tokenize.c

   1 /*
   2  * This is a really stupid C tokenizer. It doesn't do any include
   3  * files or anything complex at all. That's the preprocessor.
   4  *
   5  * Copyright (C) 2003 Transmeta Corp.
   6  *               2003 Linus Torvalds
   7  *
   8  *  Licensed under the Open Software License version 1.1
   9  */
  10 #include <stdio.h>
  11 #include <stdlib.h>
  12 #include <stdarg.h>
  13 #include <stddef.h>
  14 #include <string.h>
  15 #include <ctype.h>
  16 #include <unistd.h>
  17
  18 #include "lib.h"
  19 #include "allocate.h"
  20 #include "token.h"
  21 #include "symbol.h"
  22
  23 #define EOF (-1)
  24
  25 int input_stream_nr = 0;
  26 struct stream *input_streams;
  27 static int input_streams_allocated;
  28
  29 #define BUFSIZE (8192)
  30
  31 typedef struct {
  32         int fd, offset, size;
  33         int pos, line, nr;
  34         int newline, whitespace;
  35         struct token **tokenlist;
  36         struct token *token;
  37         unsigned char *buffer;
  38 } stream_t;
  39
  40 const char *stream_name(int stream)
  41 {
  42         if (stream < 0 || stream > input_stream_nr)
  43                 return "<bad stream>";
  44         return input_streams[stream].name;
  45 }
  46
  47 static struct position stream_pos(stream_t *stream)
  48 {
  49         struct position pos;
  50         pos.type = 0;
  51         pos.stream = stream->nr;
  52         pos.newline = stream->newline;
  53         pos.whitespace = stream->whitespace;
  54         pos.pos = stream->pos;
  55         pos.line = stream->line;
  56         pos.noexpand = 0;
  57         return pos;
  58 }
  59
  60 const char *show_special(int val)
  61 {
  62         static char buffer[4];
  63
  64         buffer[0] = val;
  65         buffer[1] = 0;
  66         if (val >= SPECIAL_BASE)
  67                 strcpy(buffer, (char *) combinations[val - SPECIAL_BASE]);
  68         return buffer;
  69 }
  70
  71 const char *show_ident(const struct ident *ident)
  72 {
  73         static char buffer[256];
  74         if (!ident)
  75                 return "<noident>";
  76         sprintf(buffer, "%.*s", ident->len, ident->name);
  77         return buffer;
  78 }
  79
  80 static char *charstr(char *ptr, unsigned char c, unsigned char escape, unsigned char next)
  81 {
  82         if (isprint(c)) {
  83                 if (c == escape || c == '\\')
  84                         *ptr++ = '\\';
  85                 *ptr++ = c;
  86                 return ptr;
  87         }
  88         *ptr++ = '\\';
  89         switch (c) {
  90         case '\n':
  91                 *ptr++ = 'n';
  92                 return ptr;
  93         case '\t':
  94                 *ptr++ = 't';
  95                 return ptr;
  96         }
  97         if (!isdigit(next))
  98                 return ptr + sprintf(ptr, "%o", c);
  99
 100         return ptr + sprintf(ptr, "%03o", c);
 101 }
 102
 103 const char *show_string(const struct string *string)
 104 {
 105         static char buffer[4 * MAX_STRING + 3];
 106         char *ptr;
 107         int i;
 108
 109         if (!string->length)
 110                 return "<bad_string>";
 111         ptr = buffer;
 112         *ptr++ = '"';
 113         for (i = 0; i < string->length-1; i++) {
 114                 const char *p = string->data + i;
 115                 ptr = charstr(ptr, p[0], '"', p[1]);
 116         }
 117         *ptr++ = '"';
 118         *ptr = '\0';
 119         return buffer;
 120 }
 121
 122 const char *show_token(const struct token *token)
 123 {
 124         static char buffer[256];
 125
 126         if (!token)
 127                 return "<no token>";
 128         switch (token_type(token)) {
 129         case TOKEN_ERROR:
 130                 return "syntax error";
 131
 132         case TOKEN_EOF:
 133                 return "end-of-input";
 134
 135         case TOKEN_IDENT:
 136                 return show_ident(token->ident);
 137
 138         case TOKEN_STRING:
 139                 return show_string(token->string);
 140
 141         case TOKEN_NUMBER:
 142                 return token->number;
 143
 144         case TOKEN_SPECIAL:
 145                 return show_special(token->special);
 146
 147         case TOKEN_CHAR: {
 148                 char *ptr = buffer;
 149                 int c = token->character;
 150                 *ptr++ = '\'';
 151                 ptr = charstr(ptr, c, '\'', 0);
 152                 *ptr++ = '\'';
 153                 *ptr++ = '\0';
 154                 return buffer;
 155         }
 156
 157         case TOKEN_STREAMBEGIN:
 158                 sprintf(buffer, "<beginning of '%s'>", stream_name(token->pos.stream));
 159                 return buffer;
 160
 161         case TOKEN_STREAMEND:
 162                 sprintf(buffer, "<end of '%s'>", stream_name(token->pos.stream));
 163                 return buffer;
 164
 165         default:
 166                 return "WTF???";
 167         }
 168 }
 169
 170 int init_stream(const char *name, int fd, const char **next_path)
 171 {
 172         int stream = input_stream_nr;
 173         struct stream *current;
 174
 175         if (stream >= input_streams_allocated) {
 176                 int newalloc = stream * 4 / 3 + 10;
 177                 input_streams = realloc(input_streams, newalloc * sizeof(struct stream));
 178                 if (!input_streams)
 179                         die("Unable to allocate more streams space");
 180                 input_streams_allocated = newalloc;
 181         }
 182         current = input_streams + stream;
 183         memset(current, 0, sizeof(*current));
 184         current->name = name;
 185         current->fd = fd;
 186         current->next_path = next_path;
 187         current->path = NULL;
 188         current->constant = CONSTANT_FILE_MAYBE;
 189         input_stream_nr = stream+1;
 190         return stream;
 191 }
 192
 193 static struct token * alloc_token(stream_t *stream)
 194 {
 195         struct token *token = __alloc_token(0);
 196         token->pos = stream_pos(stream);
 197         return token;
 198 }
 199
 200 /*
 201  *  Argh...  That was surprisingly messy - handling '\r' complicates the
 202  *  things a _lot_.
 203  */
 204 static int nextchar_slow(stream_t *stream)
 205 {
 206         int offset = stream->offset;
 207         int size = stream->size;
 208         int c;
 209         int spliced = 0, had_cr, had_backslash, complain;
 210
 211 restart:
 212         had_cr = had_backslash = complain = 0;
 213
 214 repeat:
 215         if (offset >= size) {
 216                 if (stream->fd < 0)
 217                         goto got_eof;
 218                 size = read(stream->fd, stream->buffer, BUFSIZE);
 219                 if (size <= 0)
 220                         goto got_eof;
 221                 stream->size = size;
 222                 stream->offset = offset = 0;
 223         }
 224
 225         c = stream->buffer[offset++];
 226
 227         if (had_cr && c != '\n')
 228                 complain = 1;
 229
 230         if (c == '\r') {
 231                 had_cr = 1;
 232                 goto repeat;
 233         }
 234
 235         stream->pos++;
 236
 237         if (c == '\n') {
 238                 stream->line++;
 239                 stream->pos = 0;
 240         }
 241
 242         if (!had_backslash) {
 243                 if (c == '\\') {
 244                         had_backslash = 1;
 245                         goto repeat;
 246                 }
 247                 if (c == '\n')
 248                         stream->newline = 1;
 249         } else {
 250                 if (c == '\n') {
 251                         if (complain)
 252                                 warning(stream_pos(stream), "non-ASCII data stream");
 253                         spliced = 1;
 254                         goto restart;
 255                 }
 256                 stream->pos--;
 257                 offset--;
 258                 c = '\\';
 259         }
 260
 261 out:
 262         stream->offset = offset;
 263         if (complain)
 264                 warning(stream_pos(stream), "non-ASCII data stream");
 265
 266         return c;
 267
 268 got_eof:
 269         if (had_backslash) {
 270                 c = '\\';
 271                 goto out;
 272         }
 273         if (stream->pos)
 274                 warning(stream_pos(stream), "no newline at end of file");
 275         else if (had_cr)
 276                 warning(stream_pos(stream), "non-ASCII data stream");
 277         else if (spliced)
 278                 warning(stream_pos(stream), "backslash-newline at end of file");
 279         return EOF;
 280 }
 281
 282 /*
 283  *  We want that as light as possible while covering all normal cases.
 284  *  Slow path (including the logics with line-splicing and EOF sanity
 285  *  checks) is in nextchar_slow().
 286  */
 287 static inline int nextchar(stream_t *stream)
 288 {
 289         int offset = stream->offset;
 290
 291         if (offset < stream->size) {
 292                 int c = stream->buffer[offset++];
 293                 static const char special[256] = {
 294                         ['\r'] = 1, ['\n'] = 1, ['\\'] = 1
 295                 };
 296                 if (!special[c]) {
 297                         stream->offset = offset;
 298                         stream->pos++;
 299                         return c;
 300                 }
 301         }
 302         return nextchar_slow(stream);
 303 }
 304
 305 struct token eof_token_entry;
 306
 307 static struct token *mark_eof(stream_t *stream)
 308 {
 309         struct token *end;
 310
 311         end = alloc_token(stream);
 312         token_type(end) = TOKEN_STREAMEND;
 313         end->pos.newline = 1;
 314
 315         eof_token_entry.next = &eof_token_entry;
 316         eof_token_entry.pos.newline = 1;
 317
 318         end->next =  &eof_token_entry;
 319         *stream->tokenlist = end;
 320         stream->tokenlist = NULL;
 321         return end;
 322 }
 323
 324 static void add_token(stream_t *stream)
 325 {
 326         struct token *token = stream->token;
 327
 328         stream->token = NULL;
 329         token->next = NULL;
 330         *stream->tokenlist = token;
 331         stream->tokenlist = &token->next;
 332 }
 333
 334 static void drop_token(stream_t *stream)
 335 {
 336         stream->newline |= stream->token->pos.newline;
 337         stream->whitespace |= stream->token->pos.whitespace;
 338         stream->token = NULL;
 339 }
 340
 341 enum {
 342         Letter = 1,
 343         Digit = 2,
 344         Hex = 4,
 345         Exp = 8,
 346         Dot = 16,
 347         ValidSecond = 32,
 348 };
 349
 350 static const long cclass[257] = {
 351         ['0' + 1 ... '9' + 1] = Digit | Hex,
 352         ['A' + 1 ... 'D' + 1] = Letter | Hex,
 353         ['E' + 1] = Letter | Hex | Exp,
 354         ['F' + 1] = Letter | Hex,
 355         ['G' + 1 ... 'O' + 1] = Letter,
 356         ['P' + 1] = Letter | Exp,
 357         ['Q' + 1 ... 'Z' + 1] = Letter,
 358         ['a' + 1 ... 'd' + 1] = Letter | Hex,
 359         ['e' + 1] = Letter | Hex | Exp,
 360         ['f' + 1] = Letter | Hex,
 361         ['g' + 1 ... 'o' + 1] = Letter,
 362         ['p' + 1] = Letter | Exp,
 363         ['q' + 1 ... 'z' + 1] = Letter,
 364         ['_' + 1] = Letter,
 365         ['.' + 1] = Dot | ValidSecond,
 366         ['=' + 1] = ValidSecond,
 367         ['+' + 1] = ValidSecond,
 368         ['-' + 1] = ValidSecond,
 369         ['>' + 1] = ValidSecond,
 370         ['<' + 1] = ValidSecond,
 371         ['&' + 1] = ValidSecond,
 372         ['|' + 1] = ValidSecond,
 373         ['#' + 1] = ValidSecond,
 374 };
 375
 376 /*
 377  * pp-number:
 378  *      digit
 379  *      . digit
 380  *      pp-number digit
 381  *      pp-number identifier-nodigit
 382  *      pp-number e sign
 383  *      pp-number E sign
 384  *      pp-number p sign
 385  *      pp-number P sign
 386  *      pp-number .
 387  */
 388 static int get_one_number(int c, int next, stream_t *stream)
 389 {
 390         struct token *token;
 391         static char buffer[4095];
 392         char *p = buffer, *buf, *buffer_end = buffer + sizeof (buffer);
 393         int len;
 394
 395         *p++ = c;
 396         for (;;) {
 397                 long class =  cclass[next + 1];
 398                 if (!(class & (Dot | Digit | Letter)))
 399                         break;
 400                 if (p != buffer_end)
 401                         *p++ = next;
 402                 next = nextchar(stream);
 403                 if (class & Exp) {
 404                         if (next == '-' || next == '+') {
 405                                 if (p != buffer_end)
 406                                         *p++ = next;
 407                                 next = nextchar(stream);
 408                         }
 409                 }
 410         }
 411
 412         if (p == buffer_end) {
 413                 sparse_error(stream_pos(stream), "number token exceeds %td characters",
 414                       buffer_end - buffer);
 415                 // Pretend we saw just "1".
 416                 buffer[0] = '1';
 417                 p = buffer + 1;
 418         }
 419
 420         *p++ = 0;
 421         len = p - buffer;
 422         buf = __alloc_bytes(len);
 423         memcpy(buf, buffer, len);
 424
 425         token = stream->token;
 426         token_type(token) = TOKEN_NUMBER;
 427         token->number = buf;
 428         add_token(stream);
 429
 430         return next;
 431 }
 432
 433 static int escapechar(int first, int type, stream_t *stream, int *valp)
 434 {
 435         int next, value;
 436
 437         next = nextchar(stream);
 438         value = first;
 439
 440         if (first == '\n')
 441                 warning(stream_pos(stream), "Newline in string or character constant");
 442
 443         if (first == '\\' && next != EOF) {
 444                 value = next;
 445                 next = nextchar(stream);
 446                 if (value != type) {
 447                         switch (value) {
 448                         case 'a':
 449                                 value = '\a';
 450                                 break;
 451                         case 'b':
 452                                 value = '\b';
 453                                 break;
 454                         case 't':
 455                                 value = '\t';
 456                                 break;
 457                         case 'n':
 458                                 value = '\n';
 459                                 break;
 460                         case 'v':
 461                                 value = '\v';
 462                                 break;
 463                         case 'f':
 464                                 value = '\f';
 465                                 break;
 466                         case 'r':
 467                                 value = '\r';
 468                                 break;
 469                         case 'e':
 470                                 value = '\e';
 471                                 break;
 472                         case '\\':
 473                                 break;
 474                         case '?':
 475                                 break;
 476                         case '\'':
 477                                 break;
 478                         case '"':
 479                                 break;
 480                         case '\n':
 481                                 warning(stream_pos(stream), "Newline in string or character constant");
 482                                 break;
 483                         case '0'...'7': {
 484                                 int nr = 2;
 485                                 value -= '0';
 486                                 while (next >= '0' && next <= '9') {
 487                                         value = (value << 3) + (next-'0');
 488                                         next = nextchar(stream);
 489                                         if (!--nr)
 490                                                 break;
 491                                 }
 492                                 value &= 0xff;
 493                                 break;
 494                         }
 495                         case 'x': {
 496                                 int hex = hexval(next);
 497                                 if (hex < 16) {
 498                                         value = hex;
 499                                         next = nextchar(stream);
 500                                         while ((hex = hexval(next)) < 16) {
 501                                                 value = (value << 4) + hex;
 502                                                 next = nextchar(stream);
 503                                         }
 504                                         value &= 0xff;
 505                                         break;
 506                                 }
 507                         }
 508                         /* Fall through */
 509                         default:
 510                                 warning(stream_pos(stream), "Unknown escape '%c'", value);
 511                         }
 512                 }
 513                 /* Mark it as escaped */
 514                 value |= 0x100;
 515         }
 516         *valp = value;
 517         return next;
 518 }
 519
 520 static int get_char_token(int next, stream_t *stream)
 521 {
 522         int value;
 523         struct token *token;
 524
 525         next = escapechar(next, '\'', stream, &value);
 526         if (value == '\'' || next != '\'') {
 527                 sparse_error(stream_pos(stream), "Bad character constant");
 528                 drop_token(stream);
 529                 return next;
 530         }
 531
 532         token = stream->token;
 533         token_type(token) = TOKEN_CHAR;
 534         token->character = value & 0xff;
 535
 536         add_token(stream);
 537         return nextchar(stream);
 538 }
 539
 540 static int get_string_token(int next, stream_t *stream)
 541 {
 542         static char buffer[MAX_STRING];
 543         struct string *string;
 544         struct token *token;
 545         int len = 0;
 546
 547         for (;;) {
 548                 int val;
 549                 next = escapechar(next, '"', stream, &val);
 550                 if (val == '"')
 551                         break;
 552                 if (next == EOF) {
 553                         warning(stream_pos(stream), "End of file in middle of string");
 554                         return next;
 555                 }
 556                 if (len < MAX_STRING)
 557                         buffer[len] = val;
 558                 len++;
 559         }
 560
 561         if (len > MAX_STRING) {
 562                 warning(stream_pos(stream), "string too long (%d bytes, %d bytes max)", len, MAX_STRING);
 563                 len = MAX_STRING;
 564         }
 565
 566         string = __alloc_string(len+1);
 567         memcpy(string->data, buffer, len);
 568         string->data[len] = '\0';
 569         string->length = len+1;
 570
 571         /* Pass it on.. */
 572         token = stream->token;
 573         token_type(token) = TOKEN_STRING;
 574         token->string = string;
 575         add_token(stream);
 576
 577         return next;
 578 }
 579
 580 static int drop_stream_eoln(stream_t *stream)
 581 {
 582         drop_token(stream);
 583         for (;;) {
 584                 switch (nextchar(stream)) {
 585                 case EOF:
 586                         return EOF;
 587                 case '\n':
 588                         return nextchar(stream);
 589                 }
 590         }
 591 }
 592
 593 static int drop_stream_comment(stream_t *stream)
 594 {
 595         int newline;
 596         int next;
 597         drop_token(stream);
 598         newline = stream->newline;
 599
 600         next = nextchar(stream);
 601         for (;;) {
 602                 int curr = next;
 603                 if (curr == EOF) {
 604                         warning(stream_pos(stream), "End of file in the middle of a comment");
 605                         return curr;
 606                 }
 607                 next = nextchar(stream);
 608                 if (curr == '*' && next == '/')
 609                         break;
 610         }
 611         stream->newline = newline;
 612         return nextchar(stream);
 613 }
 614
 615 unsigned char combinations[][4] = COMBINATION_STRINGS;
 616
 617 #define NR_COMBINATIONS (SPECIAL_ARG_SEPARATOR - SPECIAL_BASE)
 618
 619 /* hash function for two-character punctuators - all give unique values */
 620 #define special_hash(c0, c1) (((c0*8+c1*2)+((c0*8+c1*2)>>5))&31)
 621
 622 /*
 623  * note that we won't get false positives - special_hash(0,0) is 0 and
 624  * entry 0 is filled (by +=), so all the missing ones are OK.
 625  */
 626 static unsigned char hash_results[32][2] = {
 627 #define RES(c0, c1) [special_hash(c0, c1)] = {c0, c1}
 628         RES('+', '='), /* 00 */
 629         RES('/', '='), /* 01 */
 630         RES('^', '='), /* 05 */
 631         RES('&', '&'), /* 07 */
 632         RES('#', '#'), /* 08 */
 633         RES('<', '<'), /* 0a */
 634         RES('<', '='), /* 0c */
 635         RES('!', '='), /* 0e */
 636         RES('%', '='), /* 0f */
 637         RES('-', '-'), /* 10 */
 638         RES('-', '='), /* 11 */
 639         RES('-', '>'), /* 13 */
 640         RES('=', '='), /* 15 */
 641         RES('&', '='), /* 17 */
 642         RES('*', '='), /* 18 */
 643         RES('.', '.'), /* 1a */
 644         RES('+', '+'), /* 1b */
 645         RES('|', '='), /* 1c */
 646         RES('>', '='), /* 1d */
 647         RES('|', '|'), /* 1e */
 648         RES('>', '>')  /* 1f */
 649 #undef RES
 650 };
 651 static int code[32] = {
 652 #define CODE(c0, c1, value) [special_hash(c0, c1)] = value
 653         CODE('+', '=', SPECIAL_ADD_ASSIGN), /* 00 */
 654         CODE('/', '=', SPECIAL_DIV_ASSIGN), /* 01 */
 655         CODE('^', '=', SPECIAL_XOR_ASSIGN), /* 05 */
 656         CODE('&', '&', SPECIAL_LOGICAL_AND), /* 07 */
 657         CODE('#', '#', SPECIAL_HASHHASH), /* 08 */
 658         CODE('<', '<', SPECIAL_LEFTSHIFT), /* 0a */
 659         CODE('<', '=', SPECIAL_LTE), /* 0c */
 660         CODE('!', '=', SPECIAL_NOTEQUAL), /* 0e */
 661         CODE('%', '=', SPECIAL_MOD_ASSIGN), /* 0f */
 662         CODE('-', '-', SPECIAL_DECREMENT), /* 10 */
 663         CODE('-', '=', SPECIAL_SUB_ASSIGN), /* 11 */
 664         CODE('-', '>', SPECIAL_DEREFERENCE), /* 13 */
 665         CODE('=', '=', SPECIAL_EQUAL), /* 15 */
 666         CODE('&', '=', SPECIAL_AND_ASSIGN), /* 17 */
 667         CODE('*', '=', SPECIAL_MUL_ASSIGN), /* 18 */
 668         CODE('.', '.', SPECIAL_DOTDOT), /* 1a */
 669         CODE('+', '+', SPECIAL_INCREMENT), /* 1b */
 670         CODE('|', '=', SPECIAL_OR_ASSIGN), /* 1c */
 671         CODE('>', '=', SPECIAL_GTE), /* 1d */
 672         CODE('|', '|', SPECIAL_LOGICAL_OR), /* 1e */
 673         CODE('>', '>', SPECIAL_RIGHTSHIFT)  /* 1f */
 674 #undef CODE
 675 };
 676
 677 static int get_one_special(int c, stream_t *stream)
 678 {
 679         struct token *token;
 680         int next, value, i;
 681
 682         next = nextchar(stream);
 683
 684         /*
 685          * Check for numbers, strings, character constants, and comments
 686          */
 687         switch (c) {
 688         case '.':
 689                 if (next >= '0' && next <= '9')
 690                         return get_one_number(c, next, stream);
 691                 break;
 692         case '"':
 693                 return get_string_token(next, stream);
 694         case '\'':
 695                 return get_char_token(next, stream);
 696         case '/':
 697                 if (next == '/')
 698                         return drop_stream_eoln(stream);
 699                 if (next == '*')
 700                         return drop_stream_comment(stream);
 701         }
 702
 703         /*
 704          * Check for combinations
 705          */
 706         value = c;
 707         if (cclass[next + 1] & ValidSecond) {
 708                 i = special_hash(c, next);
 709                 if (hash_results[i][0] == c && hash_results[i][1] == next) {
 710                         value = code[i];
 711                         next = nextchar(stream);
 712                         if (value >= SPECIAL_LEFTSHIFT &&
 713                             next == "==."[value - SPECIAL_LEFTSHIFT]) {
 714                                 value += 3;
 715                                 next = nextchar(stream);
 716                         }
 717                 }
 718         }
 719
 720         /* Pass it on.. */
 721         token = stream->token;
 722         token_type(token) = TOKEN_SPECIAL;
 723         token->special = value;
 724         add_token(stream);
 725         return next;
 726 }
 727
 728 #define IDENT_HASH_BITS (13)
 729 #define IDENT_HASH_SIZE (1<<IDENT_HASH_BITS)
 730 #define IDENT_HASH_MASK (IDENT_HASH_SIZE-1)
 731
 732 #define ident_hash_init(c)              (c)
 733 #define ident_hash_add(oldhash,c)       ((oldhash)*11 + (c))
 734 #define ident_hash_end(hash)            ((((hash) >> IDENT_HASH_BITS) + (hash)) & IDENT_HASH_MASK)
 735
 736 static struct ident *hash_table[IDENT_HASH_SIZE];
 737 static int ident_hit, ident_miss, idents;
 738
 739 void show_identifier_stats(void)
 740 {
 741         int i;
 742         int distribution[100];
 743
 744         fprintf(stderr, "identifiers: %d hits, %d misses\n",
 745                 ident_hit, ident_miss);
 746
 747         for (i = 0; i < 100; i++)
 748                 distribution[i] = 0;
 749
 750         for (i = 0; i < IDENT_HASH_SIZE; i++) {
 751                 struct ident * ident = hash_table[i];
 752                 int count = 0;
 753
 754                 while (ident) {
 755                         count++;
 756                         ident = ident->next;
 757                 }
 758                 if (count > 99)
 759                         count = 99;
 760                 distribution[count]++;
 761         }
 762
 763         for (i = 0; i < 100; i++) {
 764                 if (distribution[i])
 765                         fprintf(stderr, "%2d: %d buckets\n", i, distribution[i]);
 766         }
 767 }
 768
 769 static struct ident *alloc_ident(const char *name, int len)
 770 {
 771         struct ident *ident = __alloc_ident(len);
 772         ident->symbols = NULL;
 773         ident->len = len;
 774         ident->tainted = 0;
 775         memcpy(ident->name, name, len);
 776         return ident;
 777 }
 778
 779 static struct ident * insert_hash(struct ident *ident, unsigned long hash)
 780 {
 781         ident->next = hash_table[hash];
 782         hash_table[hash] = ident;
 783         ident_miss++;
 784         return ident;
 785 }
 786
 787 static struct ident *create_hashed_ident(const char *name, int len, unsigned long hash)
 788 {
 789         struct ident *ident;
 790         struct ident **p;
 791
 792         p = &hash_table[hash];
 793         while ((ident = *p) != NULL) {
 794                 if (ident->len == (unsigned char) len) {
 795                         if (strncmp(name, ident->name, len) != 0)
 796                                 goto next;
 797
 798                         ident_hit++;
 799                         return ident;
 800                 }
 801 next:
 802                 //misses++;
 803                 p = &ident->next;
 804         }
 805         ident = alloc_ident(name, len);
 806         *p = ident;
 807         ident->next = NULL;
 808         ident_miss++;
 809         idents++;
 810         return ident;
 811 }
 812
 813 static unsigned long hash_name(const char *name, int len)
 814 {
 815         unsigned long hash;
 816         const unsigned char *p = (const unsigned char *)name;
 817
 818         hash = ident_hash_init(*p++);
 819         while (--len) {
 820                 unsigned int i = *p++;
 821                 hash = ident_hash_add(hash, i);
 822         }
 823         return ident_hash_end(hash);
 824 }
 825
 826 struct ident *hash_ident(struct ident *ident)
 827 {
 828         return insert_hash(ident, hash_name(ident->name, ident->len));
 829 }
 830
 831 struct ident *built_in_ident(const char *name)
 832 {
 833         int len = strlen(name);
 834         return create_hashed_ident(name, len, hash_name(name, len));
 835 }
 836
 837 struct token *built_in_token(int stream, const char *name)
 838 {
 839         struct token *token;
 840
 841         token = __alloc_token(0);
 842         token->pos.stream = stream;
 843         token_type(token) = TOKEN_IDENT;
 844         token->ident = built_in_ident(name);
 845         return token;
 846 }
 847
 848 static int get_one_identifier(int c, stream_t *stream)
 849 {
 850         struct token *token;
 851         struct ident *ident;
 852         unsigned long hash;
 853         char buf[256];
 854         int len = 1;
 855         int next;
 856
 857         hash = ident_hash_init(c);
 858         buf[0] = c;
 859         for (;;) {
 860                 next = nextchar(stream);
 861                 if (!(cclass[next + 1] & (Letter | Digit)))
 862                         break;
 863                 if (len >= sizeof(buf))
 864                         break;
 865                 hash = ident_hash_add(hash, next);
 866                 buf[len] = next;
 867                 len++;
 868         };
 869         hash = ident_hash_end(hash);
 870
 871         ident = create_hashed_ident(buf, len, hash);
 872
 873         /* Pass it on.. */
 874         token = stream->token;
 875         token_type(token) = TOKEN_IDENT;
 876         token->ident = ident;
 877         add_token(stream);
 878         return next;
 879 }
 880
 881 static int get_one_token(int c, stream_t *stream)
 882 {
 883         long class = cclass[c + 1];
 884         if (class & Digit)
 885                 return get_one_number(c, nextchar(stream), stream);
 886         if (class & Letter)
 887                 return get_one_identifier(c, stream);
 888         return get_one_special(c, stream);
 889 }
 890
 891 static struct token *setup_stream(stream_t *stream, int idx, int fd,
 892         unsigned char *buf, unsigned int buf_size)
 893 {
 894         struct token *begin;
 895
 896         stream->nr = idx;
 897         stream->line = 1;
 898         stream->newline = 1;
 899         stream->whitespace = 0;
 900         stream->pos = 0;
 901
 902         stream->token = NULL;
 903         stream->fd = fd;
 904         stream->offset = 0;
 905         stream->size = buf_size;
 906         stream->buffer = buf;
 907
 908         begin = alloc_token(stream);
 909         token_type(begin) = TOKEN_STREAMBEGIN;
 910         stream->tokenlist = &begin->next;
 911         return begin;
 912 }
 913
 914 static struct token *tokenize_stream(stream_t *stream)
 915 {
 916         int c = nextchar(stream);
 917         while (c != EOF) {
 918                 if (!isspace(c)) {
 919                         struct token *token = alloc_token(stream);
 920                         stream->token = token;
 921                         stream->newline = 0;
 922                         stream->whitespace = 0;
 923                         c = get_one_token(c, stream);
 924                         continue;
 925                 }
 926                 stream->whitespace = 1;
 927                 c = nextchar(stream);
 928         }
 929         return mark_eof(stream);
 930 }
 931
 932 struct token * tokenize_buffer(void *buffer, unsigned long size, struct token **endtoken)
 933 {
 934         stream_t stream;
 935         struct token *begin;
 936
 937         begin = setup_stream(&stream, 0, -1, buffer, size);
 938         *endtoken = tokenize_stream(&stream);
 939         return begin;
 940 }
 941
 942 struct token * tokenize(const char *name, int fd, struct token *endtoken, const char **next_path)
 943 {
 944         struct token *begin, *end;
 945         stream_t stream;
 946         unsigned char buffer[BUFSIZE];
 947         int idx;
 948
 949         idx = init_stream(name, fd, next_path);
 950         if (idx < 0) {
 951                 // info(endtoken->pos, "File %s is const", name);
 952                 return endtoken;
 953         }
 954
 955         begin = setup_stream(&stream, idx, fd, buffer, 0);
 956         end = tokenize_stream(&stream);
 957         if (endtoken)
 958                 end->next = endtoken;
 959         return begin;
 960 }