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 = util_strdup(lex->tok->value);
391 if (!m.name || !lex_file_frames_add(lex, m)) {
392 lexerror(lex, "out of memory");
398 static int GMQCC_WARN lex_finish_string(lex_file *lex, int quote)
406 return TOKEN_STRINGCONST;
411 lexerror(lex, "unexpected end of file");
412 lex_ungetch(lex, EOF); /* next token to be TOKEN_EOF */
413 return (lex->tok->ttype = TOKEN_ERROR);
418 case 'a': ch = '\a'; break;
419 case 'b': ch = '\b'; break;
420 case 'r': ch = '\r'; break;
421 case 'n': ch = '\n'; break;
422 case 't': ch = '\t'; break;
423 case 'f': ch = '\f'; break;
424 case 'v': ch = '\v'; break;
426 lexwarn(lex, WARN_UNKNOWN_CONTROL_SEQUENCE, "unrecognized control sequence: \\%c", ch);
427 /* so we just add the character plus backslash no matter what it actually is */
428 if (!lex_tokench(lex, '\\'))
429 return (lex->tok->ttype = TOKEN_FATAL);
431 /* add the character finally */
432 if (!lex_tokench(lex, ch))
433 return (lex->tok->ttype = TOKEN_FATAL);
435 else if (!lex_tokench(lex, ch))
436 return (lex->tok->ttype = TOKEN_FATAL);
438 lexerror(lex, "unexpected end of file within string constant");
439 lex_ungetch(lex, EOF); /* next token to be TOKEN_EOF */
440 return (lex->tok->ttype = TOKEN_ERROR);
443 static int GMQCC_WARN lex_finish_digit(lex_file *lex, int lastch)
449 /* parse a number... */
450 lex->tok->ttype = TOKEN_INTCONST;
452 if (!lex_tokench(lex, ch))
453 return (lex->tok->ttype = TOKEN_FATAL);
456 if (ch != '.' && !isdigit(ch))
458 if (lastch != '0' || ch != 'x')
460 /* end of the number or EOF */
461 lex_ungetch(lex, ch);
462 if (!lex_endtoken(lex))
463 return (lex->tok->ttype = TOKEN_FATAL);
465 lex->tok->constval.i = lastch - '0';
466 return lex->tok->ttype;
472 /* EOF would have been caught above */
476 if (!lex_tokench(lex, ch))
477 return (lex->tok->ttype = TOKEN_FATAL);
479 while (isdigit(ch) || (ishex && isxdigit_only(ch)))
481 if (!lex_tokench(lex, ch))
482 return (lex->tok->ttype = TOKEN_FATAL);
486 /* NOT else, '.' can come from above as well */
487 if (ch == '.' && !ishex)
489 /* Allow floating comma in non-hex mode */
490 lex->tok->ttype = TOKEN_FLOATCONST;
491 if (!lex_tokench(lex, ch))
492 return (lex->tok->ttype = TOKEN_FATAL);
494 /* continue digits-only */
498 if (!lex_tokench(lex, ch))
499 return (lex->tok->ttype = TOKEN_FATAL);
503 /* put back the last character */
504 /* but do not put back the trailing 'f' or a float */
505 if (lex->tok->ttype == TOKEN_FLOATCONST && ch == 'f')
508 /* generally we don't want words to follow numbers: */
510 lexerror(lex, "unexpected trailing characters after number");
511 return (lex->tok->ttype = TOKEN_ERROR);
513 lex_ungetch(lex, ch);
515 if (!lex_endtoken(lex))
516 return (lex->tok->ttype = TOKEN_FATAL);
517 if (lex->tok->ttype == TOKEN_FLOATCONST)
518 lex->tok->constval.f = strtod(lex->tok->value, NULL);
520 lex->tok->constval.i = strtol(lex->tok->value, NULL, 0);
521 return lex->tok->ttype;
524 int lex_do(lex_file *lex)
529 token_delete(lex->tok);
530 lex->tok = token_new();
534 ch = lex_skipwhite(lex);
535 lex->sline = lex->line;
536 lex->tok->ctx.line = lex->sline;
537 lex->tok->ctx.file = lex->name;
540 return (lex->tok->ttype = TOKEN_EOF);
542 /* modelgen / spiritgen commands */
546 if (!isident_start(ch)) {
547 lexerror(lex, "hanging '$' modelgen/spritegen command line");
550 if (!lex_tokench(lex, ch))
551 return (lex->tok->ttype = TOKEN_FATAL);
552 if (!lex_finish_ident(lex))
553 return (lex->tok->ttype = TOKEN_ERROR);
554 if (!lex_endtoken(lex))
555 return (lex->tok->ttype = TOKEN_FATAL);
556 /* skip the known commands */
559 if (!strcmp(v, "frame") || !strcmp(v, "framesave"))
561 /* frame/framesave command works like an enum
562 * similar to fteqcc we handle this in the lexer.
563 * The reason for this is that it is sensitive to newlines,
564 * which the parser is unaware of
566 if (!lex_finish_frames(lex))
567 return (lex->tok->ttype = TOKEN_ERROR);
571 if (!strcmp(v, "framevalue"))
574 while (ch != EOF && isspace(ch) && ch != '\n')
578 lexerror(lex, "$framevalue requires an integer parameter");
582 token_delete(lex->tok);
583 lex->tok = token_new();
584 lex->tok->ttype = lex_finish_digit(lex, ch);
585 if (!lex_endtoken(lex))
586 return (lex->tok->ttype = TOKEN_FATAL);
587 if (lex->tok->ttype != TOKEN_INTCONST) {
588 lexerror(lex, "$framevalue requires an integer parameter");
591 lex->framevalue = lex->tok->constval.i;
595 if (!strcmp(v, "flush"))
598 for (frame = 0; frame < lex->frames_count; ++frame)
599 mem_d(lex->frames[frame].name);
600 MEM_VECTOR_CLEAR(lex, frames);
601 /* skip line (fteqcc does it too) */
603 while (ch != EOF && ch != '\n')
608 if (!strcmp(v, "cd") ||
609 !strcmp(v, "origin") ||
610 !strcmp(v, "base") ||
611 !strcmp(v, "flags") ||
612 !strcmp(v, "scale") ||
617 while (ch != EOF && ch != '\n')
623 /* single-character tokens */
635 if (!lex_tokench(lex, ch) ||
638 return (lex->tok->ttype = TOKEN_FATAL);
640 return (lex->tok->ttype = ch);
645 if (lex->flags.noops)
647 /* Detect characters early which are normally
648 * operators OR PART of an operator.
666 if (!lex_tokench(lex, ch) ||
669 return (lex->tok->ttype = TOKEN_FATAL);
671 return (lex->tok->ttype = ch);
677 if (ch == ',' || ch == '.') {
678 if (!lex_tokench(lex, ch) ||
681 return (lex->tok->ttype = TOKEN_FATAL);
683 return (lex->tok->ttype = TOKEN_OPERATOR);
686 if (ch == '+' || ch == '-' || /* ++, --, +=, -= and -> as well! */
687 ch == '>' || ch == '<' || /* <<, >>, <=, >= */
688 ch == '=' || ch == '!' || /* ==, != */
689 ch == '&' || ch == '|') /* &&, ||, &=, |= */
691 if (!lex_tokench(lex, ch))
692 return (lex->tok->ttype = TOKEN_FATAL);
694 nextch = lex_getch(lex);
695 if (nextch == ch || nextch == '=') {
696 if (!lex_tokench(lex, nextch))
697 return (lex->tok->ttype = TOKEN_FATAL);
698 } else if (ch == '-' && nextch == '>') {
699 if (!lex_tokench(lex, nextch))
700 return (lex->tok->ttype = TOKEN_FATAL);
702 lex_ungetch(lex, nextch);
704 if (!lex_endtoken(lex))
705 return (lex->tok->ttype = TOKEN_FATAL);
706 return (lex->tok->ttype = TOKEN_OPERATOR);
710 if (ch == '^' || ch == '~' || ch == '!')
712 if (!lex_tokench(lex, ch) ||
715 return (lex->tok->ttype = TOKEN_FATAL);
717 return (lex->tok->ttype = TOKEN_OPERATOR);
721 if (ch == '*' || ch == '/') /* *=, /= */
723 if (!lex_tokench(lex, ch))
724 return (lex->tok->ttype = TOKEN_FATAL);
726 nextch = lex_getch(lex);
728 if (!lex_tokench(lex, nextch))
729 return (lex->tok->ttype = TOKEN_FATAL);
731 lex_ungetch(lex, nextch);
733 if (!lex_endtoken(lex))
734 return (lex->tok->ttype = TOKEN_FATAL);
735 return (lex->tok->ttype = TOKEN_OPERATOR);
738 if (isident_start(ch))
742 if (!lex_tokench(lex, ch))
743 return (lex->tok->ttype = TOKEN_FATAL);
744 if (!lex_finish_ident(lex)) {
746 return (lex->tok->ttype = TOKEN_ERROR);
748 if (!lex_endtoken(lex))
749 return (lex->tok->ttype = TOKEN_FATAL);
750 lex->tok->ttype = TOKEN_IDENT;
753 if (!strcmp(v, "void")) {
754 lex->tok->ttype = TOKEN_TYPENAME;
755 lex->tok->constval.t = TYPE_VOID;
756 } else if (!strcmp(v, "int")) {
757 lex->tok->ttype = TOKEN_TYPENAME;
758 lex->tok->constval.t = TYPE_INTEGER;
759 } else if (!strcmp(v, "float")) {
760 lex->tok->ttype = TOKEN_TYPENAME;
761 lex->tok->constval.t = TYPE_FLOAT;
762 } else if (!strcmp(v, "string")) {
763 lex->tok->ttype = TOKEN_TYPENAME;
764 lex->tok->constval.t = TYPE_STRING;
765 } else if (!strcmp(v, "entity")) {
766 lex->tok->ttype = TOKEN_TYPENAME;
767 lex->tok->constval.t = TYPE_ENTITY;
768 } else if (!strcmp(v, "vector")) {
769 lex->tok->ttype = TOKEN_TYPENAME;
770 lex->tok->constval.t = TYPE_VECTOR;
771 } else if (!strcmp(v, "for") ||
772 !strcmp(v, "while") ||
775 !strcmp(v, "else") ||
776 !strcmp(v, "local") ||
777 !strcmp(v, "return") ||
779 lex->tok->ttype = TOKEN_KEYWORD;
781 for (frame = 0; frame < lex->frames_count; ++frame) {
782 if (!strcmp(v, lex->frames[frame].name)) {
783 lex->tok->constval.i = lex->frames[frame].value;
784 return (lex->tok->ttype = TOKEN_INTCONST);
788 return lex->tok->ttype;
793 lex->tok->ttype = lex_finish_string(lex, '"');
794 while (lex->tok->ttype == TOKEN_STRINGCONST)
796 /* Allow c style "string" "continuation" */
797 ch = lex_skipwhite(lex);
799 lex_ungetch(lex, ch);
803 lex->tok->ttype = lex_finish_string(lex, '"');
805 if (!lex_endtoken(lex))
806 return (lex->tok->ttype = TOKEN_FATAL);
807 return lex->tok->ttype;
812 /* we parse character constants like string,
813 * but return TOKEN_CHARCONST, or a vector type if it fits...
814 * Likewise actual unescaping has to be done by the parser.
815 * The difference is we don't allow 'char' 'continuation'.
817 lex->tok->ttype = lex_finish_string(lex, '\'');
818 if (!lex_endtoken(lex))
819 return (lex->tok->ttype = TOKEN_FATAL);
821 /* It's a vector if we can successfully scan 3 floats */
823 if (sscanf_s(lex->tok->value, " %f %f %f ",
824 &lex->tok->constval.v.x, &lex->tok->constval.v.y, &lex->tok->constval.v.z) == 3)
826 if (sscanf(lex->tok->value, " %f %f %f ",
827 &lex->tok->constval.v.x, &lex->tok->constval.v.y, &lex->tok->constval.v.z) == 3)
830 lex->tok->ttype = TOKEN_VECTORCONST;
833 return lex->tok->ttype;
838 lex->tok->ttype = lex_finish_digit(lex, ch);
839 if (!lex_endtoken(lex))
840 return (lex->tok->ttype = TOKEN_FATAL);
841 return lex->tok->ttype;
844 lexerror(lex, "unknown token");
845 return (lex->tok->ttype = TOKEN_ERROR);