigned char s = 'A'; s != 'f' && s != 'i' && s != 'u';) { 179 unsigned char ch = (unsigned char)inp.st[0]; 180 if (ch == '\\' && inp.st[1] == '\n') { 181 inp.st++; 182 inp_skip(); 183 line_no++; 184 continue; 185 } 186 if (ch >= array_length(lex_number_row) 187 || lex_number_row[ch] == 0) 188 break; 189 190 unsigned char row = lex_number_row[ch]; 191 if (lex_number_state[row][s - 'A'] == ' ') { 192 /*- 193 * lex_number_state[0][s - 'A'] now indicates the type: 194 * f = floating, i = integer, u = unknown 195 */ 196 return; 197 } 198 199 s = lex_number_state[row][s - 'A']; 200 token_add_char(inp_next()); 201 } 202 } 203 204 static bool 205 is_identifier_start(char ch) 206 { 207 return ch_isalpha(ch) || ch == '_' || ch == '$'; 208 } 209 210 static bool 211 is_identifier_part(char ch) 212 { 213 return ch_isalnum(ch) || ch == '_' || ch == '$'; 214 } 215 216 static void 217 lex_word(void) 218 { 219 for (;;) { 220 if (is_identifier_part(inp.st[0])) 221 token_add_char(*inp.st++); 222 else if (inp.st[0] == '\\' && inp.st[1] == '\n') { 223 inp.st++; 224 inp_skip(); 225 line_no++; 226 } else 227 return; 228 } 229 } 230 231 static void 232 lex_char_or_string(void) 233 { 234 for (char delim = token.mem[token.len - 1];;) { 235 if (inp.st[0] == '\n') { 236 diag(1, "Unterminated literal"); 237 return; 238 } 239 240 token_add_char(*inp.st++); 241 if (token.mem[token.len - 1] == delim) 242 return; 243 244 if (token.mem[token.len - 1] == '\\') { 245 if (inp.st[0] == '\n') 246 ++line_no; 247 token_add_char(inp_next()); 248 } 249 } 250 } 251 252 /* Guess whether the current token is a declared type. */ 253 static bool 254 probably_typename(void) 255 { 256 if (ps.prev_lsym == lsym_modifier) 257 return true; 258 if (ps.block_init) 259 return false; 260 if (ps.in_stmt_or_decl) /* XXX: this condition looks incorrect */ 261 return false; 262 if (inp.st[0] == '*' && inp.st[1] != '=') 263 goto maybe; 264 /* XXX: is_identifier_start */ 265 if (ch_isalpha(inp.st[0])) 266 goto maybe; 267 return false; 268 maybe: 269 return ps.prev_lsym == lsym_semicolon || 270 ps.prev_lsym == lsym_lbrace || 271 ps.prev_lsym == lsym_rbrace; 272 } 273 274 static int 275 bsearch_typenames(const char *key) 276 { 277 const char **arr = typenames.items; 278 int lo = 0; 279 int hi = (int)typenames.len - 1; 280 281 while (lo <= hi) { 282 int mid = (int)((unsigned)(lo + hi) >> 1); 283 int cmp = strcmp(arr[mid], key); 284 if (cmp < 0) 285 lo = mid + 1; 286 else if (cmp > 0) 287 hi = mid - 1; 288 else 289 return mid; 290 } 291 return -(lo + 1); 292 } 293 294 static bool 295 is_typename(void) 296 { 297 if (opt.auto_typedefs && 298 token.len >= 2 && memcmp(token.mem + token.len - 2, "_t", 2) == 0) 299 return true; 300 301 return bsearch_typenames(token.st) >= 0; 302 } 303 304 static int 305 cmp_keyword_by_name(const void *key, const void *elem) 306 { 307 return strcmp(key, ((const struct keyword *)elem)->name); 308 } 309 310 /* 311 * Looking at something like 'function_name(...)' in a line, guess whether 312 * this starts a function definition or a declaration. 313 */ 314 static bool 315 probably_looking_at_definition(void) 316 { 317 int paren_level = 0; 318 for (const char *p = inp.st; *p != '\n'; p++) { 319 if (*p == '(') 320 paren_level++; 321 if (*p == ')' && --paren_level == 0) { 322 p++; 323 324 while (*p != '\n' 325 && (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_lsym == lsym_tag 376 || ps.prev_lsym == lsym_typedef; 377 378 if (ps.prev_lsym == 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 lexer_symbol lsym = lsym_word; 386 if (kw != NULL) { 387 if (kw->lsym == lsym_type) 388 lsym = lsym_type_in_parentheses; 389 ps.next_unary = true; 390 if (kw->lsym == lsym_tag || kw->lsym == lsym_type) 391 goto found_typename; 392 return kw->lsym; 393 } 394 395 if (is_typename()) { 396 lsym = lsym_type_in_parentheses; 397 ps.next_unary = true; 398 found_typename: 399 if (ps.nparen > 0) { 400 /* inside parentheses: cast, param list, offsetof or 401 * sizeof */ 402 if (ps.paren[ps.nparen - 1].cast == cast_unknown) 403 ps.paren[ps.nparen - 1].cast = cast_maybe; 404 } 405 if (ps.prev_lsym != lsym_period 406 && ps.prev_lsym != lsym_unary_op) { 407 if (kw != NULL && kw->lsym == lsym_tag) 408 return lsym_tag; 409 if (ps.nparen == 0) 410 return lsym_type_outside_parentheses; 411 } 412 } 413 414 if (inp.st[0] == '(' && ps.tos <= 1 && ps.ind_level == 0 && 415 !ps.in_func_def_params && !ps.block_init) { 416 417 if (ps.nparen == 0 && probably_looking_at_definition()) { 418 ps.is_function_definition = true; 419 if (ps.in_decl) 420 ps.in_func_def_params = true; 421 return lsym_funcname; 422 } 423 424 } else if (ps.nparen == 0 && probably_typename()) { 425 ps.next_unary = true; 426 return lsym_type_outside_parentheses; 427 } 428 429 return lsym; 430 } 431 432 static bool 433 is_asterisk_unary(void) 434 { 435 if (ps.decl_ptr == dp_word) 436 return true; 437 if (ps.next_unary || ps.in_func_def_params) 438 return true; 439 if (ps.prev_lsym == lsym_word || 440 ps.prev_lsym == lsym_rparen || 441 ps.prev_lsym == lsym_rbracket) 442 return false; 443 return ps.in_decl && ps.nparen > 0; 444 } 445 446 static bool 447 probably_in_function_definition(void) 448 { 449 for (const char *tp = inp.st; *tp != '\n';) { 450 if (ch_isspace(*tp)) 451 tp++; 452 else if (is_identifier_start(*tp)) { 453 tp++; 454 while (is_identifier_part(*tp)) 455 tp++; 456 } else 457 return *tp == '('; 458 } 459 return false; 460 } 461 462 static void 463 lex_asterisk_unary(void) 464 { 465 while (inp.st[0] == '*' || ch_isspace(inp.st[0])) { 466 if (inp.st[0] == '*') 467 token_add_char('*'); 468 inp_skip(); 469 } 470 471 if (ps.in_decl && probably_in_function_definition()) 472 ps.is_function_definition = true; 473 } 474 475 static void 476 skip_blank(const char **pp) 477 { 478 while (ch_isblank(**pp)) 479 (*pp)++; 480 } 481 482 static bool 483 skip_string(const char **pp, const char *s) 484 { 485 size_t len = strlen(s); 486 if (strncmp(*pp, s, len) == 0) { 487 *pp += len; 488 return true; 489 } 490 return false; 491 } 492 493 static void 494 lex_indent_comment(void) 495 { 496 const char *p = inp.mem; 497 498 skip_blank(&p); 499 if (!skip_string(&p, "/*")) 500 return; 501 skip_blank(&p); 502 if (!skip_string(&p, "INDENT")) 503 return; 504 505 enum indent_enabled enabled; 506 skip_blank(&p); 507 if (*p == '*' || skip_string(&p, "ON")) 508 enabled = indent_last_off_line; 509 else if (skip_string(&p, "OFF")) 510 enabled = indent_off; 511 else 512 return; 513 514 skip_blank(&p); 515 if (!skip_string(&p, "*/\n")) 516 return; 517 518 if (lab.len > 0 || code.len > 0 || com.len > 0) 519 output_line(); 520 521 indent_enabled = enabled; 522 } 523 524 /* Reads the next token, placing it in the global variable "token". */ 525 lexer_symbol 526 lexi(void) 527 { 528 token.len = 0; 529 ps.curr_col_1 = ps.next_col_1; 530 ps.next_col_1 = false; 531 532 for (;;) { 533 if (ch_isblank(inp.st[0])) { 534 ps.curr_col_1 = false; 535 inp.st++; 536 } else if (inp.st[0] == '\\' && inp.st[1] == '\n') { 537 inp.st++; 538 inp_skip(); 539 line_no++; 540 } else 541 break; 542 } 543 544 lexer_symbol alnum_lsym = lexi_alnum(); 545 if (alnum_lsym != lsym_eof) 546 return alnum_lsym; 547 548 /* Scan a non-alphanumeric token */ 549 550 token_add_char(inp_next()); 551 552 lexer_symbol lsym; 553 bool next_unary; 554 555 switch (token.mem[token.len - 1]) { 556 557 /* INDENT OFF */ 558 case '(': lsym = lsym_lparen; next_unary = true; break; 559 case '[': lsym = lsym_lbracket; next_unary = true; break; 560 case ')': lsym = lsym_rparen; next_unary = false; break; 561 case ']': lsym = lsym_rbracket; next_unary = false; break; 562 case '?': lsym = lsym_question; next_unary = true; break; 563 case ';': lsym = lsym_semicolon; next_unary = true; break; 564 case '{': lsym = lsym_lbrace; next_unary = true; break; 565 case '}': lsym = lsym_rbrace; next_unary = true; break; 566 case ',': lsym = lsym_comma; next_unary = true; break; 567 case '.': lsym = lsym_period; next_unary = false; break; 568 /* INDENT ON */ 569 570 case ':': 571 lsym = ps.quest_level > 0 572 ? (ps.quest_level--, lsym_colon_question) 573 : ps.init_or_struct 574 ? lsym_colon_other 575 : lsym_colon_label; 576 next_unary = true; 577 break; 578 579 case '\n': 580 /* if data has been exhausted, the '\n' is a dummy. */ 581 lsym = had_eof ? lsym_eof : lsym_newline; 582 next_unary = ps.next_unary; 583 ps.next_col_1 = true; 584 break; 585 586 case '#': 587 lsym = lsym_preprocessing; 588 next_unary = ps.next_unary; 589 break; 590 591 case '\'': 592 case '"': 593 lex_char_or_string(); 594 lsym = lsym_word; 595 next_unary = false; 596 break; 597 598 case '-': 599 case '+': 600 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 601 next_unary = true; 602 603 /* '++' or '--' */ 604 if (inp.st[0] == token.mem[token.len - 1]) { 605 token_add_char(*inp.st++); 606 if (ps.prev_lsym == lsym_word || 607 ps.prev_lsym == lsym_rparen || 608 ps.prev_lsym == lsym_rbracket) { 609 lsym = ps.next_unary 610 ? lsym_unary_op : lsym_postfix_op; 611 next_unary = false; 612 } 613 614 } else if (inp.st[0] == '=') { /* '+=' or '-=' */ 615 token_add_char(*inp.st++); 616 617 } else if (inp.st[0] == '>') { /* '->' */ 618 token_add_char(*inp.st++); 619 lsym = lsym_unary_op; 620 next_unary = false; 621 ps.want_blank = false; 622 } 623 break; 624 625 case '=': 626 if (ps.init_or_struct) 627 ps.block_init = true; 628 if (inp.st[0] == '=') 629 token_add_char(*inp.st++); 630 lsym = lsym_binary_op; 631 next_unary = true; 632 break; 633 634 case '>': 635 case '<': 636 case '!': /* ops like <, <<, <=, !=, etc */ 637 if (inp.st[0] == '>' || inp.st[0] == '<' || inp.st[0] == '=') 638 token_add_char(*inp.st++); 639 if (inp.st[0] == '=') 640 token_add_char(*inp.st++); 641 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 642 next_unary = true; 643 break; 644 645 case '*': 646 if (is_asterisk_unary()) { 647 lex_asterisk_unary(); 648 lsym = lsym_unary_op; 649 next_unary = true; 650 } else { 651 if (inp.st[0] == '=') 652 token_add_char(*inp.st++); 653 lsym = lsym_binary_op; 654 next_unary = true; 655 } 656 break; 657 658 default: 659 if (token.mem[token.len - 1] == '/' 660 && (inp.st[0] == '*' || inp.st[0] == '/')) { 661 enum indent_enabled prev = indent_enabled; 662 lex_indent_comment(); 663 if (prev == indent_on && indent_enabled == indent_off) 664 out.indent_off_text.len = 0; 665 token_add_char(*inp.st++); 666 lsym = lsym_comment; 667 next_unary = ps.next_unary; 668 break; 669 } 670 671 /* things like '||', '&&', '<<=', 'int *****i' */ 672 while (inp.st[0] == token.mem[token.len - 1] 673 || inp.st[0] == '=') 674 token_add_char(*inp.st++); 675 676 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 677 next_unary = true; 678 } 679 680 ps.next_unary = next_unary; 681 682 return lsym; 683 } 684 685 void 686 register_typename(const char *name) 687 { 688 if (typenames.len >= typenames.cap) { 689 typenames.cap = 16 + 2 * typenames.cap; 690 typenames.items = nonnull(realloc(typenames.items, 691 sizeof(typenames.items[0]) * typenames.cap)); 692 } 693 694 int pos = bsearch_typenames(name); 695 if (pos >= 0) 696 return; /* already in the list */ 697 698 pos = -(pos + 1); 699 memmove(typenames.items + pos + 1, typenames.items + pos, 700 sizeof(typenames.items[0]) * (typenames.len++ - (unsigned)pos)); 701 typenames.items[pos] = nonnull(strdup(name)); 702 } 703
Indexes created Wed Sep 24 15:09:51 GMT 2025