11 void lexerror(lex_file *lex, const char *fmt, ...)
17 con_vprintmsg(LVL_ERROR, lex->name, lex->sline, "parse error", fmt, ap);
19 con_vprintmsg(LVL_ERROR, "", 0, "parse error", fmt, ap);
23 bool lexwarn(lex_file *lex, int warntype, const char *fmt, ...)
26 int lvl = LVL_WARNING;
28 if (!OPTS_WARN(warntype))
35 con_vprintmsg(lvl, lex->name, lex->sline, "warning", fmt, ap);
45 token *tok = (token*)mem_a(sizeof(token));
48 memset(tok, 0, sizeof(*tok));
52 void token_delete(token *self)
54 if (self->next && self->next->prev == self)
55 self->next->prev = self->prev;
56 if (self->prev && self->prev->next == self)
57 self->prev->next = self->next;
58 MEM_VECTOR_CLEAR(self, value);
62 token* token_copy(const token *cp)
64 token* self = token_new();
68 self->value_alloc = cp->value_count + 1;
69 self->value_count = cp->value_count;
70 self->value = (char*)mem_a(self->value_alloc);
75 memcpy(self->value, cp->value, cp->value_count);
76 self->value[self->value_alloc-1] = 0;
80 self->ttype = cp->ttype;
81 memcpy(&self->constval, &cp->constval, sizeof(self->constval));
85 void token_delete_all(token *t)
96 token* token_copy_all(const token *cp)
101 out = cur = token_copy(cp);
107 cur->next = token_copy(cp);
109 token_delete_all(out);
112 cur->next->prev = cur;
119 static void lex_token_new(lex_file *lex)
123 token_delete(lex->tok);
124 lex->tok = token_new();
127 vec_shrinkto(lex->tok.value, 0);
128 lex->tok.constval.t = 0;
129 lex->tok.ctx.line = lex->sline;
130 lex->tok.ctx.file = lex->name;
135 lex_file* lex_open(const char *file)
138 FILE *in = util_fopen(file, "rb");
141 lexerror(NULL, "open failed: '%s'\n", file);
145 lex = (lex_file*)mem_a(sizeof(*lex));
148 lexerror(NULL, "out of memory\n");
152 memset(lex, 0, sizeof(*lex));
155 lex->name = util_strdup(file);
156 lex->line = 1; /* we start counting at 1 */
161 vec_push(lex_filenames, lex->name);
165 lex_file* lex_open_string(const char *str, size_t len, const char *name)
169 lex = (lex_file*)mem_a(sizeof(*lex));
171 lexerror(NULL, "out of memory\n");
175 memset(lex, 0, sizeof(*lex));
178 lex->open_string = str;
179 lex->open_string_length = len;
180 lex->open_string_pos = 0;
182 lex->name = util_strdup(name ? name : "<string-source>");
183 lex->line = 1; /* we start counting at 1 */
188 vec_push(lex_filenames, lex->name);
193 void lex_cleanup(void)
196 for (i = 0; i < vec_size(lex_filenames); ++i)
197 mem_d(lex_filenames[i]);
198 vec_free(lex_filenames);
201 void lex_close(lex_file *lex)
204 for (i = 0; i < vec_size(lex->frames); ++i)
205 mem_d(lex->frames[i].name);
206 vec_free(lex->frames);
209 vec_free(lex->modelname);
215 token_delete(lex->tok);
217 vec_free(lex->tok.value);
219 /* mem_d(lex->name); collected in lex_filenames */
223 static int lex_fgetc(lex_file *lex)
226 return fgetc(lex->file);
227 if (lex->open_string) {
228 if (lex->open_string_pos >= lex->open_string_length)
230 return lex->open_string[lex->open_string_pos++];
235 /* Get or put-back data
236 * The following to functions do NOT understand what kind of data they
238 * The are merely wrapping get/put in order to count line numbers.
240 static void lex_ungetch(lex_file *lex, int ch);
241 static int lex_try_trigraph(lex_file *lex, int old)
246 lex_ungetch(lex, c2);
252 case '=': return '#';
253 case '/': return '\\';
254 case '\'': return '^';
255 case '(': return '[';
256 case ')': return ']';
257 case '!': return '|';
258 case '<': return '{';
259 case '>': return '}';
260 case '-': return '~';
262 lex_ungetch(lex, c3);
263 lex_ungetch(lex, c2);
268 static int lex_try_digraph(lex_file *lex, int ch)
272 if (ch == '<' && c2 == ':')
274 else if (ch == ':' && c2 == '>')
276 else if (ch == '<' && c2 == '%')
278 else if (ch == '%' && c2 == '>')
280 else if (ch == '%' && c2 == ':')
282 lex_ungetch(lex, c2);
286 static int lex_getch(lex_file *lex)
292 if (lex->peek[lex->peekpos] == '\n')
294 return lex->peek[lex->peekpos];
301 return lex_try_trigraph(lex, ch);
302 else if (!lex->flags.nodigraphs && (ch == '<' || ch == ':' || ch == '%'))
303 return lex_try_digraph(lex, ch);
307 static void lex_ungetch(lex_file *lex, int ch)
309 lex->peek[lex->peekpos++] = ch;
314 /* classify characters
315 * some additions to the is*() functions of ctype.h
318 /* Idents are alphanumberic, but they start with alpha or _ */
319 static bool isident_start(int ch)
321 return isalpha(ch) || ch == '_';
324 static bool isident(int ch)
326 return isident_start(ch) || isdigit(ch);
329 /* isxdigit_only is used when we already know it's not a digit
330 * and want to see if it's a hex digit anyway.
332 static bool isxdigit_only(int ch)
334 return (ch >= 'a' && ch <= 'f') || (ch >= 'A' && ch <= 'F');
337 /* Append a character to the token buffer */
338 static void lex_tokench(lex_file *lex, int ch)
340 vec_push(lex->tok.value, ch);
343 /* Append a trailing null-byte */
344 static void lex_endtoken(lex_file *lex)
346 vec_push(lex->tok.value, 0);
347 vec_shrinkby(lex->tok.value, 1);
350 /* Skip whitespace and comments and return the first
351 * non-white character.
352 * As this makes use of the above getch() ungetch() functions,
353 * we don't need to care at all about line numbering anymore.
355 * In theory, this function should only be used at the beginning
356 * of lexing, or when we *know* the next character is part of the token.
357 * Otherwise, if the parser throws an error, the linenumber may not be
358 * the line of the error, but the line of the next token AFTER the error.
360 * This is currently only problematic when using c-like string-continuation,
361 * since comments and whitespaces are allowed between 2 such strings.
365 "A continuation of the previous string"
366 // This line is skipped
369 * In this case, if the parse decides it didn't actually want a string,
370 * and uses lex->line to print an error, it will show the ', foo);' line's
373 * On the other hand, the parser is supposed to remember the line of the next
374 * token's beginning. In this case we would want skipwhite() to be called
375 * AFTER reading a token, so that the parser, before reading the NEXT token,
376 * doesn't store teh *comment's* linenumber, but the actual token's linenumber.
379 * here is to store the line of the first character after skipping
380 * the initial whitespace in lex->sline, this happens in lex_do.
382 static int lex_skipwhite(lex_file *lex)
385 bool haswhite = false;
390 while (ch != EOF && isspace(ch)) {
391 if (lex->flags.preprocessing) {
394 /* see if there was whitespace first */
395 if (haswhite) { /* (vec_size(lex->tok.value)) { */
396 lex_ungetch(lex, ch);
400 /* otherwise return EOL */
404 lex_tokench(lex, ch);
413 /* one line comment */
416 if (lex->flags.preprocessing) {
419 lex_tokench(lex, '/');
420 lex_tokench(lex, '/');
422 lex_tokench(lex, ' ');
423 lex_tokench(lex, ' ');
426 while (ch != EOF && ch != '\n') {
427 if (lex->flags.preprocessing)
428 lex_tokench(lex, ' '); /* ch); */
431 if (lex->flags.preprocessing) {
432 lex_ungetch(lex, '\n');
440 /* multiline comment */
441 if (lex->flags.preprocessing) {
444 lex_tokench(lex, '/');
445 lex_tokench(lex, '*');
447 lex_tokench(lex, ' ');
448 lex_tokench(lex, ' ');
457 if (lex->flags.preprocessing) {
459 lex_tokench(lex, '*');
460 lex_tokench(lex, '/');
462 lex_tokench(lex, ' ');
463 lex_tokench(lex, ' ');
468 if (lex->flags.preprocessing) {
469 lex_tokench(lex, ' '); /* ch); */
472 ch = ' '; /* cause TRUE in the isspace check */
475 /* Otherwise roll back to the slash and break out of the loop */
476 lex_ungetch(lex, ch);
480 } while (ch != EOF && isspace(ch));
484 lex_ungetch(lex, ch);
491 static bool GMQCC_WARN lex_finish_ident(lex_file *lex)
496 while (ch != EOF && isident(ch))
498 lex_tokench(lex, ch);
502 /* last ch was not an ident ch: */
503 lex_ungetch(lex, ch);
508 /* read one ident for the frame list */
509 static int lex_parse_frame(lex_file *lex)
516 while (ch != EOF && ch != '\n' && isspace(ch))
522 if (!isident_start(ch)) {
523 lexerror(lex, "invalid framename, must start with one of a-z or _, got %c", ch);
527 lex_tokench(lex, ch);
528 if (!lex_finish_ident(lex))
534 /* read a list of $frames */
535 static bool lex_finish_frames(lex_file *lex)
542 rc = lex_parse_frame(lex);
543 if (rc > 0) /* end of line */
545 if (rc < 0) /* error */
548 for (i = 0; i < vec_size(lex->frames); ++i) {
549 if (!strcmp(lex->tok.value, lex->frames[i].name)) {
550 lex->frames[i].value = lex->framevalue++;
551 if (lexwarn(lex, WARN_FRAME_MACROS, "duplicate frame macro defined: `%s`", lex->tok.value))
556 if (i < vec_size(lex->frames))
559 m.value = lex->framevalue++;
560 m.name = util_strdup(lex->tok.value);
561 vec_shrinkto(lex->tok.value, 0);
562 vec_push(lex->frames, m);
566 static int GMQCC_WARN lex_finish_string(lex_file *lex, int quote)
574 return TOKEN_STRINGCONST;
576 if (lex->flags.preprocessing && ch == '\\') {
577 lex_tokench(lex, ch);
580 lexerror(lex, "unexpected end of file");
581 lex_ungetch(lex, EOF); /* next token to be TOKEN_EOF */
582 return (lex->tok.ttype = TOKEN_ERROR);
584 lex_tokench(lex, ch);
586 else if (ch == '\\') {
589 lexerror(lex, "unexpected end of file");
590 lex_ungetch(lex, EOF); /* next token to be TOKEN_EOF */
591 return (lex->tok.ttype = TOKEN_ERROR);
598 case 'a': ch = '\a'; break;
599 case 'b': ch = '\b'; break;
600 case 'r': ch = '\r'; break;
601 case 'n': ch = '\n'; break;
602 case 't': ch = '\t'; break;
603 case 'f': ch = '\f'; break;
604 case 'v': ch = '\v'; break;
606 lexwarn(lex, WARN_UNKNOWN_CONTROL_SEQUENCE, "unrecognized control sequence: \\%c", ch);
607 /* so we just add the character plus backslash no matter what it actually is */
608 lex_tokench(lex, '\\');
610 /* add the character finally */
611 lex_tokench(lex, ch);
614 lex_tokench(lex, ch);
616 lexerror(lex, "unexpected end of file within string constant");
617 lex_ungetch(lex, EOF); /* next token to be TOKEN_EOF */
618 return (lex->tok.ttype = TOKEN_ERROR);
621 static int GMQCC_WARN lex_finish_digit(lex_file *lex, int lastch)
627 /* parse a number... */
628 lex->tok.ttype = TOKEN_INTCONST;
630 lex_tokench(lex, ch);
633 if (ch != '.' && !isdigit(ch))
635 if (lastch != '0' || ch != 'x')
637 /* end of the number or EOF */
638 lex_ungetch(lex, ch);
641 lex->tok.constval.i = lastch - '0';
642 return lex->tok.ttype;
648 /* EOF would have been caught above */
652 lex_tokench(lex, ch);
654 while (isdigit(ch) || (ishex && isxdigit_only(ch)))
656 lex_tokench(lex, ch);
660 /* NOT else, '.' can come from above as well */
661 if (ch == '.' && !ishex)
663 /* Allow floating comma in non-hex mode */
664 lex->tok.ttype = TOKEN_FLOATCONST;
665 lex_tokench(lex, ch);
667 /* continue digits-only */
671 lex_tokench(lex, ch);
675 /* put back the last character */
676 /* but do not put back the trailing 'f' or a float */
677 if (lex->tok.ttype == TOKEN_FLOATCONST && ch == 'f')
680 /* generally we don't want words to follow numbers: */
682 lexerror(lex, "unexpected trailing characters after number");
683 return (lex->tok.ttype = TOKEN_ERROR);
685 lex_ungetch(lex, ch);
688 if (lex->tok.ttype == TOKEN_FLOATCONST)
689 lex->tok.constval.f = strtod(lex->tok.value, NULL);
691 lex->tok.constval.i = strtol(lex->tok.value, NULL, 0);
692 return lex->tok.ttype;
695 int lex_do(lex_file *lex)
706 ch = lex_skipwhite(lex);
707 if (!lex->flags.mergelines || ch != '\\')
711 lex_ungetch(lex, ch);
715 /* we reached a linemerge */
716 lex_tokench(lex, '\n');
720 lex->sline = lex->line;
721 lex->tok.ctx.line = lex->sline;
722 lex->tok.ctx.file = lex->name;
724 if (lex->flags.preprocessing && (ch == TOKEN_WHITE || ch == TOKEN_EOL || ch == TOKEN_FATAL)) {
725 return (lex->tok.ttype = ch);
729 return (lex->tok.ttype = TOKEN_FATAL);
733 return (lex->tok.ttype = TOKEN_EOF);
736 /* modelgen / spiritgen commands */
742 if (!isident_start(ch)) {
743 lexerror(lex, "hanging '$' modelgen/spritegen command line");
746 lex_tokench(lex, ch);
747 if (!lex_finish_ident(lex))
748 return (lex->tok.ttype = TOKEN_ERROR);
750 /* skip the known commands */
753 if (!strcmp(v, "frame") || !strcmp(v, "framesave"))
755 /* frame/framesave command works like an enum
756 * similar to fteqcc we handle this in the lexer.
757 * The reason for this is that it is sensitive to newlines,
758 * which the parser is unaware of
760 if (!lex_finish_frames(lex))
761 return (lex->tok.ttype = TOKEN_ERROR);
765 if (!strcmp(v, "framevalue"))
768 while (ch != EOF && isspace(ch) && ch != '\n')
772 lexerror(lex, "$framevalue requires an integer parameter");
777 lex->tok.ttype = lex_finish_digit(lex, ch);
779 if (lex->tok.ttype != TOKEN_INTCONST) {
780 lexerror(lex, "$framevalue requires an integer parameter");
783 lex->framevalue = lex->tok.constval.i;
787 if (!strcmp(v, "framerestore"))
793 rc = lex_parse_frame(lex);
796 lexerror(lex, "$framerestore requires a framename parameter");
800 return (lex->tok.ttype = TOKEN_FATAL);
803 for (frame = 0; frame < vec_size(lex->frames); ++frame) {
804 if (!strcmp(v, lex->frames[frame].name)) {
805 lex->framevalue = lex->frames[frame].value;
809 lexerror(lex, "unknown framename `%s`", v);
813 if (!strcmp(v, "modelname"))
819 rc = lex_parse_frame(lex);
822 lexerror(lex, "$modelname requires a parameter");
826 return (lex->tok.ttype = TOKEN_FATAL);
829 if (lex->modelname) {
831 m.value = lex->framevalue;
832 m.name = lex->modelname;
833 lex->modelname = NULL;
834 vec_push(lex->frames, m);
836 lex->modelname = lex->tok.value;
837 lex->tok.value = NULL;
841 if (!strcmp(v, "flush"))
844 for (frame = 0; frame < vec_size(lex->frames); ++frame)
845 mem_d(lex->frames[frame].name);
846 vec_free(lex->frames);
847 /* skip line (fteqcc does it too) */
849 while (ch != EOF && ch != '\n')
854 if (!strcmp(v, "cd") ||
855 !strcmp(v, "origin") ||
856 !strcmp(v, "base") ||
857 !strcmp(v, "flags") ||
858 !strcmp(v, "scale") ||
863 while (ch != EOF && ch != '\n')
868 for (frame = 0; frame < vec_size(lex->frames); ++frame) {
869 if (!strcmp(v, lex->frames[frame].name)) {
870 lex->tok.constval.i = lex->frames[frame].value;
871 return (lex->tok.ttype = TOKEN_INTCONST);
875 lexerror(lex, "invalid frame macro");
879 /* single-character tokens */
884 lex_tokench(lex, ch);
886 if (lex->flags.noops)
887 return (lex->tok.ttype = ch);
889 return (lex->tok.ttype = TOKEN_OPERATOR);
898 lex_tokench(lex, ch);
900 return (lex->tok.ttype = ch);
905 if (lex->flags.noops)
907 /* Detect characters early which are normally
908 * operators OR PART of an operator.
925 lex_tokench(lex, ch);
927 return (lex->tok.ttype = ch);
934 lex_tokench(lex, ch);
935 /* peak ahead once */
936 nextch = lex_getch(lex);
938 lex_ungetch(lex, nextch);
940 return (lex->tok.ttype = ch);
942 /* peak ahead again */
943 nextch = lex_getch(lex);
945 lex_ungetch(lex, nextch);
946 lex_ungetch(lex, nextch);
948 return (lex->tok.ttype = ch);
950 /* fill the token to be "..." */
951 lex_tokench(lex, ch);
952 lex_tokench(lex, ch);
954 return (lex->tok.ttype = TOKEN_DOTS);
958 if (ch == ',' || ch == '.') {
959 lex_tokench(lex, ch);
961 return (lex->tok.ttype = TOKEN_OPERATOR);
964 if (ch == '+' || ch == '-' || /* ++, --, +=, -= and -> as well! */
965 ch == '>' || ch == '<' || /* <<, >>, <=, >= */
966 ch == '=' || ch == '!' || /* ==, != */
967 ch == '&' || ch == '|') /* &&, ||, &=, |= */
969 lex_tokench(lex, ch);
971 nextch = lex_getch(lex);
972 if (nextch == ch || nextch == '=') {
973 lex_tokench(lex, nextch);
974 } else if (ch == '-' && nextch == '>') {
975 lex_tokench(lex, nextch);
977 lex_ungetch(lex, nextch);
980 return (lex->tok.ttype = TOKEN_OPERATOR);
984 if (ch == '^' || ch == '~' || ch == '!')
986 lex_tokench(lex, ch);
988 return (lex->tok.ttype = TOKEN_OPERATOR);
992 if (ch == '*' || ch == '/') /* *=, /= */
994 lex_tokench(lex, ch);
996 nextch = lex_getch(lex);
998 lex_tokench(lex, nextch);
1000 lex_ungetch(lex, nextch);
1003 return (lex->tok.ttype = TOKEN_OPERATOR);
1006 if (isident_start(ch))
1010 lex_tokench(lex, ch);
1011 if (!lex_finish_ident(lex)) {
1013 return (lex->tok.ttype = TOKEN_ERROR);
1016 lex->tok.ttype = TOKEN_IDENT;
1019 if (!strcmp(v, "void")) {
1020 lex->tok.ttype = TOKEN_TYPENAME;
1021 lex->tok.constval.t = TYPE_VOID;
1022 } else if (!strcmp(v, "int")) {
1023 lex->tok.ttype = TOKEN_TYPENAME;
1024 lex->tok.constval.t = TYPE_INTEGER;
1025 } else if (!strcmp(v, "float")) {
1026 lex->tok.ttype = TOKEN_TYPENAME;
1027 lex->tok.constval.t = TYPE_FLOAT;
1028 } else if (!strcmp(v, "string")) {
1029 lex->tok.ttype = TOKEN_TYPENAME;
1030 lex->tok.constval.t = TYPE_STRING;
1031 } else if (!strcmp(v, "entity")) {
1032 lex->tok.ttype = TOKEN_TYPENAME;
1033 lex->tok.constval.t = TYPE_ENTITY;
1034 } else if (!strcmp(v, "vector")) {
1035 lex->tok.ttype = TOKEN_TYPENAME;
1036 lex->tok.constval.t = TYPE_VECTOR;
1037 } else if (!strcmp(v, "for") ||
1038 !strcmp(v, "while") ||
1041 !strcmp(v, "else") ||
1042 !strcmp(v, "local") ||
1043 !strcmp(v, "return") ||
1044 !strcmp(v, "not") ||
1045 !strcmp(v, "const"))
1047 lex->tok.ttype = TOKEN_KEYWORD;
1049 else if (opts_standard != COMPILER_QCC)
1051 /* other standards reserve these keywords */
1052 if (!strcmp(v, "switch") ||
1053 !strcmp(v, "struct") ||
1054 !strcmp(v, "union") ||
1055 !strcmp(v, "break") ||
1056 !strcmp(v, "continue") ||
1059 lex->tok.ttype = TOKEN_KEYWORD;
1063 return lex->tok.ttype;
1068 lex->flags.nodigraphs = true;
1069 if (lex->flags.preprocessing)
1070 lex_tokench(lex, ch);
1071 lex->tok.ttype = lex_finish_string(lex, '"');
1072 if (lex->flags.preprocessing)
1073 lex_tokench(lex, ch);
1074 while (!lex->flags.preprocessing && lex->tok.ttype == TOKEN_STRINGCONST)
1076 /* Allow c style "string" "continuation" */
1077 ch = lex_skipwhite(lex);
1079 lex_ungetch(lex, ch);
1083 lex->tok.ttype = lex_finish_string(lex, '"');
1085 lex->flags.nodigraphs = false;
1087 return lex->tok.ttype;
1092 /* we parse character constants like string,
1093 * but return TOKEN_CHARCONST, or a vector type if it fits...
1094 * Likewise actual unescaping has to be done by the parser.
1095 * The difference is we don't allow 'char' 'continuation'.
1097 if (lex->flags.preprocessing)
1098 lex_tokench(lex, ch);
1099 lex->tok.ttype = lex_finish_string(lex, '\'');
1100 if (lex->flags.preprocessing)
1101 lex_tokench(lex, ch);
1104 /* It's a vector if we can successfully scan 3 floats */
1106 if (sscanf_s(lex->tok.value, " %f %f %f ",
1107 &lex->tok.constval.v.x, &lex->tok.constval.v.y, &lex->tok.constval.v.z) == 3)
1109 if (sscanf(lex->tok.value, " %f %f %f ",
1110 &lex->tok.constval.v.x, &lex->tok.constval.v.y, &lex->tok.constval.v.z) == 3)
1114 lex->tok.ttype = TOKEN_VECTORCONST;
1117 return lex->tok.ttype;
1122 lex->tok.ttype = lex_finish_digit(lex, ch);
1124 return lex->tok.ttype;
1127 lexerror(lex, "unknown token");
1128 return (lex->tok.ttype = TOKEN_ERROR);