9 MEM_VEC_FUNCTIONS(token, char, value)
10 MEM_VEC_FUNCTIONS(lex_file, frame_macro, frames)
12 void lexerror(lex_file *lex, const char *fmt, ...)
17 printf("error %s:%lu: ", lex->name, (unsigned long)lex->sline);
28 void lexwarn(lex_file *lex, int warn, const char *fmt, ...)
36 printf("warning %s:%lu: ", lex->name, (unsigned long)lex->sline);
49 token *tok = (token*)mem_a(sizeof(token));
52 memset(tok, 0, sizeof(*tok));
56 void token_delete(token *self)
58 if (self->next && self->next->prev == self)
59 self->next->prev = self->prev;
60 if (self->prev && self->prev->next == self)
61 self->prev->next = self->next;
62 MEM_VECTOR_CLEAR(self, value);
66 token* token_copy(const token *cp)
68 token* self = token_new();
72 self->value_alloc = cp->value_count + 1;
73 self->value_count = cp->value_count;
74 self->value = (char*)mem_a(self->value_alloc);
79 memcpy(self->value, cp->value, cp->value_count);
80 self->value[self->value_alloc-1] = 0;
84 self->ttype = cp->ttype;
85 memcpy(&self->constval, &cp->constval, sizeof(self->constval));
89 void token_delete_all(token *t)
100 token* token_copy_all(const token *cp)
105 out = cur = token_copy(cp);
111 cur->next = token_copy(cp);
113 token_delete_all(out);
116 cur->next->prev = cur;
123 lex_file* lex_open(const char *file)
126 FILE *in = util_fopen(file, "rb");
129 lexerror(NULL, "open failed: '%s'\n", file);
133 lex = (lex_file*)mem_a(sizeof(*lex));
136 lexerror(NULL, "out of memory\n");
140 memset(lex, 0, sizeof(*lex));
143 lex->name = util_strdup(file);
144 lex->line = 1; /* we start counting at 1 */
151 void lex_close(lex_file *lex)
156 token_delete(lex->tok);
161 /* Get or put-back data
162 * The following to functions do NOT understand what kind of data they
164 * The are merely wrapping get/put in order to count line numbers.
166 static int lex_getch(lex_file *lex)
172 if (lex->peek[lex->peekpos] == '\n')
174 return lex->peek[lex->peekpos];
177 ch = fgetc(lex->file);
183 static void lex_ungetch(lex_file *lex, int ch)
185 lex->peek[lex->peekpos++] = ch;
190 /* classify characters
191 * some additions to the is*() functions of ctype.h
194 /* Idents are alphanumberic, but they start with alpha or _ */
195 static bool isident_start(int ch)
197 return isalpha(ch) || ch == '_';
200 static bool isident(int ch)
202 return isident_start(ch) || isdigit(ch);
205 /* isxdigit_only is used when we already know it's not a digit
206 * and want to see if it's a hex digit anyway.
208 static bool isxdigit_only(int ch)
210 return (ch >= 'a' && ch <= 'f') || (ch >= 'A' && ch <= 'F');
213 /* Skip whitespace and comments and return the first
214 * non-white character.
215 * As this makes use of the above getch() ungetch() functions,
216 * we don't need to care at all about line numbering anymore.
218 * In theory, this function should only be used at the beginning
219 * of lexing, or when we *know* the next character is part of the token.
220 * Otherwise, if the parser throws an error, the linenumber may not be
221 * the line of the error, but the line of the next token AFTER the error.
223 * This is currently only problematic when using c-like string-continuation,
224 * since comments and whitespaces are allowed between 2 such strings.
228 "A continuation of the previous string"
229 // This line is skipped
232 * In this case, if the parse decides it didn't actually want a string,
233 * and uses lex->line to print an error, it will show the ', foo);' line's
236 * On the other hand, the parser is supposed to remember the line of the next
237 * token's beginning. In this case we would want skipwhite() to be called
238 * AFTER reading a token, so that the parser, before reading the NEXT token,
239 * doesn't store teh *comment's* linenumber, but the actual token's linenumber.
242 * here is to store the line of the first character after skipping
243 * the initial whitespace in lex->sline, this happens in lex_do.
245 static int lex_skipwhite(lex_file *lex)
252 while (ch != EOF && isspace(ch)) ch = lex_getch(lex);
258 /* one line comment */
261 /* check for special: '/', '/', '*', '/' */
270 while (ch != EOF && ch != '\n') {
277 /* multiline comment */
289 if (ch == '/') /* allow *//* direct following comment */
291 lex_ungetch(lex, ch);
292 ch = ' '; /* cause TRUE in the isspace check */
296 /* Otherwise roll back to the slash and break out of the loop */
297 lex_ungetch(lex, ch);
301 } while (ch != EOF && isspace(ch));
306 /* Append a character to the token buffer */
307 static bool GMQCC_WARN lex_tokench(lex_file *lex, int ch)
309 if (!token_value_add(lex->tok, ch)) {
310 lexerror(lex, "out of memory");
316 /* Append a trailing null-byte */
317 static bool GMQCC_WARN lex_endtoken(lex_file *lex)
319 if (!token_value_add(lex->tok, 0)) {
320 lexerror(lex, "out of memory");
323 lex->tok->value_count--;
328 static bool GMQCC_WARN lex_finish_ident(lex_file *lex)
333 while (ch != EOF && isident(ch))
335 if (!lex_tokench(lex, ch))
336 return (lex->tok->ttype = TOKEN_FATAL);
340 /* last ch was not an ident ch: */
341 lex_ungetch(lex, ch);
346 /* read one ident for the frame list */
347 static int lex_parse_frame(lex_file *lex)
352 token_delete(lex->tok);
353 lex->tok = token_new();
356 while (ch != EOF && ch != '\n' && isspace(ch))
362 if (!isident_start(ch)) {
363 lexerror(lex, "invalid framename, must start with one of a-z or _, got %c", ch);
367 if (!lex_tokench(lex, ch))
369 if (!lex_finish_ident(lex))
371 if (!lex_endtoken(lex))
376 /* read a list of $frames */
377 static bool lex_finish_frames(lex_file *lex)
383 rc = lex_parse_frame(lex);
384 if (rc > 0) /* end of line */
386 if (rc < 0) /* error */
389 m.value = lex->framevalue++;
390 m.name = lex->tok->value;
391 lex->tok->value = NULL;
392 if (!lex_file_frames_add(lex, m)) {
393 lexerror(lex, "out of memory");
399 static int GMQCC_WARN lex_finish_string(lex_file *lex, int quote)
407 return TOKEN_STRINGCONST;
412 lexerror(lex, "unexpected end of file");
413 lex_ungetch(lex, EOF); /* next token to be TOKEN_EOF */
414 return (lex->tok->ttype = TOKEN_ERROR);
419 case 'a': ch = '\a'; break;
420 case 'b': ch = '\b'; break;
421 case 'r': ch = '\r'; break;
422 case 'n': ch = '\n'; break;
423 case 't': ch = '\t'; break;
424 case 'f': ch = '\f'; break;
425 case 'v': ch = '\v'; break;
427 lexwarn(lex, WARN_UNKNOWN_CONTROL_SEQUENCE, "unrecognized control sequence: \\%c", ch);
428 /* so we just add the character plus backslash no matter what it actually is */
429 if (!lex_tokench(lex, '\\'))
430 return (lex->tok->ttype = TOKEN_FATAL);
432 /* add the character finally */
433 if (!lex_tokench(lex, ch))
434 return (lex->tok->ttype = TOKEN_FATAL);
436 else if (!lex_tokench(lex, ch))
437 return (lex->tok->ttype = TOKEN_FATAL);
439 lexerror(lex, "unexpected end of file within string constant");
440 lex_ungetch(lex, EOF); /* next token to be TOKEN_EOF */
441 return (lex->tok->ttype = TOKEN_ERROR);
444 static int GMQCC_WARN lex_finish_digit(lex_file *lex, int lastch)
450 /* parse a number... */
451 lex->tok->ttype = TOKEN_INTCONST;
453 if (!lex_tokench(lex, ch))
454 return (lex->tok->ttype = TOKEN_FATAL);
457 if (ch != '.' && !isdigit(ch))
459 if (lastch != '0' || ch != 'x')
461 /* end of the number or EOF */
462 lex_ungetch(lex, ch);
463 if (!lex_endtoken(lex))
464 return (lex->tok->ttype = TOKEN_FATAL);
466 lex->tok->constval.i = lastch - '0';
467 return lex->tok->ttype;
473 /* EOF would have been caught above */
477 if (!lex_tokench(lex, ch))
478 return (lex->tok->ttype = TOKEN_FATAL);
480 while (isdigit(ch) || (ishex && isxdigit_only(ch)))
482 if (!lex_tokench(lex, ch))
483 return (lex->tok->ttype = TOKEN_FATAL);
487 /* NOT else, '.' can come from above as well */
488 if (ch == '.' && !ishex)
490 /* Allow floating comma in non-hex mode */
491 lex->tok->ttype = TOKEN_FLOATCONST;
492 if (!lex_tokench(lex, ch))
493 return (lex->tok->ttype = TOKEN_FATAL);
495 /* continue digits-only */
499 if (!lex_tokench(lex, ch))
500 return (lex->tok->ttype = TOKEN_FATAL);
504 /* put back the last character */
505 /* but do not put back the trailing 'f' or a float */
506 if (lex->tok->ttype == TOKEN_FLOATCONST && ch == 'f')
509 /* generally we don't want words to follow numbers: */
511 lexerror(lex, "unexpected trailing characters after number");
512 return (lex->tok->ttype = TOKEN_ERROR);
514 lex_ungetch(lex, ch);
516 if (!lex_endtoken(lex))
517 return (lex->tok->ttype = TOKEN_FATAL);
518 if (lex->tok->ttype == TOKEN_FLOATCONST)
519 lex->tok->constval.f = strtod(lex->tok->value, NULL);
521 lex->tok->constval.i = strtol(lex->tok->value, NULL, 0);
522 return lex->tok->ttype;
525 int lex_do(lex_file *lex)
530 token_delete(lex->tok);
531 lex->tok = token_new();
535 ch = lex_skipwhite(lex);
536 lex->sline = lex->line;
537 lex->tok->ctx.line = lex->sline;
538 lex->tok->ctx.file = lex->name;
541 return (lex->tok->ttype = TOKEN_EOF);
543 /* modelgen / spiritgen commands */
549 if (!isident_start(ch)) {
550 lexerror(lex, "hanging '$' modelgen/spritegen command line");
553 if (!lex_tokench(lex, ch))
554 return (lex->tok->ttype = TOKEN_FATAL);
555 if (!lex_finish_ident(lex))
556 return (lex->tok->ttype = TOKEN_ERROR);
557 if (!lex_endtoken(lex))
558 return (lex->tok->ttype = TOKEN_FATAL);
559 /* skip the known commands */
562 if (!strcmp(v, "frame") || !strcmp(v, "framesave"))
564 /* frame/framesave command works like an enum
565 * similar to fteqcc we handle this in the lexer.
566 * The reason for this is that it is sensitive to newlines,
567 * which the parser is unaware of
569 if (!lex_finish_frames(lex))
570 return (lex->tok->ttype = TOKEN_ERROR);
574 if (!strcmp(v, "framevalue"))
577 while (ch != EOF && isspace(ch) && ch != '\n')
581 lexerror(lex, "$framevalue requires an integer parameter");
585 token_delete(lex->tok);
586 lex->tok = token_new();
587 lex->tok->ttype = lex_finish_digit(lex, ch);
588 if (!lex_endtoken(lex))
589 return (lex->tok->ttype = TOKEN_FATAL);
590 if (lex->tok->ttype != TOKEN_INTCONST) {
591 lexerror(lex, "$framevalue requires an integer parameter");
594 lex->framevalue = lex->tok->constval.i;
598 if (!strcmp(v, "framerestore"))
602 token_delete(lex->tok);
603 lex->tok = token_new();
605 rc = lex_parse_frame(lex);
608 lexerror(lex, "$framerestore requires a framename parameter");
612 return (lex->tok->ttype = TOKEN_FATAL);
615 for (frame = 0; frame < lex->frames_count; ++frame) {
616 if (!strcmp(v, lex->frames[frame].name)) {
617 lex->framevalue = lex->frames[frame].value;
621 lexerror(lex, "unknown framename `%s`", v);
625 if (!strcmp(v, "modelname"))
629 token_delete(lex->tok);
630 lex->tok = token_new();
632 rc = lex_parse_frame(lex);
635 lexerror(lex, "$framerestore requires a framename parameter");
639 return (lex->tok->ttype = TOKEN_FATAL);
642 if (lex->modelname) {
644 m.value = lex->framevalue;
645 m.name = lex->modelname;
646 lex->modelname = NULL;
647 if (!lex_file_frames_add(lex, m)) {
648 lexerror(lex, "out of memory");
649 return (lex->tok->ttype = TOKEN_FATAL);
652 lex->modelname = lex->tok->value;
653 lex->tok->value = NULL;
654 for (frame = 0; frame < lex->frames_count; ++frame) {
655 if (!strcmp(v, lex->frames[frame].name)) {
656 lex->framevalue = lex->frames[frame].value;
663 if (!strcmp(v, "flush"))
666 for (frame = 0; frame < lex->frames_count; ++frame)
667 mem_d(lex->frames[frame].name);
668 MEM_VECTOR_CLEAR(lex, frames);
669 /* skip line (fteqcc does it too) */
671 while (ch != EOF && ch != '\n')
676 if (!strcmp(v, "cd") ||
677 !strcmp(v, "origin") ||
678 !strcmp(v, "base") ||
679 !strcmp(v, "flags") ||
680 !strcmp(v, "scale") ||
685 while (ch != EOF && ch != '\n')
690 for (frame = 0; frame < lex->frames_count; ++frame) {
691 if (!strcmp(v, lex->frames[frame].name)) {
692 lex->tok->constval.i = lex->frames[frame].value;
693 return (lex->tok->ttype = TOKEN_INTCONST);
697 lexerror(lex, "invalid frame macro");
701 /* single-character tokens */
705 if (!lex_tokench(lex, ch) ||
708 return (lex->tok->ttype = TOKEN_FATAL);
710 if (lex->flags.noops)
711 return (lex->tok->ttype = ch);
713 return (lex->tok->ttype = TOKEN_OPERATOR);
722 if (!lex_tokench(lex, ch) ||
725 return (lex->tok->ttype = TOKEN_FATAL);
727 return (lex->tok->ttype = ch);
732 if (lex->flags.noops)
734 /* Detect characters early which are normally
735 * operators OR PART of an operator.
753 if (!lex_tokench(lex, ch) ||
756 return (lex->tok->ttype = TOKEN_FATAL);
758 return (lex->tok->ttype = ch);
764 if (ch == ',' || ch == '.') {
765 if (!lex_tokench(lex, ch) ||
768 return (lex->tok->ttype = TOKEN_FATAL);
770 return (lex->tok->ttype = TOKEN_OPERATOR);
773 if (ch == '+' || ch == '-' || /* ++, --, +=, -= and -> as well! */
774 ch == '>' || ch == '<' || /* <<, >>, <=, >= */
775 ch == '=' || ch == '!' || /* ==, != */
776 ch == '&' || ch == '|') /* &&, ||, &=, |= */
778 if (!lex_tokench(lex, ch))
779 return (lex->tok->ttype = TOKEN_FATAL);
781 nextch = lex_getch(lex);
782 if (nextch == ch || nextch == '=') {
783 if (!lex_tokench(lex, nextch))
784 return (lex->tok->ttype = TOKEN_FATAL);
785 } else if (ch == '-' && nextch == '>') {
786 if (!lex_tokench(lex, nextch))
787 return (lex->tok->ttype = TOKEN_FATAL);
789 lex_ungetch(lex, nextch);
791 if (!lex_endtoken(lex))
792 return (lex->tok->ttype = TOKEN_FATAL);
793 return (lex->tok->ttype = TOKEN_OPERATOR);
797 if (ch == '^' || ch == '~' || ch == '!')
799 if (!lex_tokench(lex, ch) ||
802 return (lex->tok->ttype = TOKEN_FATAL);
804 return (lex->tok->ttype = TOKEN_OPERATOR);
808 if (ch == '*' || ch == '/') /* *=, /= */
810 if (!lex_tokench(lex, ch))
811 return (lex->tok->ttype = TOKEN_FATAL);
813 nextch = lex_getch(lex);
815 if (!lex_tokench(lex, nextch))
816 return (lex->tok->ttype = TOKEN_FATAL);
818 lex_ungetch(lex, nextch);
820 if (!lex_endtoken(lex))
821 return (lex->tok->ttype = TOKEN_FATAL);
822 return (lex->tok->ttype = TOKEN_OPERATOR);
825 if (isident_start(ch))
829 if (!lex_tokench(lex, ch))
830 return (lex->tok->ttype = TOKEN_FATAL);
831 if (!lex_finish_ident(lex)) {
833 return (lex->tok->ttype = TOKEN_ERROR);
835 if (!lex_endtoken(lex))
836 return (lex->tok->ttype = TOKEN_FATAL);
837 lex->tok->ttype = TOKEN_IDENT;
840 if (!strcmp(v, "void")) {
841 lex->tok->ttype = TOKEN_TYPENAME;
842 lex->tok->constval.t = TYPE_VOID;
843 } else if (!strcmp(v, "int")) {
844 lex->tok->ttype = TOKEN_TYPENAME;
845 lex->tok->constval.t = TYPE_INTEGER;
846 } else if (!strcmp(v, "float")) {
847 lex->tok->ttype = TOKEN_TYPENAME;
848 lex->tok->constval.t = TYPE_FLOAT;
849 } else if (!strcmp(v, "string")) {
850 lex->tok->ttype = TOKEN_TYPENAME;
851 lex->tok->constval.t = TYPE_STRING;
852 } else if (!strcmp(v, "entity")) {
853 lex->tok->ttype = TOKEN_TYPENAME;
854 lex->tok->constval.t = TYPE_ENTITY;
855 } else if (!strcmp(v, "vector")) {
856 lex->tok->ttype = TOKEN_TYPENAME;
857 lex->tok->constval.t = TYPE_VECTOR;
858 } else if (!strcmp(v, "for") ||
859 !strcmp(v, "while") ||
862 !strcmp(v, "else") ||
863 !strcmp(v, "local") ||
864 !strcmp(v, "return") ||
866 lex->tok->ttype = TOKEN_KEYWORD;
868 return lex->tok->ttype;
873 lex->tok->ttype = lex_finish_string(lex, '"');
874 while (lex->tok->ttype == TOKEN_STRINGCONST)
876 /* Allow c style "string" "continuation" */
877 ch = lex_skipwhite(lex);
879 lex_ungetch(lex, ch);
883 lex->tok->ttype = lex_finish_string(lex, '"');
885 if (!lex_endtoken(lex))
886 return (lex->tok->ttype = TOKEN_FATAL);
887 return lex->tok->ttype;
892 /* we parse character constants like string,
893 * but return TOKEN_CHARCONST, or a vector type if it fits...
894 * Likewise actual unescaping has to be done by the parser.
895 * The difference is we don't allow 'char' 'continuation'.
897 lex->tok->ttype = lex_finish_string(lex, '\'');
898 if (!lex_endtoken(lex))
899 return (lex->tok->ttype = TOKEN_FATAL);
901 /* It's a vector if we can successfully scan 3 floats */
903 if (sscanf_s(lex->tok->value, " %f %f %f ",
904 &lex->tok->constval.v.x, &lex->tok->constval.v.y, &lex->tok->constval.v.z) == 3)
906 if (sscanf(lex->tok->value, " %f %f %f ",
907 &lex->tok->constval.v.x, &lex->tok->constval.v.y, &lex->tok->constval.v.z) == 3)
910 lex->tok->ttype = TOKEN_VECTORCONST;
913 return lex->tok->ttype;
918 lex->tok->ttype = lex_finish_digit(lex, ch);
919 if (!lex_endtoken(lex))
920 return (lex->tok->ttype = TOKEN_FATAL);
921 return lex->tok->ttype;
924 lexerror(lex, "unknown token");
925 return (lex->tok->ttype = TOKEN_ERROR);