X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/b87f8b2159820e022e3b726b8369f74d64712f8f..536545f3a45d64cad0469efa125d3d96052e723c:/src/symtab.c diff --git a/src/symtab.c b/src/symtab.c index c255a8f3..0a9d6baa 100644 --- a/src/symtab.c +++ b/src/symtab.c @@ -20,34 +20,179 @@ #include "system.h" +#include "quotearg.h" #include "hash.h" +#include "complain.h" #include "symtab.h" #include "gram.h" +/*------------------------. +| Distinguished symbols. | +`------------------------*/ + +symbol_t *errtoken = NULL; +symbol_t *undeftoken = NULL; +symbol_t *eoftoken = NULL; +symbol_t *axiom = NULL; +symbol_t *startsymbol = NULL; +location_t startsymbol_location; + /*---------------------------------. | Create a new symbol, named TAG. | `---------------------------------*/ static symbol_t * -symbol_new (const char *tag) +symbol_new (const char *tag, location_t location) { symbol_t *res = XMALLOC (symbol_t, 1); res->tag = xstrdup (tag); + res->location = location; + res->type_name = NULL; + res->destructor = NULL; + res->printer = NULL; + res->number = NUMBER_UNDEFINED; res->prec = 0; res->assoc = right_assoc; res->user_token_number = USER_NUMBER_UNDEFINED; + res->alias = NULL; res->class = unknown_sym; nsyms++; - return res; } +/*------------------------------------------------------------------. +| Set the TYPE_NAME associated to SYMBOL. Does nothing if passed 0 | +| as TYPE_NAME. | +`------------------------------------------------------------------*/ + +void +symbol_type_set (symbol_t *symbol, char *type_name, location_t location) +{ + if (type_name) + { + if (symbol->type_name) + complain_at (location, + _("type redeclaration for %s"), symbol->tag); + symbol->type_name = type_name; + } +} + + +/*-------------------------------------------------------------------. +| Set the DESTRUCTOR associated to SYMBOL. Do nothing if passed 0. | +`-------------------------------------------------------------------*/ + +void +symbol_destructor_set (symbol_t *symbol, char *destructor, location_t location) +{ + if (destructor) + { + if (symbol->destructor) + complain_at (location, + _("%s redeclaration for %s"), + "%destructor", symbol->tag); + symbol->destructor = destructor; + symbol->destructor_location = location; + } +} + + +/*----------------------------------------------------------------. +| Set the PRITNER associated to SYMBOL. Do nothing if passed 0. | +`----------------------------------------------------------------*/ + +void +symbol_printer_set (symbol_t *symbol, char *printer, location_t location) +{ + if (printer) + { + if (symbol->printer) + complain_at (location, + _("%s redeclaration for %s"), + "%printer", symbol->tag); + symbol->printer = printer; + symbol->printer_location = location; + } +} + + +/*------------------------------------------------------------------. +| Set the PRECEDENCE associated to SYMBOL. Does nothing if invoked | +| with UNDEF_ASSOC as ASSOC. | +`------------------------------------------------------------------*/ + +void +symbol_precedence_set (symbol_t *symbol, + int prec, associativity assoc, location_t location) +{ + if (assoc != undef_assoc) + { + if (symbol->prec != 0) + complain_at (location, + _("redefining precedence of %s"), + symbol->tag); + symbol->prec = prec; + symbol->assoc = assoc; + } + + /* Only terminals have a precedence. */ + symbol_class_set (symbol, token_sym, location); +} + + +/*-------------------------------------. +| Set the CLASS associated to SYMBOL. | +`-------------------------------------*/ + +void +symbol_class_set (symbol_t *symbol, symbol_class class, location_t location) +{ + if (symbol->class != unknown_sym && symbol->class != class) + complain_at (location, _("symbol %s redefined"), symbol->tag); + + if (class == nterm_sym && symbol->class != nterm_sym) + symbol->number = nvars++; + else if (class == token_sym && symbol->number == NUMBER_UNDEFINED) + symbol->number = ntokens++; + + symbol->class = class; +} + + +/*-------------------------------------------------. +| Set the USER_TOKEN_NUMBER associated to SYMBOL. | +`-------------------------------------------------*/ + +void +symbol_user_token_number_set (symbol_t *symbol, + int user_token_number, location_t location) +{ + assert (symbol->class == token_sym); + + if (symbol->user_token_number != USER_NUMBER_UNDEFINED + && symbol->user_token_number != user_token_number) + complain_at (location, _("redefining user token number of %s"), + symbol->tag); + + symbol->user_token_number = user_token_number; + /* User defined EOF token? */ + if (user_token_number == 0) + { + eoftoken = symbol; + eoftoken->number = 0; + /* It is always mapped to 0, so it was already counted in + NTOKENS. */ + --ntokens; + } +} + + /*------------. | Free THIS. | `------------*/ @@ -65,9 +210,169 @@ symbol_free (symbol_t *this) } +/*-----------------------------------------------------------. +| If THIS is not defined, report an error, and consider it a | +| nonterminal. | +`-----------------------------------------------------------*/ + +static bool +symbol_check_defined (symbol_t *this) +{ + if (this->class == unknown_sym) + { + complain_at + (this->location, + _("symbol %s is used, but is not defined as a token and has no rules"), + this->tag); + this->class = nterm_sym; + this->number = nvars++; + } + + return TRUE; +} + + +/*-------------------------------------------------------------------. +| Declare the new SYMBOL. Make it an alias of SYMVAL, and type them | +| with TYPENAME. | +`-------------------------------------------------------------------*/ + +void +symbol_make_alias (symbol_t *symbol, symbol_t *symval) +{ + if (symval->alias) + warn (_("symbol `%s' used more than once as a literal string"), + symval->tag); + else if (symbol->alias) + warn (_("symbol `%s' given more than one literal string"), + symbol->tag); + else + { + symval->class = token_sym; + symval->user_token_number = symbol->user_token_number; + symbol->user_token_number = USER_NUMBER_ALIAS; + symval->alias = symbol; + symbol->alias = symval; + /* symbol and symval combined are only one symbol */ + nsyms--; + ntokens--; + assert (ntokens == symbol->number || ntokens == symval->number); + symbol->number = symval->number = + (symval->number < symbol->number) ? symval->number : symbol->number; + } +} + + +/*---------------------------------------------------------. +| Check that THIS, and its alias, have same precedence and | +| associativity. | +`---------------------------------------------------------*/ + +static bool +symbol_check_alias_consistence (symbol_t *this) +{ + /* Check only those who _are_ the aliases. */ + if (this->alias && this->user_token_number == USER_NUMBER_ALIAS) + { + if (this->prec != this->alias->prec) + { + if (this->prec != 0 && this->alias->prec != 0) + complain (_("conflicting precedences for %s and %s"), + this->tag, this->alias->tag); + if (this->prec != 0) + this->alias->prec = this->prec; + else + this->prec = this->alias->prec; + } + + if (this->assoc != this->alias->assoc) + { + /* FIXME: For some reason (probably the S/R => keep the S), + the right assoc is chosen has the ``not set''. This is + not nice, fix this! */ + if (this->assoc != right_assoc + && this->alias->assoc != right_assoc) + complain (_("conflicting associativities for %s and %s"), + this->tag, this->alias->tag); + if (this->assoc != 0) + this->alias->assoc = this->assoc; + else + this->assoc = this->alias->assoc; + } + } + return TRUE; +} + + +/*-------------------------------------------------------------------. +| Assign a symbol number, and write the definition of the token name | +| into FDEFINES. Put in SYMBOLS. | +`-------------------------------------------------------------------*/ + +static bool +symbol_pack (symbol_t *this) +{ + if (this->class == nterm_sym) + { + this->number += ntokens; + } + else if (this->alias) + { + /* This symbol and its alias are a single token defn. + Allocate a tokno, and assign to both check agreement of + prec and assoc fields and make both the same */ + if (this->number == NUMBER_UNDEFINED) + { + if (this == eoftoken || this->alias == eoftoken) + this->number = this->alias->number = 0; + else + { + assert (this->alias->number != NUMBER_UNDEFINED); + this->number = this->alias->number; + } + } + /* Do not do processing below for USER_NUMBER_ALIASs. */ + if (this->user_token_number == USER_NUMBER_ALIAS) + return TRUE; + } + else /* this->class == token_sym */ + { + assert (this->number != NUMBER_UNDEFINED); + } + + symbols[this->number] = this; + return TRUE; +} + + + + +/*--------------------------------------------------. +| Put THIS in TOKEN_TRANSLATIONS if it is a token. | +`--------------------------------------------------*/ + +static bool +symbol_translation (symbol_t *this) +{ + /* Non-terminal? */ + if (this->class == token_sym + && this->user_token_number != USER_NUMBER_ALIAS) + { + /* A token which translation has already been set? */ + if (token_translations[this->user_token_number] != undeftoken->number) + complain (_("tokens %s and %s both assigned number %d"), + symbols[token_translations[this->user_token_number]]->tag, + this->tag, this->user_token_number); + + token_translations[this->user_token_number] = this->number; + } + + return TRUE; +} + /*----------------------. -| A symbol_t hash table. | +| A symbol hash table. | `----------------------*/ /* Initial capacity of symbols hash table. */ @@ -89,7 +394,7 @@ hash_symbol_t (const symbol_t *m, unsigned int tablesize) /*-------------------------------. -| Create the symbol_t hash table. | +| Create the symbol hash table. | `-------------------------------*/ void @@ -109,24 +414,48 @@ symbols_new (void) `----------------------------------------------------------------*/ symbol_t * -getsym (const char *key) +symbol_get (const char *key, location_t location) { symbol_t probe; symbol_t *entry; + /* Keep the symbol in a printable form. */ + key = quotearg_style (escape_quoting_style, key); (const char *) probe.tag = key; entry = hash_lookup (symbol_table, &probe); if (!entry) { /* First insertion in the hash. */ - entry = symbol_new (key); + entry = symbol_new (key, location); hash_insert (symbol_table, entry); } return entry; } +/*------------------------------------------------------------------. +| Generate a dummy nonterminal, whose name cannot conflict with the | +| user's names. | +`------------------------------------------------------------------*/ + +symbol_t * +dummy_symbol_get (location_t location) +{ + /* Incremented for each generated symbol. */ + static int dummy_count = 0; + static char buf[256]; + + symbol_t *sym; + + sprintf (buf, "@%d", ++dummy_count); + sym = symbol_get (buf, location); + sym->class = nterm_sym; + sym->number = nvars++; + return sym; +} + + /*-------------------. | Free the symbols. | `-------------------*/ @@ -135,6 +464,7 @@ void symbols_free (void) { hash_free (symbol_table); + free (symbols); } @@ -150,3 +480,95 @@ symbols_do (symbol_processor processor, void *processor_data) (Hash_processor) processor, processor_data); } + + +/*--------------------------------------------------------------. +| Check that all the symbols are defined. Report any undefined | +| symbols and consider them nonterminals. | +`--------------------------------------------------------------*/ + +void +symbols_check_defined (void) +{ + symbols_do (symbol_check_defined, NULL); +} + +/*------------------------------------------------------------------. +| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same | +| number. | +`------------------------------------------------------------------*/ + +static void +symbols_token_translations_init (void) +{ + int num_256_available_p = TRUE; + int i; + + /* Find the highest user token number, and whether 256, the POSIX + preferred user token number for the error token, is used. */ + max_user_token_number = 0; + for (i = 0; i < ntokens; ++i) + { + symbol_t *this = symbols[i]; + if (this->user_token_number != USER_NUMBER_UNDEFINED) + { + if (this->user_token_number > max_user_token_number) + max_user_token_number = this->user_token_number; + if (this->user_token_number == 256) + num_256_available_p = FALSE; + } + } + + /* If 256 is not used, assign it to error, to follow POSIX. */ + if (num_256_available_p + && errtoken->user_token_number == USER_NUMBER_UNDEFINED) + errtoken->user_token_number = 256; + + /* Set the missing user numbers. */ + if (max_user_token_number < 256) + max_user_token_number = 256; + + for (i = 0; i < ntokens; ++i) + { + symbol_t *this = symbols[i]; + if (this->user_token_number == USER_NUMBER_UNDEFINED) + this->user_token_number = ++max_user_token_number; + if (this->user_token_number > max_user_token_number) + max_user_token_number = this->user_token_number; + } + + token_translations = XCALLOC (symbol_number_t, max_user_token_number + 1); + + /* Initialize all entries for literal tokens to 2, the internal + token number for $undefined., which represents all invalid + inputs. */ + for (i = 0; i < max_user_token_number + 1; i++) + token_translations[i] = undeftoken->number; + symbols_do (symbol_translation, NULL); +} + + +/*----------------------------------------------------------------. +| Assign symbol numbers, and write definition of token names into | +| FDEFINES. Set up vectors SYMBOL_TABLE, TAGS of symbols. | +`----------------------------------------------------------------*/ + +void +symbols_pack (void) +{ + symbols = XCALLOC (symbol_t *, nsyms); + + symbols_do (symbol_check_alias_consistence, NULL); + symbols_do (symbol_pack, NULL); + + symbols_token_translations_init (); + + if (startsymbol->class == unknown_sym) + fatal_at (startsymbol_location, + _("the start symbol %s is undefined"), + startsymbol->tag); + else if (startsymbol->class == token_sym) + fatal_at (startsymbol_location, + _("the start symbol %s is a token"), + startsymbol->tag); +}