lexi.c revision 1.233
1 /* $NetBSD: lexi.c,v 1.233 2023/06/25 18:41:03 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 1.233 2023/06/25 18:41:03 rillig Exp $"); 42 43 #include <stdlib.h> 44 #include <string.h> 45 46 #include "indent.h" 47 48 /* must be sorted alphabetically, is used in binary search */ 49 static const struct keyword { 50 const char name[12]; 51 lexer_symbol lsym; 52 } keywords[] = { 53 {"_Bool", lsym_type}, 54 {"_Complex", lsym_type}, 55 {"_Imaginary", lsym_type}, 56 {"auto", lsym_modifier}, 57 {"bool", lsym_type}, 58 {"break", lsym_word}, 59 {"case", lsym_case}, 60 {"char", lsym_type}, 61 {"complex", lsym_type}, 62 {"const", lsym_modifier}, 63 {"continue", lsym_word}, 64 {"default", lsym_default}, 65 {"do", lsym_do}, 66 {"double", lsym_type}, 67 {"else", lsym_else}, 68 {"enum", lsym_tag}, 69 {"extern", lsym_modifier}, 70 {"float", lsym_type}, 71 {"for", lsym_for}, 72 {"goto", lsym_word}, 73 {"if", lsym_if}, 74 {"imaginary", lsym_type}, 75 {"inline", lsym_modifier}, 76 {"int", lsym_type}, 77 {"long", lsym_type}, 78 {"offsetof", lsym_offsetof}, 79 {"register", lsym_modifier}, 80 {"restrict", lsym_word}, 81 {"return", lsym_return}, 82 {"short", lsym_type}, 83 {"signed", lsym_type}, 84 {"sizeof", lsym_sizeof}, 85 {"static", lsym_modifier}, 86 {"struct", lsym_tag}, 87 {"switch", lsym_switch}, 88 {"typedef", lsym_typedef}, 89 {"union", lsym_tag}, 90 {"unsigned", lsym_type}, 91 {"void", lsym_type}, 92 {"volatile", lsym_modifier}, 93 {"while", lsym_while} 94 }; 95 96 static struct { 97 const char **items; 98 unsigned int len; 99 unsigned int cap; 100 } typenames; 101 102 /*- 103 * The transition table below was rewritten by hand from lx's output, given 104 * the following definitions. lx is Katherine Flavel's lexer generator. 105 * 106 * O = /[0-7]/; D = /[0-9]/; NZ = /[1-9]/; 107 * H = /[a-f0-9]/i; B = /[0-1]/; HP = /0x/i; 108 * BP = /0b/i; E = /e[+\-]?/i D+; P = /p[+\-]?/i D+; 109 * FS = /[fl]/i; IS = /u/i /(l|L|ll|LL)/? | /(l|L|ll|LL)/ /u/i?; 110 * 111 * D+ E FS? -> $float; 112 * D* "." D+ E? FS? -> $float; 113 * D+ "." E? FS? -> $float; HP H+ IS? -> $int; 114 * HP H+ P FS? -> $float; NZ D* IS? -> $int; 115 * HP H* "." H+ P FS? -> $float; "0" O* IS? -> $int; 116 * HP H+ "." P FS -> $float; BP B+ IS? -> $int; 117 */ 118 /* INDENT OFF */ 119 static const unsigned char lex_number_state[][26] = { 120 /* examples: 121 00 122 s 0xx 123 t 00xaa 124 a 11 101100xxa.. 125 r 11ee0001101lbuuxx.a.pp 126 t.01.e+008bLuxll0Ll.aa.p+0 127 states: ABCDEFGHIJKLMNOPQRSTUVWXYZ */ 128 [0] = "uuiifuufiuuiiuiiiiiuiuuuuu", /* (other) */ 129 [1] = "CEIDEHHHIJQ U Q VUVVZZZ", /* 0 */ 130 [2] = "DEIDEHHHIJQ U Q VUVVZZZ", /* 1 */ 131 [3] = "DEIDEHHHIJ U VUVVZZZ", /* 2 3 4 5 6 7 */ 132 [4] = "DEJDEHHHJJ U VUVVZZZ", /* 8 9 */ 133 [5] = " U VUVV ", /* A a C c D d */ 134 [6] = " K U VUVV ", /* B b */ 135 [7] = " FFF FF U VUVV ", /* E e */ 136 [8] = " f f U VUVV f", /* F f */ 137 [9] = " LLf fL PR Li L f", /* L */ 138 [10] = " OOf fO S P O i O f", /* l */ 139 [11] = " FFX ", /* P p */ 140 [12] = " MM M i iiM M ", /* U u */ 141 [13] = " N ", /* X x */ 142 [14] = " G Y ", /* + - */ 143 [15] = "B EE EE T W ", /* . */ 144 /* ABCDEFGHIJKLMNOPQRSTUVWXYZ */ 145 }; 146 /* INDENT ON */ 147 148 static const unsigned char lex_number_row[] = { 149 ['0'] = 1, 150 ['1'] = 2, 151 ['2'] = 3, ['3'] = 3, ['4'] = 3, ['5'] = 3, ['6'] = 3, ['7'] = 3, 152 ['8'] = 4, ['9'] = 4, 153 ['A'] = 5, ['a'] = 5, ['C'] = 5, ['c'] = 5, ['D'] = 5, ['d'] = 5, 154 ['B'] = 6, ['b'] = 6, 155 ['E'] = 7, ['e'] = 7, 156 ['F'] = 8, ['f'] = 8, 157 ['L'] = 9, 158 ['l'] = 10, 159 ['P'] = 11, ['p'] = 11, 160 ['U'] = 12, ['u'] = 12, 161 ['X'] = 13, ['x'] = 13, 162 ['+'] = 14, ['-'] = 14, 163 ['.'] = 15, 164 }; 165 166 167 static bool 168 is_identifier_start(char ch) 169 { 170 return ch_isalpha(ch) || ch == '_' || ch == '$'; 171 } 172 173 static bool 174 is_identifier_part(char ch) 175 { 176 return ch_isalnum(ch) || ch == '_' || ch == '$'; 177 } 178 179 static void 180 token_add_char(char ch) 181 { 182 buf_add_char(&token, ch); 183 } 184 185 static bool 186 skip_line_continuation(void) 187 { 188 if (inp_p[0] == '\\' && inp_p[1] == '\n') { 189 inp_p++; 190 inp_skip(); 191 line_no++; 192 return true; 193 } 194 return false; 195 } 196 197 static void 198 lex_number(void) 199 { 200 for (unsigned char s = 'A'; s != 'f' && s != 'i' && s != 'u';) { 201 unsigned char ch = (unsigned char)*inp_p; 202 if (skip_line_continuation()) 203 continue; 204 if (ch >= array_length(lex_number_row) 205 || lex_number_row[ch] == 0) 206 break; 207 208 unsigned char row = lex_number_row[ch]; 209 if (lex_number_state[row][s - 'A'] == ' ') { 210 // lex_number_state[0][s - 'A'] now indicates the type: 211 // f = floating, i = integer, u = unknown 212 return; 213 } 214 215 s = lex_number_state[row][s - 'A']; 216 token_add_char(inp_next()); 217 } 218 } 219 220 static void 221 lex_word(void) 222 { 223 for (;;) { 224 if (is_identifier_part(*inp_p)) 225 token_add_char(*inp_p++); 226 else if (skip_line_continuation()) 227 continue; 228 else 229 return; 230 } 231 } 232 233 static void 234 lex_char_or_string(void) 235 { 236 for (char delim = token.s[token.len - 1];;) { 237 if (*inp_p == '\n') { 238 diag(1, "Unterminated literal"); 239 return; 240 } 241 242 token_add_char(*inp_p++); 243 if (token.s[token.len - 1] == delim) 244 return; 245 246 if (token.s[token.len - 1] == '\\') { 247 if (*inp_p == '\n') 248 line_no++; 249 token_add_char(inp_next()); 250 } 251 } 252 } 253 254 /* Guess whether the current token is a declared type. */ 255 static bool 256 probably_typename(void) 257 { 258 if (ps.prev_lsym == lsym_modifier) 259 return true; 260 if (ps.in_init) 261 return false; 262 if (ps.in_stmt_or_decl) /* XXX: this condition looks incorrect */ 263 return false; 264 if (ps.prev_lsym == lsym_semicolon 265 || ps.prev_lsym == lsym_lbrace 266 || ps.prev_lsym == lsym_rbrace) { 267 if (inp_p[0] == '*' && inp_p[1] != '=') 268 return true; 269 /* XXX: is_identifier_start */ 270 if (ch_isalpha(inp_p[0])) 271 return true; 272 } 273 return false; 274 } 275 276 static int 277 bsearch_typenames(const char *key) 278 { 279 const char **arr = typenames.items; 280 unsigned lo = 0; 281 unsigned hi = typenames.len; 282 283 while (lo < hi) { 284 unsigned mid = (lo + hi) / 2; 285 int cmp = strcmp(arr[mid], key); 286 if (cmp < 0) 287 lo = mid + 1; 288 else if (cmp > 0) 289 hi = mid; 290 else 291 return (int)mid; 292 } 293 return -1 - (int)lo; 294 } 295 296 static bool 297 is_typename(void) 298 { 299 if (opt.auto_typedefs && 300 token.len >= 2 && memcmp(token.s + token.len - 2, "_t", 2) == 0) 301 return true; 302 303 return bsearch_typenames(token.s) >= 0; 304 } 305 306 void 307 register_typename(const char *name) 308 { 309 if (typenames.len >= typenames.cap) { 310 typenames.cap = 16 + 2 * typenames.cap; 311 typenames.items = nonnull(realloc(typenames.items, 312 sizeof(typenames.items[0]) * typenames.cap)); 313 } 314 315 int pos = bsearch_typenames(name); 316 if (pos >= 0) 317 return; /* already in the list */ 318 319 pos = -1 - pos; 320 memmove(typenames.items + pos + 1, typenames.items + pos, 321 sizeof(typenames.items[0]) * (typenames.len++ - (unsigned)pos)); 322 typenames.items[pos] = nonnull(strdup(name)); 323 } 324 325 static int 326 cmp_keyword_by_name(const void *key, const void *elem) 327 { 328 return strcmp(key, ((const struct keyword *)elem)->name); 329 } 330 331 /* 332 * Looking at the '(', guess whether this starts a function definition or a 333 * function declaration. 334 */ 335 static bool 336 probably_function_definition(void) 337 { 338 int paren_level = 0; 339 for (const char *p = inp_p; *p != '\n'; p++) { 340 if (*p == '(') 341 paren_level++; 342 if (*p == ')' && --paren_level == 0) { 343 p++; 344 345 while (*p != '\n' 346 && (ch_isspace(*p) || is_identifier_part(*p))) 347 p++; /* '__dead' or '__unused' */ 348 349 if (*p == '\n') /* func(...) */ 350 break; 351 if (*p == ';') /* func(...); */ 352 return false; 353 if (*p == ',') /* double abs(), pi; */ 354 return false; 355 if (*p == '(') /* func(...) __attribute__((...)) */ 356 paren_level++; /* func(...) __printflike(...) 357 */ 358 else 359 break; /* func(...) { ... */ 360 } 361 362 if (paren_level == 1 && p[0] == '*' && p[1] == ',') 363 return false; 364 } 365 366 /* 367 * To further reduce the cases where indent wrongly treats an 368 * incomplete function declaration as a function definition, thus 369 * adding a newline before the function name, it may be worth looking 370 * for parameter names, as these are often omitted in function 371 * declarations and only included in function definitions. Or just 372 * increase the lookahead to more than just the current line of input, 373 * until the next '{'. 374 */ 375 return true; 376 } 377 378 static lexer_symbol 379 lexi_alnum(void) 380 { 381 if (ch_isdigit(inp_p[0]) || 382 (inp_p[0] == '.' && ch_isdigit(inp_p[1]))) { 383 lex_number(); 384 } else if (is_identifier_start(inp_p[0])) { 385 lex_word(); 386 387 if (token.len == 1 && token.s[0] == 'L' && 388 (inp_p[0] == '"' || inp_p[0] == '\'')) { 389 token_add_char(*inp_p++); 390 lex_char_or_string(); 391 ps.next_unary = false; 392 return lsym_word; 393 } 394 } else 395 return lsym_eof; /* just as a placeholder */ 396 397 while (ch_isblank(*inp_p)) 398 inp_p++; 399 400 ps.next_unary = ps.prev_lsym == lsym_tag 401 || ps.prev_lsym == lsym_typedef; 402 403 if (ps.prev_lsym == lsym_tag && ps.paren.len == 0) 404 return lsym_type; 405 406 token_add_char('\0'); // Terminate in non-debug mode as well. 407 token.len--; 408 const struct keyword *kw = bsearch(token.s, keywords, 409 array_length(keywords), sizeof(keywords[0]), cmp_keyword_by_name); 410 lexer_symbol lsym = lsym_word; 411 if (kw != NULL) { 412 if (kw->lsym == lsym_type) 413 lsym = lsym_type; 414 ps.next_unary = true; 415 if (kw->lsym == lsym_tag || kw->lsym == lsym_type) 416 goto found_typename; 417 return kw->lsym; 418 } 419 420 if (is_typename()) { 421 lsym = lsym_type; 422 ps.next_unary = true; 423 found_typename: 424 if (ps.paren.len > 0) { 425 /* inside parentheses: cast, param list, offsetof or 426 * sizeof */ 427 struct paren_level *paren_level = 428 ps.paren.item + ps.paren.len - 1; 429 if (paren_level->cast == cast_unknown) 430 paren_level->cast = cast_maybe; 431 } 432 if (ps.prev_lsym != lsym_period 433 && ps.prev_lsym != lsym_unary_op) { 434 if (kw != NULL && kw->lsym == lsym_tag) 435 return lsym_tag; 436 if (ps.paren.len == 0) 437 return lsym_type; 438 } 439 } 440 441 if (*inp_p == '(' && ps.psyms.len < 3 && ps.ind_level == 0 && 442 !ps.in_func_def_params && !ps.in_init) { 443 444 if (ps.paren.len == 0 && probably_function_definition()) { 445 ps.line_has_func_def = true; 446 if (ps.in_decl) 447 ps.in_func_def_params = true; 448 return lsym_funcname; 449 } 450 451 } else if (ps.paren.len == 0 && probably_typename()) { 452 ps.next_unary = true; 453 return lsym_type; 454 } 455 456 return lsym; 457 } 458 459 static bool 460 is_asterisk_pointer(void) 461 { 462 const char *p = inp_p; 463 while (*p == '*' || ch_isblank(*p)) 464 p++; 465 if (*p == ')') 466 return true; 467 if (ps.next_unary || ps.in_func_def_params) 468 return true; 469 if (ps.prev_lsym == lsym_word || 470 ps.prev_lsym == lsym_rparen || 471 ps.prev_lsym == lsym_rbracket) 472 return false; 473 return ps.in_decl && ps.paren.len > 0; 474 } 475 476 static bool 477 probably_in_function_definition(void) 478 { 479 for (const char *p = inp_p; *p != '\n';) { 480 if (ch_isspace(*p)) 481 p++; 482 else if (is_identifier_start(*p)) { 483 p++; 484 while (is_identifier_part(*p)) 485 p++; 486 } else 487 return *p == '('; 488 } 489 return false; 490 } 491 492 static void 493 lex_asterisk_pointer(void) 494 { 495 while (*inp_p == '*' || ch_isspace(*inp_p)) { 496 if (*inp_p == '*') 497 token_add_char('*'); 498 inp_skip(); 499 } 500 501 if (ps.in_decl && probably_in_function_definition()) 502 ps.line_has_func_def = true; 503 } 504 505 static bool 506 skip(const char **pp, const char *s) 507 { 508 size_t len = strlen(s); 509 while (ch_isblank(**pp)) 510 (*pp)++; 511 if (strncmp(*pp, s, len) == 0) { 512 *pp += len; 513 return true; 514 } 515 return false; 516 } 517 518 static void 519 lex_indent_comment(void) 520 { 521 const char *p = inp.s; 522 if (skip(&p, "/*") && skip(&p, "INDENT")) { 523 enum indent_enabled enabled; 524 if (skip(&p, "ON") || *p == '*') 525 enabled = indent_last_off_line; 526 else if (skip(&p, "OFF")) 527 enabled = indent_off; 528 else 529 return; 530 if (skip(&p, "*/\n")) { 531 if (lab.len > 0 || code.len > 0 || com.len > 0) 532 output_line(); 533 indent_enabled = enabled; 534 } 535 } 536 } 537 538 /* Reads the next token, placing it in the global variable "token". */ 539 lexer_symbol 540 lexi(void) 541 { 542 buf_clear(&token); 543 544 for (;;) { 545 if (ch_isblank(*inp_p)) 546 inp_p++; 547 else if (skip_line_continuation()) 548 continue; 549 else 550 break; 551 } 552 553 lexer_symbol alnum_lsym = lexi_alnum(); 554 if (alnum_lsym != lsym_eof) 555 return alnum_lsym; 556 557 /* Scan a non-alphanumeric token */ 558 559 token_add_char(inp_next()); 560 561 lexer_symbol lsym; 562 bool next_unary; 563 564 switch (token.s[token.len - 1]) { 565 566 case '#': 567 lsym = lsym_preprocessing; 568 next_unary = ps.next_unary; 569 break; 570 571 case '\n': 572 /* if data has been exhausted, the '\n' is a dummy. */ 573 lsym = had_eof ? lsym_eof : lsym_newline; 574 next_unary = ps.next_unary; 575 break; 576 577 /* INDENT OFF */ 578 case '(': lsym = lsym_lparen; next_unary = true; break; 579 case ')': lsym = lsym_rparen; next_unary = false; break; 580 case '[': lsym = lsym_lbracket; next_unary = true; break; 581 case ']': lsym = lsym_rbracket; next_unary = false; break; 582 case '{': lsym = lsym_lbrace; next_unary = true; break; 583 case '}': lsym = lsym_rbrace; next_unary = true; break; 584 case '.': lsym = lsym_period; next_unary = false; break; 585 case '?': lsym = lsym_question; next_unary = true; break; 586 case ',': lsym = lsym_comma; next_unary = true; break; 587 case ';': lsym = lsym_semicolon; next_unary = true; break; 588 /* INDENT ON */ 589 590 case '+': 591 case '-': 592 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 593 next_unary = true; 594 595 /* '++' or '--' */ 596 if (*inp_p == token.s[token.len - 1]) { 597 token_add_char(*inp_p++); 598 if (ps.prev_lsym == lsym_word || 599 ps.prev_lsym == lsym_rparen || 600 ps.prev_lsym == lsym_rbracket) { 601 lsym = ps.next_unary 602 ? lsym_unary_op : lsym_postfix_op; 603 next_unary = false; 604 } 605 606 } else if (*inp_p == '=') { /* '+=' or '-=' */ 607 token_add_char(*inp_p++); 608 609 } else if (*inp_p == '>') { /* '->' */ 610 token_add_char(*inp_p++); 611 lsym = lsym_unary_op; 612 next_unary = false; 613 ps.want_blank = false; 614 } 615 break; 616 617 case ':': 618 lsym = ps.quest_level > 0 619 ? (ps.quest_level--, lsym_question_colon) 620 : ps.in_var_decl ? lsym_other_colon : lsym_label_colon; 621 next_unary = true; 622 break; 623 624 case '*': 625 if (*inp_p == '=') { 626 token_add_char(*inp_p++); 627 lsym = lsym_binary_op; 628 } else if (is_asterisk_pointer()) { 629 lex_asterisk_pointer(); 630 lsym = lsym_unary_op; 631 } else 632 lsym = lsym_binary_op; 633 next_unary = true; 634 break; 635 636 case '=': 637 if (ps.in_var_decl) 638 ps.in_init = true; 639 if (*inp_p == '=') 640 token_add_char(*inp_p++); 641 lsym = lsym_binary_op; 642 next_unary = true; 643 break; 644 645 case '>': 646 case '<': 647 case '!': /* ops like <, <<, <=, !=, etc. */ 648 if (*inp_p == '>' || *inp_p == '<' || *inp_p == '=') 649 token_add_char(*inp_p++); 650 if (*inp_p == '=') 651 token_add_char(*inp_p++); 652 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 653 next_unary = true; 654 break; 655 656 case '\'': 657 case '"': 658 lex_char_or_string(); 659 lsym = lsym_word; 660 next_unary = false; 661 break; 662 663 default: 664 if (token.s[token.len - 1] == '/' 665 && (*inp_p == '*' || *inp_p == '/')) { 666 enum indent_enabled prev = indent_enabled; 667 lex_indent_comment(); 668 if (prev == indent_on && indent_enabled == indent_off) 669 buf_clear(&out.indent_off_text); 670 token_add_char(*inp_p++); 671 lsym = lsym_comment; 672 next_unary = ps.next_unary; 673 break; 674 } 675 676 /* punctuation like '%', '&&', '/', '^', '||', '~' */ 677 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 678 if (*inp_p == token.s[token.len - 1]) 679 token_add_char(*inp_p++), lsym = lsym_binary_op; 680 if (*inp_p == '=') 681 token_add_char(*inp_p++), lsym = lsym_binary_op; 682 683 next_unary = true; 684 } 685 686 ps.next_unary = next_unary; 687 688 return lsym; 689 } 690
Indexes created Sun Sep 28 16:09:52 GMT 2025