1 /* parse.y - parser for flex input */
3 %token CHAR NUMBER SECTEND SCDECL XSCDECL NAME PREVCCL EOF_OP
4 %token OPTION_OP OPT_OUTFILE OPT_PREFIX OPT_YYCLASS OPT_HEADER OPT_EXTRA_TYPE
7 %token CCE_ALNUM CCE_ALPHA CCE_BLANK CCE_CNTRL CCE_DIGIT CCE_GRAPH
8 %token CCE_LOWER CCE_PRINT CCE_PUNCT CCE_SPACE CCE_UPPER CCE_XDIGIT
10 %token CCE_NEG_ALNUM CCE_NEG_ALPHA CCE_NEG_BLANK CCE_NEG_CNTRL CCE_NEG_DIGIT CCE_NEG_GRAPH
11 %token CCE_NEG_LOWER CCE_NEG_PRINT CCE_NEG_PUNCT CCE_NEG_SPACE CCE_NEG_UPPER CCE_NEG_XDIGIT
13 %left CCL_OP_DIFF CCL_OP_UNION
16 *POSIX and AT&T lex place the
17 * precedence of the repeat operator, {}, below that of concatenation.
18 * Thus, ab{3} is ababab. Most other POSIX utilities use an Extended
19 * Regular Expression (ERE) precedence that has the repeat operator
20 * higher than concatenation. This causes ab{3} to yield abbb.
22 * In order to support the POSIX and AT&T precedence and the flex
23 * precedence we define two token sets for the begin and end tokens of
24 * the repeat operator, '{' and '}'. The lexical scanner chooses
25 * which tokens to return based on whether posix_compat or lex_compat
26 * are specified. Specifying either posix_compat or lex_compat will
27 * cause flex to parse scanner files as per the AT&T and
28 * POSIX-mandated behavior.
31 %token BEGIN_REPEAT_POSIX END_REPEAT_POSIX BEGIN_REPEAT_FLEX END_REPEAT_FLEX
35 /* Copyright (c) 1990 The Regents of the University of California. */
36 /* All rights reserved. */
38 /* This code is derived from software contributed to Berkeley by */
41 /* The United States Government has rights in this work pursuant */
42 /* to contract no. DE-AC03-76SF00098 between the United States */
43 /* Department of Energy and the University of California. */
45 /* This file is part of flex. */
47 /* Redistribution and use in source and binary forms, with or without */
48 /* modification, are permitted provided that the following conditions */
51 /* 1. Redistributions of source code must retain the above copyright */
52 /* notice, this list of conditions and the following disclaimer. */
53 /* 2. Redistributions in binary form must reproduce the above copyright */
54 /* notice, this list of conditions and the following disclaimer in the */
55 /* documentation and/or other materials provided with the distribution. */
57 /* Neither the name of the University nor the names of its contributors */
58 /* may be used to endorse or promote products derived from this software */
59 /* without specific prior written permission. */
61 /* THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR */
62 /* IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED */
63 /* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR */
69 int pat, scnum, eps, headcnt, trailcnt, lastchar, i, rulelen;
70 int trlcontxt, xcluflg, currccl, cclsorted, varlength, variable_trail_rule;
75 static int madeany = false; /* whether we've made the '.' character class */
76 static int ccldot, cclany;
77 int previous_continued_action; /* whether the previous rule's action was '|' */
79 #define format_warn3(fmt, a1, a2) \
81 char fw3_msg[MAXLINE];\
82 snprintf( fw3_msg, MAXLINE,(fmt), (a1), (a2) );\
86 /* Expand a POSIX character class expression. */
87 #define CCL_EXPR(func) \
90 for ( c = 0; c < csize; ++c ) \
91 if ( isascii(c) && func(c) ) \
92 ccladd( currccl, c ); \
96 #define CCL_NEG_EXPR(func) \
99 for ( c = 0; c < csize; ++c ) \
101 ccladd( currccl, c ); \
104 /* While POSIX defines isblank(), it's not ANSI C. */
105 #define IS_BLANK(c) ((c) == ' ' || (c) == '\t')
107 /* On some over-ambitious machines, such as DEC Alpha's, the default
108 * token type is "long" instead of "int"; this leads to problems with
109 * declaring yylval in flexdef.h. But so far, all the yacc's I've seen
110 * wrap their definitions of YYSTYPE with "#ifndef YYSTYPE"'s, so the
111 * following should ensure that the default token type is "int".
118 goal : initlex sect1 sect1end sect2 initforrule
119 { /* add default rule */
125 def_rule = mkstate( -pat );
127 /* Remember the number of the default rule so we
128 * don't generate "can't match" warnings for it.
130 default_rule = num_rules;
132 finish_rule( def_rule, false, 0, 0, 0);
134 for ( i = 1; i <= lastsc; ++i )
135 scset[i] = mkbranch( scset[i], def_rule );
139 "YY_FATAL_ERROR( \"flex scanner jammed\" )" );
141 add_action( "ECHO" );
143 add_action( ";\n\tYY_BREAK\n" );
148 { /* initialize for processing rules */
150 /* Create default DFA start condition. */
151 scinstal( "INITIAL", false );
155 sect1 : sect1 startconddecl namelist1
159 { synerr( _("unknown error processing section 1") ); }
165 scon_stk = allocate_integer_array( lastsc + 1 );
170 startconddecl : SCDECL
177 namelist1 : namelist1 NAME
178 { scinstal( nmstr, xcluflg ); }
181 { scinstal( nmstr, xcluflg ); }
184 { synerr( _("bad start condition list") ); }
187 options : OPTION_OP optionlist
190 optionlist : optionlist option
194 option : OPT_OUTFILE '=' NAME
196 outfilename = copy_string( nmstr );
199 | OPT_EXTRA_TYPE '=' NAME
200 { extra_type = copy_string( nmstr ); }
201 | OPT_PREFIX '=' NAME
202 { prefix = copy_string( nmstr ); }
203 | OPT_YYCLASS '=' NAME
204 { yyclass = copy_string( nmstr ); }
205 | OPT_HEADER '=' NAME
206 { headerfilename = copy_string( nmstr ); }
207 | OPT_TABLES '=' NAME
208 { tablesext = true; tablesfilename = copy_string( nmstr ); }
211 sect2 : sect2 scon initforrule flexrule '\n'
212 { scon_stk_ptr = $2; }
213 | sect2 scon '{' sect2 '}'
214 { scon_stk_ptr = $2; }
220 /* Initialize for a parse of one rule. */
221 trlcontxt = variable_trail_rule = varlength = false;
222 trailcnt = headcnt = rulelen = 0;
223 current_state_type = STATE_NORMAL;
224 previous_continued_action = continued_action;
234 finish_rule( pat, variable_trail_rule,
235 headcnt, trailcnt , previous_continued_action);
237 if ( scon_stk_ptr > 0 )
239 for ( i = 1; i <= scon_stk_ptr; ++i )
241 mkbranch( scbol[scon_stk[i]],
247 /* Add to all non-exclusive start conditions,
248 * including the default (0) start condition.
251 for ( i = 1; i <= lastsc; ++i )
253 scbol[i] = mkbranch( scbol[i],
261 if ( performance_report > 1 )
263 "'^' operator results in sub-optimal performance" );
270 finish_rule( pat, variable_trail_rule,
271 headcnt, trailcnt , previous_continued_action);
273 if ( scon_stk_ptr > 0 )
275 for ( i = 1; i <= scon_stk_ptr; ++i )
277 mkbranch( scset[scon_stk[i]],
283 for ( i = 1; i <= lastsc; ++i )
293 if ( scon_stk_ptr > 0 )
298 /* This EOF applies to all start conditions
299 * which don't already have EOF actions.
301 for ( i = 1; i <= lastsc; ++i )
303 scon_stk[++scon_stk_ptr] = i;
305 if ( scon_stk_ptr == 0 )
307 "all start conditions already have <<EOF>> rules" );
315 { synerr( _("unrecognized rule") ); }
319 { $$ = scon_stk_ptr; }
322 scon : '<' scon_stk_ptr namelist2 '>'
329 for ( i = 1; i <= lastsc; ++i )
333 for ( j = 1; j <= scon_stk_ptr; ++j )
334 if ( scon_stk[j] == i )
337 if ( j > scon_stk_ptr )
338 scon_stk[++scon_stk_ptr] = i;
343 { $$ = scon_stk_ptr; }
346 namelist2 : namelist2 ',' sconname
351 { synerr( _("bad start condition list") ); }
356 if ( (scnum = sclookup( nmstr )) == 0 )
357 format_pinpoint_message(
358 "undeclared start condition %s",
362 for ( i = 1; i <= scon_stk_ptr; ++i )
363 if ( scon_stk[i] == scnum )
366 "<%s> specified twice",
371 if ( i > scon_stk_ptr )
372 scon_stk[++scon_stk_ptr] = scnum;
379 if ( transchar[lastst[$2]] != SYM_EPSILON )
380 /* Provide final transition \now/ so it
381 * will be marked as a trailing context
384 $2 = link_machines( $2,
385 mkstate( SYM_EPSILON ) );
387 mark_beginning_as_normal( $2 );
388 current_state_type = STATE_NORMAL;
390 if ( previous_continued_action )
392 /* We need to treat this as variable trailing
393 * context so that the backup does not happen
394 * in the action but before the action switch
395 * statement. If the backup happens in the
396 * action, then the rules "falling into" this
397 * one's action will *also* do the backup,
400 if ( ! varlength || headcnt != 0 )
402 "trailing context made variable due to preceding '|' action" );
404 /* Mark as variable. */
410 if ( lex_compat || (varlength && headcnt == 0) )
411 { /* variable trailing context rule */
412 /* Mark the first part of the rule as the
413 * accepting "head" part of a trailing
416 * By the way, we didn't do this at the
417 * beginning of this production because back
418 * then current_state_type was set up for a
419 * trail rule, and add_accept() can create
423 num_rules | YY_TRAILING_HEAD_MASK );
424 variable_trail_rule = true;
430 $$ = link_machines( $1, $2 );
434 { synerr( _("trailing context used twice") ); }
443 current_state_type = STATE_TRAILING_CONTEXT;
447 synerr( _("trailing context used twice") );
448 $$ = mkstate( SYM_EPSILON );
451 else if ( previous_continued_action )
453 /* See the comment in the rule for "re2 re"
457 "trailing context made variable due to preceding '|' action" );
462 if ( lex_compat || varlength )
464 /* Again, see the comment in the rule for
468 num_rules | YY_TRAILING_HEAD_MASK );
469 variable_trail_rule = true;
474 eps = mkstate( SYM_EPSILON );
475 $$ = link_machines( $1,
476 link_machines( eps, mkstate( '\n' ) ) );
485 if ( lex_compat || (varlength && headcnt == 0) )
486 /* Both head and trail are
489 variable_trail_rule = true;
510 /* This rule is written separately so the
511 * reduction will occur before the trailing
516 synerr( _("trailing context used twice") );
521 /* We hope the trailing context is
530 current_state_type = STATE_TRAILING_CONTEXT;
535 series : series singleton
537 /* This is where concatenation of adjacent patterns
540 $$ = link_machines( $1, $2 );
546 | series BEGIN_REPEAT_POSIX NUMBER ',' NUMBER END_REPEAT_POSIX
550 if ( $3 > $5 || $3 < 0 )
552 synerr( _("bad iteration values") );
562 _("bad iteration values") );
567 mkrep( $1, 1, $5 ) );
570 $$ = mkrep( $1, $3, $5 );
574 | series BEGIN_REPEAT_POSIX NUMBER ',' END_REPEAT_POSIX
580 synerr( _("iteration value must be positive") );
585 $$ = mkrep( $1, $3, INFINITE_REPEAT );
588 | series BEGIN_REPEAT_POSIX NUMBER END_REPEAT_POSIX
590 /* The series could be something like "(foo)",
591 * in which case we have no idea what its length
592 * is, so we punt here.
598 synerr( _("iteration value must be positive")
604 $$ = link_machines( $1,
605 copysingl( $1, $3 - 1 ) );
610 singleton : singleton '*'
629 | singleton BEGIN_REPEAT_FLEX NUMBER ',' NUMBER END_REPEAT_FLEX
633 if ( $3 > $5 || $3 < 0 )
635 synerr( _("bad iteration values") );
645 _("bad iteration values") );
650 mkrep( $1, 1, $5 ) );
653 $$ = mkrep( $1, $3, $5 );
657 | singleton BEGIN_REPEAT_FLEX NUMBER ',' END_REPEAT_FLEX
663 synerr( _("iteration value must be positive") );
668 $$ = mkrep( $1, $3, INFINITE_REPEAT );
671 | singleton BEGIN_REPEAT_FLEX NUMBER END_REPEAT_FLEX
673 /* The singleton could be something like "(foo)",
674 * in which case we have no idea what its length
675 * is, so we punt here.
681 synerr( _("iteration value must be positive") );
686 $$ = link_machines( $1,
687 copysingl( $1, $3 - 1 ) );
694 /* Create the '.' character class. */
696 ccladd( ccldot, '\n' );
700 mkeccl( ccltbl + cclmap[ccldot],
701 ccllen[ccldot], nextecm,
702 ecgroup, csize, csize );
704 /* Create the (?s:'.') character class. */
709 mkeccl( ccltbl + cclmap[cclany],
710 ccllen[cclany], nextecm,
711 ecgroup, csize, csize );
719 $$ = mkstate( -cclany );
721 $$ = mkstate( -ccldot );
726 /* Sort characters for fast searching.
728 qsort( ccltbl + cclmap[$1], ccllen[$1], sizeof (*ccltbl), cclcmp );
731 mkeccl( ccltbl + cclmap[$1], ccllen[$1],
732 nextecm, ecgroup, csize, csize );
737 rule_has_nl[num_rules] = true;
747 rule_has_nl[num_rules] = true;
763 rule_has_nl[num_rules] = true;
765 if (sf_case_ins() && has_case($1))
766 /* create an alternation, as in (a|A) */
767 $$ = mkor (mkstate($1), mkstate(reverse_case($1)));
773 fullccl CCL_OP_DIFF braceccl { $$ = ccl_set_diff ($1, $3); }
774 | fullccl CCL_OP_UNION braceccl { $$ = ccl_set_union ($1, $3); }
780 '[' ccl ']' { $$ = $2; }
789 ccl : ccl CHAR '-' CHAR
795 /* If one end of the range has case and the other
796 * does not, or the cases are different, then we're not
797 * sure what range the user is trying to express.
798 * Examples: [@-z] or [S-t]
800 if (has_case ($2) != has_case ($4)
801 || (has_case ($2) && (b_islower ($2) != b_islower ($4)))
802 || (has_case ($2) && (b_isupper ($2) != b_isupper ($4))))
804 _("the character range [%c-%c] is ambiguous in a case-insensitive scanner"),
807 /* If the range spans uppercase characters but not
808 * lowercase (or vice-versa), then should we automatically
809 * include lowercase characters in the range?
810 * Example: [@-_] spans [a-z] but not [A-Z]
812 else if (!has_case ($2) && !has_case ($4) && !range_covers_case ($2, $4))
814 _("the character range [%c-%c] is ambiguous in a case-insensitive scanner"),
819 synerr( _("negative range in character class") );
823 for ( i = $2; i <= $4; ++i )
826 /* Keep track if this ccl is staying in
827 * alphabetical order.
829 cclsorted = cclsorted && ($2 > lastchar);
832 /* Do it again for upper/lowercase */
833 if (sf_case_ins() && has_case($2) && has_case($4)){
834 $2 = reverse_case ($2);
835 $4 = reverse_case ($4);
837 for ( i = $2; i <= $4; ++i )
840 cclsorted = cclsorted && ($2 > lastchar);
852 cclsorted = cclsorted && ($2 > lastchar);
855 /* Do it again for upper/lowercase */
856 if (sf_case_ins() && has_case($2)){
857 $2 = reverse_case ($2);
860 cclsorted = cclsorted && ($2 > lastchar);
869 /* Too hard to properly maintain cclsorted. */
878 currccl = $$ = cclinit();
883 CCE_ALNUM { CCL_EXPR(isalnum); }
884 | CCE_ALPHA { CCL_EXPR(isalpha); }
885 | CCE_BLANK { CCL_EXPR(IS_BLANK); }
886 | CCE_CNTRL { CCL_EXPR(iscntrl); }
887 | CCE_DIGIT { CCL_EXPR(isdigit); }
888 | CCE_GRAPH { CCL_EXPR(isgraph); }
894 | CCE_PRINT { CCL_EXPR(isprint); }
895 | CCE_PUNCT { CCL_EXPR(ispunct); }
896 | CCE_SPACE { CCL_EXPR(isspace); }
897 | CCE_XDIGIT { CCL_EXPR(isxdigit); }
904 | CCE_NEG_ALNUM { CCL_NEG_EXPR(isalnum); }
905 | CCE_NEG_ALPHA { CCL_NEG_EXPR(isalpha); }
906 | CCE_NEG_BLANK { CCL_NEG_EXPR(IS_BLANK); }
907 | CCE_NEG_CNTRL { CCL_NEG_EXPR(iscntrl); }
908 | CCE_NEG_DIGIT { CCL_NEG_EXPR(isdigit); }
909 | CCE_NEG_GRAPH { CCL_NEG_EXPR(isgraph); }
910 | CCE_NEG_PRINT { CCL_NEG_EXPR(isprint); }
911 | CCE_NEG_PUNCT { CCL_NEG_EXPR(ispunct); }
912 | CCE_NEG_SPACE { CCL_NEG_EXPR(isspace); }
913 | CCE_NEG_XDIGIT { CCL_NEG_EXPR(isxdigit); }
916 warn(_("[:^lower:] is ambiguous in case insensitive scanner"));
918 CCL_NEG_EXPR(islower);
922 warn(_("[:^upper:] ambiguous in case insensitive scanner"));
924 CCL_NEG_EXPR(isupper);
931 rule_has_nl[num_rules] = true;
935 if (sf_case_ins() && has_case($2))
936 $$ = mkor (mkstate($2), mkstate(reverse_case($2)));
940 $$ = link_machines( $1, $$);
944 { $$ = mkstate( SYM_EPSILON ); }
950 /* build_eof_action - build the "<<EOF>>" action for the active start
954 void build_eof_action()
957 char action_text[MAXLINE];
959 for ( i = 1; i <= scon_stk_ptr; ++i )
961 if ( sceof[scon_stk[i]] )
962 format_pinpoint_message(
963 "multiple <<EOF>> rules for start condition %s",
964 scname[scon_stk[i]] );
968 sceof[scon_stk[i]] = true;
970 if (previous_continued_action /* && previous action was regular */)
971 add_action("YY_RULE_SETUP\n");
973 snprintf( action_text, sizeof(action_text), "case YY_STATE_EOF(%s):\n",
974 scname[scon_stk[i]] );
975 add_action( action_text );
979 line_directive_out( (FILE *) 0, 1 );
981 /* This isn't a normal rule after all - don't count it as
982 * such, so we don't have any holes in the rule numbering
983 * (which make generating "rule can never match" warnings
991 /* format_synerr - write out formatted syntax error */
993 void format_synerr( msg, arg )
994 const char *msg, arg[];
996 char errmsg[MAXLINE];
998 (void) snprintf( errmsg, sizeof(errmsg), msg, arg );
1003 /* synerr - report a syntax error */
1009 pinpoint_message( str );
1013 /* format_warn - write out formatted warning */
1015 void format_warn( msg, arg )
1016 const char *msg, arg[];
1018 char warn_msg[MAXLINE];
1020 snprintf( warn_msg, sizeof(warn_msg), msg, arg );
1025 /* warn - report a warning, unless -w was given */
1030 line_warning( str, linenum );
1033 /* format_pinpoint_message - write out a message formatted with one string,
1034 * pinpointing its location
1037 void format_pinpoint_message( msg, arg )
1038 const char *msg, arg[];
1040 char errmsg[MAXLINE];
1042 snprintf( errmsg, sizeof(errmsg), msg, arg );
1043 pinpoint_message( errmsg );
1047 /* pinpoint_message - write out a message, pinpointing its location */
1049 void pinpoint_message( str )
1052 line_pinpoint( str, linenum );
1056 /* line_warning - report a warning at a given line, unless -w was given */
1058 void line_warning( str, line )
1062 char warning[MAXLINE];
1066 snprintf( warning, sizeof(warning), "warning, %s", str );
1067 line_pinpoint( warning, line );
1072 /* line_pinpoint - write out a message, pinpointing it at the given line */
1074 void line_pinpoint( str, line )
1078 fprintf( stderr, "%s:%d: %s\n", infilename, line, str );
1082 /* yyerror - eat up an error message from the parser;
1083 * currently, messages are ignore