typedef struct symbol_list
{
struct symbol_list *next;
- bucket *sym;
+ symbol_t *sym;
int line;
/* The action is attached to the LHS of a rule. */
/* The guard is attached to the LHS of a rule. */
const char *guard;
int guard_line;
- bucket *ruleprec;
+ symbol_t *ruleprec;
} symbol_list;
int lineno;
-static symbol_list *grammar;
-static int start_flag;
-static bucket *startval;
+static symbol_list *grammar = NULL;
+static int start_flag = 0;
+static symbol_t *startval = NULL;
/* Nonzero if components of semantic values are used, implying
they must be unions. */
static int value_components_used;
/* Nonzero if %union has been seen. */
-static int typed;
+static int typed = 0;
/* Incremented for each %left, %right or %nonassoc seen */
-static int lastprec;
+static int lastprec = 0;
-static bucket *errtoken = NULL;
-static bucket *undeftoken = NULL;
-static bucket *eoftoken = NULL;
-static bucket *axiom = NULL;
+symbol_t *errtoken = NULL;
+symbol_t *undeftoken = NULL;
+symbol_t *eoftoken = NULL;
+symbol_t *axiom = NULL;
static symbol_list *
-symbol_list_new (bucket *sym)
+symbol_list_new (symbol_t *sym)
{
symbol_list *res = XMALLOC (symbol_list, 1);
res->next = NULL;
return res;
}
+/*------------------------.
+| Operations on symbols. |
+`------------------------*/
+
+
+/*-----------------------------------------------------------.
+| 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
+ (_("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;
+}
+
+
+/*-------------------------------------------------------------------.
+| Assign a symbol number, and write the definition of the token name |
+| into FDEFINES. Put in SYMBOLS. |
+`-------------------------------------------------------------------*/
+
+static bool
+symbol_make_alias (symbol_t *symbol, char *typename)
+{
+ 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->type_name = typename;
+ symval->user_token_number = symbol->user_token_number;
+ symbol->user_token_number = SALIAS;
+ 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;
+ }
+
+ return TRUE;
+}
+
+/*---------------------------------------------------------.
+| 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 == SALIAS)
+ {
+ 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)
+ {
+ if (this->assoc != 0 && this->alias->assoc != 0)
+ complain (_("conflicting assoc values 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 SALIASs. */
+ if (this->user_token_number == SALIAS)
+ 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 != SALIAS)
+ {
+ /* 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;
+}
\f
/*===================\
complain (_("invalid $ value"));
return NULL;
}
- i++;
+ ++i;
}
return rp->sym->type_name;
obstack_1grow (oout, c);
if (c == '\n')
- lineno++;
+ ++lineno;
}
c = getc (fin);
}
else if (c == '\n')
{
- lineno++;
+ ++lineno;
obstack_1grow (oout, c);
if (cplus_comment)
ended = 1;
{
case '\n':
obstack_1grow (&attrs_obstack, c);
- lineno++;
+ ++lineno;
break;
case '%':
char *typename = NULL;
/* The symbol being defined. */
- struct bucket *symbol = NULL;
+ symbol_t *symbol = NULL;
/* After `%token' and `%nterm', any number of symbols maybe be
defined. */
}
else if (token == tok_identifier && *symval->tag == '\"' && symbol)
{
- 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->type_name = typename;
- symval->user_token_number = symbol->user_token_number;
- symbol->user_token_number = SALIAS;
- symval->alias = symbol;
- symbol->alias = symval;
- /* symbol and symval combined are only one symbol */
- nsyms--;
- }
+ symbol_make_alias (symbol, typename);
symbol = NULL;
}
else if (token == tok_identifier)
complain (_("symbol %s redefined"), symbol->tag);
symbol->class = what_is;
if (what_is == nterm_sym && oldclass != nterm_sym)
- symbol->value = nvars++;
+ symbol->number = nvars++;
+ if (what_is == token_sym && symbol->number == NUMBER_UNDEFINED)
+ symbol->number = ntokens++;
if (typename)
{
symbol->user_token_number = numval;
/* User defined EOF token? */
if (numval == 0)
- eoftoken = symbol;
+ {
+ eoftoken = symbol;
+ eoftoken->number = 0;
+ /* It is always mapped to 0, so it was already counted in
+ NTOKENS. */
+ --ntokens;
+ }
}
else
{
char *name = NULL;
int prev = 0;
- lastprec++; /* Assign a new precedence level, never 0. */
+ /* Assign a new precedence level, never 0. */
+ ++lastprec;
for (;;)
{
symval->assoc = assoc;
if (symval->class == nterm_sym)
complain (_("symbol %s redefined"), symval->tag);
- symval->class = token_sym;
+ if (symval->number == NUMBER_UNDEFINED)
+ {
+ symval->number = ntokens++;
+ symval->class = token_sym;
+ }
if (name)
{ /* record the type, if one is specified */
if (symval->type_name == NULL)
}
else
{
- complain (_
- ("invalid text (%s) - number should be after identifier"),
-token_buffer);
+ complain
+ (_("invalid text (%s) - number should be after identifier"),
+ token_buffer);
skip_to_char ('%');
}
break;
typed = 1;
+ MUSCLE_INSERT_INT ("stype_line", lineno);
obstack_init (&union_obstack);
obstack_sgrow (&union_obstack, "union");
switch (c)
{
case '\n':
- lineno++;
+ ++lineno;
break;
case '/':
break;
case '{':
- count++;
+ ++count;
break;
case '}':
parse_thong_decl (void)
{
token_t token;
- struct bucket *symbol;
+ symbol_t *symbol;
char *typename = 0;
int usrtoknum = SUNDEF;
case tok_stropt:
case tok_intopt:
case tok_obsolete:
- abort ();
+ assert (0);
break;
case tok_illegal:
{
case '\n':
obstack_1grow (&action_obstack, c);
- lineno++;
+ ++lineno;
break;
case '{':
obstack_1grow (&action_obstack, c);
- count++;
+ ++count;
break;
case '\'':
| with the user's names. |
`-------------------------------------------------------------------*/
-static bucket *
+static symbol_t *
gensym (void)
{
/* Incremented for each generated symbol */
static int gensym_count = 0;
static char buf[256];
- bucket *sym;
+ symbol_t *sym;
sprintf (buf, "@%d", ++gensym_count);
token_buffer = buf;
sym = getsym (token_buffer);
sym->class = nterm_sym;
- sym->value = nvars++;
+ sym->number = nvars++;
return sym;
}
\f
readgram (void)
{
token_t t;
- bucket *lhs = NULL;
+ symbol_t *lhs = NULL;
symbol_list *p = NULL;
symbol_list *p1 = NULL;
- bucket *bp;
/* Points to first symbol_list of current rule. its symbol is the
lhs of the rule. */
/* Number of symbols in rhs of this rule so far */
int rulelength = 0;
int xactions = 0; /* JF for error checking */
- bucket *first_rhs = 0;
+ symbol_t *first_rhs = 0;
if (t == tok_identifier)
{
}
/* start a new rule and record its lhs. */
- nrules++;
- nitems++;
+ ++nrules;
+ ++nritems;
p = symbol_list_new (lhs);
if (lhs->class == unknown_sym)
{
lhs->class = nterm_sym;
- lhs->value = nvars;
- nvars++;
+ lhs->number = nvars;
+ ++nvars;
}
else if (lhs->class == token_sym)
complain (_("rule given for %s, which is a token"), lhs->tag);
If one does, exit this rule now. */
if (t == tok_identifier)
{
- bucket *ssave;
+ symbol_t *ssave;
token_t t1;
ssave = symval;
symval = ssave;
if (t1 == tok_colon)
{
- complain (_("previous rule lacks an ending `;'"));
+ warn (_("previous rule lacks an ending `;'"));
break;
}
inserting the new rule before it. */
/* Make a dummy nonterminal, a gensym. */
- bucket *sdummy = gensym ();
+ symbol_t *sdummy = gensym ();
/* Make a new rule, whose body is empty, before the
current one, so that the action just read can
belong to it. */
- nrules++;
- nitems++;
+ ++nrules;
+ ++nritems;
p = symbol_list_new (sdummy);
/* Attach its lineno to that of the host rule. */
p->line = crule->line;
/* Insert the dummy generated by that rule into this
rule. */
- nitems++;
+ ++nritems;
p = symbol_list_new (sdummy);
p1->next = p;
p1 = p;
if (t == tok_identifier)
{
- nitems++;
+ ++nritems;
p = symbol_list_new (symval);
p1->next = p;
p1 = p;
{
parse_action (crule, rulelength);
action_flag = 1;
- xactions++; /* JF */
+ ++xactions; /* JF */
}
- rulelength++;
+ ++rulelength;
} /* end of read rhs of rule */
/* Put an empty link in the list to mark the end of this rule */
complain (_("two actions at end of one rule"));
parse_action (crule, rulelength);
action_flag = 1;
- xactions++; /* -wjh */
+ ++xactions; /* -wjh */
t = lex ();
}
/* If $$ is being set in default way, report if any type
else if (!xactions && !first_rhs && lhs->type_name != 0)
complain (_("empty rule for typed nonterminal, and no action"));
if (t == tok_two_percents || t == tok_eof)
- complain (_("previous rule lacks an ending `;'"));
+ warn (_("previous rule lacks an ending `;'"));
if (t == tok_semicolon)
t = lex ();
}
fatal (_("no rules in the input grammar"));
/* Report any undefined symbols and consider them nonterminals. */
-
- for (bp = firstsymbol; bp; bp = bp->next)
- if (bp->class == unknown_sym)
- {
- complain (_
- ("symbol %s is used, but is not defined as a token and has no rules"),
- bp->tag);
- bp->class = nterm_sym;
- bp->value = nvars++;
- }
+ symbols_do (symbol_check_defined, NULL);
/* Insert the initial rule, which line is that of the first rule
(not that of the start symbol):
p->next->next->next = symbol_list_new (NULL);
p->next->next->next->next = grammar;
nrules += 1;
- nitems += 3;
+ nritems += 3;
grammar = p;
startval = axiom;
- if (nsyms > MAXSHORT)
+ if (nsyms > SHRT_MAX)
fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
- MAXSHORT);
+ SHRT_MAX);
- ntokens = nsyms - nvars;
+ assert (nsyms == ntokens + nvars);
}
/* At the end of the grammar file, some C source code must
static void
read_additionnal_code (void)
{
- char c;
+ int c;
struct obstack el_obstack;
obstack_init (&el_obstack);
static void
token_translations_init (void)
{
- bucket *bp = NULL;
+ int num_256_available_p = TRUE;
int i;
- token_translations = XCALLOC (short, max_user_token_number + 1);
+ /* 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 != SUNDEF)
+ {
+ 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;
+ }
+ }
- /* 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; i++)
- token_translations[i] = 2;
+ /* If 256 is not used, assign it to error, to follow POSIX. */
+ if (num_256_available_p && errtoken->user_token_number == SUNDEF)
+ errtoken->user_token_number = 256;
+
+ /* Set the missing user numbers. */
+ if (max_user_token_number < 256)
+ max_user_token_number = 256;
- for (bp = firstsymbol; bp; bp = bp->next)
+ for (i = 0; i < ntokens; ++i)
{
- /* Non-terminal? */
- if (bp->value >= ntokens)
- continue;
- /* A token string alias? */
- if (bp->user_token_number == SALIAS)
- continue;
+ symbol_t *this = symbols[i];
+ if (this->user_token_number == SUNDEF)
+ 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;
+ }
- assert (bp->user_token_number != SUNDEF);
+ token_translations = XCALLOC (token_number_t, max_user_token_number + 1);
- /* A token which translation has already been set? */
- if (token_translations[bp->user_token_number] != 2)
- complain (_("tokens %s and %s both assigned number %d"),
- symbols[token_translations[bp->user_token_number]]->tag,
- bp->tag, bp->user_token_number);
- token_translations[bp->user_token_number] = bp->value;
- }
+ /* 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);
}
static void
packsymbols (void)
{
- bucket *bp = NULL;
- int tokno = 1;
- int last_user_token_number;
-
- symbols = XCALLOC (bucket *, nsyms);
+ symbols = XCALLOC (symbol_t *, nsyms);
- max_user_token_number = 256;
- last_user_token_number = 256;
-
- for (bp = firstsymbol; bp; bp = bp->next)
- {
- if (bp->class == nterm_sym)
- {
- bp->value += ntokens;
- }
- else if (bp->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 (bp->value == -1)
- {
- if (bp == eoftoken || bp->alias == eoftoken)
- bp->value = bp->alias->value = 0;
- else
- {
- bp->value = bp->alias->value = tokno++;
- }
- }
-
- if (bp->prec != bp->alias->prec)
- {
- if (bp->prec != 0 && bp->alias->prec != 0
- && bp->user_token_number == SALIAS)
- complain (_("conflicting precedences for %s and %s"),
- bp->tag, bp->alias->tag);
- if (bp->prec != 0)
- bp->alias->prec = bp->prec;
- else
- bp->prec = bp->alias->prec;
- }
-
- if (bp->assoc != bp->alias->assoc)
- {
- if (bp->assoc != 0 && bp->alias->assoc != 0
- && bp->user_token_number == SALIAS)
- complain (_("conflicting assoc values for %s and %s"),
- bp->tag, bp->alias->tag);
- if (bp->assoc != 0)
- bp->alias->assoc = bp->assoc;
- else
- bp->assoc = bp->alias->assoc;
- }
-
- /* Do not do processing below for SALIASs. */
- if (bp->user_token_number == SALIAS)
- continue;
-
- }
- else /* bp->class == token_sym */
- {
- if (bp == eoftoken)
- bp->value = 0;
- else
- bp->value = tokno++;
- }
-
- if (bp->class == token_sym)
- {
- if (bp->user_token_number == SUNDEF)
- bp->user_token_number = ++last_user_token_number;
- if (bp->user_token_number > max_user_token_number)
- max_user_token_number = bp->user_token_number;
- }
-
- symbols[bp->value] = bp;
- }
+ symbols_do (symbol_check_alias_consistence, NULL);
+ symbols_do (symbol_pack, NULL);
token_translations_init ();
- error_token_number = errtoken->value;
-
if (startval->class == unknown_sym)
fatal (_("the start symbol %s is undefined"), startval->tag);
else if (startval->class == token_sym)
fatal (_("the start symbol %s is a token"), startval->tag);
- start_symbol = startval->value;
+ start_symbol = startval->number;
}
/*---------------------------------------------------------------.
| Convert the rules into the representation using RRHS, RLHS and |
-| RITEMS. |
+| RITEM. |
`---------------------------------------------------------------*/
static void
int ruleno;
symbol_list *p;
- /* We use short to index items. */
- if (nitems >= MAXSHORT)
- fatal (_("too many items (max %d)"), MAXSHORT);
-
- ritem = XCALLOC (short, nitems + 1);
+ ritem = XCALLOC (item_number_t, nritems + 1);
rules = XCALLOC (rule_t, nrules) - 1;
itemno = 0;
p = grammar;
while (p)
{
- bucket *ruleprec = p->ruleprec;
- rules[ruleno].lhs = p->sym->value;
- rules[ruleno].rhs = itemno;
+ symbol_t *ruleprec = p->ruleprec;
+ rules[ruleno].user_number = ruleno;
+ rules[ruleno].number = ruleno;
+ rules[ruleno].lhs = p->sym;
+ rules[ruleno].rhs = ritem + itemno;
rules[ruleno].line = p->line;
rules[ruleno].useful = TRUE;
rules[ruleno].action = p->action;
p = p->next;
while (p && p->sym)
{
- ritem[itemno++] = p->sym->value;
+ /* item_number_t = token_number_t.
+ But the former needs to contain more: negative rule numbers. */
+ ritem[itemno++] = token_number_as_item_number (p->sym->number);
/* A rule gets by default the precedence and associativity
of the last token in it. */
if (p->sym->class == token_sym)
- {
- rules[ruleno].prec = p->sym->prec;
- rules[ruleno].assoc = p->sym->assoc;
- }
+ rules[ruleno].prec = p->sym;
if (p)
p = p->next;
}
the specified symbol's precedence replaces the default. */
if (ruleprec)
{
- rules[ruleno].prec = ruleprec->prec;
- rules[ruleno].assoc = ruleprec->assoc;
- rules[ruleno].precsym = ruleprec->value;
+ rules[ruleno].precsym = ruleprec;
+ rules[ruleno].prec = ruleprec;
}
-
ritem[itemno++] = -ruleno;
- ruleno++;
+ ++ruleno;
if (p)
p = p->next;
}
ritem[itemno] = 0;
- nritems = itemno;
- assert (nritems == nitems);
+ assert (itemno == nritems);
if (trace_flag)
ritem_print (stderr);
void
reader (void)
{
- start_flag = 0;
- startval = NULL; /* start symbol not specified yet. */
-
- nsyms = 0;
- nvars = 0;
- nrules = 0;
- nitems = 0;
-
- typed = 0;
- lastprec = 0;
-
- semantic_parser = 0;
- pure_parser = 0;
-
- grammar = NULL;
-
lex_init ();
lineno = 1;
obstack_init (&muscle_obstack);
/* Initialize the symbol table. */
- tabinit ();
+ symbols_new ();
/* Construct the axiom symbol. */
axiom = getsym ("$axiom");
axiom->class = nterm_sym;
- axiom->value = nvars++;
+ axiom->number = nvars++;
/* Construct the error token */
errtoken = getsym ("error");
errtoken->class = token_sym;
- errtoken->user_token_number = 256; /* Value specified by POSIX. */
+ errtoken->number = ntokens++;
/* Construct a token that represents all undefined literal tokens.
It is always token number 2. */
undeftoken = getsym ("$undefined.");
undeftoken->class = token_sym;
- undeftoken->user_token_number = 2;
+ undeftoken->number = ntokens++;
/* Initialize the obstacks. */
obstack_init (&action_obstack);
{
eoftoken = getsym ("$");
eoftoken->class = token_sym;
+ eoftoken->number = 0;
/* Value specified by POSIX. */
eoftoken->user_token_number = 0;
}
XFREE (ritem);
free (rules + 1);
/* Free the symbol table data structure. */
- free_symtab ();
+ symbols_free ();
}