-/* Input parser for bison
- Copyright 1984, 1986, 1989, 1992, 1998, 2000, 2001
+/* Input parser for Bison
+
+ Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002
Free Software Foundation, Inc.
This file is part of Bison, the GNU Compiler Compiler.
the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
Boston, MA 02111-1307, USA. */
-
#include "system.h"
-#include "quotearg.h"
-#include "quote.h"
-#include "getargs.h"
+
+#include <quotearg.h>
+
+#include "complain.h"
+#include "conflicts.h"
#include "files.h"
-#include "symtab.h"
-#include "options.h"
-#include "lex.h"
+#include "getargs.h"
#include "gram.h"
-#include "complain.h"
+#include "muscle_tab.h"
#include "output.h"
#include "reader.h"
-#include "conflicts.h"
-#include "muscle_tab.h"
+#include "symlist.h"
+#include "symtab.h"
-typedef struct symbol_list
-{
- struct symbol_list *next;
- bucket *sym;
- int line;
-
- /* The action is attached to the LHS of a rule. */
- const char *action;
- int action_line;
-
- /* The guard is attached to the LHS of a rule. */
- const char *guard;
- int guard_line;
- bucket *ruleprec;
-} symbol_list;
-
-int lineno;
-char **tags;
-short *user_toknums;
-static symbol_list *grammar;
-static int start_flag;
-static bucket *startval;
-
-/* Nonzero if components of semantic values are used, implying
- they must be unions. */
-static int value_components_used;
+static symbol_list *grammar = NULL;
+static int start_flag = 0;
+merger_list *merge_functions;
/* Nonzero if %union has been seen. */
-static int typed;
-
-/* Incremented for each %left, %right or %nonassoc seen */
-static int lastprec;
-
-static bucket *errtoken = NULL;
-static bucket *undeftoken = NULL;
-static bucket *eoftoken = NULL;
-static bucket *axiom = NULL;
-
-static symbol_list *
-symbol_list_new (bucket *sym)
-{
- symbol_list *res = XMALLOC (symbol_list, 1);
- res->next = NULL;
- res->sym = sym;
- res->line = lineno;
- res->action = NULL;
- res->action_line = 0;
- res->guard = NULL;
- res->guard_line = 0;
- res->ruleprec = NULL;
- return res;
-}
-
+int typed = 0;
\f
+/*-----------------------.
+| Set the start symbol. |
+`-----------------------*/
-/*===================\
-| Low level lexing. |
-\===================*/
-
-static void
-skip_to_char (int target)
+void
+grammar_start_symbol_set (symbol *s, location loc)
{
- int c;
- if (target == '\n')
- complain (_(" Skipping to next \\n"));
+ if (start_flag)
+ complain_at (loc, _("multiple %s declarations"), "%start");
else
- complain (_(" Skipping to next %c"), target);
-
- do
- c = skip_white_space ();
- while (c != target && c != EOF);
- if (c != EOF)
- ungetc (c, finput);
-}
-
-
-/*---------------------------------------------------------.
-| Read a signed integer from STREAM and return its value. |
-`---------------------------------------------------------*/
-
-static inline int
-read_signed_integer (FILE *stream)
-{
- int c = getc (stream);
- int sign = 1;
- int n = 0;
-
- if (c == '-')
- {
- c = getc (stream);
- sign = -1;
- }
-
- while (isdigit (c))
{
- n = 10 * n + (c - '0');
- c = getc (stream);
+ start_flag = 1;
+ startsymbol = s;
+ startsymbol_location = loc;
}
-
- ungetc (c, stream);
-
- return sign * n;
}
-\f
-/*--------------------------------------------------------------.
-| Get the data type (alternative in the union) of the value for |
-| symbol N in rule RULE. |
-`--------------------------------------------------------------*/
-static char *
-get_type_name (int n, symbol_list *rule)
-{
- int i;
- symbol_list *rp;
- if (n < 0)
- {
- complain (_("invalid $ value"));
- return NULL;
- }
-
- rp = rule;
- i = 0;
-
- while (i < n)
- {
- rp = rp->next;
- if (rp == NULL || rp->sym == NULL)
- {
- complain (_("invalid $ value"));
- return NULL;
- }
- i++;
- }
-
- return rp->sym->type_name;
-}
-\f
-/*------------------------------------------------------------.
-| Dump the string from FIN to OOUT if non null. MATCH is the |
-| delimiter of the string (either ' or "). |
-`------------------------------------------------------------*/
+/*----------------------------------------------------------------.
+| There are two prologues: one before %union, one after. Augment |
+| the current one. |
+`----------------------------------------------------------------*/
-static inline void
-copy_string2 (FILE *fin, struct obstack *oout, int match, int store)
+void
+prologue_augment (const char *prologue, location loc)
{
- int c;
-
- if (store)
- obstack_1grow (oout, match);
+ struct obstack *oout =
+ !typed ? &pre_prologue_obstack : &post_prologue_obstack;
- c = getc (fin);
-
- while (c != match)
- {
- if (c == EOF)
- fatal (_("unterminated string at end of file"));
- if (c == '\n')
- {
- complain (_("unterminated string"));
- ungetc (c, fin);
- c = match; /* invent terminator */
- continue;
- }
+ obstack_fgrow1 (oout, "]b4_syncline([[%d]], [[", loc.start.line);
+ MUSCLE_OBSTACK_SGROW (oout,
+ quotearg_style (c_quoting_style, loc.start.file));
+ obstack_sgrow (oout, "]])[\n");
+ obstack_sgrow (oout, prologue);
+}
- obstack_1grow (oout, c);
- if (c == '\\')
- {
- c = getc (fin);
- if (c == EOF)
- fatal (_("unterminated string at end of file"));
- obstack_1grow (oout, c);
- if (c == '\n')
- lineno++;
- }
- c = getc (fin);
- }
+/*----------------------.
+| Handle the epilogue. |
+`----------------------*/
- if (store)
- obstack_1grow (oout, c);
+void
+epilogue_augment (const char *epilogue, location loc)
+{
+ char *extension = NULL;
+ obstack_fgrow1 (&muscle_obstack, "]b4_syncline([[%d]], [[", loc.start.line);
+ MUSCLE_OBSTACK_SGROW (&muscle_obstack,
+ quotearg_style (c_quoting_style, loc.start.file));
+ obstack_sgrow (&muscle_obstack, "]])[\n");
+ obstack_sgrow (&muscle_obstack, epilogue);
+ obstack_1grow (&muscle_obstack, 0);
+ extension = obstack_finish (&muscle_obstack);
+ muscle_grow ("epilogue", extension, "");
+ obstack_free (&muscle_obstack, extension);
}
-/* FIXME. */
-static inline void
-copy_string (FILE *fin, struct obstack *oout, int match)
-{
- copy_string2 (fin, oout, match, 1);
-}
+\f
-/* FIXME. */
+/*-------------------------------------------------------------------.
+| Return the merger index for a merging function named NAME, whose |
+| arguments have type TYPE. Records the function, if new, in |
+| MERGER_LIST. |
+`-------------------------------------------------------------------*/
-static inline void
-copy_identifier (FILE *fin, struct obstack *oout)
+static int
+get_merge_function (uniqstr name, uniqstr type, location loc)
{
- int c;
-
- while (isalnum (c = getc (fin)) || c == '_')
- obstack_1grow (oout, c);
-
- ungetc (c, fin);
-}
-
-
-/*------------------------------------------------------------------.
-| Dump the wannabee comment from IN to OOUT. In fact we just saw a |
-| `/', which might or might not be a comment. In any case, copy |
-| what we saw. |
-`------------------------------------------------------------------*/
+ merger_list *syms;
+ merger_list head;
+ int n;
-static inline void
-copy_comment (FILE *fin, struct obstack *oout)
-{
- int cplus_comment;
- int ended;
- int c;
+ if (! glr_parser)
+ return 0;
- /* We read a `/', output it. */
- obstack_1grow (oout, '/');
+ if (type == NULL)
+ type = uniqstr_new ("");
- switch ((c = getc (fin)))
- {
- case '/':
- cplus_comment = 1;
- break;
- case '*':
- cplus_comment = 0;
+ head.next = merge_functions;
+ for (syms = &head, n = 1; syms->next != NULL; syms = syms->next, n += 1)
+ if (UNIQSTR_EQ (name, syms->next->name))
break;
- default:
- ungetc (c, fin);
- return;
- }
-
- obstack_1grow (oout, c);
- c = getc (fin);
-
- ended = 0;
- while (!ended)
+ if (syms->next == NULL)
{
- if (!cplus_comment && c == '*')
- {
- while (c == '*')
- {
- obstack_1grow (oout, c);
- c = getc (fin);
- }
-
- if (c == '/')
- {
- obstack_1grow (oout, c);
- ended = 1;
- }
- }
- else if (c == '\n')
- {
- lineno++;
- obstack_1grow (oout, c);
- if (cplus_comment)
- ended = 1;
- else
- c = getc (fin);
- }
- else if (c == EOF)
- fatal (_("unterminated comment"));
- else
- {
- obstack_1grow (oout, c);
- c = getc (fin);
- }
+ MALLOC (syms->next, 1);
+ syms->next->name = uniqstr_new (name);
+ syms->next->type = uniqstr_new (type);
+ syms->next->next = NULL;
+ merge_functions = head.next;
}
+ else if (!UNIQSTR_EQ (type, syms->next->type))
+ warn_at (loc, _("result type clash on merge function %s: <%s> != <%s>"),
+ name, type, syms->next->type);
+ return n;
}
+/*--------------------------------------.
+| Free all merge-function definitions. |
+`--------------------------------------*/
-/*-----------------------------------------------------------------.
-| FIN is pointing to a location (i.e., a `@'). Output to OOUT a |
-| reference to this location. STACK_OFFSET is the number of values |
-| in the current rule so far, which says where to find `$0' with |
-| respect to the top of the stack. |
-`-----------------------------------------------------------------*/
-
-static inline void
-copy_at (FILE *fin, struct obstack *oout, int stack_offset)
+void
+free_merger_functions (void)
{
- int c;
-
- c = getc (fin);
- if (c == '$')
- {
- obstack_sgrow (oout, "yyloc");
- locations_flag = 1;
- }
- else if (isdigit (c) || c == '-')
+ merger_list *L0;
+ if (! glr_parser)
+ return;
+ L0 = merge_functions;
+ while (L0 != NULL)
{
- int n;
-
- ungetc (c, fin);
- n = read_signed_integer (fin);
- if (n > stack_offset)
- complain (_("invalid value: %s%d"), "@", n);
- else
- {
- /* Offset is always 0 if parser has already popped the stack
- pointer. */
- obstack_fgrow1 (oout, "yylsp[%d]",
- n - (semantic_parser ? 0 : stack_offset));
- locations_flag = 1;
- }
- }
- else
- {
- char buf[] = "@c";
- buf[1] = c;
- complain (_("%s is invalid"), quote (buf));
+ merger_list *L1 = L0->next;
+ free (L0);
+ L0 = L1;
}
}
-
+\f
/*-------------------------------------------------------------------.
-| FIN is pointing to a wannabee semantic value (i.e., a `$'). |
+| Parse the input grammar into a one symbol_list structure. Each |
+| rule is represented by a sequence of symbols: the left hand side |
+| followed by the contents of the right hand side, followed by a |
+| null pointer instead of a symbol to terminate the rule. The next |
+| symbol is the lhs of the following rule. |
| |
-| Possible inputs: $[<TYPENAME>]($|integer) |
+| All actions are copied out, labelled by the rule number they apply |
+| to. |
| |
-| Output to OOUT a reference to this semantic value. STACK_OFFSET is |
-| the number of values in the current rule so far, which says where |
-| to find `$0' with respect to the top of the stack. |
+| Bison used to allow some %directives in the rules sections, but |
+| this is no longer consider appropriate: (i) the documented grammar |
+| doesn't claim it, (ii), it would promote bad style, (iii), error |
+| recovery for %directives consists in skipping the junk until a `%' |
+| is seen and helrp synchronizing. This scheme is definitely wrong |
+| in the rules section. |
`-------------------------------------------------------------------*/
-static inline void
-copy_dollar (FILE *fin, struct obstack *oout,
- symbol_list *rule, int stack_offset)
-{
- int c = getc (fin);
- const char *type_name = NULL;
-
- /* Get the type name if explicit. */
- if (c == '<')
- {
- read_type_name (fin);
- type_name = token_buffer;
- value_components_used = 1;
- c = getc (fin);
- }
-
- if (c == '$')
- {
- obstack_sgrow (oout, "yyval");
-
- if (!type_name)
- type_name = get_type_name (0, rule);
- if (type_name)
- obstack_fgrow1 (oout, ".%s", type_name);
- if (!type_name && typed)
- complain (_("$$ of `%s' has no declared type"),
- rule->sym->tag);
- }
- else if (isdigit (c) || c == '-')
- {
- int n;
- ungetc (c, fin);
- n = read_signed_integer (fin);
-
- if (n > stack_offset)
- complain (_("invalid value: %s%d"), "$", n);
- else
- {
- if (!type_name && n > 0)
- type_name = get_type_name (n, rule);
-
- /* Offset is always 0 if parser has already popped the stack
- pointer. */
- obstack_fgrow1 (oout, "yyvsp[%d]",
- n - (semantic_parser ? 0 : stack_offset));
-
- if (type_name)
- obstack_fgrow1 (oout, ".%s", type_name);
- if (!type_name && typed)
- complain (_("$%d of `%s' has no declared type"),
- n, rule->sym->tag);
- }
- }
- else
- {
- char buf[] = "$c";
- buf[1] = c;
- complain (_("%s is invalid"), quote (buf));
- }
-}
-\f
-/*-------------------------------------------------------------------.
-| Copy the contents of a `%{ ... %}' into the definitions file. The |
-| `%{' has already been read. Return after reading the `%}'. |
-`-------------------------------------------------------------------*/
+/* The (currently) last symbol of GRAMMAR. */
+symbol_list *grammar_end = NULL;
-static void
-copy_definition (void)
+/* Append SYM to the grammar. */
+void
+grammar_symbol_append (symbol *sym, location loc)
{
- int c;
- /* -1 while reading a character if prev char was %. */
- int after_percent;
-
- if (!no_lines_flag)
- {
- obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"),
- lineno, quotearg_style (c_quoting_style,
- muscle_find ("filename")));
- }
-
- after_percent = 0;
+ symbol_list *p = symbol_list_new (sym, loc);
- c = getc (finput);
-
- for (;;)
- {
- switch (c)
- {
- case '\n':
- obstack_1grow (&attrs_obstack, c);
- lineno++;
- break;
-
- case '%':
- after_percent = -1;
- break;
-
- case '\'':
- case '"':
- copy_string (finput, &attrs_obstack, c);
- break;
-
- case '/':
- copy_comment (finput, &attrs_obstack);
- break;
-
- case EOF:
- fatal ("%s", _("unterminated `%{' definition"));
-
- default:
- obstack_1grow (&attrs_obstack, c);
- }
-
- c = getc (finput);
+ if (grammar_end)
+ grammar_end->next = p;
+ else
+ grammar = p;
- if (after_percent)
- {
- if (c == '}')
- return;
- obstack_1grow (&attrs_obstack, '%');
- }
- after_percent = 0;
- }
+ grammar_end = p;
}
-
-/*-------------------------------------------------------------------.
-| Parse what comes after %token or %nterm. For %token, WHAT_IS is |
-| token_sym and WHAT_IS_NOT is nterm_sym. For %nterm, the arguments |
-| are reversed. |
-`-------------------------------------------------------------------*/
-
-static void
-parse_token_decl (symbol_class what_is, symbol_class what_is_not)
-{
- token_t token = tok_undef;
- char *typename = NULL;
-
- /* The symbol being defined. */
- struct bucket *symbol = NULL;
-
- /* After `%token' and `%nterm', any number of symbols maybe be
- defined. */
- for (;;)
- {
- int tmp_char = ungetc (skip_white_space (), finput);
-
- /* `%' (for instance from `%token', or from `%%' etc.) is the
- only valid means to end this declaration. */
- if (tmp_char == '%')
- return;
- if (tmp_char == EOF)
- fatal (_("Premature EOF after %s"), token_buffer);
-
- token = lex ();
- if (token == tok_comma)
- {
- symbol = NULL;
- continue;
- }
- if (token == tok_typename)
- {
- typename = xstrdup (token_buffer);
- value_components_used = 1;
- symbol = NULL;
- }
- 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 = NULL;
- }
- else if (token == tok_identifier)
- {
- int oldclass = symval->class;
- symbol = symval;
-
- if (symbol->class == what_is_not)
- complain (_("symbol %s redefined"), symbol->tag);
- symbol->class = what_is;
- if (what_is == nterm_sym && oldclass != nterm_sym)
- symbol->value = nvars++;
-
- if (typename)
- {
- if (symbol->type_name == NULL)
- symbol->type_name = typename;
- else if (strcmp (typename, symbol->type_name) != 0)
- complain (_("type redeclaration for %s"), symbol->tag);
- }
- }
- else if (symbol && token == tok_number)
- {
- symbol->user_token_number = numval;
- /* User defined EOF token? */
- if (numval == 0)
- eoftoken = symbol;
- }
- else
- {
- complain (_("`%s' is invalid in %s"),
- token_buffer,
- (what_is == token_sym) ? "%token" : "%nterm");
- skip_to_char ('%');
- }
- }
-
-}
+/* The rule currently being defined, and the previous rule.
+ CURRENT_RULE points to the first LHS of the current rule, while
+ PREVIOUS_RULE_END points to the *end* of the previous rule (NULL). */
+symbol_list *current_rule = NULL;
+symbol_list *previous_rule_end = NULL;
-/*------------------------------.
-| Parse what comes after %start |
-`------------------------------*/
+/*----------------------------------------------.
+| Create a new rule for LHS in to the GRAMMAR. |
+`----------------------------------------------*/
-static void
-parse_start_decl (void)
+void
+grammar_rule_begin (symbol *lhs, location loc)
{
- if (start_flag)
- complain (_("multiple %s declarations"), "%start");
- if (lex () != tok_identifier)
- complain (_("invalid %s declaration"), "%start");
- else
+ if (!start_flag)
{
+ startsymbol = lhs;
+ startsymbol_location = loc;
start_flag = 1;
- startval = symval;
}
-}
-/*-----------------------------------------------------------.
-| read in a %type declaration and record its information for |
-| get_type_name to access |
-`-----------------------------------------------------------*/
-
-static void
-parse_type_decl (void)
-{
- char *name;
+ /* Start a new rule and record its lhs. */
+ ++nrules;
+ ++nritems;
- if (lex () != tok_typename)
- {
- complain ("%s", _("%type declaration has no <typename>"));
- skip_to_char ('%');
- return;
- }
+ previous_rule_end = grammar_end;
+ grammar_symbol_append (lhs, loc);
+ current_rule = grammar_end;
- name = xstrdup (token_buffer);
+ /* Mark the rule's lhs as a nonterminal if not already so. */
- for (;;)
+ if (lhs->class == unknown_sym)
{
- token_t t;
- int tmp_char = ungetc (skip_white_space (), finput);
-
- if (tmp_char == '%')
- return;
- if (tmp_char == EOF)
- fatal (_("Premature EOF after %s"), token_buffer);
-
- t = lex ();
-
- switch (t)
- {
-
- case tok_comma:
- case tok_semicolon:
- break;
-
- case tok_identifier:
- if (symval->type_name == NULL)
- symval->type_name = name;
- else if (strcmp (name, symval->type_name) != 0)
- complain (_("type redeclaration for %s"), symval->tag);
-
- break;
-
- default:
- complain (_("invalid %%type declaration due to item: %s"),
- token_buffer);
- skip_to_char ('%');
- }
+ lhs->class = nterm_sym;
+ lhs->number = nvars;
+ ++nvars;
}
+ else if (lhs->class == token_sym)
+ complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
}
-
-
-/*----------------------------------------------------------------.
-| Read in a %left, %right or %nonassoc declaration and record its |
-| information. |
-`----------------------------------------------------------------*/
+/* Check that the last rule (CURRENT_RULE) is properly defined. For
+ instance, there should be no type clash on the default action. */
static void
-parse_assoc_decl (associativity assoc)
+grammar_current_rule_check (void)
{
- char *name = NULL;
- int prev = 0;
-
- lastprec++; /* Assign a new precedence level, never 0. */
-
- for (;;)
- {
- token_t t;
- int tmp_char = ungetc (skip_white_space (), finput);
+ symbol *lhs = current_rule->sym;
+ char const *lhs_type = lhs->type_name;
+ symbol *first_rhs = current_rule->next->sym;
- if (tmp_char == '%')
- return;
- if (tmp_char == EOF)
- fatal (_("Premature EOF after %s"), token_buffer);
+ /* If there is an action, then there is nothing we can do: the user
+ is allowed to shoot herself in the foot. */
+ if (current_rule->action)
+ return;
- t = lex ();
-
- switch (t)
- {
- case tok_typename:
- name = xstrdup (token_buffer);
- break;
-
- case tok_comma:
- break;
-
- case tok_identifier:
- if (symval->prec != 0)
- complain (_("redefining precedence of %s"), symval->tag);
- symval->prec = lastprec;
- symval->assoc = assoc;
- if (symval->class == nterm_sym)
- complain (_("symbol %s redefined"), symval->tag);
- symval->class = token_sym;
- if (name)
- { /* record the type, if one is specified */
- if (symval->type_name == NULL)
- symval->type_name = name;
- else if (strcmp (name, symval->type_name) != 0)
- complain (_("type redeclaration for %s"), symval->tag);
- }
- break;
-
- case tok_number:
- if (prev == tok_identifier)
- {
- symval->user_token_number = numval;
- }
- else
- {
- complain (_
- ("invalid text (%s) - number should be after identifier"),
-token_buffer);
- skip_to_char ('%');
- }
- break;
-
- case tok_semicolon:
- return;
-
- default:
- complain (_("unexpected item: %s"), token_buffer);
- skip_to_char ('%');
- }
+ /* Don't worry about the default action if $$ is untyped, since $$'s
+ value can't be used. */
+ if (! lhs_type)
+ return;
- prev = t;
- }
-}
-
-
-
-/*--------------------------------------------------------------.
-| Copy the union declaration into the stype muscle |
-| (and fdefines), where it is made into the definition of |
-| YYSTYPE, the type of elements of the parser value stack. |
-`--------------------------------------------------------------*/
-
-static void
-parse_union_decl (void)
-{
- int c;
- int count = 0;
- bool done = FALSE;
- struct obstack union_obstack;
- if (typed)
- complain (_("multiple %s declarations"), "%union");
-
- typed = 1;
-
- obstack_init (&union_obstack);
- obstack_sgrow (&union_obstack, "union");
-
- while (!done)
+ /* If $$ is being set in default way, report if any type mismatch. */
+ if (first_rhs)
{
- c = xgetc (finput);
-
- /* If C contains '/', it is output by copy_comment (). */
- if (c != '/')
- obstack_1grow (&union_obstack, c);
-
- switch (c)
- {
- case '\n':
- lineno++;
- break;
-
- case '/':
- copy_comment (finput, &union_obstack);
- break;
-
- case '{':
- count++;
- break;
-
- case '}':
- /* FIXME: Errr. How could this happen???. --akim */
- if (count == 0)
- complain (_("unmatched %s"), "`}'");
- count--;
- if (!count)
- done = TRUE;
- break;
- }
+ const char *rhs_type = first_rhs->type_name ? first_rhs->type_name : "";
+ if (!UNIQSTR_EQ (lhs_type, rhs_type))
+ warn_at (current_rule->location,
+ _("type clash on default action: <%s> != <%s>"),
+ lhs_type, rhs_type);
}
-
- /* JF don't choke on trailing semi */
- c = skip_white_space ();
- if (c != ';')
- ungetc (c, finput);
- obstack_1grow (&union_obstack, 0);
- muscle_insert ("stype", obstack_finish (&union_obstack));
-}
-
-
-/*-------------------------------------------------------.
-| Parse the declaration %expect N which says to expect N |
-| shift-reduce conflicts. |
-`-------------------------------------------------------*/
-
-static void
-parse_expect_decl (void)
-{
- int c = skip_white_space ();
- ungetc (c, finput);
-
- if (!isdigit (c))
- complain (_("argument of %%expect is not an integer"));
+ /* Warn if there is no default for $$ but we need one. */
else
- expected_conflicts = read_signed_integer (finput);
-}
-
-
-/*-------------------------------------------------------------------.
-| Parse what comes after %thong. the full syntax is |
-| |
-| %thong <type> token number literal |
-| |
-| the <type> or number may be omitted. The number specifies the |
-| user_token_number. |
-| |
-| Two symbols are entered in the table, one for the token symbol and |
-| one for the literal. Both are given the <type>, if any, from the |
-| declaration. The ->user_token_number of the first is SALIAS and |
-| the ->user_token_number of the second is set to the number, if |
-| any, from the declaration. The two symbols are linked via |
-| pointers in their ->alias fields. |
-| |
-| During OUTPUT_DEFINES_TABLE, the symbol is reported thereafter, |
-| only the literal string is retained it is the literal string that |
-| is output to yytname |
-`-------------------------------------------------------------------*/
-
-static void
-parse_thong_decl (void)
-{
- token_t token;
- struct bucket *symbol;
- char *typename = 0;
- int usrtoknum = SUNDEF;
-
- token = lex (); /* fetch typename or first token */
- if (token == tok_typename)
- {
- typename = xstrdup (token_buffer);
- value_components_used = 1;
- token = lex (); /* fetch first token */
- }
-
- /* process first token */
-
- if (token != tok_identifier)
- {
- complain (_("unrecognized item %s, expected an identifier"),
- token_buffer);
- skip_to_char ('%');
- return;
- }
- symval->class = token_sym;
- symval->type_name = typename;
- symval->user_token_number = SALIAS;
- symbol = symval;
-
- token = lex (); /* get number or literal string */
-
- if (token == tok_number)
- {
- usrtoknum = numval;
- token = lex (); /* okay, did number, now get literal */
- }
-
- /* process literal string token */
-
- if (token != tok_identifier || *symval->tag != '\"')
- {
- complain (_("expected string constant instead of %s"), token_buffer);
- skip_to_char ('%');
- return;
- }
- symval->class = token_sym;
- symval->type_name = typename;
- symval->user_token_number = usrtoknum;
-
- symval->alias = symbol;
- symbol->alias = symval;
-
- /* symbol and symval combined are only one symbol. */
- nsyms--;
-}
-
-
-static void
-parse_muscle_decl (void)
-{
- int ch = ungetc (skip_white_space (), finput);
- char *muscle_key;
- char *muscle_value;
-
- /* Read key. */
- if (!isalpha (ch) && ch != '_')
- {
- complain (_("invalid %s declaration"), "%define");
- skip_to_char ('%');
- return;
- }
- copy_identifier (finput, &muscle_obstack);
- obstack_1grow (&muscle_obstack, 0);
- muscle_key = obstack_finish (&muscle_obstack);
-
- /* Read value. */
- ch = skip_white_space ();
- if (ch != '"')
- {
- ungetc (ch, finput);
- if (ch != EOF)
- {
- complain (_("invalid %s declaration"), "%define");
- skip_to_char ('%');
- return;
- }
- else
- fatal (_("Premature EOF after %s"), "\"");
- }
- copy_string2 (finput, &muscle_obstack, '"', 0);
- obstack_1grow (&muscle_obstack, 0);
- muscle_value = obstack_finish (&muscle_obstack);
-
- /* Store the (key, value) pair in the environment. */
- muscle_insert (muscle_key, muscle_value);
+ warn_at (current_rule->location,
+ _("empty rule for typed nonterminal, and no action"));
}
+/*-------------------------------------.
+| End the currently being grown rule. |
+`-------------------------------------*/
-/*---------------------------------.
-| Parse a double quoted parameter. |
-`---------------------------------*/
-
-static const char *
-parse_dquoted_param (const char *from)
-{
- struct obstack param_obstack;
- const char *param = NULL;
- int c;
-
- obstack_init (¶m_obstack);
- c = skip_white_space ();
-
- if (c != '"')
- {
- complain (_("invalid %s declaration"), from);
- ungetc (c, finput);
- skip_to_char ('%');
- return NULL;
- }
-
- while ((c = literalchar ()) != '"')
- obstack_1grow (¶m_obstack, c);
-
- obstack_1grow (¶m_obstack, '\0');
- param = obstack_finish (¶m_obstack);
-
- if (c != '"' || strlen (param) == 0)
- {
- complain (_("invalid %s declaration"), from);
- if (c != '"')
- ungetc (c, finput);
- skip_to_char ('%');
- return NULL;
- }
-
- return param;
-}
-
-/*----------------------------------.
-| Parse what comes after %skeleton. |
-`----------------------------------*/
-
-static void
-parse_skel_decl (void)
+void
+grammar_rule_end (location loc)
{
- skeleton = parse_dquoted_param ("%skeleton");
+ /* Put an empty link in the list to mark the end of this rule */
+ grammar_symbol_append (NULL, grammar_end->location);
+ current_rule->location = loc;
+ grammar_current_rule_check ();
}
-/*----------------------------------------------------------------.
-| Read from finput until `%%' is seen. Discard the `%%'. Handle |
-| any `%' declarations, and copy the contents of any `%{ ... %}' |
-| groups to ATTRS_OBSTACK. |
-`----------------------------------------------------------------*/
-
-static void
-read_declarations (void)
-{
- for (;;)
- {
- int c = skip_white_space ();
- if (c == '%')
- {
- token_t tok = parse_percent_token ();
-
- switch (tok)
- {
- case tok_two_percents:
- return;
-
- case tok_percent_left_curly:
- copy_definition ();
- break;
-
- case tok_token:
- parse_token_decl (token_sym, nterm_sym);
- break;
-
- case tok_nterm:
- parse_token_decl (nterm_sym, token_sym);
- break;
-
- case tok_type:
- parse_type_decl ();
- break;
-
- case tok_start:
- parse_start_decl ();
- break;
-
- case tok_union:
- parse_union_decl ();
- break;
-
- case tok_expect:
- parse_expect_decl ();
- break;
-
- case tok_thong:
- parse_thong_decl ();
- break;
-
- case tok_left:
- parse_assoc_decl (left_assoc);
- break;
-
- case tok_right:
- parse_assoc_decl (right_assoc);
- break;
-
- case tok_nonassoc:
- parse_assoc_decl (non_assoc);
- break;
-
- case tok_define:
- parse_muscle_decl ();
- break;
-
- case tok_skel:
- parse_skel_decl ();
- break;
-
- case tok_noop:
- break;
-
- case tok_stropt:
- case tok_intopt:
- case tok_obsolete:
- abort ();
- break;
-
- case tok_illegal:
- default:
- complain (_("unrecognized: %s"), token_buffer);
- skip_to_char ('%');
- }
- }
- else if (c == EOF)
- fatal (_("no input grammar"));
- else
- {
- char buf[] = "c";
- buf[0] = c;
- complain (_("unknown character: %s"), quote (buf));
- skip_to_char ('%');
- }
- }
-}
-\f
/*-------------------------------------------------------------------.
-| Assuming that a `{' has just been seen, copy everything up to the |
-| matching `}' into the actions file. STACK_OFFSET is the number of |
-| values in the current rule so far, which says where to find `$0' |
-| with respect to the top of the stack. |
-| |
-| This routine is used both for actions and guards. Only |
-| ACTION_OBSTACK is used, but this is fine, since we use only |
-| pointers to relevant portions inside this obstack. |
+| The previous action turns out the be a mid-rule action. Attach it |
+| to the current rule, i.e., create a dummy symbol, attach it this |
+| mid-rule action, and append this dummy nonterminal to the current |
+| rule. |
`-------------------------------------------------------------------*/
-static void
-parse_braces (symbol_list *rule, int stack_offset)
-{
- int c;
- int count;
-
- count = 1;
- while (count > 0)
- {
- while ((c = getc (finput)) != '}')
- switch (c)
- {
- case '\n':
- obstack_1grow (&action_obstack, c);
- lineno++;
- break;
-
- case '{':
- obstack_1grow (&action_obstack, c);
- count++;
- break;
-
- case '\'':
- case '"':
- copy_string (finput, &action_obstack, c);
- break;
-
- case '/':
- copy_comment (finput, &action_obstack);
- break;
-
- case '$':
- copy_dollar (finput, &action_obstack,
- rule, stack_offset);
- break;
-
- case '@':
- copy_at (finput, &action_obstack,
- stack_offset);
- break;
-
- case EOF:
- fatal (_("unmatched %s"), "`{'");
-
- default:
- obstack_1grow (&action_obstack, c);
- }
-
- /* Above loop exits when C is '}'. */
- if (--count)
- {
- obstack_1grow (&action_obstack, c);
- c = getc (finput);
- }
- }
-
- obstack_1grow (&action_obstack, '\0');
-}
-
+void
+grammar_midrule_action (void)
+{
+ /* Since the action was written out with this rule's number, we must
+ give the new rule this number by inserting the new rule before
+ it. */
+
+ /* Make a DUMMY nonterminal, whose location is that of the midrule
+ action. Create the MIDRULE. */
+ location dummy_location = current_rule->action_location;
+ symbol *dummy = dummy_symbol_get (dummy_location);
+ symbol_list *midrule = symbol_list_new (dummy, dummy_location);
+
+ /* Make a new rule, whose body is empty, before the current one, so
+ that the action just read can belong to it. */
+ ++nrules;
+ ++nritems;
+ /* Attach its location and actions to that of the DUMMY. */
+ midrule->location = dummy_location;
+ midrule->action = current_rule->action;
+ midrule->action_location = dummy_location;
+ current_rule->action = NULL;
+
+ if (previous_rule_end)
+ previous_rule_end->next = midrule;
+ else
+ grammar = midrule;
-static void
-parse_action (symbol_list *rule, int stack_offset)
-{
- rule->action_line = lineno;
- parse_braces (rule, stack_offset);
- rule->action = obstack_finish (&action_obstack);
-}
+ /* End the dummy's rule. */
+ previous_rule_end = symbol_list_new (NULL, dummy_location);
+ previous_rule_end->next = current_rule;
+ midrule->next = previous_rule_end;
-static void
-parse_guard (symbol_list *rule, int stack_offset)
-{
- token_t t = lex ();
- if (t != tok_left_curly)
- complain (_("invalid %s declaration"), "%guard");
- rule->guard_line = lineno;
- parse_braces (rule, stack_offset);
- rule->guard = obstack_finish (&action_obstack);
+ /* Insert the dummy nonterminal replacing the midrule action into
+ the current rule. */
+ grammar_current_rule_symbol_append (dummy, dummy_location);
}
-\f
+/* Set the precedence symbol of the current rule to PRECSYM. */
-/*-------------------------------------------------------------------.
-| Generate a dummy symbol, a nonterminal, whose name cannot conflict |
-| with the user's names. |
-`-------------------------------------------------------------------*/
-
-static bucket *
-gensym (void)
+void
+grammar_current_rule_prec_set (symbol *precsym, location loc)
{
- /* Incremented for each generated symbol */
- static int gensym_count = 0;
- static char buf[256];
-
- bucket *sym;
-
- sprintf (buf, "@%d", ++gensym_count);
- token_buffer = buf;
- sym = getsym (token_buffer);
- sym->class = nterm_sym;
- sym->value = nvars++;
- return sym;
+ if (current_rule->ruleprec)
+ complain_at (loc, _("only one %s allowed per rule"), "%prec");
+ current_rule->ruleprec = precsym;
}
-\f
-/*-------------------------------------------------------------------.
-| Parse the input grammar into a one symbol_list structure. Each |
-| rule is represented by a sequence of symbols: the left hand side |
-| followed by the contents of the right hand side, followed by a |
-| null pointer instead of a symbol to terminate the rule. The next |
-| symbol is the lhs of the following rule. |
-| |
-| All guards and actions are copied out to the appropriate files, |
-| labelled by the rule number they apply to. |
-| |
-| Bison used to allow some %directives in the rules sections, but |
-| this is no longer consider appropriate: (i) the documented grammar |
-| doesn't claim it, (ii), it would promote bad style, (iii), error |
-| recovery for %directives consists in skipping the junk until a `%' |
-| is seen and helrp synchronizing. This scheme is definitely wrong |
-| in the rules section. |
-`-------------------------------------------------------------------*/
-static void
-readgram (void)
-{
- token_t t;
- bucket *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. */
- symbol_list *crule = NULL;
- /* Points to the symbol_list preceding crule. */
- symbol_list *crule1 = NULL;
-
- t = lex ();
-
- while (t != tok_two_percents && t != tok_eof)
- if (t == tok_identifier || t == tok_bar)
- {
- int action_flag = 0;
- /* Number of symbols in rhs of this rule so far */
- int rulelength = 0;
- int xactions = 0; /* JF for error checking */
- bucket *first_rhs = 0;
-
- if (t == tok_identifier)
- {
- lhs = symval;
-
- if (!start_flag)
- {
- startval = lhs;
- start_flag = 1;
- }
-
- t = lex ();
- if (t != tok_colon)
- {
- complain (_("ill-formed rule: initial symbol not followed by colon"));
- unlex (t);
- }
- }
-
- if (nrules == 0 && t == tok_bar)
- {
- complain (_("grammar starts with vertical bar"));
- lhs = symval; /* BOGUS: use a random symval */
- }
- /* start a new rule and record its lhs. */
-
- nrules++;
- nitems++;
-
- p = symbol_list_new (lhs);
-
- crule1 = p1;
- if (p1)
- p1->next = p;
- else
- grammar = p;
-
- p1 = p;
- crule = p;
-
- /* mark the rule's lhs as a nonterminal if not already so. */
-
- if (lhs->class == unknown_sym)
- {
- lhs->class = nterm_sym;
- lhs->value = nvars;
- nvars++;
- }
- else if (lhs->class == token_sym)
- complain (_("rule given for %s, which is a token"), lhs->tag);
-
- /* read the rhs of the rule. */
-
- for (;;)
- {
- t = lex ();
- if (t == tok_prec)
- {
- t = lex ();
- crule->ruleprec = symval;
- t = lex ();
- }
-
- if (!(t == tok_identifier || t == tok_left_curly))
- break;
-
- /* If next token is an identifier, see if a colon follows it.
- If one does, exit this rule now. */
- if (t == tok_identifier)
- {
- bucket *ssave;
- token_t t1;
-
- ssave = symval;
- t1 = lex ();
- unlex (t1);
- symval = ssave;
- if (t1 == tok_colon)
- break;
-
- if (!first_rhs) /* JF */
- first_rhs = symval;
- /* Not followed by colon =>
- process as part of this rule's rhs. */
- }
-
- /* If we just passed an action, that action was in the middle
- of a rule, so make a dummy rule to reduce it to a
- non-terminal. */
- if (action_flag)
- {
- /* Since the action was written out with this rule's
- number, we must give the new rule this number by
- inserting the new rule before it. */
-
- /* Make a dummy nonterminal, a gensym. */
- bucket *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++;
- p = symbol_list_new (sdummy);
- /* Attach its lineno to that of the host rule. */
- p->line = crule->line;
- if (crule1)
- crule1->next = p;
- else
- grammar = p;
- /* End of the rule. */
- crule1 = symbol_list_new (NULL);
- crule1->next = crule;
-
- p->next = crule1;
-
- /* Insert the dummy generated by that rule into this
- rule. */
- nitems++;
- p = symbol_list_new (sdummy);
- p1->next = p;
- p1 = p;
-
- action_flag = 0;
- }
-
- if (t == tok_identifier)
- {
- nitems++;
- p = symbol_list_new (symval);
- p1->next = p;
- p1 = p;
- }
- else /* handle an action. */
- {
- parse_action (crule, rulelength);
- action_flag = 1;
- xactions++; /* JF */
- }
- rulelength++;
- } /* end of read rhs of rule */
-
- /* Put an empty link in the list to mark the end of this rule */
- p = symbol_list_new (NULL);
- p1->next = p;
- p1 = p;
-
- if (t == tok_prec)
- {
- complain (_("two @prec's in a row"));
- t = lex ();
- crule->ruleprec = symval;
- t = lex ();
- }
-
- if (t == tok_guard)
- {
- if (!semantic_parser)
- complain (_("%%guard present but %%semantic_parser not specified"));
-
- parse_guard (crule, rulelength);
- t = lex ();
- }
-
- if (t == tok_left_curly)
- {
- /* This case never occurs -wjh */
- if (action_flag)
- complain (_("two actions at end of one rule"));
- parse_action (crule, rulelength);
- action_flag = 1;
- xactions++; /* -wjh */
- t = lex ();
- }
- /* If $$ is being set in default way, report if any type
- mismatch. */
- else if (!xactions
- && first_rhs && lhs->type_name != first_rhs->type_name)
- {
- if (lhs->type_name == 0
- || first_rhs->type_name == 0
- || strcmp (lhs->type_name, first_rhs->type_name))
- complain (_("type clash (`%s' `%s') on default action"),
- lhs->type_name ? lhs->type_name : "",
- first_rhs->type_name ? first_rhs->type_name : "");
- }
- /* Warn if there is no default for $$ but we need one. */
- else if (!xactions && !first_rhs && lhs->type_name != 0)
- complain (_("empty rule for typed nonterminal, and no action"));
- if (t == tok_semicolon)
- t = lex ();
- }
- else
- {
- complain (_("invalid input: %s"), quote (token_buffer));
- t = lex ();
- }
-
- /* grammar has been read. Do some checking */
-
- if (nrules == 0)
- fatal (_("no rules in the input grammar"));
-
- /* Report any undefined symbols and consider them nonterminals. */
+/* Attach dynamic precedence DPREC to the current rule. */
- 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++;
- }
-
- /* Insert the initial rule, which line is that of the first rule
- (not that of the start symbol):
-
- axiom: %start EOF. */
- p = symbol_list_new (axiom);
- p->line = grammar->line;
- p->next = symbol_list_new (startval);
- p->next->next = symbol_list_new (eoftoken);
- p->next->next->next = symbol_list_new (NULL);
- p->next->next->next->next = grammar;
- nrules += 1;
- nitems += 3;
- grammar = p;
- startval = axiom;
-
- if (nsyms > MAXSHORT)
- fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
- MAXSHORT);
-
- ntokens = nsyms - nvars;
-}
-
-/* At the end of the grammar file, some C source code must
- be stored. It is going to be associated to the epilogue
- directive. */
-static void
-read_additionnal_code (void)
+void
+grammar_current_rule_dprec_set (int dprec, location loc)
{
- char c;
- struct obstack el_obstack;
-
- obstack_init (&el_obstack);
-
- if (!no_lines_flag)
- {
- obstack_fgrow2 (&el_obstack, muscle_find ("linef"),
- lineno, quotearg_style (c_quoting_style,
- muscle_find ("filename")));
- }
-
- while ((c = getc (finput)) != EOF)
- obstack_1grow (&el_obstack, c);
-
- obstack_1grow (&el_obstack, 0);
- muscle_insert ("epilogue", obstack_finish (&el_obstack));
+ if (! glr_parser)
+ warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
+ if (dprec <= 0)
+ complain_at (loc, _("%s must be followed by positive number"), "%dprec");
+ else if (current_rule->dprec != 0)
+ complain_at (loc, _("only one %s allowed per rule"), "%dprec");
+ current_rule->dprec = dprec;
}
-\f
-/*------------------------------------------------------------------.
-| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same |
-| number. |
-`------------------------------------------------------------------*/
+/* Attach a merge function NAME with argument type TYPE to current
+ rule. */
-static void
-token_translations_init (void)
+void
+grammar_current_rule_merge_set (uniqstr name, location loc)
{
- bucket *bp = NULL;
- int i;
-
- token_translations = XCALLOC (short, 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; i++)
- token_translations[i] = 2;
-
- for (bp = firstsymbol; bp; bp = bp->next)
- {
- /* Non-terminal? */
- if (bp->value >= ntokens)
- continue;
- /* A token string alias? */
- if (bp->user_token_number == SALIAS)
- continue;
-
- assert (bp->user_token_number != SUNDEF);
-
- /* 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"),
- tags[token_translations[bp->user_token_number]],
- bp->tag, bp->user_token_number);
- token_translations[bp->user_token_number] = bp->value;
- }
+ if (! glr_parser)
+ warn_at (loc, _("%s affects only GLR parsers"), "%merge");
+ if (current_rule->merger != 0)
+ complain_at (loc, _("only one %s allowed per rule"), "%merge");
+ current_rule->merger =
+ get_merge_function (name, current_rule->sym->type_name, loc);
}
+/* Attach SYM to the current rule. If needed, move the previous
+ action as a mid-rule action. */
-/*----------------------------------------------------------------.
-| Assign symbol numbers, and write definition of token names into |
-| FDEFINES. Set up vectors SYMBOL_TABLE, TAGS of symbols. |
-`----------------------------------------------------------------*/
-
-static void
-packsymbols (void)
+void
+grammar_current_rule_symbol_append (symbol *sym, location loc)
{
- bucket *bp = NULL;
- int tokno = 1;
- int last_user_token_number;
-
- tags = XCALLOC (char *, nsyms + 1);
- user_toknums = XCALLOC (short, nsyms + 1);
- symbols = XCALLOC (bucket *, nsyms);
-
- sassoc = XCALLOC (short, 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;
- tags[bp->value] = bp->tag;
- user_toknums[bp->value] = bp->user_token_number;
- sassoc[bp->value] = bp->assoc;
- }
-
- 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;
+ if (current_rule->action)
+ grammar_midrule_action ();
+ ++nritems;
+ grammar_symbol_append (sym, loc);
}
+/* Attach an ACTION to the current rule. If needed, move the previous
+ action as a mid-rule action. */
-/*---------------------------------------------------------------.
-| Save the definition of token names in the `TOKENDEFS' muscle. |
-`---------------------------------------------------------------*/
-
-static void
-symbols_save (void)
+void
+grammar_current_rule_action_append (const char *action, location loc)
{
- struct obstack tokendefs;
- bucket *bp;
- obstack_init (&tokendefs);
-
- for (bp = firstsymbol; bp; bp = bp->next)
- {
- char *symbol = bp->tag; /* get symbol */
-
- if (bp->value >= ntokens)
- continue;
- if (bp->user_token_number == SALIAS)
- continue;
- if ('\'' == *symbol)
- continue; /* skip literal character */
- if (bp == errtoken)
- continue; /* skip error token */
- if ('\"' == *symbol)
- {
- /* use literal string only if given a symbol with an alias */
- if (bp->alias)
- symbol = bp->alias->tag;
- else
- continue;
- }
-
- /* Don't #define nonliteral tokens whose names contain periods. */
- if (strchr (symbol, '.'))
- continue;
-
- obstack_fgrow2 (&tokendefs, "# define %s\t%d\n",
- symbol, bp->user_token_number);
- if (semantic_parser)
- /* FIXME: This is probably wrong, and should be just as
- above. --akim. */
- obstack_fgrow2 (&tokendefs, "# define T%s\t%d\n", symbol, bp->value);
- }
-
- obstack_1grow (&tokendefs, 0);
- muscle_insert ("tokendef", xstrdup (obstack_finish (&tokendefs)));
- obstack_free (&tokendefs, NULL);
+ if (current_rule->action)
+ grammar_midrule_action ();
+ current_rule->action = action;
+ current_rule->action_location = loc;
}
-
+\f
/*---------------------------------------------------------------.
| Convert the rules into the representation using RRHS, RLHS and |
-| RITEMS. |
+| RITEM. |
`---------------------------------------------------------------*/
static void
packgram (void)
{
- int itemno;
- int ruleno;
- symbol_list *p;
-
- /* We use short to index items. */
- if (nitems >= MAXSHORT)
- fatal (_("too many items (max %d)"), MAXSHORT);
+ unsigned int itemno = 0;
+ rule_number ruleno = 0;
+ symbol_list *p = grammar;
- ritem = XCALLOC (short, nitems + 1);
- rule_table = XCALLOC (rule_t, nrules) - 1;
+ CALLOC (ritem, nritems);
+ CALLOC (rules, nrules);
- itemno = 0;
- ruleno = 1;
-
- p = grammar;
while (p)
{
- bucket *ruleprec = p->ruleprec;
- rule_table[ruleno].lhs = p->sym->value;
- rule_table[ruleno].rhs = itemno;
- rule_table[ruleno].line = p->line;
- rule_table[ruleno].useful = TRUE;
- rule_table[ruleno].action = p->action;
- rule_table[ruleno].action_line = p->action_line;
- rule_table[ruleno].guard = p->guard;
- rule_table[ruleno].guard_line = p->guard_line;
+ symbol *ruleprec = p->ruleprec;
+ rules[ruleno].user_number = ruleno;
+ rules[ruleno].number = ruleno;
+ rules[ruleno].lhs = p->sym;
+ rules[ruleno].rhs = ritem + itemno;
+ rules[ruleno].location = p->location;
+ rules[ruleno].useful = true;
+ rules[ruleno].action = p->action;
+ rules[ruleno].action_location = p->action_location;
+ rules[ruleno].dprec = p->dprec;
+ rules[ruleno].merger = p->merger;
p = p->next;
while (p && p->sym)
{
- ritem[itemno++] = p->sym->value;
+ /* item_number = symbol_number.
+ But the former needs to contain more: negative rule numbers. */
+ ritem[itemno++] = symbol_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)
- {
- rule_table[ruleno].prec = p->sym->prec;
- rule_table[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)
{
- rule_table[ruleno].prec = ruleprec->prec;
- rule_table[ruleno].assoc = ruleprec->assoc;
- rule_table[ruleno].precsym = ruleprec->value;
+ rules[ruleno].precsym = ruleprec;
+ rules[ruleno].prec = ruleprec;
}
-
- ritem[itemno++] = -ruleno;
- ruleno++;
+ ritem[itemno++] = rule_number_as_item_number (ruleno);
+ ++ruleno;
if (p)
p = p->next;
}
- ritem[itemno] = 0;
- nritems = itemno;
- assert (nritems == nitems);
+ if (itemno != nritems)
+ abort ();
- if (trace_flag)
+ if (trace_flag & trace_sets)
ritem_print (stderr);
}
\f
-/*-------------------------------------------------------------------.
-| Read in the grammar specification and record it in the format |
-| described in gram.h. All guards are copied into the GUARD_OBSTACK |
-| and all actions into ACTION_OBSTACK, in each case forming the body |
-| of a C function (YYGUARD or YYACTION) which contains a switch |
-| statement to decide which guard or action to execute. |
-`-------------------------------------------------------------------*/
+/*------------------------------------------------------------------.
+| Read in the grammar specification and record it in the format |
+| described in gram.h. All actions are copied into ACTION_OBSTACK, |
+| in each case forming the body of a C function (YYACTION) which |
+| contains a switch statement to decide which action to execute. |
+`------------------------------------------------------------------*/
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;
-
- /* Initialize the muscle obstack. */
- 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++;
+ /* Construct the accept symbol. */
+ accept = symbol_get ("$accept", empty_location);
+ accept->class = nterm_sym;
+ accept->number = nvars++;
/* Construct the error token */
- errtoken = getsym ("error");
+ errtoken = symbol_get ("error", empty_location);
errtoken->class = token_sym;
- errtoken->user_token_number = 256; /* Value specified by POSIX. */
+ errtoken->number = ntokens++;
/* Construct a token that represents all undefined literal tokens.
It is always token number 2. */
- undeftoken = getsym ("$undefined.");
+ undeftoken = symbol_get ("$undefined", empty_location);
undeftoken->class = token_sym;
- undeftoken->user_token_number = 2;
+ undeftoken->number = ntokens++;
/* Initialize the obstacks. */
- obstack_init (&action_obstack);
- obstack_init (&attrs_obstack);
- obstack_init (&output_obstack);
+ obstack_init (&pre_prologue_obstack);
+ obstack_init (&post_prologue_obstack);
+
+ finput = xfopen (grammar_file, "r");
+ gram_in = finput;
- finput = xfopen (infile, "r");
+ gram__flex_debug = trace_flag & trace_scan;
+ gram_debug = trace_flag & trace_parse;
+ scanner_initialize ();
+ gram_parse ();
- /* Read the declaration section. Copy %{ ... %} groups to
- TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left,
- etc. found there. */
- read_declarations ();
+ /* If something went wrong during the parsing, don't try to
+ continue. */
+ if (complaint_issued)
+ return;
+
+ /* Grammar has been read. Do some checking */
+ if (nrules == 0)
+ fatal (_("no rules in the input grammar"));
- /* If the user did not define her EOFTOKEN, do it now. */
- if (!eoftoken)
+ /* Report any undefined symbols and consider them nonterminals. */
+ symbols_check_defined ();
+
+ /* If the user did not define her ENDTOKEN, do it now. */
+ if (!endtoken)
{
- eoftoken = getsym ("$");
- eoftoken->class = token_sym;
+ endtoken = symbol_get ("$end", empty_location);
+ endtoken->class = token_sym;
+ endtoken->number = 0;
/* Value specified by POSIX. */
- eoftoken->user_token_number = 0;
+ endtoken->user_token_number = 0;
}
- /* Read in the grammar, build grammar in list form. Write out
- guards and actions. */
- readgram ();
- /* Some C code is given at the end of the grammar file. */
- read_additionnal_code ();
+ /* Insert the initial rule, which line is that of the first rule
+ (not that of the start symbol):
+
+ accept: %start EOF. */
+ {
+ symbol_list *p = symbol_list_new (accept, empty_location);
+ p->location = grammar->location;
+ p->next = symbol_list_new (startsymbol, empty_location);
+ p->next->next = symbol_list_new (endtoken, empty_location);
+ p->next->next->next = symbol_list_new (NULL, empty_location);
+ p->next->next->next->next = grammar;
+ nrules += 1;
+ nritems += 3;
+ grammar = p;
+ }
+
+ if (! (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars))
+ abort ();
- lex_free ();
xfclose (finput);
/* Assign the symbols their symbol numbers. Write #defines for the
token symbols into FDEFINES if requested. */
- packsymbols ();
- /* Save them. */
- symbols_save ();
+ symbols_pack ();
/* Convert the grammar into the format described in gram.h. */
packgram ();
+
+ /* The grammar as a symbol_list is no longer needed. */
+ LIST_FREE (symbol_list, grammar);
}
projects / bison.git / blobdiff