3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * "It all comes from here, the stench and the peril." --Frodo
16 * This file is the lexer for Perl. It's closely linked to the
19 * The main routine is yylex(), which returns the next token.
23 #define PERL_IN_TOKE_C
26 #define yylval (PL_parser->yylval)
28 /* YYINITDEPTH -- initial size of the parser's stacks. */
29 #define YYINITDEPTH 200
31 /* XXX temporary backwards compatibility */
32 #define PL_lex_brackets (PL_parser->lex_brackets)
33 #define PL_lex_brackstack (PL_parser->lex_brackstack)
34 #define PL_lex_casemods (PL_parser->lex_casemods)
35 #define PL_lex_casestack (PL_parser->lex_casestack)
36 #define PL_lex_defer (PL_parser->lex_defer)
37 #define PL_lex_dojoin (PL_parser->lex_dojoin)
38 #define PL_lex_expect (PL_parser->lex_expect)
39 #define PL_lex_formbrack (PL_parser->lex_formbrack)
40 #define PL_lex_inpat (PL_parser->lex_inpat)
41 #define PL_lex_inwhat (PL_parser->lex_inwhat)
42 #define PL_lex_op (PL_parser->lex_op)
43 #define PL_lex_repl (PL_parser->lex_repl)
44 #define PL_lex_starts (PL_parser->lex_starts)
45 #define PL_lex_stuff (PL_parser->lex_stuff)
46 #define PL_multi_start (PL_parser->multi_start)
47 #define PL_multi_open (PL_parser->multi_open)
48 #define PL_multi_close (PL_parser->multi_close)
49 #define PL_pending_ident (PL_parser->pending_ident)
50 #define PL_preambled (PL_parser->preambled)
51 #define PL_sublex_info (PL_parser->sublex_info)
54 # define PL_endwhite (PL_parser->endwhite)
55 # define PL_faketokens (PL_parser->faketokens)
56 # define PL_lasttoke (PL_parser->lasttoke)
57 # define PL_nextwhite (PL_parser->nextwhite)
58 # define PL_realtokenstart (PL_parser->realtokenstart)
59 # define PL_skipwhite (PL_parser->skipwhite)
60 # define PL_thisclose (PL_parser->thisclose)
61 # define PL_thismad (PL_parser->thismad)
62 # define PL_thisopen (PL_parser->thisopen)
63 # define PL_thisstuff (PL_parser->thisstuff)
64 # define PL_thistoken (PL_parser->thistoken)
65 # define PL_thiswhite (PL_parser->thiswhite)
69 S_pending_ident(pTHX);
71 static const char ident_too_long[] = "Identifier too long";
72 static const char commaless_variable_list[] = "comma-less variable list";
74 static void restore_rsfp(pTHX_ void *f);
75 #ifndef PERL_NO_UTF16_FILTER
76 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
77 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
81 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
82 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
84 # define CURMAD(slot,sv)
85 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
88 #define XFAKEBRACK 128
91 #ifdef USE_UTF8_SCRIPTS
92 # define UTF (!IN_BYTES)
94 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
97 /* In variables named $^X, these are the legal values for X.
98 * 1999-02-27 mjd-perl-patch@plover.com */
99 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
101 /* On MacOS, respect nonbreaking spaces */
102 #ifdef MACOS_TRADITIONAL
103 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
105 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
108 /* LEX_* are values for PL_lex_state, the state of the lexer.
109 * They are arranged oddly so that the guard on the switch statement
110 * can get by with a single comparison (if the compiler is smart enough).
113 /* #define LEX_NOTPARSING 11 is done in perl.h. */
115 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
116 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
117 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
118 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
119 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
121 /* at end of code, eg "$x" followed by: */
122 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
123 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
125 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
126 string or after \E, $foo, etc */
127 #define LEX_INTERPCONST 2 /* NOT USED */
128 #define LEX_FORMLINE 1 /* expecting a format line */
129 #define LEX_KNOWNEXT 0 /* next token known; just return it */
133 static const char* const lex_state_names[] = {
152 #include "keywords.h"
154 /* CLINE is a macro that ensures PL_copline has a sane value */
159 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
162 # define SKIPSPACE0(s) skipspace0(s)
163 # define SKIPSPACE1(s) skipspace1(s)
164 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
165 # define PEEKSPACE(s) skipspace2(s,0)
167 # define SKIPSPACE0(s) skipspace(s)
168 # define SKIPSPACE1(s) skipspace(s)
169 # define SKIPSPACE2(s,tsv) skipspace(s)
170 # define PEEKSPACE(s) skipspace(s)
174 * Convenience functions to return different tokens and prime the
175 * lexer for the next token. They all take an argument.
177 * TOKEN : generic token (used for '(', DOLSHARP, etc)
178 * OPERATOR : generic operator
179 * AOPERATOR : assignment operator
180 * PREBLOCK : beginning the block after an if, while, foreach, ...
181 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
182 * PREREF : *EXPR where EXPR is not a simple identifier
183 * TERM : expression term
184 * LOOPX : loop exiting command (goto, last, dump, etc)
185 * FTST : file test operator
186 * FUN0 : zero-argument function
187 * FUN1 : not used, except for not, which isn't a UNIOP
188 * BOop : bitwise or or xor
190 * SHop : shift operator
191 * PWop : power operator
192 * PMop : pattern-matching operator
193 * Aop : addition-level operator
194 * Mop : multiplication-level operator
195 * Eop : equality-testing operator
196 * Rop : relational operator <= != gt
198 * Also see LOP and lop() below.
201 #ifdef DEBUGGING /* Serve -DT. */
202 # define REPORT(retval) tokereport((I32)retval)
204 # define REPORT(retval) (retval)
207 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
208 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
209 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
210 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
211 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
212 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
213 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
214 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
215 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
216 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
217 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
218 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
219 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
220 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
221 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
222 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
223 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
224 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
225 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
226 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
228 /* This bit of chicanery makes a unary function followed by
229 * a parenthesis into a function with one argument, highest precedence.
230 * The UNIDOR macro is for unary functions that can be followed by the //
231 * operator (such as C<shift // 0>).
233 #define UNI2(f,x) { \
237 PL_last_uni = PL_oldbufptr; \
238 PL_last_lop_op = f; \
240 return REPORT( (int)FUNC1 ); \
242 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
244 #define UNI(f) UNI2(f,XTERM)
245 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
247 #define UNIBRACK(f) { \
250 PL_last_uni = PL_oldbufptr; \
252 return REPORT( (int)FUNC1 ); \
254 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
257 /* grandfather return to old style */
258 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
262 /* how to interpret the yylval associated with the token */
266 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
272 static struct debug_tokens {
274 enum token_type type;
276 } const debug_tokens[] =
278 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
279 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
280 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
281 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
282 { ARROW, TOKENTYPE_NONE, "ARROW" },
283 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
284 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
285 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
286 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
287 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
288 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
289 { DO, TOKENTYPE_NONE, "DO" },
290 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
291 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
292 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
293 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
294 { ELSE, TOKENTYPE_NONE, "ELSE" },
295 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
296 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
297 { FOR, TOKENTYPE_IVAL, "FOR" },
298 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
299 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
300 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
301 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
302 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
303 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
304 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
305 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
306 { IF, TOKENTYPE_IVAL, "IF" },
307 { LABEL, TOKENTYPE_PVAL, "LABEL" },
308 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
309 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
310 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
311 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
312 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
313 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
314 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
315 { MY, TOKENTYPE_IVAL, "MY" },
316 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
317 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
318 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
319 { OROP, TOKENTYPE_IVAL, "OROP" },
320 { OROR, TOKENTYPE_NONE, "OROR" },
321 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
322 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
323 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
324 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
325 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
326 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
327 { PREINC, TOKENTYPE_NONE, "PREINC" },
328 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
329 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
330 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
331 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
332 { SUB, TOKENTYPE_NONE, "SUB" },
333 { THING, TOKENTYPE_OPVAL, "THING" },
334 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
335 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
336 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
337 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
338 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
339 { USE, TOKENTYPE_IVAL, "USE" },
340 { WHEN, TOKENTYPE_IVAL, "WHEN" },
341 { WHILE, TOKENTYPE_IVAL, "WHILE" },
342 { WORD, TOKENTYPE_OPVAL, "WORD" },
343 { 0, TOKENTYPE_NONE, NULL }
346 /* dump the returned token in rv, plus any optional arg in yylval */
349 S_tokereport(pTHX_ I32 rv)
353 const char *name = NULL;
354 enum token_type type = TOKENTYPE_NONE;
355 const struct debug_tokens *p;
356 SV* const report = newSVpvs("<== ");
358 for (p = debug_tokens; p->token; p++) {
359 if (p->token == (int)rv) {
366 Perl_sv_catpv(aTHX_ report, name);
367 else if ((char)rv > ' ' && (char)rv < '~')
368 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
370 sv_catpvs(report, "EOF");
372 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
375 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
378 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
380 case TOKENTYPE_OPNUM:
381 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
382 PL_op_name[yylval.ival]);
385 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
387 case TOKENTYPE_OPVAL:
389 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
390 PL_op_name[yylval.opval->op_type]);
391 if (yylval.opval->op_type == OP_CONST) {
392 Perl_sv_catpvf(aTHX_ report, " %s",
393 SvPEEK(cSVOPx_sv(yylval.opval)));
398 sv_catpvs(report, "(opval=null)");
401 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
407 /* print the buffer with suitable escapes */
410 S_printbuf(pTHX_ const char* fmt, const char* s)
412 SV* const tmp = newSVpvs("");
413 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
422 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
423 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
427 S_ao(pTHX_ int toketype)
430 if (*PL_bufptr == '=') {
432 if (toketype == ANDAND)
433 yylval.ival = OP_ANDASSIGN;
434 else if (toketype == OROR)
435 yylval.ival = OP_ORASSIGN;
436 else if (toketype == DORDOR)
437 yylval.ival = OP_DORASSIGN;
445 * When Perl expects an operator and finds something else, no_op
446 * prints the warning. It always prints "<something> found where
447 * operator expected. It prints "Missing semicolon on previous line?"
448 * if the surprise occurs at the start of the line. "do you need to
449 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
450 * where the compiler doesn't know if foo is a method call or a function.
451 * It prints "Missing operator before end of line" if there's nothing
452 * after the missing operator, or "... before <...>" if there is something
453 * after the missing operator.
457 S_no_op(pTHX_ const char *what, char *s)
460 char * const oldbp = PL_bufptr;
461 const bool is_first = (PL_oldbufptr == PL_linestart);
467 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
468 if (ckWARN_d(WARN_SYNTAX)) {
470 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
471 "\t(Missing semicolon on previous line?)\n");
472 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
474 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
476 if (t < PL_bufptr && isSPACE(*t))
477 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
478 "\t(Do you need to predeclare %.*s?)\n",
479 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
483 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
484 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
492 * Complain about missing quote/regexp/heredoc terminator.
493 * If it's called with NULL then it cauterizes the line buffer.
494 * If we're in a delimited string and the delimiter is a control
495 * character, it's reformatted into a two-char sequence like ^C.
500 S_missingterm(pTHX_ char *s)
506 char * const nl = strrchr(s,'\n');
512 iscntrl(PL_multi_close)
514 PL_multi_close < 32 || PL_multi_close == 127
518 tmpbuf[1] = (char)toCTRL(PL_multi_close);
523 *tmpbuf = (char)PL_multi_close;
527 q = strchr(s,'"') ? '\'' : '"';
528 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
531 #define FEATURE_IS_ENABLED(name) \
532 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
533 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
535 * S_feature_is_enabled
536 * Check whether the named feature is enabled.
539 S_feature_is_enabled(pTHX_ const char *name, STRLEN namelen)
542 HV * const hinthv = GvHV(PL_hintgv);
543 char he_name[32] = "feature_";
544 (void) my_strlcpy(&he_name[8], name, 24);
546 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
554 Perl_deprecate(pTHX_ const char *s)
556 if (ckWARN(WARN_DEPRECATED))
557 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
561 Perl_deprecate_old(pTHX_ const char *s)
563 /* This function should NOT be called for any new deprecated warnings */
564 /* Use Perl_deprecate instead */
566 /* It is here to maintain backward compatibility with the pre-5.8 */
567 /* warnings category hierarchy. The "deprecated" category used to */
568 /* live under the "syntax" category. It is now a top-level category */
569 /* in its own right. */
571 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
572 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
573 "Use of %s is deprecated", s);
577 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
578 * utf16-to-utf8-reversed.
581 #ifdef PERL_CR_FILTER
585 register const char *s = SvPVX_const(sv);
586 register const char * const e = s + SvCUR(sv);
587 /* outer loop optimized to do nothing if there are no CR-LFs */
589 if (*s++ == '\r' && *s == '\n') {
590 /* hit a CR-LF, need to copy the rest */
591 register char *d = s - 1;
594 if (*s == '\r' && s[1] == '\n')
605 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
607 const I32 count = FILTER_READ(idx+1, sv, maxlen);
608 if (count > 0 && !maxlen)
618 * Initialize variables. Uses the Perl save_stack to save its state (for
619 * recursive calls to the parser).
623 Perl_lex_start(pTHX_ SV *line)
626 const char *s = NULL;
630 /* create and initialise a parser */
632 Newxz(parser, 1, yy_parser);
633 parser->old_parser = PL_parser;
636 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
637 parser->ps = parser->stack;
638 parser->stack_size = YYINITDEPTH;
640 parser->stack->state = 0;
641 parser->yyerrstatus = 0;
642 parser->yychar = YYEMPTY; /* Cause a token to be read. */
644 /* initialise lexer state */
646 SAVEI32(PL_lex_state);
648 if (PL_lex_state == LEX_KNOWNEXT) {
649 I32 toke = parser->old_parser->lasttoke;
650 while (--toke >= 0) {
651 SAVEI32(PL_nexttoke[toke].next_type);
652 SAVEVPTR(PL_nexttoke[toke].next_val);
654 SAVEVPTR(PL_nexttoke[toke].next_mad);
657 SAVEI32(PL_curforce);
659 if (PL_lex_state == LEX_KNOWNEXT) {
660 I32 toke = PL_nexttoke;
661 while (--toke >= 0) {
662 SAVEI32(PL_nexttype[toke]);
663 SAVEVPTR(PL_nextval[toke]);
665 SAVEI32(PL_nexttoke);
668 SAVECOPLINE(PL_curcop);
671 SAVEPPTR(PL_oldbufptr);
672 SAVEPPTR(PL_oldoldbufptr);
673 SAVEPPTR(PL_last_lop);
674 SAVEPPTR(PL_last_uni);
675 SAVEPPTR(PL_linestart);
676 SAVESPTR(PL_linestr);
677 SAVEDESTRUCTOR_X(restore_rsfp, PL_rsfp);
680 PL_lex_state = LEX_NORMAL;
682 Newx(parser->lex_brackstack, 120, char);
683 Newx(parser->lex_casestack, 12, char);
684 *parser->lex_casestack = '\0';
690 s = SvPV_const(line, len);
695 PL_linestr = newSVpvs("\n;");
696 } else if (SvREADONLY(line) || s[len-1] != ';') {
697 PL_linestr = newSVsv(line);
699 sv_catpvs(PL_linestr, "\n;");
702 SvREFCNT_inc_simple_void_NN(line);
705 /* PL_linestr needs to survive until end of scope, not just the next
706 FREETMPS. See changes 17505 and 17546 which fixed the symptoms only. */
707 SAVEFREESV(PL_linestr);
708 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
709 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
710 PL_last_lop = PL_last_uni = NULL;
716 * Finalizer for lexing operations. Must be called when the parser is
717 * done with the lexer.
724 PL_doextract = FALSE;
729 * This subroutine has nothing to do with tilting, whether at windmills
730 * or pinball tables. Its name is short for "increment line". It
731 * increments the current line number in CopLINE(PL_curcop) and checks
732 * to see whether the line starts with a comment of the form
733 * # line 500 "foo.pm"
734 * If so, it sets the current line number and file to the values in the comment.
738 S_incline(pTHX_ const char *s)
745 CopLINE_inc(PL_curcop);
748 while (SPACE_OR_TAB(*s))
750 if (strnEQ(s, "line", 4))
754 if (SPACE_OR_TAB(*s))
758 while (SPACE_OR_TAB(*s))
766 while (SPACE_OR_TAB(*s))
768 if (*s == '"' && (t = strchr(s+1, '"'))) {
778 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
780 if (*e != '\n' && *e != '\0')
781 return; /* false alarm */
784 const STRLEN len = t - s;
786 const char * const cf = CopFILE(PL_curcop);
787 STRLEN tmplen = cf ? strlen(cf) : 0;
788 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
789 /* must copy *{"::_<(eval N)[oldfilename:L]"}
790 * to *{"::_<newfilename"} */
791 /* However, the long form of evals is only turned on by the
792 debugger - usually they're "(eval %lu)" */
796 STRLEN tmplen2 = len;
797 if (tmplen + 2 <= sizeof smallbuf)
800 Newx(tmpbuf, tmplen + 2, char);
803 memcpy(tmpbuf + 2, cf, tmplen);
805 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
810 if (tmplen2 + 2 <= sizeof smallbuf)
813 Newx(tmpbuf2, tmplen2 + 2, char);
815 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
816 /* Either they malloc'd it, or we malloc'd it,
817 so no prefix is present in ours. */
822 memcpy(tmpbuf2 + 2, s, tmplen2);
825 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
827 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
828 /* adjust ${"::_<newfilename"} to store the new file name */
829 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
830 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
831 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
834 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
836 if (tmpbuf != smallbuf) Safefree(tmpbuf);
839 CopFILE_free(PL_curcop);
840 CopFILE_setn(PL_curcop, s, len);
842 CopLINE_set(PL_curcop, atoi(n)-1);
846 /* skip space before PL_thistoken */
849 S_skipspace0(pTHX_ register char *s)
856 PL_thiswhite = newSVpvs("");
857 sv_catsv(PL_thiswhite, PL_skipwhite);
858 sv_free(PL_skipwhite);
861 PL_realtokenstart = s - SvPVX(PL_linestr);
865 /* skip space after PL_thistoken */
868 S_skipspace1(pTHX_ register char *s)
870 const char *start = s;
871 I32 startoff = start - SvPVX(PL_linestr);
876 start = SvPVX(PL_linestr) + startoff;
877 if (!PL_thistoken && PL_realtokenstart >= 0) {
878 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
879 PL_thistoken = newSVpvn(tstart, start - tstart);
881 PL_realtokenstart = -1;
884 PL_nextwhite = newSVpvs("");
885 sv_catsv(PL_nextwhite, PL_skipwhite);
886 sv_free(PL_skipwhite);
893 S_skipspace2(pTHX_ register char *s, SV **svp)
896 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
897 const I32 startoff = s - SvPVX(PL_linestr);
900 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
901 if (!PL_madskills || !svp)
903 start = SvPVX(PL_linestr) + startoff;
904 if (!PL_thistoken && PL_realtokenstart >= 0) {
905 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
906 PL_thistoken = newSVpvn(tstart, start - tstart);
907 PL_realtokenstart = -1;
912 sv_setsv(*svp, PL_skipwhite);
913 sv_free(PL_skipwhite);
922 S_update_debugger_info(pTHX_ SV *orig_sv, const char *buf, STRLEN len)
924 AV *av = CopFILEAVx(PL_curcop);
926 SV * const sv = newSV(0);
927 sv_upgrade(sv, SVt_PVMG);
929 sv_setsv(sv, orig_sv);
931 sv_setpvn(sv, buf, len);
934 av_store(av, (I32)CopLINE(PL_curcop), sv);
940 * Called to gobble the appropriate amount and type of whitespace.
941 * Skips comments as well.
945 S_skipspace(pTHX_ register char *s)
950 int startoff = s - SvPVX(PL_linestr);
953 sv_free(PL_skipwhite);
958 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
959 while (s < PL_bufend && SPACE_OR_TAB(*s))
969 SSize_t oldprevlen, oldoldprevlen;
970 SSize_t oldloplen = 0, oldunilen = 0;
971 while (s < PL_bufend && isSPACE(*s)) {
972 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
977 if (s < PL_bufend && *s == '#') {
978 while (s < PL_bufend && *s != '\n')
982 if (PL_in_eval && !PL_rsfp) {
989 /* only continue to recharge the buffer if we're at the end
990 * of the buffer, we're not reading from a source filter, and
991 * we're in normal lexing mode
993 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
994 PL_lex_state == LEX_FORMLINE)
1001 /* try to recharge the buffer */
1003 curoff = s - SvPVX(PL_linestr);
1006 if ((s = filter_gets(PL_linestr, PL_rsfp,
1007 (prevlen = SvCUR(PL_linestr)))) == NULL)
1010 if (PL_madskills && curoff != startoff) {
1012 PL_skipwhite = newSVpvs("");
1013 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1017 /* mustn't throw out old stuff yet if madpropping */
1018 SvCUR(PL_linestr) = curoff;
1019 s = SvPVX(PL_linestr) + curoff;
1021 if (curoff && s[-1] == '\n')
1025 /* end of file. Add on the -p or -n magic */
1026 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1029 sv_catpvs(PL_linestr,
1030 ";}continue{print or die qq(-p destination: $!\\n);}");
1032 sv_setpvs(PL_linestr,
1033 ";}continue{print or die qq(-p destination: $!\\n);}");
1035 PL_minus_n = PL_minus_p = 0;
1037 else if (PL_minus_n) {
1039 sv_catpvn(PL_linestr, ";}", 2);
1041 sv_setpvn(PL_linestr, ";}", 2);
1047 sv_catpvn(PL_linestr,";", 1);
1049 sv_setpvn(PL_linestr,";", 1);
1052 /* reset variables for next time we lex */
1053 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1059 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1060 PL_last_lop = PL_last_uni = NULL;
1062 /* Close the filehandle. Could be from -P preprocessor,
1063 * STDIN, or a regular file. If we were reading code from
1064 * STDIN (because the commandline held no -e or filename)
1065 * then we don't close it, we reset it so the code can
1066 * read from STDIN too.
1069 if (PL_preprocess && !PL_in_eval)
1070 (void)PerlProc_pclose(PL_rsfp);
1071 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1072 PerlIO_clearerr(PL_rsfp);
1074 (void)PerlIO_close(PL_rsfp);
1079 /* not at end of file, so we only read another line */
1080 /* make corresponding updates to old pointers, for yyerror() */
1081 oldprevlen = PL_oldbufptr - PL_bufend;
1082 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1084 oldunilen = PL_last_uni - PL_bufend;
1086 oldloplen = PL_last_lop - PL_bufend;
1087 PL_linestart = PL_bufptr = s + prevlen;
1088 PL_bufend = s + SvCUR(PL_linestr);
1090 PL_oldbufptr = s + oldprevlen;
1091 PL_oldoldbufptr = s + oldoldprevlen;
1093 PL_last_uni = s + oldunilen;
1095 PL_last_lop = s + oldloplen;
1098 /* debugger active and we're not compiling the debugger code,
1099 * so store the line into the debugger's array of lines
1101 if (PERLDB_LINE && PL_curstash != PL_debstash)
1102 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1109 PL_skipwhite = newSVpvs("");
1110 curoff = s - SvPVX(PL_linestr);
1111 if (curoff - startoff)
1112 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1121 * Check the unary operators to ensure there's no ambiguity in how they're
1122 * used. An ambiguous piece of code would be:
1124 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1125 * the +5 is its argument.
1135 if (PL_oldoldbufptr != PL_last_uni)
1137 while (isSPACE(*PL_last_uni))
1140 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1142 if ((t = strchr(s, '(')) && t < PL_bufptr)
1145 if (ckWARN_d(WARN_AMBIGUOUS)){
1146 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1147 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1148 (int)(s - PL_last_uni), PL_last_uni);
1153 * LOP : macro to build a list operator. Its behaviour has been replaced
1154 * with a subroutine, S_lop() for which LOP is just another name.
1157 #define LOP(f,x) return lop(f,x,s)
1161 * Build a list operator (or something that might be one). The rules:
1162 * - if we have a next token, then it's a list operator [why?]
1163 * - if the next thing is an opening paren, then it's a function
1164 * - else it's a list operator
1168 S_lop(pTHX_ I32 f, int x, char *s)
1175 PL_last_lop = PL_oldbufptr;
1176 PL_last_lop_op = (OPCODE)f;
1179 return REPORT(LSTOP);
1182 return REPORT(LSTOP);
1185 return REPORT(FUNC);
1188 return REPORT(FUNC);
1190 return REPORT(LSTOP);
1196 * Sets up for an eventual force_next(). start_force(0) basically does
1197 * an unshift, while start_force(-1) does a push. yylex removes items
1202 S_start_force(pTHX_ int where)
1206 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1207 where = PL_lasttoke;
1208 assert(PL_curforce < 0 || PL_curforce == where);
1209 if (PL_curforce != where) {
1210 for (i = PL_lasttoke; i > where; --i) {
1211 PL_nexttoke[i] = PL_nexttoke[i-1];
1215 if (PL_curforce < 0) /* in case of duplicate start_force() */
1216 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1217 PL_curforce = where;
1220 curmad('^', newSVpvs(""));
1221 CURMAD('_', PL_nextwhite);
1226 S_curmad(pTHX_ char slot, SV *sv)
1232 if (PL_curforce < 0)
1233 where = &PL_thismad;
1235 where = &PL_nexttoke[PL_curforce].next_mad;
1238 sv_setpvn(sv, "", 0);
1241 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1243 else if (PL_encoding) {
1244 sv_recode_to_utf8(sv, PL_encoding);
1249 /* keep a slot open for the head of the list? */
1250 if (slot != '_' && *where && (*where)->mad_key == '^') {
1251 (*where)->mad_key = slot;
1252 sv_free((*where)->mad_val);
1253 (*where)->mad_val = (void*)sv;
1256 addmad(newMADsv(slot, sv), where, 0);
1259 # define start_force(where) NOOP
1260 # define curmad(slot, sv) NOOP
1265 * When the lexer realizes it knows the next token (for instance,
1266 * it is reordering tokens for the parser) then it can call S_force_next
1267 * to know what token to return the next time the lexer is called. Caller
1268 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1269 * and possibly PL_expect to ensure the lexer handles the token correctly.
1273 S_force_next(pTHX_ I32 type)
1277 if (PL_curforce < 0)
1278 start_force(PL_lasttoke);
1279 PL_nexttoke[PL_curforce].next_type = type;
1280 if (PL_lex_state != LEX_KNOWNEXT)
1281 PL_lex_defer = PL_lex_state;
1282 PL_lex_state = LEX_KNOWNEXT;
1283 PL_lex_expect = PL_expect;
1286 PL_nexttype[PL_nexttoke] = type;
1288 if (PL_lex_state != LEX_KNOWNEXT) {
1289 PL_lex_defer = PL_lex_state;
1290 PL_lex_expect = PL_expect;
1291 PL_lex_state = LEX_KNOWNEXT;
1297 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1300 SV * const sv = newSVpvn(start,len);
1301 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1308 * When the lexer knows the next thing is a word (for instance, it has
1309 * just seen -> and it knows that the next char is a word char, then
1310 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1314 * char *start : buffer position (must be within PL_linestr)
1315 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1316 * int check_keyword : if true, Perl checks to make sure the word isn't
1317 * a keyword (do this if the word is a label, e.g. goto FOO)
1318 * int allow_pack : if true, : characters will also be allowed (require,
1319 * use, etc. do this)
1320 * int allow_initial_tick : used by the "sub" lexer only.
1324 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1330 start = SKIPSPACE1(start);
1332 if (isIDFIRST_lazy_if(s,UTF) ||
1333 (allow_pack && *s == ':') ||
1334 (allow_initial_tick && *s == '\'') )
1336 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1337 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1339 start_force(PL_curforce);
1341 curmad('X', newSVpvn(start,s-start));
1342 if (token == METHOD) {
1347 PL_expect = XOPERATOR;
1350 NEXTVAL_NEXTTOKE.opval
1351 = (OP*)newSVOP(OP_CONST,0,
1352 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1353 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1361 * Called when the lexer wants $foo *foo &foo etc, but the program
1362 * text only contains the "foo" portion. The first argument is a pointer
1363 * to the "foo", and the second argument is the type symbol to prefix.
1364 * Forces the next token to be a "WORD".
1365 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1369 S_force_ident(pTHX_ register const char *s, int kind)
1373 const STRLEN len = strlen(s);
1374 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1375 start_force(PL_curforce);
1376 NEXTVAL_NEXTTOKE.opval = o;
1379 o->op_private = OPpCONST_ENTERED;
1380 /* XXX see note in pp_entereval() for why we forgo typo
1381 warnings if the symbol must be introduced in an eval.
1383 gv_fetchpvn_flags(s, len,
1384 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1386 kind == '$' ? SVt_PV :
1387 kind == '@' ? SVt_PVAV :
1388 kind == '%' ? SVt_PVHV :
1396 Perl_str_to_version(pTHX_ SV *sv)
1401 const char *start = SvPV_const(sv,len);
1402 const char * const end = start + len;
1403 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1404 while (start < end) {
1408 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1413 retval += ((NV)n)/nshift;
1422 * Forces the next token to be a version number.
1423 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1424 * and if "guessing" is TRUE, then no new token is created (and the caller
1425 * must use an alternative parsing method).
1429 S_force_version(pTHX_ char *s, int guessing)
1435 I32 startoff = s - SvPVX(PL_linestr);
1444 while (isDIGIT(*d) || *d == '_' || *d == '.')
1448 start_force(PL_curforce);
1449 curmad('X', newSVpvn(s,d-s));
1452 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1454 s = scan_num(s, &yylval);
1455 version = yylval.opval;
1456 ver = cSVOPx(version)->op_sv;
1457 if (SvPOK(ver) && !SvNIOK(ver)) {
1458 SvUPGRADE(ver, SVt_PVNV);
1459 SvNV_set(ver, str_to_version(ver));
1460 SvNOK_on(ver); /* hint that it is a version */
1463 else if (guessing) {
1466 sv_free(PL_nextwhite); /* let next token collect whitespace */
1468 s = SvPVX(PL_linestr) + startoff;
1476 if (PL_madskills && !version) {
1477 sv_free(PL_nextwhite); /* let next token collect whitespace */
1479 s = SvPVX(PL_linestr) + startoff;
1482 /* NOTE: The parser sees the package name and the VERSION swapped */
1483 start_force(PL_curforce);
1484 NEXTVAL_NEXTTOKE.opval = version;
1492 * Tokenize a quoted string passed in as an SV. It finds the next
1493 * chunk, up to end of string or a backslash. It may make a new
1494 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1499 S_tokeq(pTHX_ SV *sv)
1503 register char *send;
1511 s = SvPV_force(sv, len);
1512 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1515 while (s < send && *s != '\\')
1520 if ( PL_hints & HINT_NEW_STRING ) {
1521 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1527 if (s + 1 < send && (s[1] == '\\'))
1528 s++; /* all that, just for this */
1533 SvCUR_set(sv, d - SvPVX_const(sv));
1535 if ( PL_hints & HINT_NEW_STRING )
1536 return new_constant(NULL, 0, "q", sv, pv, "q");
1541 * Now come three functions related to double-quote context,
1542 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1543 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1544 * interact with PL_lex_state, and create fake ( ... ) argument lists
1545 * to handle functions and concatenation.
1546 * They assume that whoever calls them will be setting up a fake
1547 * join call, because each subthing puts a ',' after it. This lets
1550 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1552 * (I'm not sure whether the spurious commas at the end of lcfirst's
1553 * arguments and join's arguments are created or not).
1558 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1560 * Pattern matching will set PL_lex_op to the pattern-matching op to
1561 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1563 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1565 * Everything else becomes a FUNC.
1567 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1568 * had an OP_CONST or OP_READLINE). This just sets us up for a
1569 * call to S_sublex_push().
1573 S_sublex_start(pTHX)
1576 register const I32 op_type = yylval.ival;
1578 if (op_type == OP_NULL) {
1579 yylval.opval = PL_lex_op;
1583 if (op_type == OP_CONST || op_type == OP_READLINE) {
1584 SV *sv = tokeq(PL_lex_stuff);
1586 if (SvTYPE(sv) == SVt_PVIV) {
1587 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1589 const char * const p = SvPV_const(sv, len);
1590 SV * const nsv = newSVpvn(p, len);
1596 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1597 PL_lex_stuff = NULL;
1598 /* Allow <FH> // "foo" */
1599 if (op_type == OP_READLINE)
1600 PL_expect = XTERMORDORDOR;
1603 else if (op_type == OP_BACKTICK && PL_lex_op) {
1604 /* readpipe() vas overriden */
1605 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1606 yylval.opval = PL_lex_op;
1608 PL_lex_stuff = NULL;
1612 PL_sublex_info.super_state = PL_lex_state;
1613 PL_sublex_info.sub_inwhat = op_type;
1614 PL_sublex_info.sub_op = PL_lex_op;
1615 PL_lex_state = LEX_INTERPPUSH;
1619 yylval.opval = PL_lex_op;
1629 * Create a new scope to save the lexing state. The scope will be
1630 * ended in S_sublex_done. Returns a '(', starting the function arguments
1631 * to the uc, lc, etc. found before.
1632 * Sets PL_lex_state to LEX_INTERPCONCAT.
1641 PL_lex_state = PL_sublex_info.super_state;
1642 SAVEI32(PL_lex_dojoin);
1643 SAVEI32(PL_lex_brackets);
1644 SAVEI32(PL_lex_casemods);
1645 SAVEI32(PL_lex_starts);
1646 SAVEI32(PL_lex_state);
1647 SAVEVPTR(PL_lex_inpat);
1648 SAVEI32(PL_lex_inwhat);
1649 SAVECOPLINE(PL_curcop);
1650 SAVEPPTR(PL_bufptr);
1651 SAVEPPTR(PL_bufend);
1652 SAVEPPTR(PL_oldbufptr);
1653 SAVEPPTR(PL_oldoldbufptr);
1654 SAVEPPTR(PL_last_lop);
1655 SAVEPPTR(PL_last_uni);
1656 SAVEPPTR(PL_linestart);
1657 SAVESPTR(PL_linestr);
1658 SAVEGENERICPV(PL_lex_brackstack);
1659 SAVEGENERICPV(PL_lex_casestack);
1661 PL_linestr = PL_lex_stuff;
1662 PL_lex_stuff = NULL;
1664 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1665 = SvPVX(PL_linestr);
1666 PL_bufend += SvCUR(PL_linestr);
1667 PL_last_lop = PL_last_uni = NULL;
1668 SAVEFREESV(PL_linestr);
1670 PL_lex_dojoin = FALSE;
1671 PL_lex_brackets = 0;
1672 Newx(PL_lex_brackstack, 120, char);
1673 Newx(PL_lex_casestack, 12, char);
1674 PL_lex_casemods = 0;
1675 *PL_lex_casestack = '\0';
1677 PL_lex_state = LEX_INTERPCONCAT;
1678 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1680 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1681 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1682 PL_lex_inpat = PL_sublex_info.sub_op;
1684 PL_lex_inpat = NULL;
1691 * Restores lexer state after a S_sublex_push.
1698 if (!PL_lex_starts++) {
1699 SV * const sv = newSVpvs("");
1700 if (SvUTF8(PL_linestr))
1702 PL_expect = XOPERATOR;
1703 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1707 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1708 PL_lex_state = LEX_INTERPCASEMOD;
1712 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1713 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1714 PL_linestr = PL_lex_repl;
1716 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1717 PL_bufend += SvCUR(PL_linestr);
1718 PL_last_lop = PL_last_uni = NULL;
1719 SAVEFREESV(PL_linestr);
1720 PL_lex_dojoin = FALSE;
1721 PL_lex_brackets = 0;
1722 PL_lex_casemods = 0;
1723 *PL_lex_casestack = '\0';
1725 if (SvEVALED(PL_lex_repl)) {
1726 PL_lex_state = LEX_INTERPNORMAL;
1728 /* we don't clear PL_lex_repl here, so that we can check later
1729 whether this is an evalled subst; that means we rely on the
1730 logic to ensure sublex_done() is called again only via the
1731 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1734 PL_lex_state = LEX_INTERPCONCAT;
1744 PL_endwhite = newSVpvs("");
1745 sv_catsv(PL_endwhite, PL_thiswhite);
1749 sv_setpvn(PL_thistoken,"",0);
1751 PL_realtokenstart = -1;
1755 PL_bufend = SvPVX(PL_linestr);
1756 PL_bufend += SvCUR(PL_linestr);
1757 PL_expect = XOPERATOR;
1758 PL_sublex_info.sub_inwhat = 0;
1766 Extracts a pattern, double-quoted string, or transliteration. This
1769 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1770 processing a pattern (PL_lex_inpat is true), a transliteration
1771 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1773 Returns a pointer to the character scanned up to. If this is
1774 advanced from the start pointer supplied (i.e. if anything was
1775 successfully parsed), will leave an OP for the substring scanned
1776 in yylval. Caller must intuit reason for not parsing further
1777 by looking at the next characters herself.
1781 double-quoted style: \r and \n
1782 regexp special ones: \D \s
1785 case and quoting: \U \Q \E
1786 stops on @ and $, but not for $ as tail anchor
1788 In transliterations:
1789 characters are VERY literal, except for - not at the start or end
1790 of the string, which indicates a range. If the range is in bytes,
1791 scan_const expands the range to the full set of intermediate
1792 characters. If the range is in utf8, the hyphen is replaced with
1793 a certain range mark which will be handled by pmtrans() in op.c.
1795 In double-quoted strings:
1797 double-quoted style: \r and \n
1799 deprecated backrefs: \1 (in substitution replacements)
1800 case and quoting: \U \Q \E
1803 scan_const does *not* construct ops to handle interpolated strings.
1804 It stops processing as soon as it finds an embedded $ or @ variable
1805 and leaves it to the caller to work out what's going on.
1807 embedded arrays (whether in pattern or not) could be:
1808 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1810 $ in double-quoted strings must be the symbol of an embedded scalar.
1812 $ in pattern could be $foo or could be tail anchor. Assumption:
1813 it's a tail anchor if $ is the last thing in the string, or if it's
1814 followed by one of "()| \r\n\t"
1816 \1 (backreferences) are turned into $1
1818 The structure of the code is
1819 while (there's a character to process) {
1820 handle transliteration ranges
1821 skip regexp comments /(?#comment)/ and codes /(?{code})/
1822 skip #-initiated comments in //x patterns
1823 check for embedded arrays
1824 check for embedded scalars
1826 leave intact backslashes from leaveit (below)
1827 deprecate \1 in substitution replacements
1828 handle string-changing backslashes \l \U \Q \E, etc.
1829 switch (what was escaped) {
1830 handle \- in a transliteration (becomes a literal -)
1831 handle \132 (octal characters)
1832 handle \x15 and \x{1234} (hex characters)
1833 handle \N{name} (named characters)
1834 handle \cV (control characters)
1835 handle printf-style backslashes (\f, \r, \n, etc)
1837 } (end if backslash)
1838 } (end while character to read)
1843 S_scan_const(pTHX_ char *start)
1846 register char *send = PL_bufend; /* end of the constant */
1847 SV *sv = newSV(send - start); /* sv for the constant */
1848 register char *s = start; /* start of the constant */
1849 register char *d = SvPVX(sv); /* destination for copies */
1850 bool dorange = FALSE; /* are we in a translit range? */
1851 bool didrange = FALSE; /* did we just finish a range? */
1852 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1853 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1856 UV literal_endpoint = 0;
1857 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1860 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1861 /* If we are doing a trans and we know we want UTF8 set expectation */
1862 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1863 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1867 while (s < send || dorange) {
1868 /* get transliterations out of the way (they're most literal) */
1869 if (PL_lex_inwhat == OP_TRANS) {
1870 /* expand a range A-Z to the full set of characters. AIE! */
1872 I32 i; /* current expanded character */
1873 I32 min; /* first character in range */
1874 I32 max; /* last character in range */
1885 char * const c = (char*)utf8_hop((U8*)d, -1);
1889 *c = (char)UTF_TO_NATIVE(0xff);
1890 /* mark the range as done, and continue */
1896 i = d - SvPVX_const(sv); /* remember current offset */
1899 SvLEN(sv) + (has_utf8 ?
1900 (512 - UTF_CONTINUATION_MARK +
1903 /* How many two-byte within 0..255: 128 in UTF-8,
1904 * 96 in UTF-8-mod. */
1906 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1908 d = SvPVX(sv) + i; /* refresh d after realloc */
1912 for (j = 0; j <= 1; j++) {
1913 char * const c = (char*)utf8_hop((U8*)d, -1);
1914 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1920 max = (U8)0xff; /* only to \xff */
1921 uvmax = uv; /* \x{100} to uvmax */
1923 d = c; /* eat endpoint chars */
1928 d -= 2; /* eat the first char and the - */
1929 min = (U8)*d; /* first char in range */
1930 max = (U8)d[1]; /* last char in range */
1937 "Invalid range \"%c-%c\" in transliteration operator",
1938 (char)min, (char)max);
1942 if (literal_endpoint == 2 &&
1943 ((isLOWER(min) && isLOWER(max)) ||
1944 (isUPPER(min) && isUPPER(max)))) {
1946 for (i = min; i <= max; i++)
1948 *d++ = NATIVE_TO_NEED(has_utf8,i);
1950 for (i = min; i <= max; i++)
1952 *d++ = NATIVE_TO_NEED(has_utf8,i);
1957 for (i = min; i <= max; i++)
1960 const U8 ch = (U8)NATIVE_TO_UTF(i);
1961 if (UNI_IS_INVARIANT(ch))
1964 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
1965 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
1974 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
1976 *d++ = (char)UTF_TO_NATIVE(0xff);
1978 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
1982 /* mark the range as done, and continue */
1986 literal_endpoint = 0;
1991 /* range begins (ignore - as first or last char) */
1992 else if (*s == '-' && s+1 < send && s != start) {
1994 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2001 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2011 literal_endpoint = 0;
2012 native_range = TRUE;
2017 /* if we get here, we're not doing a transliteration */
2019 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2020 except for the last char, which will be done separately. */
2021 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2023 while (s+1 < send && *s != ')')
2024 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2026 else if (s[2] == '{' /* This should match regcomp.c */
2027 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
2030 char *regparse = s + (s[2] == '{' ? 3 : 4);
2033 while (count && (c = *regparse)) {
2034 if (c == '\\' && regparse[1])
2042 if (*regparse != ')')
2043 regparse--; /* Leave one char for continuation. */
2044 while (s < regparse)
2045 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2049 /* likewise skip #-initiated comments in //x patterns */
2050 else if (*s == '#' && PL_lex_inpat &&
2051 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2052 while (s+1 < send && *s != '\n')
2053 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2056 /* check for embedded arrays
2057 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2059 else if (*s == '@' && s[1]) {
2060 if (isALNUM_lazy_if(s+1,UTF))
2062 if (strchr(":'{$", s[1]))
2064 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2065 break; /* in regexp, neither @+ nor @- are interpolated */
2068 /* check for embedded scalars. only stop if we're sure it's a
2071 else if (*s == '$') {
2072 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2074 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2075 break; /* in regexp, $ might be tail anchor */
2078 /* End of else if chain - OP_TRANS rejoin rest */
2081 if (*s == '\\' && s+1 < send) {
2084 /* deprecate \1 in strings and substitution replacements */
2085 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2086 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2088 if (ckWARN(WARN_SYNTAX))
2089 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2094 /* string-change backslash escapes */
2095 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2099 /* skip any other backslash escapes in a pattern */
2100 else if (PL_lex_inpat) {
2101 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2102 goto default_action;
2105 /* if we get here, it's either a quoted -, or a digit */
2108 /* quoted - in transliterations */
2110 if (PL_lex_inwhat == OP_TRANS) {
2117 if ((isALPHA(*s) || isDIGIT(*s)) &&
2119 Perl_warner(aTHX_ packWARN(WARN_MISC),
2120 "Unrecognized escape \\%c passed through",
2122 /* default action is to copy the quoted character */
2123 goto default_action;
2126 /* \132 indicates an octal constant */
2127 case '0': case '1': case '2': case '3':
2128 case '4': case '5': case '6': case '7':
2132 uv = grok_oct(s, &len, &flags, NULL);
2135 goto NUM_ESCAPE_INSERT;
2137 /* \x24 indicates a hex constant */
2141 char* const e = strchr(s, '}');
2142 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2143 PERL_SCAN_DISALLOW_PREFIX;
2148 yyerror("Missing right brace on \\x{}");
2152 uv = grok_hex(s, &len, &flags, NULL);
2158 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2159 uv = grok_hex(s, &len, &flags, NULL);
2165 /* Insert oct or hex escaped character.
2166 * There will always enough room in sv since such
2167 * escapes will be longer than any UTF-8 sequence
2168 * they can end up as. */
2170 /* We need to map to chars to ASCII before doing the tests
2173 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2174 if (!has_utf8 && uv > 255) {
2175 /* Might need to recode whatever we have
2176 * accumulated so far if it contains any
2179 * (Can't we keep track of that and avoid
2180 * this rescan? --jhi)
2184 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2185 if (!NATIVE_IS_INVARIANT(*c)) {
2190 const STRLEN offset = d - SvPVX_const(sv);
2192 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2196 while (src >= (const U8 *)SvPVX_const(sv)) {
2197 if (!NATIVE_IS_INVARIANT(*src)) {
2198 const U8 ch = NATIVE_TO_ASCII(*src);
2199 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2200 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2210 if (has_utf8 || uv > 255) {
2211 d = (char*)uvchr_to_utf8((U8*)d, uv);
2213 if (PL_lex_inwhat == OP_TRANS &&
2214 PL_sublex_info.sub_op) {
2215 PL_sublex_info.sub_op->op_private |=
2216 (PL_lex_repl ? OPpTRANS_FROM_UTF
2220 if (uv > 255 && !dorange)
2221 native_range = FALSE;
2233 /* \N{LATIN SMALL LETTER A} is a named character */
2237 char* e = strchr(s, '}');
2244 yyerror("Missing right brace on \\N{}");
2248 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2250 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2251 PERL_SCAN_DISALLOW_PREFIX;
2254 uv = grok_hex(s, &len, &flags, NULL);
2255 if ( e > s && len != (STRLEN)(e - s) ) {
2259 goto NUM_ESCAPE_INSERT;
2261 res = newSVpvn(s + 1, e - s - 1);
2262 type = newSVpvn(s - 2,e - s + 3);
2263 res = new_constant( NULL, 0, "charnames",
2264 res, NULL, SvPVX(type) );
2267 sv_utf8_upgrade(res);
2268 str = SvPV_const(res,len);
2269 #ifdef EBCDIC_NEVER_MIND
2270 /* charnames uses pack U and that has been
2271 * recently changed to do the below uni->native
2272 * mapping, so this would be redundant (and wrong,
2273 * the code point would be doubly converted).
2274 * But leave this in just in case the pack U change
2275 * gets revoked, but the semantics is still
2276 * desireable for charnames. --jhi */
2278 UV uv = utf8_to_uvchr((const U8*)str, 0);
2281 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2283 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2284 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2285 str = SvPV_const(res, len);
2289 if (!has_utf8 && SvUTF8(res)) {
2290 const char * const ostart = SvPVX_const(sv);
2291 SvCUR_set(sv, d - ostart);
2294 sv_utf8_upgrade(sv);
2295 /* this just broke our allocation above... */
2296 SvGROW(sv, (STRLEN)(send - start));
2297 d = SvPVX(sv) + SvCUR(sv);
2300 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2301 const char * const odest = SvPVX_const(sv);
2303 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2304 d = SvPVX(sv) + (d - odest);
2308 native_range = FALSE; /* \N{} is guessed to be Unicode */
2310 Copy(str, d, len, char);
2317 yyerror("Missing braces on \\N{}");
2320 /* \c is a control character */
2329 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2332 yyerror("Missing control char name in \\c");
2336 /* printf-style backslashes, formfeeds, newlines, etc */
2338 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2341 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2344 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2347 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2350 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2353 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2356 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2362 } /* end if (backslash) */
2369 /* If we started with encoded form, or already know we want it
2370 and then encode the next character */
2371 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2373 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2374 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2377 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2378 const STRLEN off = d - SvPVX_const(sv);
2379 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2381 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2384 if (uv > 255 && !dorange)
2385 native_range = FALSE;
2389 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2391 } /* while loop to process each character */
2393 /* terminate the string and set up the sv */
2395 SvCUR_set(sv, d - SvPVX_const(sv));
2396 if (SvCUR(sv) >= SvLEN(sv))
2397 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2400 if (PL_encoding && !has_utf8) {
2401 sv_recode_to_utf8(sv, PL_encoding);
2407 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2408 PL_sublex_info.sub_op->op_private |=
2409 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2413 /* shrink the sv if we allocated more than we used */
2414 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2415 SvPV_shrink_to_cur(sv);
2418 /* return the substring (via yylval) only if we parsed anything */
2419 if (s > PL_bufptr) {
2420 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2421 sv = new_constant(start, s - start,
2422 (const char *)(PL_lex_inpat ? "qr" : "q"),
2425 (( PL_lex_inwhat == OP_TRANS
2427 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2430 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2437 * Returns TRUE if there's more to the expression (e.g., a subscript),
2440 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2442 * ->[ and ->{ return TRUE
2443 * { and [ outside a pattern are always subscripts, so return TRUE
2444 * if we're outside a pattern and it's not { or [, then return FALSE
2445 * if we're in a pattern and the first char is a {
2446 * {4,5} (any digits around the comma) returns FALSE
2447 * if we're in a pattern and the first char is a [
2449 * [SOMETHING] has a funky algorithm to decide whether it's a
2450 * character class or not. It has to deal with things like
2451 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2452 * anything else returns TRUE
2455 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2458 S_intuit_more(pTHX_ register char *s)
2461 if (PL_lex_brackets)
2463 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2465 if (*s != '{' && *s != '[')
2470 /* In a pattern, so maybe we have {n,m}. */
2487 /* On the other hand, maybe we have a character class */
2490 if (*s == ']' || *s == '^')
2493 /* this is terrifying, and it works */
2494 int weight = 2; /* let's weigh the evidence */
2496 unsigned char un_char = 255, last_un_char;
2497 const char * const send = strchr(s,']');
2498 char tmpbuf[sizeof PL_tokenbuf * 4];
2500 if (!send) /* has to be an expression */
2503 Zero(seen,256,char);
2506 else if (isDIGIT(*s)) {
2508 if (isDIGIT(s[1]) && s[2] == ']')
2514 for (; s < send; s++) {
2515 last_un_char = un_char;
2516 un_char = (unsigned char)*s;
2521 weight -= seen[un_char] * 10;
2522 if (isALNUM_lazy_if(s+1,UTF)) {
2524 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2525 len = (int)strlen(tmpbuf);
2526 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2531 else if (*s == '$' && s[1] &&
2532 strchr("[#!%*<>()-=",s[1])) {
2533 if (/*{*/ strchr("])} =",s[2]))
2542 if (strchr("wds]",s[1]))
2544 else if (seen[(U8)'\''] || seen[(U8)'"'])
2546 else if (strchr("rnftbxcav",s[1]))
2548 else if (isDIGIT(s[1])) {
2550 while (s[1] && isDIGIT(s[1]))
2560 if (strchr("aA01! ",last_un_char))
2562 if (strchr("zZ79~",s[1]))
2564 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2565 weight -= 5; /* cope with negative subscript */
2568 if (!isALNUM(last_un_char)
2569 && !(last_un_char == '$' || last_un_char == '@'
2570 || last_un_char == '&')
2571 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2576 if (keyword(tmpbuf, d - tmpbuf, 0))
2579 if (un_char == last_un_char + 1)
2581 weight -= seen[un_char];
2586 if (weight >= 0) /* probably a character class */
2596 * Does all the checking to disambiguate
2598 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2599 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2601 * First argument is the stuff after the first token, e.g. "bar".
2603 * Not a method if bar is a filehandle.
2604 * Not a method if foo is a subroutine prototyped to take a filehandle.
2605 * Not a method if it's really "Foo $bar"
2606 * Method if it's "foo $bar"
2607 * Not a method if it's really "print foo $bar"
2608 * Method if it's really "foo package::" (interpreted as package->foo)
2609 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2610 * Not a method if bar is a filehandle or package, but is quoted with
2615 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2618 char *s = start + (*start == '$');
2619 char tmpbuf[sizeof PL_tokenbuf];
2627 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2631 const char *proto = SvPVX_const(cv);
2642 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2643 /* start is the beginning of the possible filehandle/object,
2644 * and s is the end of it
2645 * tmpbuf is a copy of it
2648 if (*start == '$') {
2649 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2652 len = start - SvPVX(PL_linestr);
2656 start = SvPVX(PL_linestr) + len;
2660 return *s == '(' ? FUNCMETH : METHOD;
2662 if (!keyword(tmpbuf, len, 0)) {
2663 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2667 soff = s - SvPVX(PL_linestr);
2671 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2672 if (indirgv && GvCVu(indirgv))
2674 /* filehandle or package name makes it a method */
2675 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2677 soff = s - SvPVX(PL_linestr);
2680 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2681 return 0; /* no assumptions -- "=>" quotes bearword */
2683 start_force(PL_curforce);
2684 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2685 newSVpvn(tmpbuf,len));
2686 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2688 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2693 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2695 return *s == '(' ? FUNCMETH : METHOD;
2703 * Return a string of Perl code to load the debugger. If PERL5DB
2704 * is set, it will return the contents of that, otherwise a
2705 * compile-time require of perl5db.pl.
2713 const char * const pdb = PerlEnv_getenv("PERL5DB");
2717 SETERRNO(0,SS_NORMAL);
2718 return "BEGIN { require 'perl5db.pl' }";
2724 /* Encoded script support. filter_add() effectively inserts a
2725 * 'pre-processing' function into the current source input stream.
2726 * Note that the filter function only applies to the current source file
2727 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2729 * The datasv parameter (which may be NULL) can be used to pass
2730 * private data to this instance of the filter. The filter function
2731 * can recover the SV using the FILTER_DATA macro and use it to
2732 * store private buffers and state information.
2734 * The supplied datasv parameter is upgraded to a PVIO type
2735 * and the IoDIRP/IoANY field is used to store the function pointer,
2736 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2737 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2738 * private use must be set using malloc'd pointers.
2742 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2748 if (!PL_rsfp_filters)
2749 PL_rsfp_filters = newAV();
2752 SvUPGRADE(datasv, SVt_PVIO);
2753 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2754 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2755 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2756 FPTR2DPTR(void *, IoANY(datasv)),
2757 SvPV_nolen(datasv)));
2758 av_unshift(PL_rsfp_filters, 1);
2759 av_store(PL_rsfp_filters, 0, datasv) ;
2764 /* Delete most recently added instance of this filter function. */
2766 Perl_filter_del(pTHX_ filter_t funcp)
2772 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2773 FPTR2DPTR(void*, funcp)));
2775 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2777 /* if filter is on top of stack (usual case) just pop it off */
2778 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2779 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2780 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2781 IoANY(datasv) = (void *)NULL;
2782 sv_free(av_pop(PL_rsfp_filters));
2786 /* we need to search for the correct entry and clear it */
2787 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2791 /* Invoke the idxth filter function for the current rsfp. */
2792 /* maxlen 0 = read one text line */
2794 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2799 /* This API is bad. It should have been using unsigned int for maxlen.
2800 Not sure if we want to change the API, but if not we should sanity
2801 check the value here. */
2802 const unsigned int correct_length
2811 if (!PL_rsfp_filters)
2813 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2814 /* Provide a default input filter to make life easy. */
2815 /* Note that we append to the line. This is handy. */
2816 DEBUG_P(PerlIO_printf(Perl_debug_log,
2817 "filter_read %d: from rsfp\n", idx));
2818 if (correct_length) {
2821 const int old_len = SvCUR(buf_sv);
2823 /* ensure buf_sv is large enough */
2824 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2825 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2826 correct_length)) <= 0) {
2827 if (PerlIO_error(PL_rsfp))
2828 return -1; /* error */
2830 return 0 ; /* end of file */
2832 SvCUR_set(buf_sv, old_len + len) ;
2835 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2836 if (PerlIO_error(PL_rsfp))
2837 return -1; /* error */
2839 return 0 ; /* end of file */
2842 return SvCUR(buf_sv);
2844 /* Skip this filter slot if filter has been deleted */
2845 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2846 DEBUG_P(PerlIO_printf(Perl_debug_log,
2847 "filter_read %d: skipped (filter deleted)\n",
2849 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2851 /* Get function pointer hidden within datasv */
2852 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2853 DEBUG_P(PerlIO_printf(Perl_debug_log,
2854 "filter_read %d: via function %p (%s)\n",
2855 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2856 /* Call function. The function is expected to */
2857 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2858 /* Return: <0:error, =0:eof, >0:not eof */
2859 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2863 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2866 #ifdef PERL_CR_FILTER
2867 if (!PL_rsfp_filters) {
2868 filter_add(S_cr_textfilter,NULL);
2871 if (PL_rsfp_filters) {
2873 SvCUR_set(sv, 0); /* start with empty line */
2874 if (FILTER_READ(0, sv, 0) > 0)
2875 return ( SvPVX(sv) ) ;
2880 return (sv_gets(sv, fp, append));
2884 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2889 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2893 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2894 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2896 return GvHV(gv); /* Foo:: */
2899 /* use constant CLASS => 'MyClass' */
2900 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2901 if (gv && GvCV(gv)) {
2902 SV * const sv = cv_const_sv(GvCV(gv));
2904 pkgname = SvPV_nolen_const(sv);
2907 return gv_stashpv(pkgname, 0);
2911 * S_readpipe_override
2912 * Check whether readpipe() is overriden, and generates the appropriate
2913 * optree, provided sublex_start() is called afterwards.
2916 S_readpipe_override(pTHX)
2919 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2920 yylval.ival = OP_BACKTICK;
2922 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2924 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2925 && (gv_readpipe = *gvp) != (GV*)&PL_sv_undef
2926 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2928 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2929 append_elem(OP_LIST,
2930 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2931 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2941 * The intent of this yylex wrapper is to minimize the changes to the
2942 * tokener when we aren't interested in collecting madprops. It remains
2943 * to be seen how successful this strategy will be...
2950 char *s = PL_bufptr;
2952 /* make sure PL_thiswhite is initialized */
2956 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2957 if (PL_pending_ident)
2958 return S_pending_ident(aTHX);
2960 /* previous token ate up our whitespace? */
2961 if (!PL_lasttoke && PL_nextwhite) {
2962 PL_thiswhite = PL_nextwhite;
2966 /* isolate the token, and figure out where it is without whitespace */
2967 PL_realtokenstart = -1;
2971 assert(PL_curforce < 0);
2973 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2974 if (!PL_thistoken) {
2975 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2976 PL_thistoken = newSVpvs("");
2978 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2979 PL_thistoken = newSVpvn(tstart, s - tstart);
2982 if (PL_thismad) /* install head */
2983 CURMAD('X', PL_thistoken);
2986 /* last whitespace of a sublex? */
2987 if (optype == ')' && PL_endwhite) {
2988 CURMAD('X', PL_endwhite);
2993 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2994 if (!PL_thiswhite && !PL_endwhite && !optype) {
2995 sv_free(PL_thistoken);
3000 /* put off final whitespace till peg */
3001 if (optype == ';' && !PL_rsfp) {
3002 PL_nextwhite = PL_thiswhite;
3005 else if (PL_thisopen) {
3006 CURMAD('q', PL_thisopen);
3008 sv_free(PL_thistoken);
3012 /* Store actual token text as madprop X */
3013 CURMAD('X', PL_thistoken);
3017 /* add preceding whitespace as madprop _ */
3018 CURMAD('_', PL_thiswhite);
3022 /* add quoted material as madprop = */
3023 CURMAD('=', PL_thisstuff);
3027 /* add terminating quote as madprop Q */
3028 CURMAD('Q', PL_thisclose);
3032 /* special processing based on optype */
3036 /* opval doesn't need a TOKEN since it can already store mp */
3047 append_madprops(PL_thismad, yylval.opval, 0);
3055 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3064 /* remember any fake bracket that lexer is about to discard */
3065 if (PL_lex_brackets == 1 &&
3066 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3069 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3072 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3073 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3076 break; /* don't bother looking for trailing comment */
3085 /* attach a trailing comment to its statement instead of next token */
3089 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3091 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3093 if (*s == '\n' || *s == '#') {
3094 while (s < PL_bufend && *s != '\n')
3098 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3099 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3116 /* Create new token struct. Note: opvals return early above. */
3117 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3124 S_tokenize_use(pTHX_ int is_use, char *s) {
3126 if (PL_expect != XSTATE)
3127 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3128 is_use ? "use" : "no"));
3130 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3131 s = force_version(s, TRUE);
3132 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3133 start_force(PL_curforce);
3134 NEXTVAL_NEXTTOKE.opval = NULL;
3137 else if (*s == 'v') {
3138 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3139 s = force_version(s, FALSE);
3143 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3144 s = force_version(s, FALSE);
3146 yylval.ival = is_use;
3150 static const char* const exp_name[] =
3151 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3152 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3159 Works out what to call the token just pulled out of the input
3160 stream. The yacc parser takes care of taking the ops we return and
3161 stitching them into a tree.
3167 if read an identifier
3168 if we're in a my declaration
3169 croak if they tried to say my($foo::bar)
3170 build the ops for a my() declaration
3171 if it's an access to a my() variable
3172 are we in a sort block?
3173 croak if my($a); $a <=> $b
3174 build ops for access to a my() variable
3175 if in a dq string, and they've said @foo and we can't find @foo
3177 build ops for a bareword
3178 if we already built the token before, use it.
3183 #pragma segment Perl_yylex
3189 register char *s = PL_bufptr;
3194 /* orig_keyword, gvp, and gv are initialized here because
3195 * jump to the label just_a_word_zero can bypass their
3196 * initialization later. */
3197 I32 orig_keyword = 0;
3202 SV* tmp = newSVpvs("");
3203 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3204 (IV)CopLINE(PL_curcop),
3205 lex_state_names[PL_lex_state],
3206 exp_name[PL_expect],
3207 pv_display(tmp, s, strlen(s), 0, 60));
3210 /* check if there's an identifier for us to look at */
3211 if (PL_pending_ident)
3212 return REPORT(S_pending_ident(aTHX));
3214 /* no identifier pending identification */
3216 switch (PL_lex_state) {
3218 case LEX_NORMAL: /* Some compilers will produce faster */
3219 case LEX_INTERPNORMAL: /* code if we comment these out. */
3223 /* when we've already built the next token, just pull it out of the queue */
3227 yylval = PL_nexttoke[PL_lasttoke].next_val;
3229 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3230 PL_nexttoke[PL_lasttoke].next_mad = 0;
3231 if (PL_thismad && PL_thismad->mad_key == '_') {
3232 PL_thiswhite = (SV*)PL_thismad->mad_val;
3233 PL_thismad->mad_val = 0;
3234 mad_free(PL_thismad);
3239 PL_lex_state = PL_lex_defer;
3240 PL_expect = PL_lex_expect;
3241 PL_lex_defer = LEX_NORMAL;
3242 if (!PL_nexttoke[PL_lasttoke].next_type)
3247 yylval = PL_nextval[PL_nexttoke];
3249 PL_lex_state = PL_lex_defer;
3250 PL_expect = PL_lex_expect;
3251 PL_lex_defer = LEX_NORMAL;
3255 /* FIXME - can these be merged? */
3256 return(PL_nexttoke[PL_lasttoke].next_type);
3258 return REPORT(PL_nexttype[PL_nexttoke]);
3261 /* interpolated case modifiers like \L \U, including \Q and \E.
3262 when we get here, PL_bufptr is at the \
3264 case LEX_INTERPCASEMOD:
3266 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3267 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3269 /* handle \E or end of string */
3270 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3272 if (PL_lex_casemods) {
3273 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3274 PL_lex_casestack[PL_lex_casemods] = '\0';
3276 if (PL_bufptr != PL_bufend
3277 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3279 PL_lex_state = LEX_INTERPCONCAT;
3282 PL_thistoken = newSVpvs("\\E");
3288 while (PL_bufptr != PL_bufend &&
3289 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3291 PL_thiswhite = newSVpvs("");
3292 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3296 if (PL_bufptr != PL_bufend)
3299 PL_lex_state = LEX_INTERPCONCAT;
3303 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3304 "### Saw case modifier\n"); });
3306 if (s[1] == '\\' && s[2] == 'E') {
3309 PL_thiswhite = newSVpvs("");
3310 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3313 PL_lex_state = LEX_INTERPCONCAT;
3318 if (!PL_madskills) /* when just compiling don't need correct */
3319 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3320 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3321 if ((*s == 'L' || *s == 'U') &&
3322 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3323 PL_lex_casestack[--PL_lex_casemods] = '\0';
3326 if (PL_lex_casemods > 10)
3327 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3328 PL_lex_casestack[PL_lex_casemods++] = *s;
3329 PL_lex_casestack[PL_lex_casemods] = '\0';
3330 PL_lex_state = LEX_INTERPCONCAT;
3331 start_force(PL_curforce);
3332 NEXTVAL_NEXTTOKE.ival = 0;
3334 start_force(PL_curforce);
3336 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3338 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3340 NEXTVAL_NEXTTOKE.ival = OP_LC;
3342 NEXTVAL_NEXTTOKE.ival = OP_UC;
3344 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3346 Perl_croak(aTHX_ "panic: yylex");
3348 SV* const tmpsv = newSVpvs("");
3349 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3355 if (PL_lex_starts) {
3361 sv_free(PL_thistoken);
3362 PL_thistoken = newSVpvs("");
3365 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3366 if (PL_lex_casemods == 1 && PL_lex_inpat)
3375 case LEX_INTERPPUSH:
3376 return REPORT(sublex_push());
3378 case LEX_INTERPSTART:
3379 if (PL_bufptr == PL_bufend)
3380 return REPORT(sublex_done());
3381 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3382 "### Interpolated variable\n"); });
3384 PL_lex_dojoin = (*PL_bufptr == '@');
3385 PL_lex_state = LEX_INTERPNORMAL;
3386 if (PL_lex_dojoin) {
3387 start_force(PL_curforce);
3388 NEXTVAL_NEXTTOKE.ival = 0;
3390 start_force(PL_curforce);
3391 force_ident("\"", '$');
3392 start_force(PL_curforce);
3393 NEXTVAL_NEXTTOKE.ival = 0;
3395 start_force(PL_curforce);
3396 NEXTVAL_NEXTTOKE.ival = 0;
3398 start_force(PL_curforce);
3399 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3402 if (PL_lex_starts++) {
3407 sv_free(PL_thistoken);
3408 PL_thistoken = newSVpvs("");
3411 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3412 if (!PL_lex_casemods && PL_lex_inpat)
3419 case LEX_INTERPENDMAYBE:
3420 if (intuit_more(PL_bufptr)) {
3421 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3427 if (PL_lex_dojoin) {
3428 PL_lex_dojoin = FALSE;
3429 PL_lex_state = LEX_INTERPCONCAT;
3433 sv_free(PL_thistoken);
3434 PL_thistoken = newSVpvs("");
3439 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3440 && SvEVALED(PL_lex_repl))
3442 if (PL_bufptr != PL_bufend)
3443 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3447 case LEX_INTERPCONCAT:
3449 if (PL_lex_brackets)
3450 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3452 if (PL_bufptr == PL_bufend)
3453 return REPORT(sublex_done());
3455 if (SvIVX(PL_linestr) == '\'') {
3456 SV *sv = newSVsv(PL_linestr);
3459 else if ( PL_hints & HINT_NEW_RE )
3460 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3461 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3465 s = scan_const(PL_bufptr);
3467 PL_lex_state = LEX_INTERPCASEMOD;
3469 PL_lex_state = LEX_INTERPSTART;
3472 if (s != PL_bufptr) {
3473 start_force(PL_curforce);
3475 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3477 NEXTVAL_NEXTTOKE = yylval;
3480 if (PL_lex_starts++) {
3484 sv_free(PL_thistoken);
3485 PL_thistoken = newSVpvs("");
3488 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3489 if (!PL_lex_casemods && PL_lex_inpat)
3502 PL_lex_state = LEX_NORMAL;
3503 s = scan_formline(PL_bufptr);
3504 if (!PL_lex_formbrack)
3510 PL_oldoldbufptr = PL_oldbufptr;
3516 sv_free(PL_thistoken);
3519 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3523 if (isIDFIRST_lazy_if(s,UTF))
3525 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3528 goto fake_eof; /* emulate EOF on ^D or ^Z */
3537 if (PL_lex_brackets) {
3538 yyerror((const char *)
3540 ? "Format not terminated"
3541 : "Missing right curly or square bracket"));
3543 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3544 "### Tokener got EOF\n");
3548 if (s++ < PL_bufend)
3549 goto retry; /* ignore stray nulls */
3552 if (!PL_in_eval && !PL_preambled) {
3553 PL_preambled = TRUE;
3558 sv_setpv(PL_linestr,incl_perldb());
3559 if (SvCUR(PL_linestr))
3560 sv_catpvs(PL_linestr,";");
3562 while(AvFILLp(PL_preambleav) >= 0) {
3563 SV *tmpsv = av_shift(PL_preambleav);
3564 sv_catsv(PL_linestr, tmpsv);
3565 sv_catpvs(PL_linestr, ";");
3568 sv_free((SV*)PL_preambleav);
3569 PL_preambleav = NULL;
3571 if (PL_minus_n || PL_minus_p) {
3572 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3574 sv_catpvs(PL_linestr,"chomp;");
3577 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3578 || *PL_splitstr == '"')
3579 && strchr(PL_splitstr + 1, *PL_splitstr))
3580 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3582 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3583 bytes can be used as quoting characters. :-) */
3584 const char *splits = PL_splitstr;
3585 sv_catpvs(PL_linestr, "our @F=split(q\0");
3588 if (*splits == '\\')
3589 sv_catpvn(PL_linestr, splits, 1);
3590 sv_catpvn(PL_linestr, splits, 1);
3591 } while (*splits++);
3592 /* This loop will embed the trailing NUL of
3593 PL_linestr as the last thing it does before
3595 sv_catpvs(PL_linestr, ");");
3599 sv_catpvs(PL_linestr,"our @F=split(' ');");
3603 sv_catpvs(PL_linestr,"use feature ':5.10';");
3604 sv_catpvs(PL_linestr, "\n");
3605 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3606 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3607 PL_last_lop = PL_last_uni = NULL;
3608 if (PERLDB_LINE && PL_curstash != PL_debstash)
3609 update_debugger_info(PL_linestr, NULL, 0);
3613 bof = PL_rsfp ? TRUE : FALSE;
3614 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3617 PL_realtokenstart = -1;
3620 if (PL_preprocess && !PL_in_eval)
3621 (void)PerlProc_pclose(PL_rsfp);
3622 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3623 PerlIO_clearerr(PL_rsfp);
3625 (void)PerlIO_close(PL_rsfp);
3627 PL_doextract = FALSE;
3629 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3634 sv_setpv(PL_linestr,
3637 ? ";}continue{print;}" : ";}"));
3638 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3639 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3640 PL_last_lop = PL_last_uni = NULL;
3641 PL_minus_n = PL_minus_p = 0;
3644 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3645 PL_last_lop = PL_last_uni = NULL;
3646 sv_setpvn(PL_linestr,"",0);
3647 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3649 /* If it looks like the start of a BOM or raw UTF-16,
3650 * check if it in fact is. */
3656 #ifdef PERLIO_IS_STDIO
3657 # ifdef __GNU_LIBRARY__
3658 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3659 # define FTELL_FOR_PIPE_IS_BROKEN
3663 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3664 # define FTELL_FOR_PIPE_IS_BROKEN
3669 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3670 /* This loses the possibility to detect the bof
3671 * situation on perl -P when the libc5 is being used.
3672 * Workaround? Maybe attach some extra state to PL_rsfp?
3675 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3677 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3680 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3681 s = swallow_bom((U8*)s);
3685 /* Incest with pod. */
3688 sv_catsv(PL_thiswhite, PL_linestr);
3690 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3691 sv_setpvn(PL_linestr, "", 0);
3692 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3693 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3694 PL_last_lop = PL_last_uni = NULL;
3695 PL_doextract = FALSE;
3699 } while (PL_doextract);
3700 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3701 if (PERLDB_LINE && PL_curstash != PL_debstash)
3702 update_debugger_info(PL_linestr, NULL, 0);
3703 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3704 PL_last_lop = PL_last_uni = NULL;
3705 if (CopLINE(PL_curcop) == 1) {
3706 while (s < PL_bufend && isSPACE(*s))
3708 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3712 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3716 if (*s == '#' && *(s+1) == '!')
3718 #ifdef ALTERNATE_SHEBANG
3720 static char const as[] = ALTERNATE_SHEBANG;
3721 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3722 d = s + (sizeof(as) - 1);
3724 #endif /* ALTERNATE_SHEBANG */
3733 while (*d && !isSPACE(*d))
3737 #ifdef ARG_ZERO_IS_SCRIPT
3738 if (ipathend > ipath) {
3740 * HP-UX (at least) sets argv[0] to the script name,
3741 * which makes $^X incorrect. And Digital UNIX and Linux,
3742 * at least, set argv[0] to the basename of the Perl
3743 * interpreter. So, having found "#!", we'll set it right.
3745 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3747 assert(SvPOK(x) || SvGMAGICAL(x));
3748 if (sv_eq(x, CopFILESV(PL_curcop))) {
3749 sv_setpvn(x, ipath, ipathend - ipath);
3755 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3756 const char * const lstart = SvPV_const(x,llen);
3758 bstart += blen - llen;
3759 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3760 sv_setpvn(x, ipath, ipathend - ipath);
3765 TAINT_NOT; /* $^X is always tainted, but that's OK */
3767 #endif /* ARG_ZERO_IS_SCRIPT */
3772 d = instr(s,"perl -");
3774 d = instr(s,"perl");
3776 /* avoid getting into infinite loops when shebang
3777 * line contains "Perl" rather than "perl" */
3779 for (d = ipathend-4; d >= ipath; --d) {
3780 if ((*d == 'p' || *d == 'P')
3781 && !ibcmp(d, "perl", 4))
3791 #ifdef ALTERNATE_SHEBANG
3793 * If the ALTERNATE_SHEBANG on this system starts with a
3794 * character that can be part of a Perl expression, then if
3795 * we see it but not "perl", we're probably looking at the
3796 * start of Perl code, not a request to hand off to some
3797 * other interpreter. Similarly, if "perl" is there, but
3798 * not in the first 'word' of the line, we assume the line
3799 * contains the start of the Perl program.
3801 if (d && *s != '#') {
3802 const char *c = ipath;
3803 while (*c && !strchr("; \t\r\n\f\v#", *c))
3806 d = NULL; /* "perl" not in first word; ignore */
3808 *s = '#'; /* Don't try to parse shebang line */
3810 #endif /* ALTERNATE_SHEBANG */
3811 #ifndef MACOS_TRADITIONAL
3816 !instr(s,"indir") &&
3817 instr(PL_origargv[0],"perl"))
3824 while (s < PL_bufend && isSPACE(*s))
3826 if (s < PL_bufend) {
3827 Newxz(newargv,PL_origargc+3,char*);
3829 while (s < PL_bufend && !isSPACE(*s))
3832 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3835 newargv = PL_origargv;
3838 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3840 Perl_croak(aTHX_ "Can't exec %s", ipath);
3844 while (*d && !isSPACE(*d))
3846 while (SPACE_OR_TAB(*d))
3850 const bool switches_done = PL_doswitches;
3851 const U32 oldpdb = PL_perldb;
3852 const bool oldn = PL_minus_n;
3853 const bool oldp = PL_minus_p;
3856 if (*d == 'M' || *d == 'm' || *d == 'C') {
3857 const char * const m = d;
3858 while (*d && !isSPACE(*d))
3860 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3863 d = moreswitches(d);
3865 if (PL_doswitches && !switches_done) {
3866 int argc = PL_origargc;
3867 char **argv = PL_origargv;
3870 } while (argc && argv[0][0] == '-' && argv[0][1]);
3871 init_argv_symbols(argc,argv);
3873 if ((PERLDB_LINE && !oldpdb) ||
3874 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3875 /* if we have already added "LINE: while (<>) {",
3876 we must not do it again */
3878 sv_setpvn(PL_linestr, "", 0);
3879 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3880 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3881 PL_last_lop = PL_last_uni = NULL;
3882 PL_preambled = FALSE;
3884 (void)gv_fetchfile(PL_origfilename);
3891 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3893 PL_lex_state = LEX_FORMLINE;
3898 #ifdef PERL_STRICT_CR
3899 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3901 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3903 case ' ': case '\t': case '\f': case 013:
3904 #ifdef MACOS_TRADITIONAL
3908 PL_realtokenstart = -1;
3917 PL_realtokenstart = -1;
3921 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3922 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3923 /* handle eval qq[#line 1 "foo"\n ...] */
3924 CopLINE_dec(PL_curcop);
3927 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3929 if (!PL_in_eval || PL_rsfp)
3934 while (d < PL_bufend && *d != '\n')
3938 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3939 Perl_croak(aTHX_ "panic: input overflow");
3942 PL_thiswhite = newSVpvn(s, d - s);
3947 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3949 PL_lex_state = LEX_FORMLINE;
3955 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3956 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3959 TOKEN(PEG); /* make sure any #! line is accessible */
3964 /* if (PL_madskills && PL_lex_formbrack) { */
3966 while (d < PL_bufend && *d != '\n')
3970 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3971 Perl_croak(aTHX_ "panic: input overflow");
3972 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3974 PL_thiswhite = newSVpvs("");
3975 if (CopLINE(PL_curcop) == 1) {
3976 sv_setpvn(PL_thiswhite, "", 0);
3979 sv_catpvn(PL_thiswhite, s, d - s);
3993 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4001 while (s < PL_bufend && SPACE_OR_TAB(*s))
4004 if (strnEQ(s,"=>",2)) {
4005 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4006 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4007 OPERATOR('-'); /* unary minus */
4009 PL_last_uni = PL_oldbufptr;
4011 case 'r': ftst = OP_FTEREAD; break;
4012 case 'w': ftst = OP_FTEWRITE; break;
4013 case 'x': ftst = OP_FTEEXEC; break;
4014 case 'o': ftst = OP_FTEOWNED; break;
4015 case 'R': ftst = OP_FTRREAD; break;
4016 case 'W': ftst = OP_FTRWRITE; break;
4017 case 'X': ftst = OP_FTREXEC; break;
4018 case 'O': ftst = OP_FTROWNED; break;
4019 case 'e': ftst = OP_FTIS; break;
4020 case 'z': ftst = OP_FTZERO; break;
4021 case 's': ftst = OP_FTSIZE; break;
4022 case 'f': ftst = OP_FTFILE; break;
4023 case 'd': ftst = OP_FTDIR; break;
4024 case 'l': ftst = OP_FTLINK; break;
4025 case 'p': ftst = OP_FTPIPE; break;
4026 case 'S': ftst = OP_FTSOCK; break;
4027 case 'u': ftst = OP_FTSUID; break;
4028 case 'g': ftst = OP_FTSGID; break;
4029 case 'k': ftst = OP_FTSVTX; break;
4030 case 'b': ftst = OP_FTBLK; break;
4031 case 'c': ftst = OP_FTCHR; break;
4032 case 't': ftst = OP_FTTTY; break;
4033 case 'T': ftst = OP_FTTEXT; break;
4034 case 'B': ftst = OP_FTBINARY; break;
4035 case 'M': case 'A': case 'C':
4036 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4038 case 'M': ftst = OP_FTMTIME; break;
4039 case 'A': ftst = OP_FTATIME; break;
4040 case 'C': ftst = OP_FTCTIME; break;
4048 PL_last_lop_op = (OPCODE)ftst;
4049 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4050 "### Saw file test %c\n", (int)tmp);
4055 /* Assume it was a minus followed by a one-letter named
4056 * subroutine call (or a -bareword), then. */
4057 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4058 "### '-%c' looked like a file test but was not\n",
4065 const char tmp = *s++;
4068 if (PL_expect == XOPERATOR)
4073 else if (*s == '>') {
4076 if (isIDFIRST_lazy_if(s,UTF)) {
4077 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4085 if (PL_expect == XOPERATOR)
4088 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4090 OPERATOR('-'); /* unary minus */
4096 const char tmp = *s++;
4099 if (PL_expect == XOPERATOR)
4104 if (PL_expect == XOPERATOR)
4107 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4114 if (PL_expect != XOPERATOR) {
4115 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4116 PL_expect = XOPERATOR;
4117 force_ident(PL_tokenbuf, '*');
4130 if (PL_expect == XOPERATOR) {
4134 PL_tokenbuf[0] = '%';
4135 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
4136 if (!PL_tokenbuf[1]) {
4139 PL_pending_ident = '%';
4150 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4157 const char tmp = *s++;
4163 goto just_a_word_zero_gv;
4166 switch (PL_expect) {
4172 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4174 PL_bufptr = s; /* update in case we back off */
4180 PL_expect = XTERMBLOCK;
4183 stuffstart = s - SvPVX(PL_linestr) - 1;
4187 while (isIDFIRST_lazy_if(s,UTF)) {
4190 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4191 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4192 if (tmp < 0) tmp = -tmp;
4207 sv = newSVpvn(s, len);
4209 d = scan_str(d,TRUE,TRUE);
4211 /* MUST advance bufptr here to avoid bogus
4212 "at end of line" context messages from yyerror().
4214 PL_bufptr = s + len;
4215 yyerror("Unterminated attribute parameter in attribute list");
4219 return REPORT(0); /* EOF indicator */
4223 sv_catsv(sv, PL_lex_stuff);
4224 attrs = append_elem(OP_LIST, attrs,
4225 newSVOP(OP_CONST, 0, sv));
4226 SvREFCNT_dec(PL_lex_stuff);
4227 PL_lex_stuff = NULL;
4230 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4232 if (PL_in_my == KEY_our) {
4234 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4236 /* skip to avoid loading attributes.pm */
4238 deprecate(":unique");
4241 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4244 /* NOTE: any CV attrs applied here need to be part of
4245 the CVf_BUILTIN_ATTRS define in cv.h! */
4246 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4248 CvLVALUE_on(PL_compcv);
4250 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4252 CvLOCKED_on(PL_compcv);
4254 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4256 CvMETHOD_on(PL_compcv);
4258 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4260 CvASSERTION_on(PL_compcv);
4262 /* After we've set the flags, it could be argued that
4263 we don't need to do the attributes.pm-based setting
4264 process, and shouldn't bother appending recognized
4265 flags. To experiment with that, uncomment the
4266 following "else". (Note that's already been
4267 uncommented. That keeps the above-applied built-in
4268 attributes from being intercepted (and possibly
4269 rejected) by a package's attribute routines, but is
4270 justified by the performance win for the common case
4271 of applying only built-in attributes.) */
4273 attrs = append_elem(OP_LIST, attrs,
4274 newSVOP(OP_CONST, 0,
4278 if (*s == ':' && s[1] != ':')
4281 break; /* require real whitespace or :'s */
4282 /* XXX losing whitespace on sequential attributes here */
4286 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4287 if (*s != ';' && *s != '}' && *s != tmp
4288 && (tmp != '=' || *s != ')')) {
4289 const char q = ((*s == '\'') ? '"' : '\'');
4290 /* If here for an expression, and parsed no attrs, back
4292 if (tmp == '=' && !attrs) {
4296 /* MUST advance bufptr here to avoid bogus "at end of line"
4297 context messages from yyerror().
4300 yyerror( (const char *)
4302 ? Perl_form(aTHX_ "Invalid separator character "
4303 "%c%c%c in attribute list", q, *s, q)
4304 : "Unterminated attribute list" ) );
4312 start_force(PL_curforce);
4313 NEXTVAL_NEXTTOKE.opval = attrs;
4314 CURMAD('_', PL_nextwhite);
4319 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4320 (s - SvPVX(PL_linestr)) - stuffstart);
4328 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4329 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4337 const char tmp = *s++;
4342 const char tmp = *s++;
4350 if (PL_lex_brackets <= 0)
4351 yyerror("Unmatched right square bracket");
4354 if (PL_lex_state == LEX_INTERPNORMAL) {
4355 if (PL_lex_brackets == 0) {
4356 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4357 PL_lex_state = LEX_INTERPEND;
4364 if (PL_lex_brackets > 100) {
4365 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4367 switch (PL_expect) {
4369 if (PL_lex_formbrack) {
4373 if (PL_oldoldbufptr == PL_last_lop)
4374 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4376 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4377 OPERATOR(HASHBRACK);
4379 while (s < PL_bufend && SPACE_OR_TAB(*s))
4382 PL_tokenbuf[0] = '\0';
4383 if (d < PL_bufend && *d == '-') {
4384 PL_tokenbuf[0] = '-';
4386 while (d < PL_bufend && SPACE_OR_TAB(*d))
4389 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4390 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4392 while (d < PL_bufend && SPACE_OR_TAB(*d))
4395 const char minus = (PL_tokenbuf[0] == '-');
4396 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4404 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4409 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4414 if (PL_oldoldbufptr == PL_last_lop)
4415 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4417 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;