8387c24d79
Reviewed by: Geoff.
818 lines
16 KiB
Plaintext
818 lines
16 KiB
Plaintext
/* parse.y - parser for flex input */
|
|
|
|
%token CHAR NUMBER SECTEND SCDECL XSCDECL WHITESPACE NAME PREVCCL EOF_OP
|
|
|
|
%{
|
|
/*-
|
|
* Copyright (c) 1990 The Regents of the University of California.
|
|
* All rights reserved.
|
|
*
|
|
* This code is derived from software contributed to Berkeley by
|
|
* Vern Paxson.
|
|
*
|
|
* The United States Government has rights in this work pursuant
|
|
* to contract no. DE-AC03-76SF00098 between the United States
|
|
* Department of Energy and the University of California.
|
|
*
|
|
* Redistribution and use in source and binary forms are permitted provided
|
|
* that: (1) source distributions retain this entire copyright notice and
|
|
* comment, and (2) distributions including binaries display the following
|
|
* acknowledgement: ``This product includes software developed by the
|
|
* University of California, Berkeley and its contributors'' in the
|
|
* documentation or other materials provided with the distribution and in
|
|
* all advertising materials mentioning features or use of this software.
|
|
* Neither the name of the University nor the names of its contributors may
|
|
* be used to endorse or promote products derived from this software without
|
|
* specific prior written permission.
|
|
* THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED
|
|
* WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
|
|
* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
|
|
*/
|
|
|
|
/* $Header: /home/daffy/u0/vern/flex/RCS/parse.y,v 2.15 93/12/09 13:57:23 vern Exp $ */
|
|
|
|
|
|
/* Some versions of bison are broken in that they use alloca() but don't
|
|
* declare it properly. The following is the patented (just kidding!)
|
|
* #ifdef chud to fix the problem, courtesy of Francois Pinard.
|
|
*/
|
|
#ifdef YYBISON
|
|
/* AIX requires this to be the first thing in the file. */
|
|
#ifdef __GNUC__
|
|
#define alloca __builtin_alloca
|
|
#else /* not __GNUC__ */
|
|
#if HAVE_ALLOCA_H
|
|
#include <alloca.h>
|
|
#else /* not HAVE_ALLOCA_H */
|
|
#ifdef _AIX
|
|
#pragma alloca
|
|
#else /* not _AIX */
|
|
char *alloca ();
|
|
#endif /* not _AIX */
|
|
#endif /* not HAVE_ALLOCA_H */
|
|
#endif /* not __GNUC__ */
|
|
#endif /* YYBISON */
|
|
|
|
/* Bletch, ^^^^ that was ugly! */
|
|
|
|
|
|
#include "flexdef.h"
|
|
|
|
int pat, scnum, eps, headcnt, trailcnt, anyccl, lastchar, i, actvp, rulelen;
|
|
int trlcontxt, xcluflg, cclsorted, varlength, variable_trail_rule;
|
|
int *active_ss;
|
|
Char clower();
|
|
void build_eof_action();
|
|
void yyerror();
|
|
|
|
static int madeany = false; /* whether we've made the '.' character class */
|
|
int previous_continued_action; /* whether the previous rule's action was '|' */
|
|
|
|
/* On some over-ambitious machines, such as DEC Alpha's, the default
|
|
* token type is "long" instead of "int"; this leads to problems with
|
|
* declaring yylval in flexdef.h. But so far, all the yacc's I've seen
|
|
* wrap their definitions of YYSTYPE with "#ifndef YYSTYPE"'s, so the
|
|
* following should ensure that the default token type is "int".
|
|
*/
|
|
#define YYSTYPE int
|
|
|
|
%}
|
|
|
|
%%
|
|
goal : initlex sect1 sect1end sect2 initforrule
|
|
{ /* add default rule */
|
|
int def_rule;
|
|
|
|
pat = cclinit();
|
|
cclnegate( pat );
|
|
|
|
def_rule = mkstate( -pat );
|
|
|
|
/* Remember the number of the default rule so we
|
|
* don't generate "can't match" warnings for it.
|
|
*/
|
|
default_rule = num_rules;
|
|
|
|
finish_rule( def_rule, false, 0, 0 );
|
|
|
|
for ( i = 1; i <= lastsc; ++i )
|
|
scset[i] = mkbranch( scset[i], def_rule );
|
|
|
|
if ( spprdflt )
|
|
add_action(
|
|
"YY_FATAL_ERROR( \"flex scanner jammed\" )" );
|
|
else
|
|
add_action( "ECHO" );
|
|
|
|
add_action( ";\n\tYY_BREAK\n" );
|
|
}
|
|
;
|
|
|
|
initlex :
|
|
{ /* initialize for processing rules */
|
|
|
|
/* Create default DFA start condition. */
|
|
scinstal( "INITIAL", false );
|
|
|
|
/* Initially, the start condition scoping is
|
|
* "no start conditions active".
|
|
*/
|
|
actvp = 0;
|
|
}
|
|
;
|
|
|
|
sect1 : sect1 startconddecl WHITESPACE namelist1 '\n'
|
|
|
|
|
| error '\n'
|
|
{ synerr( "unknown error processing section 1" ); }
|
|
;
|
|
|
|
sect1end : SECTEND
|
|
{
|
|
/* We now know how many start conditions there
|
|
* are, so create the "activity" map indicating
|
|
* which conditions are active.
|
|
*/
|
|
active_ss = allocate_integer_array( lastsc + 1 );
|
|
|
|
for ( i = 1; i <= lastsc; ++i )
|
|
active_ss[i] = 0;
|
|
}
|
|
;
|
|
|
|
startconddecl : SCDECL
|
|
{ xcluflg = false; }
|
|
|
|
| XSCDECL
|
|
{ xcluflg = true; }
|
|
;
|
|
|
|
namelist1 : namelist1 WHITESPACE NAME
|
|
{ scinstal( nmstr, xcluflg ); }
|
|
|
|
| NAME
|
|
{ scinstal( nmstr, xcluflg ); }
|
|
|
|
| error
|
|
{ synerr( "bad start condition list" ); }
|
|
;
|
|
|
|
sect2 : sect2 initforrule flexrule '\n'
|
|
|
|
|
;
|
|
|
|
initforrule :
|
|
{
|
|
/* Initialize for a parse of one rule. */
|
|
trlcontxt = variable_trail_rule = varlength = false;
|
|
trailcnt = headcnt = rulelen = 0;
|
|
current_state_type = STATE_NORMAL;
|
|
previous_continued_action = continued_action;
|
|
new_rule();
|
|
}
|
|
;
|
|
|
|
flexrule : scon '^' rule
|
|
{
|
|
pat = $3;
|
|
finish_rule( pat, variable_trail_rule,
|
|
headcnt, trailcnt );
|
|
|
|
for ( i = 1; i <= actvp; ++i )
|
|
scbol[actvsc[i]] =
|
|
mkbranch( scbol[actvsc[i]], pat );
|
|
|
|
if ( ! bol_needed )
|
|
{
|
|
bol_needed = true;
|
|
|
|
if ( performance_report > 1 )
|
|
pinpoint_message(
|
|
"'^' operator results in sub-optimal performance" );
|
|
}
|
|
}
|
|
|
|
| scon rule
|
|
{
|
|
pat = $2;
|
|
finish_rule( pat, variable_trail_rule,
|
|
headcnt, trailcnt );
|
|
|
|
for ( i = 1; i <= actvp; ++i )
|
|
scset[actvsc[i]] =
|
|
mkbranch( scset[actvsc[i]], pat );
|
|
}
|
|
|
|
| '^' rule
|
|
{
|
|
pat = $2;
|
|
finish_rule( pat, variable_trail_rule,
|
|
headcnt, trailcnt );
|
|
|
|
/* Add to all non-exclusive start conditions,
|
|
* including the default (0) start condition.
|
|
*/
|
|
|
|
for ( i = 1; i <= lastsc; ++i )
|
|
if ( ! scxclu[i] )
|
|
scbol[i] = mkbranch( scbol[i], pat );
|
|
|
|
if ( ! bol_needed )
|
|
{
|
|
bol_needed = true;
|
|
|
|
if ( performance_report > 1 )
|
|
pinpoint_message(
|
|
"'^' operator results in sub-optimal performance" );
|
|
}
|
|
}
|
|
|
|
| rule
|
|
{
|
|
pat = $1;
|
|
finish_rule( pat, variable_trail_rule,
|
|
headcnt, trailcnt );
|
|
|
|
for ( i = 1; i <= lastsc; ++i )
|
|
if ( ! scxclu[i] )
|
|
scset[i] = mkbranch( scset[i], pat );
|
|
}
|
|
|
|
| scon EOF_OP
|
|
{ build_eof_action(); }
|
|
|
|
| EOF_OP
|
|
{
|
|
/* This EOF applies to all start conditions
|
|
* which don't already have EOF actions.
|
|
*/
|
|
actvp = 0;
|
|
|
|
for ( i = 1; i <= lastsc; ++i )
|
|
if ( ! sceof[i] )
|
|
actvsc[++actvp] = i;
|
|
|
|
if ( actvp == 0 )
|
|
warn(
|
|
"all start conditions already have <<EOF>> rules" );
|
|
|
|
else
|
|
build_eof_action();
|
|
}
|
|
|
|
| error
|
|
{ synerr( "unrecognized rule" ); }
|
|
;
|
|
|
|
scon : '<' namelist2 '>'
|
|
|
|
| '<' '*' '>'
|
|
{
|
|
actvp = 0;
|
|
|
|
for ( i = 1; i <= lastsc; ++i )
|
|
actvsc[++actvp] = i;
|
|
}
|
|
;
|
|
|
|
namelist2 : namelist2 ',' sconname
|
|
|
|
| { actvp = 0; } sconname
|
|
|
|
| error
|
|
{ synerr( "bad start condition list" ); }
|
|
;
|
|
|
|
sconname : NAME
|
|
{
|
|
if ( (scnum = sclookup( nmstr )) == 0 )
|
|
format_pinpoint_message(
|
|
"undeclared start condition %s",
|
|
nmstr );
|
|
else
|
|
{
|
|
if ( ++actvp >= current_max_scs )
|
|
/* Some bozo has included multiple
|
|
* instances of start condition names.
|
|
*/
|
|
pinpoint_message(
|
|
"too many start conditions in <> construct!" );
|
|
|
|
else
|
|
actvsc[actvp] = scnum;
|
|
}
|
|
}
|
|
;
|
|
|
|
rule : re2 re
|
|
{
|
|
if ( transchar[lastst[$2]] != SYM_EPSILON )
|
|
/* Provide final transition \now/ so it
|
|
* will be marked as a trailing context
|
|
* state.
|
|
*/
|
|
$2 = link_machines( $2,
|
|
mkstate( SYM_EPSILON ) );
|
|
|
|
mark_beginning_as_normal( $2 );
|
|
current_state_type = STATE_NORMAL;
|
|
|
|
if ( previous_continued_action )
|
|
{
|
|
/* We need to treat this as variable trailing
|
|
* context so that the backup does not happen
|
|
* in the action but before the action switch
|
|
* statement. If the backup happens in the
|
|
* action, then the rules "falling into" this
|
|
* one's action will *also* do the backup,
|
|
* erroneously.
|
|
*/
|
|
if ( ! varlength || headcnt != 0 )
|
|
warn(
|
|
"trailing context made variable due to preceding '|' action" );
|
|
|
|
/* Mark as variable. */
|
|
varlength = true;
|
|
headcnt = 0;
|
|
}
|
|
|
|
if ( lex_compat || (varlength && headcnt == 0) )
|
|
{ /* variable trailing context rule */
|
|
/* Mark the first part of the rule as the
|
|
* accepting "head" part of a trailing
|
|
* context rule.
|
|
*
|
|
* By the way, we didn't do this at the
|
|
* beginning of this production because back
|
|
* then current_state_type was set up for a
|
|
* trail rule, and add_accept() can create
|
|
* a new state ...
|
|
*/
|
|
add_accept( $1,
|
|
num_rules | YY_TRAILING_HEAD_MASK );
|
|
variable_trail_rule = true;
|
|
}
|
|
|
|
else
|
|
trailcnt = rulelen;
|
|
|
|
$$ = link_machines( $1, $2 );
|
|
}
|
|
|
|
| re2 re '$'
|
|
{ synerr( "trailing context used twice" ); }
|
|
|
|
| re '$'
|
|
{
|
|
headcnt = 0;
|
|
trailcnt = 1;
|
|
rulelen = 1;
|
|
varlength = false;
|
|
|
|
current_state_type = STATE_TRAILING_CONTEXT;
|
|
|
|
if ( trlcontxt )
|
|
{
|
|
synerr( "trailing context used twice" );
|
|
$$ = mkstate( SYM_EPSILON );
|
|
}
|
|
|
|
else if ( previous_continued_action )
|
|
{
|
|
/* See the comment in the rule for "re2 re"
|
|
* above.
|
|
*/
|
|
warn(
|
|
"trailing context made variable due to preceding '|' action" );
|
|
|
|
varlength = true;
|
|
}
|
|
|
|
if ( lex_compat || varlength )
|
|
{
|
|
/* Again, see the comment in the rule for
|
|
* "re2 re" above.
|
|
*/
|
|
add_accept( $1,
|
|
num_rules | YY_TRAILING_HEAD_MASK );
|
|
variable_trail_rule = true;
|
|
}
|
|
|
|
trlcontxt = true;
|
|
|
|
eps = mkstate( SYM_EPSILON );
|
|
$$ = link_machines( $1,
|
|
link_machines( eps, mkstate( '\n' ) ) );
|
|
}
|
|
|
|
| re
|
|
{
|
|
$$ = $1;
|
|
|
|
if ( trlcontxt )
|
|
{
|
|
if ( lex_compat || (varlength && headcnt == 0) )
|
|
/* Both head and trail are
|
|
* variable-length.
|
|
*/
|
|
variable_trail_rule = true;
|
|
else
|
|
trailcnt = rulelen;
|
|
}
|
|
}
|
|
;
|
|
|
|
|
|
re : re '|' series
|
|
{
|
|
varlength = true;
|
|
$$ = mkor( $1, $3 );
|
|
}
|
|
|
|
| series
|
|
{ $$ = $1; }
|
|
;
|
|
|
|
|
|
re2 : re '/'
|
|
{
|
|
/* This rule is written separately so the
|
|
* reduction will occur before the trailing
|
|
* series is parsed.
|
|
*/
|
|
|
|
if ( trlcontxt )
|
|
synerr( "trailing context used twice" );
|
|
else
|
|
trlcontxt = true;
|
|
|
|
if ( varlength )
|
|
/* We hope the trailing context is
|
|
* fixed-length.
|
|
*/
|
|
varlength = false;
|
|
else
|
|
headcnt = rulelen;
|
|
|
|
rulelen = 0;
|
|
|
|
current_state_type = STATE_TRAILING_CONTEXT;
|
|
$$ = $1;
|
|
}
|
|
;
|
|
|
|
series : series singleton
|
|
{
|
|
/* This is where concatenation of adjacent patterns
|
|
* gets done.
|
|
*/
|
|
$$ = link_machines( $1, $2 );
|
|
}
|
|
|
|
| singleton
|
|
{ $$ = $1; }
|
|
;
|
|
|
|
singleton : singleton '*'
|
|
{
|
|
varlength = true;
|
|
|
|
$$ = mkclos( $1 );
|
|
}
|
|
|
|
| singleton '+'
|
|
{
|
|
varlength = true;
|
|
$$ = mkposcl( $1 );
|
|
}
|
|
|
|
| singleton '?'
|
|
{
|
|
varlength = true;
|
|
$$ = mkopt( $1 );
|
|
}
|
|
|
|
| singleton '{' NUMBER ',' NUMBER '}'
|
|
{
|
|
varlength = true;
|
|
|
|
if ( $3 > $5 || $3 < 0 )
|
|
{
|
|
synerr( "bad iteration values" );
|
|
$$ = $1;
|
|
}
|
|
else
|
|
{
|
|
if ( $3 == 0 )
|
|
{
|
|
if ( $5 <= 0 )
|
|
{
|
|
synerr(
|
|
"bad iteration values" );
|
|
$$ = $1;
|
|
}
|
|
else
|
|
$$ = mkopt(
|
|
mkrep( $1, 1, $5 ) );
|
|
}
|
|
else
|
|
$$ = mkrep( $1, $3, $5 );
|
|
}
|
|
}
|
|
|
|
| singleton '{' NUMBER ',' '}'
|
|
{
|
|
varlength = true;
|
|
|
|
if ( $3 <= 0 )
|
|
{
|
|
synerr( "iteration value must be positive" );
|
|
$$ = $1;
|
|
}
|
|
|
|
else
|
|
$$ = mkrep( $1, $3, INFINITY );
|
|
}
|
|
|
|
| singleton '{' NUMBER '}'
|
|
{
|
|
/* The singleton could be something like "(foo)",
|
|
* in which case we have no idea what its length
|
|
* is, so we punt here.
|
|
*/
|
|
varlength = true;
|
|
|
|
if ( $3 <= 0 )
|
|
{
|
|
synerr( "iteration value must be positive" );
|
|
$$ = $1;
|
|
}
|
|
|
|
else
|
|
$$ = link_machines( $1,
|
|
copysingl( $1, $3 - 1 ) );
|
|
}
|
|
|
|
| '.'
|
|
{
|
|
if ( ! madeany )
|
|
{
|
|
/* Create the '.' character class. */
|
|
anyccl = cclinit();
|
|
ccladd( anyccl, '\n' );
|
|
cclnegate( anyccl );
|
|
|
|
if ( useecs )
|
|
mkeccl( ccltbl + cclmap[anyccl],
|
|
ccllen[anyccl], nextecm,
|
|
ecgroup, csize, csize );
|
|
|
|
madeany = true;
|
|
}
|
|
|
|
++rulelen;
|
|
|
|
$$ = mkstate( -anyccl );
|
|
}
|
|
|
|
| fullccl
|
|
{
|
|
if ( ! cclsorted )
|
|
/* Sort characters for fast searching. We
|
|
* use a shell sort since this list could
|
|
* be large.
|
|
*/
|
|
cshell( ccltbl + cclmap[$1], ccllen[$1], true );
|
|
|
|
if ( useecs )
|
|
mkeccl( ccltbl + cclmap[$1], ccllen[$1],
|
|
nextecm, ecgroup, csize, csize );
|
|
|
|
++rulelen;
|
|
|
|
$$ = mkstate( -$1 );
|
|
}
|
|
|
|
| PREVCCL
|
|
{
|
|
++rulelen;
|
|
|
|
$$ = mkstate( -$1 );
|
|
}
|
|
|
|
| '"' string '"'
|
|
{ $$ = $2; }
|
|
|
|
| '(' re ')'
|
|
{ $$ = $2; }
|
|
|
|
| CHAR
|
|
{
|
|
++rulelen;
|
|
|
|
if ( caseins && $1 >= 'A' && $1 <= 'Z' )
|
|
$1 = clower( $1 );
|
|
|
|
$$ = mkstate( $1 );
|
|
}
|
|
;
|
|
|
|
fullccl : '[' ccl ']'
|
|
{ $$ = $2; }
|
|
|
|
| '[' '^' ccl ']'
|
|
{
|
|
cclnegate( $3 );
|
|
$$ = $3;
|
|
}
|
|
;
|
|
|
|
ccl : ccl CHAR '-' CHAR
|
|
{
|
|
if ( caseins )
|
|
{
|
|
if ( $2 >= 'A' && $2 <= 'Z' )
|
|
$2 = clower( $2 );
|
|
if ( $4 >= 'A' && $4 <= 'Z' )
|
|
$4 = clower( $4 );
|
|
}
|
|
|
|
if ( $2 > $4 )
|
|
synerr( "negative range in character class" );
|
|
|
|
else
|
|
{
|
|
for ( i = $2; i <= $4; ++i )
|
|
ccladd( $1, i );
|
|
|
|
/* Keep track if this ccl is staying in
|
|
* alphabetical order.
|
|
*/
|
|
cclsorted = cclsorted && ($2 > lastchar);
|
|
lastchar = $4;
|
|
}
|
|
|
|
$$ = $1;
|
|
}
|
|
|
|
| ccl CHAR
|
|
{
|
|
if ( caseins && $2 >= 'A' && $2 <= 'Z' )
|
|
$2 = clower( $2 );
|
|
|
|
ccladd( $1, $2 );
|
|
cclsorted = cclsorted && ($2 > lastchar);
|
|
lastchar = $2;
|
|
$$ = $1;
|
|
}
|
|
|
|
|
|
|
{
|
|
cclsorted = true;
|
|
lastchar = 0;
|
|
$$ = cclinit();
|
|
}
|
|
;
|
|
|
|
string : string CHAR
|
|
{
|
|
if ( caseins && $2 >= 'A' && $2 <= 'Z' )
|
|
$2 = clower( $2 );
|
|
|
|
++rulelen;
|
|
|
|
$$ = link_machines( $1, mkstate( $2 ) );
|
|
}
|
|
|
|
|
|
|
{ $$ = mkstate( SYM_EPSILON ); }
|
|
;
|
|
|
|
%%
|
|
|
|
|
|
/* build_eof_action - build the "<<EOF>>" action for the active start
|
|
* conditions
|
|
*/
|
|
|
|
void build_eof_action()
|
|
{
|
|
register int i;
|
|
char action_text[MAXLINE];
|
|
|
|
for ( i = 1; i <= actvp; ++i )
|
|
{
|
|
if ( sceof[actvsc[i]] )
|
|
format_pinpoint_message(
|
|
"multiple <<EOF>> rules for start condition %s",
|
|
scname[actvsc[i]] );
|
|
|
|
else
|
|
{
|
|
sceof[actvsc[i]] = true;
|
|
sprintf( action_text, "case YY_STATE_EOF(%s):\n",
|
|
scname[actvsc[i]] );
|
|
add_action( action_text );
|
|
}
|
|
}
|
|
|
|
line_directive_out( (FILE *) 0 );
|
|
|
|
/* This isn't a normal rule after all - don't count it as
|
|
* such, so we don't have any holes in the rule numbering
|
|
* (which make generating "rule can never match" warnings
|
|
* more difficult.
|
|
*/
|
|
--num_rules;
|
|
++num_eof_rules;
|
|
}
|
|
|
|
|
|
/* format_synerr - write out formatted syntax error */
|
|
|
|
void format_synerr( msg, arg )
|
|
char msg[], arg[];
|
|
{
|
|
char errmsg[MAXLINE];
|
|
|
|
(void) sprintf( errmsg, msg, arg );
|
|
synerr( errmsg );
|
|
}
|
|
|
|
|
|
/* synerr - report a syntax error */
|
|
|
|
void synerr( str )
|
|
char str[];
|
|
{
|
|
syntaxerror = true;
|
|
pinpoint_message( str );
|
|
}
|
|
|
|
|
|
/* warn - report a warning, unless -w was given */
|
|
|
|
void warn( str )
|
|
char str[];
|
|
{
|
|
line_warning( str, linenum );
|
|
}
|
|
|
|
/* format_pinpoint_message - write out a message formatted with one string,
|
|
* pinpointing its location
|
|
*/
|
|
|
|
void format_pinpoint_message( msg, arg )
|
|
char msg[], arg[];
|
|
{
|
|
char errmsg[MAXLINE];
|
|
|
|
(void) sprintf( errmsg, msg, arg );
|
|
pinpoint_message( errmsg );
|
|
}
|
|
|
|
|
|
/* pinpoint_message - write out a message, pinpointing its location */
|
|
|
|
void pinpoint_message( str )
|
|
char str[];
|
|
{
|
|
line_pinpoint( str, linenum );
|
|
}
|
|
|
|
|
|
/* line_warning - report a warning at a given line, unless -w was given */
|
|
|
|
void line_warning( str, line )
|
|
char str[];
|
|
int line;
|
|
{
|
|
char warning[MAXLINE];
|
|
|
|
if ( ! nowarn )
|
|
{
|
|
sprintf( warning, "warning, %s", str );
|
|
line_pinpoint( warning, line );
|
|
}
|
|
}
|
|
|
|
|
|
/* line_pinpoint - write out a message, pinpointing it at the given line */
|
|
|
|
void line_pinpoint( str, line )
|
|
char str[];
|
|
int line;
|
|
{
|
|
fprintf( stderr, "\"%s\", line %d: %s\n", infilename, line, str );
|
|
}
|
|
|
|
|
|
/* yyerror - eat up an error message from the parser;
|
|
* currently, messages are ignore
|
|
*/
|
|
|
|
void yyerror( msg )
|
|
char msg[];
|
|
{
|
|
}
|