X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/676385e29c4aedfc05d20daf1ef20cd4ccc84856..bfdcc3a0b273c116e01d289ce52d400f22e4da14:/src/parse-gram.y?ds=sidebyside
diff --git a/src/parse-gram.y b/src/parse-gram.y
index 2d00e035..5f77a5bd 100644
--- a/src/parse-gram.y
+++ b/src/parse-gram.y
@@ -1,11 +1,12 @@
-/* Bison Grammar Parser -*- C -*-
- Copyright (C) 2002 Free Software Foundation, Inc.
+%{/* Bison Grammar Parser -*- C -*-
+
+ Copyright (C) 2002-2012 Free Software Foundation, Inc.
This file is part of Bison, the GNU Compiler Compiler.
- This program is free software; you can redistribute it and/or modify
+ This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
- the Free Software Foundation; either version 2 of the License, or
+ the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
@@ -14,89 +15,119 @@
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
- along with this program; if not, write to the Free Software
- Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
- 02111-1307 USA
-*/
-
+ along with this program. If not, see . */
-%debug
-%defines
-%locations
-%pure-parser
-// %error-verbose
-%defines
-%name-prefix="gram_"
-
-%{
+#include
#include "system.h"
-#include "muscle_tab.h"
+
+#include "c-ctype.h"
+#include "complain.h"
+#include "conflicts.h"
#include "files.h"
#include "getargs.h"
-#include "output.h"
-#include "symlist.h"
#include "gram.h"
+#include "muscle-tab.h"
+#include "named-ref.h"
+#include "quotearg.h"
#include "reader.h"
-#include "conflicts.h"
+#include "symlist.h"
+#include "scan-gram.h"
+#include "scan-code.h"
+#include "xmemdup0.h"
+
+#define YYLLOC_DEFAULT(Current, Rhs, N) (Current) = lloc_default (Rhs, N)
+static YYLTYPE lloc_default (YYLTYPE const *, int);
-/* Produce verbose parse errors. */
-#define YYERROR_VERBOSE 1
-#define YYLLOC_DEFAULT(Current, Rhs, N) \
-do { \
- if (N) \
- { \
- Current.first_column = Rhs[1].first_column; \
- Current.first_line = Rhs[1].first_line; \
- Current.last_column = Rhs[N].last_column; \
- Current.last_line = Rhs[N].last_line; \
- } \
- else \
- { \
- Current = Rhs[0]; \
- } \
-} while (0)
-
-/* Pass the control structure to YYPARSE and YYLEX. */
-#define YYPARSE_PARAM gram_control
-#define YYLEX_PARAM gram_control
-/* YYPARSE receives GRAM_CONTROL as a void *. Provide a
- correctly typed access to it. */
-#define yycontrol ((gram_control_t *) gram_control)
-
-/* Request detailed parse error messages, and pass them to
- GRAM_ERROR. */
+#define YY_LOCATION_PRINT(File, Loc) \
+ location_print (File, Loc)
+
+static void version_check (location const *loc, char const *version);
+
+/* Request detailed syntax error messages, and pass them to GRAM_ERROR.
+ FIXME: depends on the undocumented availability of YYLLOC. */
#undef yyerror
#define yyerror(Msg) \
- gram_error (yycontrol, &yylloc, Msg)
-
-#define YYPRINT(File, Type, Value) \
- yyprint (File, Type, &Value)
-static void yyprint (FILE *file, int type, const yystype *value);
-
-symbol_class current_class = unknown_sym;
-char *current_type = 0;
-symbol_t *current_lhs;
-location_t current_lhs_location;
-associativity current_assoc;
-int current_prec = 0;
-braced_code_t current_braced_code = action_braced_code;
+ gram_error (&yylloc, Msg)
+static void gram_error (location const *, char const *);
+
+static char const *char_name (char);
+
+/** Add a lex-param or a parse-param.
+ *
+ * \param type \a lex_param or \a parse_param
+ * \param decl the formal argument
+ * \param loc the location in the source.
+ */
+static void add_param (char const *type, char *decl, location loc);
+
+
+static symbol_class current_class = unknown_sym;
+static uniqstr current_type = NULL;
+static symbol *current_lhs_symbol;
+static location current_lhs_location;
+static named_ref *current_lhs_named_ref;
+static int current_prec = 0;
+
+/** Set the new current left-hand side symbol, possibly common
+ * to several right-hand side parts of rule.
+ */
+static
+void
+current_lhs(symbol *sym, location loc, named_ref *ref)
+{
+ current_lhs_symbol = sym;
+ current_lhs_location = loc;
+ /* In order to simplify memory management, named references for lhs
+ are always assigned by deep copy into the current symbol_list
+ node. This is because a single named-ref in the grammar may
+ result in several uses when the user factors lhs between several
+ rules using "|". Therefore free the parser's original copy. */
+ free (current_lhs_named_ref);
+ current_lhs_named_ref = ref;
+}
+
+
+#define YYTYPE_INT16 int_fast16_t
+#define YYTYPE_INT8 int_fast8_t
+#define YYTYPE_UINT16 uint_fast16_t
+#define YYTYPE_UINT8 uint_fast8_t
%}
+%debug
+%define api.prefix "gram_"
+%define api.pure
+%define parse.lac full
+%defines
+%error-verbose
+%expect 0
+%locations
+%verbose
+
+%initial-action
+{
+ /* Bison's grammar can initial empty locations, hence a default
+ location is needed. */
+ boundary_set (&@$.start, current_file, 1, 1);
+ boundary_set (&@$.end, current_file, 1, 1);
+}
-/* Only NUMBERS have a value. */
%union
{
- symbol_t *symbol;
- symbol_list_t *list;
+ symbol *symbol;
+ symbol_list *list;
int integer;
- char *string;
- associativity assoc;
+ char const *chars;
+ char *code;
+ assoc assoc;
+ uniqstr uniqstr;
+ unsigned char character;
+ named_ref *named_ref;
};
-/* Define the tokens together with there human representation. */
-%token GRAM_EOF 0 "end of string"
-%token STRING CHARACTER
-%token INT
+/* Define the tokens together with their human representation. */
+%token GRAM_EOF 0 "end of file"
+%token STRING "string"
+%token INT "integer"
%token PERCENT_TOKEN "%token"
%token PERCENT_NTERM "%nterm"
@@ -105,62 +136,98 @@ braced_code_t current_braced_code = action_braced_code;
%token PERCENT_DESTRUCTOR "%destructor"
%token PERCENT_PRINTER "%printer"
-%token PERCENT_UNION "%union"
-
%token PERCENT_LEFT "%left"
%token PERCENT_RIGHT "%right"
%token PERCENT_NONASSOC "%nonassoc"
-%token PERCENT_EXPECT "%expect"
-%token PERCENT_START "%start"
-%token PERCENT_PREC "%prec"
-%token PERCENT_DPREC "%dprec"
-%token PERCENT_MERGE "%merge"
-%token PERCENT_VERBOSE "%verbose"
-%token PERCENT_ERROR_VERBOSE "%error-verbose"
-
-%token PERCENT_OUTPUT "%output"
-%token PERCENT_FILE_PREFIX "%file-prefix"
-%token PERCENT_NAME_PREFIX "%name-prefix"
-
-%token PERCENT_DEFINE "%define"
-%token PERCENT_PURE_PARSER "%pure-parser"
-%token PERCENT_GLR_PARSER "%glr-parser"
-
-%token PERCENT_DEFINES "%defines"
-
-%token PERCENT_YACC "%yacc"
-
-%token PERCENT_DEBUG "%debug"
-%token PERCENT_LOCATIONS "%locations"
-%token PERCENT_NO_LINES "%no-lines"
-%token PERCENT_SKELETON "%skeleton"
-%token PERCENT_TOKEN_TABLE "%token-table"
-
-%token TYPE
-%token EQUAL "="
-%token SEMICOLON ";"
-%token COLON ":"
-%token PIPE "|"
-%token ID "identifier"
+%token PERCENT_PREC "%prec"
+%token PERCENT_DPREC "%dprec"
+%token PERCENT_MERGE "%merge"
+
+
+/*----------------------.
+| Global Declarations. |
+`----------------------*/
+
+%token
+ PERCENT_CODE "%code"
+ PERCENT_DEBUG "%debug"
+ PERCENT_DEFAULT_PREC "%default-prec"
+ PERCENT_DEFINE "%define"
+ PERCENT_DEFINES "%defines"
+ PERCENT_ERROR_VERBOSE "%error-verbose"
+ PERCENT_EXPECT "%expect"
+ PERCENT_EXPECT_RR "%expect-rr"
+ PERCENT_FILE_PREFIX "%file-prefix"
+ PERCENT_GLR_PARSER "%glr-parser"
+ PERCENT_INITIAL_ACTION "%initial-action"
+ PERCENT_LANGUAGE "%language"
+ PERCENT_LEX_PARAM "%lex-param"
+ PERCENT_LOCATIONS "%locations"
+ PERCENT_NAME_PREFIX "%name-prefix"
+ PERCENT_NO_DEFAULT_PREC "%no-default-prec"
+ PERCENT_NO_LINES "%no-lines"
+ PERCENT_NONDETERMINISTIC_PARSER
+ "%nondeterministic-parser"
+ PERCENT_OUTPUT "%output"
+ PERCENT_PARSE_PARAM "%parse-param"
+ PERCENT_PURE_PARSER "%pure-parser"
+ PERCENT_REQUIRE "%require"
+ PERCENT_SKELETON "%skeleton"
+ PERCENT_START "%start"
+ PERCENT_TOKEN_TABLE "%token-table"
+ PERCENT_VERBOSE "%verbose"
+ PERCENT_YACC "%yacc"
+;
+
+%token BRACED_CODE "{...}"
+%token BRACKETED_ID "[identifier]"
+%token CHAR "char"
+%token EPILOGUE "epilogue"
+%token EQUAL "="
+%token ID "identifier"
+%token ID_COLON "identifier:"
%token PERCENT_PERCENT "%%"
-%token PROLOGUE EPILOGUE
-%token BRACED_CODE
+%token PIPE "|"
+%token PROLOGUE "%{...%}"
+%token SEMICOLON ";"
+%token TYPE "type"
+%token TYPE_TAG_ANY "<*>"
+%token TYPE_TAG_NONE "<>"
+
+%type CHAR
+%printer { fputs (char_name ($$), stderr); } CHAR
+
+/* braceless is not to be used for rule or symbol actions, as it
+ calls code_props_plain_init. */
+%type STRING "%{...%}" EPILOGUE braceless content.opt
+%type "{...}"
+%printer { fputs (quotearg_style (c_quoting_style, $$), stderr); }
+ STRING
+%printer { fprintf (stderr, "{\n%s\n}", $$); }
+ braceless content.opt "{...}" "%{...%}" EPILOGUE
+
+%type BRACKETED_ID ID ID_COLON TYPE variable
+%printer { fputs ($$, stderr); }
+%printer { fprintf (stderr, "[%s]", $$); } BRACKETED_ID
+%printer { fprintf (stderr, "%s:", $$); } ID_COLON
+%printer { fprintf (stderr, "<%s>", $$); } TYPE
-%type CHARACTER TYPE STRING string_content
- BRACED_CODE PROLOGUE EPILOGUE epilogue.opt action
%type INT
-%type ID symbol string_as_id
+%printer { fprintf (stderr, "%d", $$); }
+
+%type id id_colon string_as_id symbol symbol.prec
+%printer { fprintf (stderr, "%s", $$->tag); }
+%printer { fprintf (stderr, "%s:", $$->tag); } id_colon
+
%type precedence_declarator
-%type symbols.1
+%type symbols.1 symbols.prec generic_symlist generic_symlist_item
+%type named_ref.opt
+
%%
input:
- declarations "%%" grammar epilogue.opt
- {
- yycontrol->errcode = 0;
- epilogue_set ($4, @4);
- }
+ prologue_declarations "%%" grammar epilogue.opt
;
@@ -168,30 +235,108 @@ input:
| Declarations: before the first %%. |
`------------------------------------*/
-declarations:
+prologue_declarations:
/* Nothing */
-| declarations declaration semi_colon.opt
+| prologue_declarations prologue_declaration
;
-declaration:
+prologue_declaration:
grammar_declaration
-| PROLOGUE { prologue_augment ($1, @1); }
-| "%debug" { debug_flag = 1; }
-| "%define" string_content string_content { muscle_insert ($2, $3); }
-| "%defines" { defines_flag = 1; }
-| "%error-verbose" { error_verbose = 1; }
-| "%expect" INT { expected_conflicts = $2; }
-| "%file-prefix" "=" string_content { spec_file_prefix = $3; }
-| "%locations" { locations_flag = 1; }
-| "%name-prefix" "=" string_content { spec_name_prefix = $3; }
-| "%no-lines" { no_lines_flag = 1; }
-| "%output" "=" string_content { spec_outfile = $3; }
-| "%pure-parser" { pure_parser = 1; }
-| "%glr-parser" { glr_parser = 1; }
-| "%skeleton" string_content { skeleton = $2; }
-| "%token-table" { token_table_flag = 1; }
-| "%verbose" { report_flag = 1; }
-| "%yacc" { yacc_flag = 1; }
+| "%{...%}"
+ {
+ code_props plain_code;
+ code_props_plain_init (&plain_code, $1, @1);
+ code_props_translate_code (&plain_code);
+ gram_scanner_last_string_free ();
+ muscle_code_grow (union_seen ? "post_prologue" : "pre_prologue",
+ plain_code.code, @1);
+ code_scanner_last_string_free ();
+ }
+| "%debug" { debug = true; }
+| "%define" variable content.opt
+ {
+ muscle_percent_define_insert ($2, @2, $3,
+ MUSCLE_PERCENT_DEFINE_GRAMMAR_FILE);
+ }
+| "%defines" { defines_flag = true; }
+| "%defines" STRING
+ {
+ defines_flag = true;
+ spec_defines_file = xstrdup ($2);
+ }
+| "%error-verbose" { error_verbose = true; }
+| "%expect" INT { expected_sr_conflicts = $2; }
+| "%expect-rr" INT { expected_rr_conflicts = $2; }
+| "%file-prefix" STRING { spec_file_prefix = $2; }
+| "%file-prefix" "=" STRING { spec_file_prefix = $3; } /* deprecated */
+| "%glr-parser"
+ {
+ nondeterministic_parser = true;
+ glr_parser = true;
+ }
+| "%initial-action" "{...}"
+ {
+ code_props action;
+ code_props_symbol_action_init (&action, $2, @2);
+ code_props_translate_code (&action);
+ gram_scanner_last_string_free ();
+ muscle_code_grow ("initial_action", action.code, @2);
+ code_scanner_last_string_free ();
+ }
+| "%language" STRING { language_argmatch ($2, grammar_prio, @1); }
+| "%lex-param" "{...}" { add_param ("lex_param", $2, @2); }
+| "%locations" { locations_flag = true; }
+| "%name-prefix" STRING { spec_name_prefix = $2; }
+| "%name-prefix" "=" STRING { spec_name_prefix = $3; } /* deprecated */
+| "%no-lines" { no_lines_flag = true; }
+| "%nondeterministic-parser" { nondeterministic_parser = true; }
+| "%output" STRING { spec_outfile = $2; }
+| "%output" "=" STRING { spec_outfile = $3; } /* deprecated */
+| "%parse-param" "{...}" { add_param ("parse_param", $2, @2); }
+| "%pure-parser"
+ {
+ /* %pure-parser is deprecated in favor of `%define api.pure', so use
+ `%define api.pure' in a backward-compatible manner here. First, don't
+ complain if %pure-parser is specified multiple times. */
+ if (!muscle_find_const ("percent_define(api.pure)"))
+ muscle_percent_define_insert ("api.pure", @1, "",
+ MUSCLE_PERCENT_DEFINE_GRAMMAR_FILE);
+ /* In all cases, use api.pure now so that the backend doesn't complain if
+ the skeleton ignores api.pure, but do warn now if there's a previous
+ conflicting definition from an actual %define. */
+ if (!muscle_percent_define_flag_if ("api.pure"))
+ muscle_percent_define_insert ("api.pure", @1, "",
+ MUSCLE_PERCENT_DEFINE_GRAMMAR_FILE);
+ }
+| "%require" STRING { version_check (&@2, $2); }
+| "%skeleton" STRING
+ {
+ char const *skeleton_user = $2;
+ if (mbschr (skeleton_user, '/'))
+ {
+ size_t dir_length = strlen (current_file);
+ char *skeleton_build;
+ while (dir_length && current_file[dir_length - 1] != '/')
+ --dir_length;
+ while (dir_length && current_file[dir_length - 1] == '/')
+ --dir_length;
+ skeleton_build =
+ xmalloc (dir_length + 1 + strlen (skeleton_user) + 1);
+ if (dir_length > 0)
+ {
+ memcpy (skeleton_build, current_file, dir_length);
+ skeleton_build[dir_length++] = '/';
+ }
+ strcpy (skeleton_build + dir_length, skeleton_user);
+ skeleton_user = uniqstr_new (skeleton_build);
+ free (skeleton_build);
+ }
+ skeleton_arg (skeleton_user, grammar_prio, @1);
+ }
+| "%token-table" { token_table_flag = true; }
+| "%verbose" { report_flag |= report_states; }
+| "%yacc" { yacc_flag = true; }
+| /*FIXME: Err? What is this horror doing here? */ ";"
;
grammar_declaration:
@@ -201,34 +346,76 @@ grammar_declaration:
{
grammar_start_symbol_set ($2, @2);
}
-| "%union" BRACED_CODE
+| "%destructor" "{...}" generic_symlist
+ {
+ code_props code;
+ code_props_symbol_action_init (&code, $2, @2);
+ code_props_translate_code (&code);
+ {
+ symbol_list *list;
+ for (list = $3; list; list = list->next)
+ symbol_list_destructor_set (list, &code);
+ symbol_list_free ($3);
+ }
+ }
+| "%printer" "{...}" generic_symlist
+ {
+ code_props code;
+ code_props_symbol_action_init (&code, $2, @2);
+ code_props_translate_code (&code);
+ {
+ symbol_list *list;
+ for (list = $3; list; list = list->next)
+ symbol_list_printer_set (list, &code);
+ symbol_list_free ($3);
+ }
+ }
+| "%default-prec"
+ {
+ default_prec = true;
+ }
+| "%no-default-prec"
+ {
+ default_prec = false;
+ }
+| "%code" braceless
{
- typed = 1;
- MUSCLE_INSERT_INT ("stype_line", @2.first_line);
- muscle_insert ("stype", $2);
+ /* Do not invoke muscle_percent_code_grow here since it invokes
+ muscle_user_name_list_grow. */
+ muscle_code_grow ("percent_code()", $2, @2);
+ code_scanner_last_string_free ();
}
-| "%destructor"
- { current_braced_code = destructor_braced_code; }
- BRACED_CODE symbols.1
+| "%code" ID braceless
{
- symbol_list_t *list;
- for (list = $4; list; list = list->next)
- symbol_destructor_set (list->sym, $3, @3);
- symbol_list_free ($4);
- current_braced_code = action_braced_code;
+ muscle_percent_code_grow ($2, @2, $3, @3);
+ code_scanner_last_string_free ();
}
-| "%printer"
- { current_braced_code = printer_braced_code; }
- BRACED_CODE symbols.1
+;
+
+
+/*----------*
+ | %union. |
+ *----------*/
+
+%token PERCENT_UNION "%union";
+
+union_name:
+ /* Nothing. */ {}
+| ID { muscle_code_grow ("union_name", $1, @1); }
+;
+
+grammar_declaration:
+ "%union" union_name braceless
{
- symbol_list_t *list;
- for (list = $4; list; list = list->next)
- symbol_printer_set (list->sym, $3, list->location);
- symbol_list_free ($4);
- current_braced_code = action_braced_code;
+ union_seen = true;
+ muscle_code_grow ("stype", $3, @3);
+ code_scanner_last_string_free ();
}
;
+
+
+
symbol_declaration:
"%nterm" { current_class = nterm_sym; } symbol_defs.1
{
@@ -242,22 +429,23 @@ symbol_declaration:
}
| "%type" TYPE symbols.1
{
- symbol_list_t *list;
+ symbol_list *list;
+ tag_seen = true;
for (list = $3; list; list = list->next)
- symbol_type_set (list->sym, $2, @2);
+ symbol_type_set (list->content.sym, $2, @2);
symbol_list_free ($3);
}
;
precedence_declaration:
- precedence_declarator type.opt symbols.1
+ precedence_declarator type.opt symbols.prec
{
- symbol_list_t *list;
+ symbol_list *list;
++current_prec;
for (list = $3; list; list = list->next)
{
- symbol_type_set (list->sym, current_type, @2);
- symbol_precedence_set (list->sym, current_prec, $1, @1);
+ symbol_type_set (list->content.sym, current_type, @2);
+ symbol_precedence_set (list->content.sym, current_prec, $1, @1);
}
symbol_list_free ($3);
current_type = NULL;
@@ -271,15 +459,41 @@ precedence_declarator:
;
type.opt:
- /* Nothing. */ { current_type = NULL;}
-| TYPE { current_type = $1; }
+ /* Nothing. */ { current_type = NULL; }
+| TYPE { current_type = $1; tag_seen = true; }
;
-/* One or more nonterminals to be %typed. */
+/* Just like symbols.1 but accept INT for the sake of POSIX. */
+symbols.prec:
+ symbol.prec
+ { $$ = symbol_list_sym_new ($1, @1); }
+| symbols.prec symbol.prec
+ { $$ = symbol_list_prepend ($1, symbol_list_sym_new ($2, @2)); }
+;
+symbol.prec:
+ symbol { $$ = $1; }
+ | symbol INT { $$ = $1; symbol_user_token_number_set ($1, $2, @2); }
+ ;
+
+/* One or more symbols to be %typed. */
symbols.1:
- symbol { $$ = symbol_list_new ($1, @1); }
-| symbols.1 symbol { $$ = symbol_list_prepend ($1, $2, @2); }
+ symbol
+ { $$ = symbol_list_sym_new ($1, @1); }
+| symbols.1 symbol
+ { $$ = symbol_list_prepend ($1, symbol_list_sym_new ($2, @2)); }
+;
+
+generic_symlist:
+ generic_symlist_item { $$ = $1; }
+| generic_symlist generic_symlist_item { $$ = symbol_list_prepend ($1, $2); }
+;
+
+generic_symlist_item:
+ symbol { $$ = symbol_list_sym_new ($1, @1); }
+| TYPE { $$ = symbol_list_type_new ($1, @1); }
+| "<*>" { $$ = symbol_list_default_tagged_new (@1); }
+| "<>" { $$ = symbol_list_default_tagless_new (@1); }
;
/* One token definition. */
@@ -287,39 +501,38 @@ symbol_def:
TYPE
{
current_type = $1;
+ tag_seen = true;
}
-| ID
+| id
{
- symbol_class_set ($1, current_class, @1);
+ symbol_class_set ($1, current_class, @1, true);
symbol_type_set ($1, current_type, @1);
}
-| ID INT
+| id INT
{
- symbol_class_set ($1, current_class, @1);
+ symbol_class_set ($1, current_class, @1, true);
symbol_type_set ($1, current_type, @1);
symbol_user_token_number_set ($1, $2, @2);
}
-| ID string_as_id
+| id string_as_id
{
- symbol_class_set ($1, current_class, @1);
+ symbol_class_set ($1, current_class, @1, true);
symbol_type_set ($1, current_type, @1);
- symbol_make_alias ($1, $2);
+ symbol_make_alias ($1, $2, @$);
}
-| ID INT string_as_id
+| id INT string_as_id
{
- symbol_class_set ($1, current_class, @1);
+ symbol_class_set ($1, current_class, @1, true);
symbol_type_set ($1, current_type, @1);
symbol_user_token_number_set ($1, $2, @2);
- symbol_make_alias ($1, $3);
+ symbol_make_alias ($1, $3, @$);
}
;
/* One or more symbol definitions. */
symbol_defs.1:
symbol_def
- {;}
| symbol_defs.1 symbol_def
- {;}
;
@@ -333,30 +546,38 @@ grammar:
;
/* As a Bison extension, one can use the grammar declarations in the
- body of the grammar. But to remain LALR(1), they must be ended
- with a semi-colon. */
+ body of the grammar. */
rules_or_grammar_declaration:
rules
| grammar_declaration ";"
+| error ";"
+ {
+ yyerrok;
+ }
;
rules:
- ID ":" { current_lhs = $1; current_lhs_location = @1; } rhses.1 ";"
- {;}
+ id_colon named_ref.opt { current_lhs ($1, @1, $2); } rhses.1
+ {
+ /* Free the current lhs. */
+ current_lhs (0, @1, 0);
+ }
;
rhses.1:
- rhs { grammar_rule_end (@1); }
-| rhses.1 "|" rhs { grammar_rule_end (@3); }
+ rhs { grammar_current_rule_end (@1); }
+| rhses.1 "|" rhs { grammar_current_rule_end (@3); }
+| rhses.1 ";"
;
rhs:
/* Nothing. */
- { grammar_rule_begin (current_lhs, current_lhs_location); }
-| rhs symbol
- { grammar_current_rule_symbol_append ($2, @2); }
-| rhs action
- { grammar_current_rule_action_append ($2, @2); }
+ { grammar_current_rule_begin (current_lhs_symbol, current_lhs_location,
+ current_lhs_named_ref); }
+| rhs symbol named_ref.opt
+ { grammar_current_rule_symbol_append ($2, @2, $3); }
+| rhs "{...}" named_ref.opt
+ { grammar_current_rule_action_append ($2, @2, $3); }
| rhs "%prec" symbol
{ grammar_current_rule_prec_set ($3, @3); }
| rhs "%dprec" INT
@@ -365,94 +586,208 @@ rhs:
{ grammar_current_rule_merge_set ($3, @3); }
;
-symbol:
- ID { $$ = $1; }
-| string_as_id { $$ = $1; }
-| CHARACTER { $$ = getsym ($1, @1); }
+named_ref.opt:
+ /* Nothing. */ { $$ = 0; }
+|
+ BRACKETED_ID { $$ = named_ref_new($1, @1); }
+;
+
+
+/*----------------------------*
+ | variable and content.opt. |
+ *---------------------------*/
+
+/* The STRING form of variable is deprecated and is not M4-friendly.
+ For example, M4 fails for `%define "[" "value"'. */
+variable:
+ ID
+| STRING { $$ = uniqstr_new ($1); }
;
-action:
- BRACED_CODE
- { $$ = $1; }
+/* Some content or empty by default. */
+content.opt:
+ /* Nothing. */ { $$ = ""; }
+| ID { $$ = $1; }
+| STRING
;
-/* A string used as an ID: we have to keep the quotes. */
-string_as_id:
- STRING
+
+/*-------------*
+ | braceless. |
+ *-------------*/
+
+braceless:
+ "{...}"
{
- $$ = getsym ($1, @1);
- symbol_class_set ($$, token_sym, @1);
+ code_props plain_code;
+ $1[strlen ($1) - 1] = '\n';
+ code_props_plain_init (&plain_code, $1+1, @1);
+ code_props_translate_code (&plain_code);
+ gram_scanner_last_string_free ();
+ $$ = plain_code.code;
}
;
-/* A string used for its contents. Strip the quotes. */
-string_content:
- STRING
- {
- $$ = $1 + 1;
- $$[strlen ($$) - 1] = '\0';
- };
+/*---------------*
+ | Identifiers. |
+ *---------------*/
-epilogue.opt:
- /* Nothing. */
+/* Identifiers are returned as uniqstr values by the scanner.
+ Depending on their use, we may need to make them genuine symbols. */
+
+id:
+ ID
+ { $$ = symbol_from_uniqstr ($1, @1); }
+| CHAR
{
- $$ = xstrdup ("");
+ $$ = symbol_get (char_name ($1), @1);
+ symbol_class_set ($$, token_sym, @1, false);
+ symbol_user_token_number_set ($$, $1, @1);
}
-| "%%" EPILOGUE
+;
+
+id_colon:
+ ID_COLON { $$ = symbol_from_uniqstr ($1, @1); }
+;
+
+
+symbol:
+ id
+| string_as_id
+;
+
+/* A string used as an ID: quote it. */
+string_as_id:
+ STRING
{
- $$ = $2;
+ $$ = symbol_get (quotearg_style (c_quoting_style, $1), @1);
+ symbol_class_set ($$, token_sym, @1, false);
}
;
-semi_colon.opt:
+epilogue.opt:
/* Nothing. */
-| ";"
+| "%%" EPILOGUE
+ {
+ code_props plain_code;
+ code_props_plain_init (&plain_code, $2, @2);
+ code_props_translate_code (&plain_code);
+ gram_scanner_last_string_free ();
+ muscle_code_grow ("epilogue", plain_code.code, @2);
+ code_scanner_last_string_free ();
+ }
;
+
%%
-/*------------------------------------------------------------------.
-| When debugging the parser, display tokens' locations and values. |
-`------------------------------------------------------------------*/
+
+
+/* Return the location of the left-hand side of a rule whose
+ right-hand side is RHS[1] ... RHS[N]. Ignore empty nonterminals in
+ the right-hand side, and return an empty location equal to the end
+ boundary of RHS[0] if the right-hand side is empty. */
+
+static YYLTYPE
+lloc_default (YYLTYPE const *rhs, int n)
+{
+ int i;
+ YYLTYPE loc;
+
+ /* SGI MIPSpro 7.4.1m miscompiles "loc.start = loc.end = rhs[n].end;".
+ The bug is fixed in 7.4.2m, but play it safe for now. */
+ loc.start = rhs[n].end;
+ loc.end = rhs[n].end;
+
+ /* Ignore empty nonterminals the start of the right-hand side.
+ Do not bother to ignore them at the end of the right-hand side,
+ since empty nonterminals have the same end as their predecessors. */
+ for (i = 1; i <= n; i++)
+ if (! equal_boundaries (rhs[i].start, rhs[i].end))
+ {
+ loc.start = rhs[i].start;
+ break;
+ }
+
+ return loc;
+}
+
+
+/* Add a lex-param or a parse-param (depending on TYPE) with
+ declaration DECL and location LOC. */
static void
-yyprint (FILE *file,
- int type, const yystype *value)
+add_param (char const *type, char *decl, location loc)
{
- fputc (' ', file);
- switch (type)
+ static char const alphanum[26 + 26 + 1 + 10] =
+ "abcdefghijklmnopqrstuvwxyz"
+ "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
+ "_"
+ "0123456789";
+ char const *name_start = NULL;
+ char *p;
+
+ /* Stop on last actual character. */
+ for (p = decl; p[1]; p++)
+ if ((p == decl
+ || ! memchr (alphanum, p[-1], sizeof alphanum))
+ && memchr (alphanum, p[0], sizeof alphanum - 10))
+ name_start = p;
+
+ /* Strip the surrounding '{' and '}', and any blanks just inside
+ the braces. */
+ --p;
+ while (c_isspace ((unsigned char) *p))
+ --p;
+ p[1] = '\0';
+ ++decl;
+ while (c_isspace ((unsigned char) *decl))
+ ++decl;
+
+ if (! name_start)
+ complain_at (loc, _("missing identifier in parameter declaration"));
+ else
{
- case CHARACTER:
- fprintf (file, " = '%s'", value->string);
- break;
-
- case ID:
- fprintf (file, " = %s", value->symbol->tag);
- break;
-
- case INT:
- fprintf (file, " = %d", value->integer);
- break;
-
- case STRING:
- fprintf (file, " = \"%s\"", value->string);
- break;
-
- case TYPE:
- fprintf (file, " = <%s>", value->string);
- break;
-
- case BRACED_CODE:
- case PROLOGUE:
- case EPILOGUE:
- fprintf (file, " = {{ %s }}", value->string);
- break;
+ char *name = xmemdup0 (name_start, strspn (name_start, alphanum));
+ muscle_pair_list_grow (type, decl, name);
+ free (name);
}
+
+ gram_scanner_last_string_free ();
}
-void
-gram_error (gram_control_t *control ATTRIBUTE_UNUSED,
- location_t *yylloc, const char *msg)
+
+static void
+version_check (location const *loc, char const *version)
+{
+ if (strverscmp (version, PACKAGE_VERSION) > 0)
+ {
+ complain_at (*loc, "require bison %s, but have %s",
+ version, PACKAGE_VERSION);
+ exit (EX_MISMATCH);
+ }
+}
+
+static void
+gram_error (location const *loc, char const *msg)
{
- LOCATION_PRINT (stderr, *yylloc);
- fprintf (stderr, ": %s\n", msg);
+ complain_at (*loc, "%s", msg);
+}
+
+char const *
+token_name (int type)
+{
+ return yytname[YYTRANSLATE (type)];
+}
+
+static char const *
+char_name (char c)
+{
+ if (c == '\'')
+ return "'\\''";
+ else
+ {
+ char buf[4];
+ buf[0] = '\''; buf[1] = c; buf[2] = '\''; buf[3] = '\0';
+ return quotearg_style (escape_quoting_style, buf);
+ }
}