This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Objective C lexer - small cleanup
- To: gcc-patches at gcc dot gnu dot org, Stan Shebs <shebs at apple dot com>
- Subject: Objective C lexer - small cleanup
- From: "Zack Weinberg" <zackw at stanford dot edu>
- Date: Thu, 28 Jun 2001 17:03:07 -0700
This patch cleans up the way Objective C @-prefixed tokens are
handled. @"string" is fed to the parser as <@> <"string"> and dealt
with in the grammar. @keyword is handled by cpplib's lookahead
mechanism and the same context trick we're using for protocol
qualifiers.
This approach should be less of an obstacle to improvements in the
handling of string constant concatenation. It's also conceptually
tidier.
It compiles and passes the Objective C testsuite with no regressions.
I am running a complete bootstrap now.
--
zw But your argument is simply post hoc ergo ante hoc.
-- Umberto Eco, _Foucault's Pendulum_
* c-common.h (enum rid): Add RID_FIRST_AT, RID_LAST_AT,
RID_LAST_PQ. Move RID_FIRST_PQ down with the other FIRST/LAST
enumerators.
(OBJC_IS_AT_KEYWORD, OBJC_IS_PQ_KEYWORD): New macros.
* c-parse.in (OBJC_STRING): Kill.
(objc_string): Decompose to [objc_string] '@' STRING.
(reswords): Take the leading '@' off all the Objective C
keywords.
(objc_rid_sans_at): Kill.
(init_reswords): Don't initialize it.
(yylexname): Use OBJC_IS_AT_KEYWORD and OBJC_IS_PQ_KEYWORD.
(_yylex): Kill reconsider label. Look ahead one token after
an '@'; if we get an identifier, check whether it's an
Objective C @-keyword. If so, return the keyword. Otherwise,
put back the token and return the '@' as a terminal.
===================================================================
Index: c-common.h
--- c-common.h 2001/06/18 16:15:01 1.75
+++ c-common.h 2001/06/28 23:58:54
@@ -63,7 +63,7 @@ enum rid
RID_FRIEND, RID_VIRTUAL, RID_EXPLICIT, RID_EXPORT, RID_MUTABLE,
/* ObjC */
- RID_FIRST_PQ, RID_IN = RID_FIRST_PQ, RID_OUT, RID_INOUT, RID_BYCOPY, RID_BYREF, RID_ONEWAY,
+ RID_IN, RID_OUT, RID_INOUT, RID_BYCOPY, RID_BYREF, RID_ONEWAY,
/* C */
RID_INT, RID_CHAR, RID_FLOAT, RID_DOUBLE, RID_VOID,
@@ -107,8 +107,21 @@ enum rid
RID_MAX,
RID_FIRST_MODIFIER = RID_STATIC,
- RID_LAST_MODIFIER = RID_ONEWAY
+ RID_LAST_MODIFIER = RID_ONEWAY,
+
+ RID_FIRST_AT = RID_AT_ENCODE,
+ RID_LAST_AT = RID_AT_IMPLEMENTATION,
+ RID_FIRST_PQ = RID_IN,
+ RID_LAST_PQ = RID_ONEWAY
};
+
+#define OBJC_IS_AT_KEYWORD(rid) \
+ ((unsigned int)(rid) >= (unsigned int)RID_FIRST_AT && \
+ (unsigned int)(rid) <= (unsigned int)RID_LAST_AT)
+
+#define OBJC_IS_PQ_KEYWORD(rid) \
+ ((unsigned int)(rid) >= (unsigned int)RID_FIRST_PQ && \
+ (unsigned int)(rid) <= (unsigned int)RID_LAST_PQ)
/* The elements of `ridpointers' are identifier nodes for the reserved
type names and storage classes. It is indexed by a RID_... value. */
===================================================================
Index: c-parse.in
--- c-parse.in 2001/06/10 13:47:54 1.94
+++ c-parse.in 2001/06/28 23:58:54
@@ -157,11 +157,6 @@ end ifc
%token INTERFACE IMPLEMENTATION END SELECTOR DEFS ENCODE
%token CLASSNAME PUBLIC PRIVATE PROTECTED PROTOCOL OBJECTNAME CLASS ALIAS
-/* Objective-C string constants in raw form.
- yylval is an STRING_CST node. */
-%token OBJC_STRING
-
-
%type <code> unop
%type <ttype> ENUM STRUCT UNION IF ELSE WHILE DO FOR SWITCH CASE DEFAULT
%type <ttype> BREAK CONTINUE RETURN GOTO ASM_KEYWORD SIZEOF TYPEOF ALIGNOF
@@ -231,7 +226,7 @@ ifobjc
%type <ttype> selectorarg keywordnamelist keywordname objcencodeexpr
%type <ttype> objc_string non_empty_protocolrefs protocolrefs identifier_list objcprotocolexpr
-%type <ttype> CLASSNAME OBJC_STRING OBJECTNAME
+%type <ttype> CLASSNAME OBJECTNAME
end ifobjc
%{
@@ -767,9 +762,10 @@ ifobjc
/* Produces an STRING_CST with perhaps more STRING_CSTs chained
onto it, which is to be read as an ObjC string object. */
objc_string:
- OBJC_STRING
- | objc_string OBJC_STRING
- { $$ = chainon ($1, $2); }
+ '@' STRING
+ { $$ = $2; }
+ | objc_string '@' STRING
+ { $$ = chainon ($1, $3); }
;
end ifobjc
@@ -3407,19 +3403,25 @@ static const struct resword reswords[] =
{ "volatile", RID_VOLATILE, D_TRAD },
{ "while", RID_WHILE, 0 },
ifobjc
- { "@class", RID_AT_CLASS, D_OBJC },
- { "@compatibility_alias", RID_AT_ALIAS, D_OBJC },
- { "@defs", RID_AT_DEFS, D_OBJC },
- { "@encode", RID_AT_ENCODE, D_OBJC },
- { "@end", RID_AT_END, D_OBJC },
- { "@implementation", RID_AT_IMPLEMENTATION, D_OBJC },
- { "@interface", RID_AT_INTERFACE, D_OBJC },
- { "@private", RID_AT_PRIVATE, D_OBJC },
- { "@protected", RID_AT_PROTECTED, D_OBJC },
- { "@protocol", RID_AT_PROTOCOL, D_OBJC },
- { "@public", RID_AT_PUBLIC, D_OBJC },
- { "@selector", RID_AT_SELECTOR, D_OBJC },
{ "id", RID_ID, D_OBJC },
+
+ /* These objc keywords are recognized only immediately after
+ an '@'. */
+ { "class", RID_AT_CLASS, D_OBJC },
+ { "compatibility_alias", RID_AT_ALIAS, D_OBJC },
+ { "defs", RID_AT_DEFS, D_OBJC },
+ { "encode", RID_AT_ENCODE, D_OBJC },
+ { "end", RID_AT_END, D_OBJC },
+ { "implementation", RID_AT_IMPLEMENTATION, D_OBJC },
+ { "interface", RID_AT_INTERFACE, D_OBJC },
+ { "private", RID_AT_PRIVATE, D_OBJC },
+ { "protected", RID_AT_PROTECTED, D_OBJC },
+ { "protocol", RID_AT_PROTOCOL, D_OBJC },
+ { "public", RID_AT_PUBLIC, D_OBJC },
+ { "selector", RID_AT_SELECTOR, D_OBJC },
+
+ /* These are recognized only in protocol-qualifier context
+ (see above) */
{ "bycopy", RID_BYCOPY, D_OBJC },
{ "byref", RID_BYREF, D_OBJC },
{ "in", RID_IN, D_OBJC },
@@ -3568,13 +3570,6 @@ static const short rid_to_yy[RID_MAX] =
/* RID_AT_IMPLEMENTATION */ IMPLEMENTATION
};
-ifobjc
-/* Lookup table for ObjC keywords beginning with '@'. Crude but
- hopefully effective. */
-#define N_at_reswords ((int) RID_AT_IMPLEMENTATION - (int)RID_AT_ENCODE + 1)
-static tree objc_rid_sans_at[N_at_reswords];
-end ifobjc
-
static void
init_reswords ()
{
@@ -3602,16 +3597,6 @@ init_reswords ()
C_RID_CODE (id) = reswords[i].rid;
C_IS_RESERVED_WORD (id) = 1;
ridpointers [(int) reswords[i].rid] = id;
-
-ifobjc
- /* Enter ObjC @-prefixed keywords into the "sans" table
- _without_ their leading at-sign. Again, all these
- identifiers are reachable by the get_identifer table, so it's
- not necessary to make objc_rid_sans_at a GC root. */
- if (reswords[i].word[0] == '@')
- objc_rid_sans_at[(int) reswords[i].rid - (int) RID_AT_ENCODE]
- = get_identifier (reswords[i].word + 1);
-end ifobjc
}
}
@@ -3680,8 +3665,8 @@ yylexname ()
enum rid rid_code = C_RID_CODE (yylval.ttype);
ifobjc
- if (!((unsigned int) rid_code - (unsigned int) RID_FIRST_PQ < 6)
- || objc_pq_context)
+ if (!OBJC_IS_AT_KEYWORD (rid_code)
+ && (!OBJC_IS_PQ_KEYWORD (rid_code) || objc_pq_context))
end ifobjc
{
int yycode = rid_to_yy[(int) rid_code];
@@ -3730,9 +3715,6 @@ _yylex ()
{
get_next:
last_token = c_lex (&yylval.ttype);
-ifobjc
- reconsider:
-end ifobjc
switch (last_token)
{
case CPP_EQ: return '=';
@@ -3809,23 +3791,27 @@ end ifobjc
token special significance. */
case CPP_ATSIGN:
ifobjc
- last_token = c_lex (&yylval.ttype);
- if (last_token == CPP_STRING)
- return OBJC_STRING;
- else if (last_token == CPP_NAME)
- {
- int i;
- for (i = 0; i < N_at_reswords; i++)
- if (objc_rid_sans_at[i] == yylval.ttype)
- {
- int rid_code = i + (int) RID_AT_ENCODE;
- yylval.ttype = ridpointers[rid_code];
- return rid_to_yy[rid_code];
- }
- }
- error ("syntax error at '@' token");
- goto reconsider;
+ {
+ tree after_at;
+ enum cpp_ttype after_at_type;
+
+ cpp_start_lookahead (parse_in);
+ after_at_type = c_lex (&after_at);
+
+ if (after_at_type == CPP_NAME
+ && C_IS_RESERVED_WORD (after_at)
+ && OBJC_IS_AT_KEYWORD (C_RID_CODE (after_at)))
+ {
+ cpp_stop_lookahead (parse_in, 1); /* accept this token */
+ yylval.ttype = after_at;
+ last_token = after_at_type;
+ return rid_to_yy [(int) C_RID_CODE (after_at)];
+ }
+ cpp_stop_lookahead (parse_in, 0); /* put back this token */
+ return '@';
+ }
end ifobjc
+
/* These tokens are C++ specific (and will not be generated
in C mode, but let's be cautious). */
case CPP_SCOPE: