+/*
+ * This file is part of cparser.
+ * Copyright (C) 2007-2009 Matthias Braun <matze@braunis.de>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
+ * 02111-1307, USA.
+ */
#include <config.h>
#include "token_t.h"
+#include "symbol_t.h"
#include <assert.h>
#include <stdio.h>
#include "symbol.h"
+#include "lang_features.h"
#include "adt/array.h"
static symbol_t *token_symbols[T_LAST_TOKEN];
+static symbol_t *pp_token_symbols[TP_LAST_TOKEN];
-void init_tokens(void)
+const source_position_t builtin_source_position = { "<built-in>", 0, 0 };
+
+static int last_id;
+
+static symbol_t *intern_register_token(token_kind_t id, const char *string)
+{
+ assert(0 <= id && id < T_LAST_TOKEN);
+ symbol_t *symbol = symbol_table_insert(string);
+ if (token_symbols[id] == NULL)
+ token_symbols[id] = symbol;
+ return symbol;
+}
+
+static symbol_t *intern_register_pp_token(preprocessor_token_kind_t id, const char *string)
{
- symbol_t *symbol;
+ assert(0 <= id && id < TP_LAST_TOKEN);
+ symbol_t *symbol = symbol_table_insert(string);
+ if (pp_token_symbols[id] == NULL)
+ pp_token_symbols[id] = symbol;
+ return symbol;
+}
- memset(token_symbols, 0, T_LAST_TOKEN * sizeof(token_symbols[0]));
+static void register_token(unsigned mode, token_kind_t id, const char *string)
+{
+ if (id > 255) {
+ assert(id >= last_id);
+ last_id = id;
+ }
+ if (c_mode & mode) {
+ symbol_t *symbol = intern_register_token(id, string);
+ symbol->ID = id;
+ }
+}
+
+static void register_pp_token(unsigned mode, token_kind_t id,
+ const char *string)
+{
+ if (! (c_mode & mode))
+ return;
+
+ symbol_t *symbol = intern_register_pp_token(id, string);
+ symbol->pp_ID = id;
+}
-#define T(x,str,val) \
- assert(T_##x >= 0 && T_##x < T_LAST_TOKEN); \
- symbol = symbol_table_insert(str); \
- symbol->ID = T_##x; \
- token_symbols[T_##x] = symbol;
+void init_tokens(void)
+{
+ memset(token_symbols, 0, T_LAST_TOKEN * sizeof(token_symbols[0]));
+ memset(pp_token_symbols, 0, TP_LAST_TOKEN * sizeof(pp_token_symbols[0]));
-#define TS(x,str,val) \
- assert(T_##x >= 0 && T_##x < T_LAST_TOKEN); \
- symbol = symbol_table_insert(str); \
- token_symbols[T_##x] = symbol;
+ last_id = -2;
+#define T(mode,x,str,val) register_token(mode, T_##x, str);
+#define TS(x,str,val) intern_register_token(T_##x, str);
#include "tokens.inc"
+#undef TS
+#undef T
+#define T(mode,x,str,val) register_pp_token(mode, TP_##x, str);
+#define TS(x,str,val) intern_register_pp_token(TP_##x, str);
+#include "tokens_preprocessor.inc"
#undef TS
#undef T
}
{
}
-void print_token_type(FILE *f, token_type_t token_type)
+void print_token_kind(FILE *f, token_kind_t token_kind)
{
- if(token_type >= 0 && token_type < 256) {
- fprintf(f, "'%c'", token_type);
+ if(token_kind == T_EOF) {
+ fputs("end of file", f);
return;
}
- if(token_type == T_EOF) {
- fputs("end of file", f);
+ if(token_kind == T_ERROR) {
+ fputs("error", f);
return;
}
int token_symbols_len = T_LAST_TOKEN;
- if(token_type < 0 || token_type >= token_symbols_len) {
+ if(token_kind < 0 || token_kind >= token_symbols_len) {
fputs("invalid token", f);
return;
}
- const symbol_t *symbol = token_symbols[token_type];
+ const symbol_t *symbol = token_symbols[token_kind];
if(symbol != NULL) {
fputs(symbol->string, f);
} else {
+ if(token_kind >= 0 && token_kind < 256) {
+ fputc(token_kind, f);
+ return;
+ }
fputs("unknown token", f);
}
}
+symbol_t *get_token_kind_symbol(int kind)
+{
+ return token_symbols[kind];
+}
+
+static void print_stringrep(const string_t *string, FILE *f)
+{
+ for (size_t i = 0; i < string->size; ++i) {
+ fputc(string->begin[i], f);
+ }
+}
+
void print_token(FILE *f, const token_t *token)
{
- switch(token->type) {
+ switch(token->kind) {
case T_IDENTIFIER:
- fprintf(f, "symbol '%s'", token->v.symbol->string);
+ fprintf(f, "identifier '%s'", token->identifier.symbol->string);
break;
case T_INTEGER:
- fprintf(f, "integer number %d", token->v.intvalue);
+ case T_INTEGER_OCTAL:
+ case T_INTEGER_HEXADECIMAL:
+ case T_FLOATINGPOINT:
+ case T_FLOATINGPOINT_HEXADECIMAL:
+ print_token_kind(f, (token_kind_t)token->kind);
+ fputs(" '", f);
+ print_stringrep(&token->number.number, f);
+ if (token->number.suffix.size > 0)
+ print_stringrep(&token->number.suffix, f);
+ fputc('\'', f);
break;
+ case T_WIDE_STRING_LITERAL:
case T_STRING_LITERAL:
- fprintf(f, "string '%s'", token->v.string);
+ print_token_kind(f, (token_kind_t)token->kind);
+ fprintf(f, " \"%s\"", token->string.string.begin);
+ break;
+ case T_CHARACTER_CONSTANT:
+ case T_WIDE_CHARACTER_CONSTANT:
+ print_token_kind(f, (token_kind_t)token->kind);
+ fputs(" \'", f);
+ print_stringrep(&token->string.string, f);
+ fputs("'", f);
break;
default:
- print_token_type(f, token->type);
+ fputc('\'', f);
+ print_token_kind(f, (token_kind_t)token->kind);
+ fputc('\'', f);
break;
}
}
+
+void print_pp_token_kind(FILE *f, int token_kind)
+{
+ if (token_kind == TP_EOF) {
+ fputs("end of file", f);
+ return;
+ }
+ if (token_kind == TP_ERROR) {
+ fputs("error", f);
+ return;
+ }
+
+ int token_symbols_len = TP_LAST_TOKEN;
+ if (token_kind < 0 || token_kind >= token_symbols_len) {
+ fputs("invalid token", f);
+ return;
+ }
+
+ const symbol_t *symbol = pp_token_symbols[token_kind];
+ if (symbol != NULL) {
+ fputs(symbol->string, f);
+ } else {
+ if(token_kind >= 0 && token_kind < 256) {
+ fputc(token_kind, f);
+ return;
+ }
+ fputs("unknown token", f);
+ }
+}
+
+void print_pp_token(FILE *f, const token_t *token)
+{
+ switch((preprocessor_token_kind_t) token->kind) {
+ case TP_IDENTIFIER:
+ fprintf(f, "identifier '%s'", token->identifier.symbol->string);
+ break;
+ case TP_NUMBER:
+ fprintf(f, "number '%s'", token->number.number.begin);
+ break;
+ case TP_STRING_LITERAL:
+ fprintf(f, "string \"%s\"", token->string.string.begin);
+ break;
+ default:
+ print_pp_token_kind(f, (preprocessor_token_kind_t) token->kind);
+ break;
+ }
+}
+
+bool tokens_would_paste(preprocessor_token_kind_t token1,
+ preprocessor_token_kind_t token2)
+{
+ char c = token2 < 256 ? (char) token2 : pp_token_symbols[token2]->string[0];
+
+ switch (token1) {
+ case '>': return c == '>' || c == '=';
+ case '<': return c == '<' || c == '=' || c == '%' || c == ':';
+ case '+': return c == '+' || c == '=';
+ case '-': return c == '-' || c == '>';
+ case '/': return c == '/' || c == '=' || c == '*';
+ case '%': return c == ':' || c == '=' || c == '>';
+ case '&': return c == '&' || c == '=';
+ case '|': return c == '|' || c == '=';
+ case ':': return c == ':' || c == '>';
+ case '*': return c == '*' || c == '=';
+ case '.': return c == '.' || c == '%' || token2 == TP_NUMBER;
+ case '#': return c == '#' || c == '%';
+ case TP_GREATERGREATER: return c == '=';
+ case TP_LESSLESS: return c == '=';
+ case '^': return c == '=';
+ case '!': return c == '=';
+ case TP_IDENTIFIER:
+ return token2 == TP_IDENTIFIER || token2 == TP_NUMBER ||
+ token2 == TP_CHARACTER_CONSTANT ||
+ token2 == TP_WIDE_CHARACTER_CONSTANT ||
+ token2 == TP_WIDE_STRING_LITERAL ||
+ token2 == TP_STRING_LITERAL; /* L */
+ case TP_NUMBER:
+ return token2 == TP_NUMBER || token2 == TP_IDENTIFIER ||
+ token2 == '.' || token2 == '+' || token2 == '-';
+ default:
+ return false;
+ }
+}