/* Input parser for bison
- Copyright 1984, 1986, 1989, 1992, 1998, 2000
+ Copyright 1984, 1986, 1989, 1992, 1998, 2000, 2001
Free Software Foundation, Inc.
This file is part of Bison, the GNU Compiler Compiler.
#include "quote.h"
#include "getargs.h"
#include "files.h"
-#include "xalloc.h"
#include "symtab.h"
+#include "options.h"
#include "lex.h"
#include "gram.h"
#include "complain.h"
#include "output.h"
#include "reader.h"
#include "conflicts.h"
+#include "muscle_tab.h"
/* Number of slots allocated (but not necessarily used yet) in `rline' */
static int rline_allocated;
/* Incremented for each %left, %right or %nonassoc seen */
static int lastprec;
-/* Incremented for each generated symbol */
-static int gensym_count;
-
static bucket *errtoken;
static bucket *undeftoken;
\f
`------------------------------------------------------------*/
static inline void
-copy_string (FILE *fin, struct obstack *oout, int match)
+copy_string2 (FILE *fin, struct obstack *oout, int match, int store)
{
int c;
- obstack_1grow (oout, match);
+ if (store)
+ obstack_1grow (oout, match);
c = getc (fin);
c = getc (fin);
}
- obstack_1grow (oout, c);
+ if (store)
+ obstack_1grow (oout, c);
+}
+
+/* FIXME. */
+
+static inline void
+copy_string (FILE *fin, struct obstack *oout, int match)
+{
+ copy_string2 (fin, oout, match, 1);
}
+/* FIXME. */
+
+static inline void
+copy_identifier (FILE *fin, struct obstack *oout)
+{
+ int c;
+
+ while (isalnum (c = getc (fin)) || c == '_')
+ obstack_1grow (oout, c);
+
+ ungetc (c, fin);
+}
/*-----------------------------------------------------------------.
| Dump the wannabee comment from IN to OUT1 and OUT2 (which can be |
else if (isdigit (c) || c == '-')
{
int n;
- char buf[4096];
ungetc (c, fin);
n = read_signed_integer (fin);
symbol_list *rule, int stack_offset)
{
int c = getc (fin);
- char *type_name = NULL;
+ const char *type_name = NULL;
/* Get the type name if explicit. */
if (c == '<')
/* -1 while reading a character if prev char was %. */
int after_percent;
+#if 0
if (!no_lines_flag)
- obstack_fgrow2 (&attrs_obstack, "#line %d %s\n",
- lineno, quotearg_style (c_quoting_style, infile));
+ {
+ obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"),
+ lineno, quotearg_style (c_quoting_style,
+ muscle_find("filename")));
+ }
+#endif
after_percent = 0;
static void
parse_token_decl (symbol_class what_is, symbol_class what_is_not)
{
- token_t token = 0;
- char *typename = 0;
+ token_t token = tok_undef;
+ char *typename = NULL;
/* The symbol being defined. */
struct bucket *symbol = NULL;
/* symbol and symval combined are only one symbol */
nsyms--;
}
- translations = 1;
symbol = NULL;
}
else if (token == tok_identifier)
else if (symbol && token == tok_number)
{
symbol->user_token_number = numval;
- translations = 1;
}
else
{
if (prev == tok_identifier)
{
symval->user_token_number = numval;
- translations = 1;
}
else
{
}
prev = t;
-
}
}
/*--------------------------------------------------------------.
-| Copy the union declaration into ATTRS_OBSTACK (and fdefines), |
-| where it is made into the definition of YYSTYPE, the type of |
-| elements of the parser value stack. |
+| 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
{
int c;
int count = 0;
+ struct obstack union_obstack;
+ const char *prologue = "\
+#ifndef YYSTYPE\n\
+typedef union";
+ const char *epilogue = "\
+ yystype;\n\
+# define YYSTYPE yystype\n\
+#endif\n";
if (typed)
complain (_("multiple %s declarations"), "%union");
typed = 1;
- if (!no_lines_flag)
- obstack_fgrow2 (&attrs_obstack, "\n#line %d %s\n",
- lineno, quotearg_style (c_quoting_style, infile));
- else
+ /* FIXME: I'm worried: are you sure attrs_obstack is properly
+ filled? */
+ /* I don't see any reasons to keep this line, because we should
+ create a special skeleton for this option. */
+ if (no_lines_flag)
obstack_1grow (&attrs_obstack, '\n');
- obstack_sgrow (&attrs_obstack, "typedef union");
+ obstack_init (&union_obstack);
+ obstack_sgrow (&union_obstack, "union");
if (defines_flag)
- obstack_sgrow (&defines_obstack, "typedef union");
+ obstack_sgrow (&defines_obstack, prologue);
c = getc (finput);
while (c != EOF)
{
- obstack_1grow (&attrs_obstack, c);
- if (defines_flag)
- obstack_1grow (&defines_obstack, c);
+ /* If C contains '/', it is output by copy_comment (). */
+ if (c != '/')
+ {
+ obstack_1grow (&union_obstack, c);
+ if (defines_flag)
+ obstack_1grow (&defines_obstack, c);
+ }
switch (c)
{
break;
case '/':
- copy_comment2 (finput, &defines_obstack, &attrs_obstack);
+ copy_comment2 (finput, &defines_obstack, &union_obstack);
break;
case '{':
count--;
if (count <= 0)
{
- obstack_sgrow (&attrs_obstack, " YYSTYPE;\n");
if (defines_flag)
- obstack_sgrow (&defines_obstack, " YYSTYPE;\n");
+ obstack_sgrow (&defines_obstack, epilogue);
/* 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));
return;
}
}
c = getc (finput);
}
+
}
token_t token;
struct bucket *symbol;
char *typename = 0;
- int usrtoknum;
+ int usrtoknum = SUNDEF;
- translations = 1;
token = lex (); /* fetch typename or first token */
if (token == tok_typename)
{
usrtoknum = numval;
token = lex (); /* okay, did number, now get literal */
}
- else
- usrtoknum = 0;
/* process literal string token */
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);
+}
+
+
+/*----------------------------------.
+| Parse what comes after %skeleton. |
+`----------------------------------*/
+
+void
+parse_skel_decl (void)
+{
+ /* Complete with parse_dquoted_param () on the CVS branch 1.29. */
+}
/*----------------------------------------------------------------.
| Read from finput until `%%' is seen. Discard the `%%'. Handle |
static void
read_declarations (void)
{
- int c;
- int tok;
-
for (;;)
{
- c = skip_white_space ();
+ int c = skip_white_space ();
if (c == '%')
{
- tok = parse_percent_token ();
+ token_t tok = parse_percent_token ();
switch (tok)
{
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:
+ case tok_illegal:
+ abort ();
+ break;
+
default:
complain (_("unrecognized: %s"), token_buffer);
skip_to_char ('%');
if (semantic_parser)
stack_offset = 0;
- sprintf (buf, "\ncase %d:\n", nrules);
- obstack_grow (&action_obstack, buf, strlen (buf));
+ obstack_fgrow1 (&action_obstack, "\ncase %d:\n", nrules);
if (!no_lines_flag)
{
- sprintf (buf, "#line %d %s\n",
- lineno, quotearg_style (c_quoting_style, infile));
- obstack_grow (&action_obstack, buf, strlen (buf));
+ obstack_fgrow2 (&action_obstack, muscle_find ("linef"),
+ lineno, quotearg_style (c_quoting_style,
+ muscle_find ("filename")));
}
obstack_1grow (&action_obstack, '{');
obstack_fgrow1 (&guard_obstack, "\ncase %d:\n", nrules);
if (!no_lines_flag)
- obstack_fgrow2 (&guard_obstack, "#line %d %s\n",
- lineno, quotearg_style (c_quoting_style, infile));
+ obstack_fgrow2 (&guard_obstack, muscle_find ("linef"),
+ lineno, quotearg_style (c_quoting_style,
+ muscle_find ("filename")));
obstack_1grow (&guard_obstack, '{');
count = 0;
}
\f
-static void
-record_rule_line (void)
-{
- /* Record each rule's source line number in rline table. */
-
- if (nrules >= rline_allocated)
- {
- rline_allocated = nrules * 2;
- rline = XREALLOC (rline, short, rline_allocated);
- }
- rline[nrules] = lineno;
-}
-
-
/*-------------------------------------------------------------------.
| Generate a dummy symbol, a nonterminal, whose name cannot conflict |
| with the user's names. |
static bucket *
gensym (void)
{
+ /* Incremented for each generated symbol */
+ static int gensym_count = 0;
+ static char buf[256];
+
bucket *sym;
- sprintf (token_buffer, "@%d", ++gensym_count);
+ sprintf (buf, "@%d", ++gensym_count);
+ token_buffer = buf;
sym = getsym (token_buffer);
sym->class = nterm_sym;
sym->value = nvars++;
nrules++;
nitems++;
- record_rule_line ();
-
p = XCALLOC (symbol_list, 1);
p->sym = lhs;
just read can belong to it. */
nrules++;
nitems++;
- record_rule_line ();
p = XCALLOC (symbol_list, 1);
if (crule1)
crule1->next = p;
grammar = p;
p->sym = sdummy;
crule1 = XCALLOC (symbol_list, 1);
- p->next = crule1;
crule1->next = crule;
+ p->next = crule1;
+
/* Insert the dummy generated by that rule into this
rule. */
nitems++;
else
{
- complain (_("invalid input: %s"), token_buffer);
+ complain (_("invalid input: %s"), quote (token_buffer));
t = lex ();
}
}
if (nrules == 0)
fatal (_("no rules in the input grammar"));
- /* JF put out same default YYSTYPE as YACC does */
- if (typed == 0
- && !value_components_used)
- {
- /* We used to use `unsigned long' as YYSTYPE on MSDOS,
- but it seems better to be consistent.
- Most programs should declare their own type anyway. */
- obstack_sgrow (&attrs_obstack,
- "#ifndef YYSTYPE\n#define YYSTYPE int\n#endif\n");
- if (defines_flag)
- obstack_sgrow (&defines_obstack, "\
-#ifndef YYSTYPE\n\
-# define YYSTYPE int\n\
-#endif\n");
- }
-
/* Report any undefined symbols and consider them nonterminals. */
for (bp = firstsymbol; bp; bp = bp->next)
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)
+{
+ char c;
+ struct obstack el_obstack;
+
+ obstack_init (&el_obstack);
+
+ while ((c = getc (finput)) != EOF)
+ obstack_1grow (&el_obstack, c);
+
+ obstack_1grow (&el_obstack, 0);
+ muscle_insert ("epilogue", obstack_finish (&el_obstack));
+}
+
\f
/*--------------------------------------------------------------.
| For named tokens, but not literal ones, define the name. The |
if (c != '\0')
continue;
- obstack_fgrow2 (oout, "#define\t%s\t%d\n",
- symbol,
- ((translations && !raw_flag)
- ? bp->user_token_number : bp->value));
+ obstack_fgrow2 (oout, "# define\t%s\t%d\n",
+ symbol, bp->user_token_number);
if (semantic_parser)
- obstack_fgrow2 (oout, "#define\tT%s\t%d\n", symbol, bp->value);
+ /* FIXME: This is certainly dead wrong, and should be just as
+ above. --akim. */
+ obstack_fgrow2 (oout, "# define\tT%s\t%d\n", symbol, bp->value);
}
+}
+
+
+/*------------------------------------------------------------------.
+| Set TOKEN_TRANSLATIONS. Check that no two symbols share the same |
+| number. |
+`------------------------------------------------------------------*/
+
+static void
+token_translations_init (void)
+{
+ 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 $undefined., which represents all invalid
+ inputs. */
+ for (i = 0; i <= max_user_token_number; i++)
+ token_translations[i] = 2;
+
+ for (bp = firstsymbol; bp; bp = bp->next)
+ {
+ /* Non-terminal? */
+ if (bp->value >= ntokens)
+ continue;
+ /* A token string alias? */
+ if (bp->user_token_number == SALIAS)
+ continue;
- obstack_1grow (oout, '\n');
+ assert (bp->user_token_number != SUNDEF);
+
+ /* A token which translation has already been set? */
+ if (token_translations[bp->user_token_number] != 2)
+ complain (_("tokens %s and %s both assigned number %d"),
+ tags[token_translations[bp->user_token_number]],
+ bp->tag, bp->user_token_number);
+ token_translations[bp->user_token_number] = bp->value;
+ }
}
static void
packsymbols (void)
{
- bucket *bp;
+ bucket *bp = NULL;
int tokno = 1;
- int i;
int last_user_token_number;
static char DOLLAR[] = "$";
- /* int lossage = 0; JF set but not used */
-
tags = XCALLOC (char *, nsyms + 1);
- tags[0] = DOLLAR;
user_toknums = XCALLOC (short, nsyms + 1);
- user_toknums[0] = 0;
sprec = XCALLOC (short, nsyms);
sassoc = XCALLOC (short, nsyms);
+ /* The EOF token. */
+ tags[0] = DOLLAR;
+ user_toknums[0] = 0;
+
max_user_token_number = 256;
last_user_token_number = 256;
if (bp->class == token_sym)
{
- if (translations && !(bp->user_token_number))
+ 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;
user_toknums[bp->value] = bp->user_token_number;
sprec[bp->value] = bp->prec;
sassoc[bp->value] = bp->assoc;
-
}
- if (translations)
- {
- int j;
-
- 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 (j = 0; j <= max_user_token_number; j++)
- token_translations[j] = 2;
-
- for (bp = firstsymbol; bp; bp = bp->next)
- {
- if (bp->value >= ntokens)
- continue; /* non-terminal */
- if (bp->user_token_number == SALIAS)
- continue;
- 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;
- }
- }
+ token_translations_init ();
error_token_number = errtoken->value;
- if (!no_parser_flag)
- output_token_defines (&table_obstack);
-
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;
+}
+
+
+/*-----------------------------------.
+| Output definition of token names. |
+`-----------------------------------*/
+
+static void
+symbols_output (void)
+{
+ {
+ struct obstack tokendefs;
+ obstack_init (&tokendefs);
+ output_token_defines (&tokendefs);
+ obstack_1grow (&tokendefs, 0);
+ muscle_insert ("tokendef", xstrdup (obstack_finish (&tokendefs)));
+ obstack_free (&tokendefs, NULL);
+ }
+
+#if 0
+ if (!no_parser_flag)
+ output_token_defines (&table_obstack);
+#endif
if (defines_flag)
{
}
if (semantic_parser)
- for (i = ntokens; i < nsyms; i++)
- {
- /* don't make these for dummy nonterminals made by gensym. */
- if (*tags[i] != '@')
- obstack_fgrow2 (&defines_obstack,
- "#define\tNT%s\t%d\n", tags[i], i);
- }
+ {
+ int i;
+
+ for (i = ntokens; i < nsyms; i++)
+ {
+ /* don't make these for dummy nonterminals made by gensym. */
+ if (*tags[i] != '@')
+ obstack_fgrow2 (&defines_obstack,
+ "# define\tNT%s\t%d\n", tags[i], i);
+ }
#if 0
- /* `fdefines' is now a temporary file, so we need to copy its
- contents in `done', so we can't close it here. */
- fclose (fdefines);
- fdefines = NULL;
+ /* `fdefines' is now a temporary file, so we need to copy its
+ contents in `done', so we can't close it here. */
+ fclose (fdefines);
+ fdefines = NULL;
#endif
+ }
}
}
bucket *ruleprec;
ritem = XCALLOC (short, nitems + 1);
- rlhs = XCALLOC (short, nrules) - 1;
- rrhs = XCALLOC (short, nrules) - 1;
- rprec = XCALLOC (short, nrules) - 1;
- rprecsym = XCALLOC (short, nrules) - 1;
- rassoc = XCALLOC (short, nrules) - 1;
+ rule_table = XCALLOC (rule_t, nrules) - 1;
itemno = 0;
ruleno = 1;
p = grammar;
while (p)
{
- rlhs[ruleno] = p->sym->value;
- rrhs[ruleno] = itemno;
+ rule_table[ruleno].lhs = p->sym->value;
+ rule_table[ruleno].rhs = itemno;
+ rule_table[ruleno].line = p->sym->line;
ruleprec = p->ruleprec;
p = p->next;
of the last token in it. */
if (p->sym->class == token_sym)
{
- rprec[ruleno] = p->sym->prec;
- rassoc[ruleno] = p->sym->assoc;
+ rule_table[ruleno].prec = p->sym->prec;
+ rule_table[ruleno].assoc = p->sym->assoc;
}
if (p)
p = p->next;
the specified symbol's precedence replaces the default. */
if (ruleprec)
{
- rprec[ruleno] = ruleprec->prec;
- rassoc[ruleno] = ruleprec->assoc;
- rprecsym[ruleno] = ruleprec->value;
+ rule_table[ruleno].prec = ruleprec->prec;
+ rule_table[ruleno].assoc = ruleprec->assoc;
+ rule_table[ruleno].precsym = ruleprec->value;
}
ritem[itemno++] = -ruleno;
start_flag = 0;
startval = NULL; /* start symbol not specified yet. */
-#if 0
- /* initially assume token number translation not needed. */
- translations = 0;
-#endif
- /* Nowadays translations is always set to 1, since we give `error' a
- user-token-number to satisfy the Posix demand for YYERRCODE==256.
- */
- translations = 1;
-
nsyms = 1;
nvars = 0;
nrules = 0;
nitems = 0;
- rline_allocated = 10;
- rline = XCALLOC (short, rline_allocated);
typed = 0;
lastprec = 0;
- gensym_count = 0;
-
semantic_parser = 0;
pure_parser = 0;
grammar = NULL;
- init_lex ();
+ lex_init ();
lineno = 1;
+ /* Initialize the muscle obstack. */
+ obstack_init (&muscle_obstack);
+
/* Initialize the symbol table. */
tabinit ();
+
/* Construct the error token */
errtoken = getsym ("error");
errtoken->class = token_sym;
errtoken->user_token_number = 256; /* Value specified by POSIX. */
+
/* Construct a token that represents all undefined literal tokens.
It is always token number 2. */
undeftoken = getsym ("$undefined.");
/* Read the declaration section. Copy %{ ... %} groups to
TABLE_OBSTACK and FDEFINES file. Also notice any %token, %left,
etc. found there. */
- obstack_1grow (&table_obstack, '\n');
- obstack_fgrow3 (&table_obstack, "\
-/* %s, made from %s\n\
- by GNU bison %s. */\n\
-\n",
- no_parser_flag ? "Bison-generated parse tables" : "A Bison parser",
- infile, VERSION);
-
- obstack_sgrow (&table_obstack,
- "#define YYBISON 1 /* Identify Bison output. */\n\n");
read_declarations ();
- /* Start writing the guard and action files, if they are needed. */
- output_headers ();
/* 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 ();
+
/* Now we know whether we need the line-number stack. If we do,
- write its type into the .tab.h file. */
- if (defines_flag)
- reader_output_yylsp (&defines_obstack);
- /* Write closing delimiters for actions and guards. */
- output_trailers ();
- if (locations_flag)
- obstack_sgrow (&table_obstack, "#define YYLSP_NEEDED 1\n\n");
+ write its type into the .tab.h file.
+ This is no longer need with header skeleton. */
+
/* Assign the symbols their symbol numbers. Write #defines for the
token symbols into FDEFINES if requested. */
packsymbols ();
+ symbols_output ();
/* Convert the grammar into the format described in gram.h. */
packgram ();
- /* Free the symbol table data structure since symbols are now all
- referred to by symbol number. */
- free_symtab ();
-}
-
-
-/*------------------------------------------------------------------.
-| Define YYLTYPE. Cannot be in the skeleton since we might have to |
-| output it in the headers if --defines is used. |
-`------------------------------------------------------------------*/
-
-void
-reader_output_yylsp (struct obstack *oout)
-{
- if (locations_flag)
- obstack_sgrow (oout, "\
-\n\
-#ifndef YYLTYPE\n\
-typedef struct yyltype\n\
-{\n\
- int first_line;\n\
- int first_column;\n\
-\n\
- int last_line;\n\
- int last_column;\n\
-} yyltype;\n\
-\n\
-# define YYLTYPE yyltype\n\
-#endif\n\
-\n");
+ /* Output the headers. */
+ symbols_output ();
}
projects / bison.git / blobdiff