lexi.c revision 1.196
1 /* $NetBSD: lexi.c,v 1.196 2023/05/16 12:46:43 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 1.196 2023/05/16 12:46:43 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_peek(); 182 if (ch == '\\' && inp_lookahead(1) == '\n') { 183 inp_skip(); 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_peek())) 222 token_add_char(inp_next()); 223 else if (inp_peek() == '\\' && inp_lookahead(1) == '\n') { 224 inp_skip(); 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_peek() == '\n') { 237 diag(1, "Unterminated literal"); 238 return; 239 } 240 241 token_add_char(inp_next()); 242 if (token.mem[token.len - 1] == delim) 243 return; 244 245 if (token.mem[token.len - 1] == '\\') { 246 if (inp_peek() == '\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_peek() == '*' && inp_lookahead(1) != '=') 264 goto maybe; 265 /* XXX: is_identifier_start */ 266 if (ch_isalpha(inp_peek())) 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_p(); *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 else 337 break; /* func(...) { ... */ 338 } 339 } 340 341 /* 342 * To further reduce the cases where indent wrongly treats an incomplete 343 * function declaration as a function definition, thus adding a newline 344 * before the function name, it may be worth looking for parameter names, 345 * as these are often omitted in function declarations and only included 346 * in function definitions. Or just increase the lookahead to more than 347 * just the current line of input, until the next '{'. 348 */ 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_peek()) || 357 (inp_peek() == '.' && ch_isdigit(inp_lookahead(1)))) { 358 lex_number(); 359 } else if (is_identifier_start(inp_peek())) { 360 lex_word(); 361 362 if (token.len == 1 && token.st[0] == 'L' && 363 (inp_peek() == '"' || inp_peek() == '\'')) { 364 token_add_char(inp_next()); 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_peek())) 373 inp_skip(); 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 sizeof */ 404 if (ps.paren[ps.nparen - 1].cast == cast_unknown) 405 ps.paren[ps.nparen - 1].cast = cast_maybe; 406 } 407 if (ps.prev_token != lsym_period && ps.prev_token != lsym_unary_op) { 408 if (kw != NULL && kw->lsym == lsym_tag) { 409 if (token.st[0] == 'e' /* enum */) 410 ps.in_enum = in_enum_enum; 411 return lsym_tag; 412 } 413 if (ps.nparen == 0) 414 return lsym_type_outside_parentheses; 415 } 416 } 417 418 if (inp_peek() == '(' && ps.tos <= 1 && ps.ind_level == 0 && 419 !ps.in_func_def_params && !ps.block_init) { 420 421 if (ps.nparen == 0 && probably_looking_at_definition()) { 422 ps.is_function_definition = true; 423 if (ps.in_decl) 424 ps.in_func_def_params = true; 425 return lsym_funcname; 426 } 427 428 } else if (ps.nparen == 0 && probably_typename()) { 429 ps.next_unary = true; 430 return lsym_type_outside_parentheses; 431 } 432 433 return is_type ? lsym_type_in_parentheses : lsym_word; 434 } 435 436 static bool 437 is_asterisk_unary(void) 438 { 439 if (ps.next_unary || ps.in_func_def_params) 440 return true; 441 if (ps.prev_token == lsym_word || 442 ps.prev_token == lsym_rparen_or_rbracket) 443 return false; 444 return ps.in_decl && ps.nparen > 0; 445 } 446 447 static void 448 lex_asterisk_unary(void) 449 { 450 while (inp_peek() == '*' || ch_isspace(inp_peek())) { 451 if (inp_peek() == '*') 452 token_add_char('*'); 453 inp_skip(); 454 } 455 456 if (ps.in_decl) { 457 for (const char *tp = inp_p(); *tp != '\n';) { 458 if (ch_isspace(*tp)) 459 tp++; 460 else if (is_identifier_start(*tp)) { 461 tp++; 462 while (is_identifier_part(*tp)) 463 tp++; 464 } else { 465 if (*tp == '(') 466 ps.is_function_definition = true; 467 break; 468 } 469 } 470 } 471 } 472 473 static void 474 skip_blank(const char **pp) 475 { 476 while (ch_isblank(**pp)) 477 (*pp)++; 478 } 479 480 static bool 481 skip_string(const char **pp, const char *s) 482 { 483 size_t len = strlen(s); 484 if (strncmp(*pp, s, len) == 0) { 485 *pp += len; 486 return true; 487 } 488 return false; 489 } 490 491 static void 492 lex_indent_comment(void) 493 { 494 const char *p = inp_line_start(); 495 496 skip_blank(&p); 497 if (!skip_string(&p, "/*")) 498 return; 499 skip_blank(&p); 500 if (!skip_string(&p, "INDENT")) 501 return; 502 503 enum indent_enabled enabled; 504 skip_blank(&p); 505 if (*p == '*' || skip_string(&p, "ON")) 506 enabled = indent_last_off_line; 507 else if (skip_string(&p, "OFF")) 508 enabled = indent_off; 509 else 510 return; 511 512 skip_blank(&p); 513 if (!skip_string(&p, "*/\n")) 514 return; 515 516 if (lab.len > 0 || code.len > 0 || com.len > 0) 517 output_line(); 518 519 indent_enabled = enabled; 520 } 521 522 /* Reads the next token, placing it in the global variable "token". */ 523 lexer_symbol 524 lexi(void) 525 { 526 token.len = 0; 527 ps.curr_col_1 = ps.next_col_1; 528 ps.next_col_1 = false; 529 530 for (;;) { 531 if (ch_isblank(inp_peek())) { 532 ps.curr_col_1 = false; 533 inp_skip(); 534 } else if (inp_peek() == '\\' && inp_lookahead(1) == '\n') { 535 inp_skip(); 536 inp_skip(); 537 line_no++; 538 } else 539 break; 540 } 541 542 lexer_symbol alnum_lsym = lexi_alnum(); 543 if (alnum_lsym != lsym_eof) { 544 debug_parser_state(alnum_lsym); 545 return alnum_lsym; 546 } 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 '(': 559 case '[': lsym = lsym_lparen_or_lbracket; next_unary = true; break; 560 case ')': 561 case ']': lsym = lsym_rparen_or_rbracket; next_unary = false; break; 562 case '?': lsym = lsym_question; next_unary = true; break; 563 case ':': lsym = lsym_colon; next_unary = true; break; 564 case ';': lsym = lsym_semicolon; next_unary = true; break; 565 case '{': lsym = lsym_lbrace; next_unary = true; break; 566 case '}': lsym = lsym_rbrace; next_unary = true; break; 567 case ',': lsym = lsym_comma; next_unary = true; break; 568 case '.': lsym = lsym_period; next_unary = false; break; 569 /* INDENT ON */ 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 ps.next_col_1 = true; 576 break; 577 578 case '#': 579 lsym = lsym_preprocessing; 580 next_unary = ps.next_unary; 581 break; 582 583 case '\'': 584 case '"': 585 lex_char_or_string(); 586 lsym = lsym_word; 587 next_unary = false; 588 break; 589 590 case '-': 591 case '+': 592 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 593 next_unary = true; 594 595 if (inp_peek() == token.mem[token.len - 1]) { /* '++' or '--' */ 596 token_add_char(inp_next()); 597 if (ps.prev_token == lsym_word || 598 ps.prev_token == lsym_rparen_or_rbracket) { 599 lsym = ps.next_unary ? lsym_unary_op : lsym_postfix_op; 600 next_unary = false; 601 } 602 603 } else if (inp_peek() == '=') { /* '+=' or '-=' */ 604 token_add_char(inp_next()); 605 606 } else if (inp_peek() == '>') { /* '->' */ 607 token_add_char(inp_next()); 608 lsym = lsym_unary_op; 609 next_unary = false; 610 ps.want_blank = false; 611 } 612 break; 613 614 case '=': 615 if (ps.init_or_struct) 616 ps.block_init = true; 617 if (inp_peek() == '=') 618 token_add_char(inp_next()); 619 lsym = lsym_binary_op; 620 next_unary = true; 621 break; 622 623 case '>': 624 case '<': 625 case '!': /* ops like <, <<, <=, !=, etc */ 626 if (inp_peek() == '>' || inp_peek() == '<' || inp_peek() == '=') 627 token_add_char(inp_next()); 628 if (inp_peek() == '=') 629 token_add_char(inp_next()); 630 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 631 next_unary = true; 632 break; 633 634 case '*': 635 if (is_asterisk_unary()) { 636 lex_asterisk_unary(); 637 lsym = lsym_unary_op; 638 next_unary = true; 639 } else { 640 if (inp_peek() == '=') 641 token_add_char(inp_next()); 642 lsym = lsym_binary_op; 643 next_unary = true; 644 } 645 break; 646 647 default: 648 if (token.mem[token.len - 1] == '/' 649 && (inp_peek() == '*' || inp_peek() == '/')) { 650 enum indent_enabled prev = indent_enabled; 651 lex_indent_comment(); 652 if (prev == indent_on && indent_enabled == indent_off) 653 clear_indent_off_text(); 654 token_add_char(inp_next()); 655 lsym = lsym_comment; 656 next_unary = ps.next_unary; 657 break; 658 } 659 660 /* handle '||', '&&', etc., and also things as in 'int *****i' */ 661 while (token.mem[token.len - 1] == inp_peek() || inp_peek() == '=') 662 token_add_char(inp_next()); 663 664 lsym = ps.next_unary ? lsym_unary_op : lsym_binary_op; 665 next_unary = true; 666 } 667 668 if (ps.in_enum == in_enum_enum || ps.in_enum == in_enum_type) 669 ps.in_enum = lsym == lsym_lbrace ? in_enum_brace : in_enum_no; 670 if (lsym == lsym_rbrace) 671 ps.in_enum = in_enum_no; 672 673 ps.next_unary = next_unary; 674 675 debug_parser_state(lsym); 676 return lsym; 677 } 678 679 void 680 register_typename(const char *name) 681 { 682 if (typenames.len >= typenames.cap) { 683 typenames.cap = 16 + 2 * typenames.cap; 684 typenames.items = nonnull(realloc(typenames.items, 685 sizeof(typenames.items[0]) * typenames.cap)); 686 } 687 688 int pos = bsearch_typenames(name); 689 if (pos >= 0) 690 return; /* already in the list */ 691 692 pos = -(pos + 1); 693 memmove(typenames.items + pos + 1, typenames.items + pos, 694 sizeof(typenames.items[0]) * (typenames.len++ - (unsigned)pos)); 695 typenames.items[pos] = nonnull(strdup(name)); 696 } 697
Indexes created Mon Sep 22 13:09:51 GMT 2025