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_ char *s)
746 CopLINE_inc(PL_curcop);
749 while (SPACE_OR_TAB(*s))
751 if (strnEQ(s, "line", 4))
755 if (SPACE_OR_TAB(*s))
759 while (SPACE_OR_TAB(*s))
767 while (SPACE_OR_TAB(*s))
769 if (*s == '"' && (t = strchr(s+1, '"'))) {
779 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
781 if (*e != '\n' && *e != '\0')
782 return; /* false alarm */
788 const char * const cf = CopFILE(PL_curcop);
789 STRLEN tmplen = cf ? strlen(cf) : 0;
790 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
791 /* must copy *{"::_<(eval N)[oldfilename:L]"}
792 * to *{"::_<newfilename"} */
793 char smallbuf[256], smallbuf2[256];
794 char *tmpbuf, *tmpbuf2;
796 STRLEN tmplen2 = strlen(s);
797 if (tmplen + 3 < sizeof smallbuf)
800 Newx(tmpbuf, tmplen + 3, char);
801 if (tmplen2 + 3 < sizeof smallbuf2)
804 Newx(tmpbuf2, tmplen2 + 3, char);
805 tmpbuf[0] = tmpbuf2[0] = '_';
806 tmpbuf[1] = tmpbuf2[1] = '<';
807 memcpy(tmpbuf + 2, cf, ++tmplen);
808 memcpy(tmpbuf2 + 2, s, ++tmplen2);
810 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
812 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
814 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
815 /* adjust ${"::_<newfilename"} to store the new file name */
816 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
817 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
818 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
821 if (tmpbuf != smallbuf) Safefree(tmpbuf);
822 if (tmpbuf2 != smallbuf2) Safefree(tmpbuf2);
825 CopFILE_free(PL_curcop);
826 CopFILE_set(PL_curcop, s);
829 CopLINE_set(PL_curcop, atoi(n)-1);
833 /* skip space before PL_thistoken */
836 S_skipspace0(pTHX_ register char *s)
843 PL_thiswhite = newSVpvs("");
844 sv_catsv(PL_thiswhite, PL_skipwhite);
845 sv_free(PL_skipwhite);
848 PL_realtokenstart = s - SvPVX(PL_linestr);
852 /* skip space after PL_thistoken */
855 S_skipspace1(pTHX_ register char *s)
857 const char *start = s;
858 I32 startoff = start - SvPVX(PL_linestr);
863 start = SvPVX(PL_linestr) + startoff;
864 if (!PL_thistoken && PL_realtokenstart >= 0) {
865 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
866 PL_thistoken = newSVpvn(tstart, start - tstart);
868 PL_realtokenstart = -1;
871 PL_nextwhite = newSVpvs("");
872 sv_catsv(PL_nextwhite, PL_skipwhite);
873 sv_free(PL_skipwhite);
880 S_skipspace2(pTHX_ register char *s, SV **svp)
883 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
884 const I32 startoff = s - SvPVX(PL_linestr);
887 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
888 if (!PL_madskills || !svp)
890 start = SvPVX(PL_linestr) + startoff;
891 if (!PL_thistoken && PL_realtokenstart >= 0) {
892 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
893 PL_thistoken = newSVpvn(tstart, start - tstart);
894 PL_realtokenstart = -1;
899 sv_setsv(*svp, PL_skipwhite);
900 sv_free(PL_skipwhite);
909 S_update_debugger_info(pTHX_ SV *orig_sv, const char *buf, STRLEN len)
911 AV *av = CopFILEAVx(PL_curcop);
913 SV * const sv = newSV(0);
914 sv_upgrade(sv, SVt_PVMG);
916 sv_setsv(sv, orig_sv);
918 sv_setpvn(sv, buf, len);
921 av_store(av, (I32)CopLINE(PL_curcop), sv);
927 * Called to gobble the appropriate amount and type of whitespace.
928 * Skips comments as well.
932 S_skipspace(pTHX_ register char *s)
937 int startoff = s - SvPVX(PL_linestr);
940 sv_free(PL_skipwhite);
945 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
946 while (s < PL_bufend && SPACE_OR_TAB(*s))
956 SSize_t oldprevlen, oldoldprevlen;
957 SSize_t oldloplen = 0, oldunilen = 0;
958 while (s < PL_bufend && isSPACE(*s)) {
959 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
964 if (s < PL_bufend && *s == '#') {
965 while (s < PL_bufend && *s != '\n')
969 if (PL_in_eval && !PL_rsfp) {
976 /* only continue to recharge the buffer if we're at the end
977 * of the buffer, we're not reading from a source filter, and
978 * we're in normal lexing mode
980 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
981 PL_lex_state == LEX_FORMLINE)
988 /* try to recharge the buffer */
990 curoff = s - SvPVX(PL_linestr);
993 if ((s = filter_gets(PL_linestr, PL_rsfp,
994 (prevlen = SvCUR(PL_linestr)))) == NULL)
997 if (PL_madskills && curoff != startoff) {
999 PL_skipwhite = newSVpvs("");
1000 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1004 /* mustn't throw out old stuff yet if madpropping */
1005 SvCUR(PL_linestr) = curoff;
1006 s = SvPVX(PL_linestr) + curoff;
1008 if (curoff && s[-1] == '\n')
1012 /* end of file. Add on the -p or -n magic */
1013 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1016 sv_catpv(PL_linestr,
1017 ";}continue{print or die qq(-p destination: $!\\n);}");
1019 sv_setpv(PL_linestr,
1020 ";}continue{print or die qq(-p destination: $!\\n);}");
1022 PL_minus_n = PL_minus_p = 0;
1024 else if (PL_minus_n) {
1026 sv_catpvn(PL_linestr, ";}", 2);
1028 sv_setpvn(PL_linestr, ";}", 2);
1034 sv_catpvn(PL_linestr,";", 1);
1036 sv_setpvn(PL_linestr,";", 1);
1039 /* reset variables for next time we lex */
1040 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1046 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1047 PL_last_lop = PL_last_uni = NULL;
1049 /* Close the filehandle. Could be from -P preprocessor,
1050 * STDIN, or a regular file. If we were reading code from
1051 * STDIN (because the commandline held no -e or filename)
1052 * then we don't close it, we reset it so the code can
1053 * read from STDIN too.
1056 if (PL_preprocess && !PL_in_eval)
1057 (void)PerlProc_pclose(PL_rsfp);
1058 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1059 PerlIO_clearerr(PL_rsfp);
1061 (void)PerlIO_close(PL_rsfp);
1066 /* not at end of file, so we only read another line */
1067 /* make corresponding updates to old pointers, for yyerror() */
1068 oldprevlen = PL_oldbufptr - PL_bufend;
1069 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1071 oldunilen = PL_last_uni - PL_bufend;
1073 oldloplen = PL_last_lop - PL_bufend;
1074 PL_linestart = PL_bufptr = s + prevlen;
1075 PL_bufend = s + SvCUR(PL_linestr);
1077 PL_oldbufptr = s + oldprevlen;
1078 PL_oldoldbufptr = s + oldoldprevlen;
1080 PL_last_uni = s + oldunilen;
1082 PL_last_lop = s + oldloplen;
1085 /* debugger active and we're not compiling the debugger code,
1086 * so store the line into the debugger's array of lines
1088 if (PERLDB_LINE && PL_curstash != PL_debstash)
1089 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1096 PL_skipwhite = newSVpvs("");
1097 curoff = s - SvPVX(PL_linestr);
1098 if (curoff - startoff)
1099 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1108 * Check the unary operators to ensure there's no ambiguity in how they're
1109 * used. An ambiguous piece of code would be:
1111 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1112 * the +5 is its argument.
1122 if (PL_oldoldbufptr != PL_last_uni)
1124 while (isSPACE(*PL_last_uni))
1127 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1129 if ((t = strchr(s, '(')) && t < PL_bufptr)
1132 if (ckWARN_d(WARN_AMBIGUOUS)){
1133 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1134 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1135 (int)(s - PL_last_uni), PL_last_uni);
1140 * LOP : macro to build a list operator. Its behaviour has been replaced
1141 * with a subroutine, S_lop() for which LOP is just another name.
1144 #define LOP(f,x) return lop(f,x,s)
1148 * Build a list operator (or something that might be one). The rules:
1149 * - if we have a next token, then it's a list operator [why?]
1150 * - if the next thing is an opening paren, then it's a function
1151 * - else it's a list operator
1155 S_lop(pTHX_ I32 f, int x, char *s)
1162 PL_last_lop = PL_oldbufptr;
1163 PL_last_lop_op = (OPCODE)f;
1166 return REPORT(LSTOP);
1169 return REPORT(LSTOP);
1172 return REPORT(FUNC);
1175 return REPORT(FUNC);
1177 return REPORT(LSTOP);
1183 * Sets up for an eventual force_next(). start_force(0) basically does
1184 * an unshift, while start_force(-1) does a push. yylex removes items
1189 S_start_force(pTHX_ int where)
1193 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1194 where = PL_lasttoke;
1195 assert(PL_curforce < 0 || PL_curforce == where);
1196 if (PL_curforce != where) {
1197 for (i = PL_lasttoke; i > where; --i) {
1198 PL_nexttoke[i] = PL_nexttoke[i-1];
1202 if (PL_curforce < 0) /* in case of duplicate start_force() */
1203 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1204 PL_curforce = where;
1207 curmad('^', newSVpvs(""));
1208 CURMAD('_', PL_nextwhite);
1213 S_curmad(pTHX_ char slot, SV *sv)
1219 if (PL_curforce < 0)
1220 where = &PL_thismad;
1222 where = &PL_nexttoke[PL_curforce].next_mad;
1225 sv_setpvn(sv, "", 0);
1228 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1230 else if (PL_encoding) {
1231 sv_recode_to_utf8(sv, PL_encoding);
1236 /* keep a slot open for the head of the list? */
1237 if (slot != '_' && *where && (*where)->mad_key == '^') {
1238 (*where)->mad_key = slot;
1239 sv_free((*where)->mad_val);
1240 (*where)->mad_val = (void*)sv;
1243 addmad(newMADsv(slot, sv), where, 0);
1246 # define start_force(where) NOOP
1247 # define curmad(slot, sv) NOOP
1252 * When the lexer realizes it knows the next token (for instance,
1253 * it is reordering tokens for the parser) then it can call S_force_next
1254 * to know what token to return the next time the lexer is called. Caller
1255 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1256 * and possibly PL_expect to ensure the lexer handles the token correctly.
1260 S_force_next(pTHX_ I32 type)
1264 if (PL_curforce < 0)
1265 start_force(PL_lasttoke);
1266 PL_nexttoke[PL_curforce].next_type = type;
1267 if (PL_lex_state != LEX_KNOWNEXT)
1268 PL_lex_defer = PL_lex_state;
1269 PL_lex_state = LEX_KNOWNEXT;
1270 PL_lex_expect = PL_expect;
1273 PL_nexttype[PL_nexttoke] = type;
1275 if (PL_lex_state != LEX_KNOWNEXT) {
1276 PL_lex_defer = PL_lex_state;
1277 PL_lex_expect = PL_expect;
1278 PL_lex_state = LEX_KNOWNEXT;
1284 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1287 SV * const sv = newSVpvn(start,len);
1288 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1295 * When the lexer knows the next thing is a word (for instance, it has
1296 * just seen -> and it knows that the next char is a word char, then
1297 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1301 * char *start : buffer position (must be within PL_linestr)
1302 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1303 * int check_keyword : if true, Perl checks to make sure the word isn't
1304 * a keyword (do this if the word is a label, e.g. goto FOO)
1305 * int allow_pack : if true, : characters will also be allowed (require,
1306 * use, etc. do this)
1307 * int allow_initial_tick : used by the "sub" lexer only.
1311 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1317 start = SKIPSPACE1(start);
1319 if (isIDFIRST_lazy_if(s,UTF) ||
1320 (allow_pack && *s == ':') ||
1321 (allow_initial_tick && *s == '\'') )
1323 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1324 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1326 start_force(PL_curforce);
1328 curmad('X', newSVpvn(start,s-start));
1329 if (token == METHOD) {
1334 PL_expect = XOPERATOR;
1337 NEXTVAL_NEXTTOKE.opval
1338 = (OP*)newSVOP(OP_CONST,0,
1339 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1340 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1348 * Called when the lexer wants $foo *foo &foo etc, but the program
1349 * text only contains the "foo" portion. The first argument is a pointer
1350 * to the "foo", and the second argument is the type symbol to prefix.
1351 * Forces the next token to be a "WORD".
1352 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1356 S_force_ident(pTHX_ register const char *s, int kind)
1360 const STRLEN len = strlen(s);
1361 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1362 start_force(PL_curforce);
1363 NEXTVAL_NEXTTOKE.opval = o;
1366 o->op_private = OPpCONST_ENTERED;
1367 /* XXX see note in pp_entereval() for why we forgo typo
1368 warnings if the symbol must be introduced in an eval.
1370 gv_fetchpvn_flags(s, len,
1371 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1373 kind == '$' ? SVt_PV :
1374 kind == '@' ? SVt_PVAV :
1375 kind == '%' ? SVt_PVHV :
1383 Perl_str_to_version(pTHX_ SV *sv)
1388 const char *start = SvPV_const(sv,len);
1389 const char * const end = start + len;
1390 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1391 while (start < end) {
1395 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1400 retval += ((NV)n)/nshift;
1409 * Forces the next token to be a version number.
1410 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1411 * and if "guessing" is TRUE, then no new token is created (and the caller
1412 * must use an alternative parsing method).
1416 S_force_version(pTHX_ char *s, int guessing)
1422 I32 startoff = s - SvPVX(PL_linestr);
1431 while (isDIGIT(*d) || *d == '_' || *d == '.')
1435 start_force(PL_curforce);
1436 curmad('X', newSVpvn(s,d-s));
1439 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1441 s = scan_num(s, &yylval);
1442 version = yylval.opval;
1443 ver = cSVOPx(version)->op_sv;
1444 if (SvPOK(ver) && !SvNIOK(ver)) {
1445 SvUPGRADE(ver, SVt_PVNV);
1446 SvNV_set(ver, str_to_version(ver));
1447 SvNOK_on(ver); /* hint that it is a version */
1450 else if (guessing) {
1453 sv_free(PL_nextwhite); /* let next token collect whitespace */
1455 s = SvPVX(PL_linestr) + startoff;
1463 if (PL_madskills && !version) {
1464 sv_free(PL_nextwhite); /* let next token collect whitespace */
1466 s = SvPVX(PL_linestr) + startoff;
1469 /* NOTE: The parser sees the package name and the VERSION swapped */
1470 start_force(PL_curforce);
1471 NEXTVAL_NEXTTOKE.opval = version;
1479 * Tokenize a quoted string passed in as an SV. It finds the next
1480 * chunk, up to end of string or a backslash. It may make a new
1481 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1486 S_tokeq(pTHX_ SV *sv)
1490 register char *send;
1498 s = SvPV_force(sv, len);
1499 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1502 while (s < send && *s != '\\')
1507 if ( PL_hints & HINT_NEW_STRING ) {
1508 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1514 if (s + 1 < send && (s[1] == '\\'))
1515 s++; /* all that, just for this */
1520 SvCUR_set(sv, d - SvPVX_const(sv));
1522 if ( PL_hints & HINT_NEW_STRING )
1523 return new_constant(NULL, 0, "q", sv, pv, "q");
1528 * Now come three functions related to double-quote context,
1529 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1530 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1531 * interact with PL_lex_state, and create fake ( ... ) argument lists
1532 * to handle functions and concatenation.
1533 * They assume that whoever calls them will be setting up a fake
1534 * join call, because each subthing puts a ',' after it. This lets
1537 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1539 * (I'm not sure whether the spurious commas at the end of lcfirst's
1540 * arguments and join's arguments are created or not).
1545 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1547 * Pattern matching will set PL_lex_op to the pattern-matching op to
1548 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1550 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1552 * Everything else becomes a FUNC.
1554 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1555 * had an OP_CONST or OP_READLINE). This just sets us up for a
1556 * call to S_sublex_push().
1560 S_sublex_start(pTHX)
1563 register const I32 op_type = yylval.ival;
1565 if (op_type == OP_NULL) {
1566 yylval.opval = PL_lex_op;
1570 if (op_type == OP_CONST || op_type == OP_READLINE) {
1571 SV *sv = tokeq(PL_lex_stuff);
1573 if (SvTYPE(sv) == SVt_PVIV) {
1574 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1576 const char * const p = SvPV_const(sv, len);
1577 SV * const nsv = newSVpvn(p, len);
1583 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1584 PL_lex_stuff = NULL;
1585 /* Allow <FH> // "foo" */
1586 if (op_type == OP_READLINE)
1587 PL_expect = XTERMORDORDOR;
1590 else if (op_type == OP_BACKTICK && PL_lex_op) {
1591 /* readpipe() vas overriden */
1592 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1593 yylval.opval = PL_lex_op;
1595 PL_lex_stuff = NULL;
1599 PL_sublex_info.super_state = PL_lex_state;
1600 PL_sublex_info.sub_inwhat = op_type;
1601 PL_sublex_info.sub_op = PL_lex_op;
1602 PL_lex_state = LEX_INTERPPUSH;
1606 yylval.opval = PL_lex_op;
1616 * Create a new scope to save the lexing state. The scope will be
1617 * ended in S_sublex_done. Returns a '(', starting the function arguments
1618 * to the uc, lc, etc. found before.
1619 * Sets PL_lex_state to LEX_INTERPCONCAT.
1628 PL_lex_state = PL_sublex_info.super_state;
1629 SAVEI32(PL_lex_dojoin);
1630 SAVEI32(PL_lex_brackets);
1631 SAVEI32(PL_lex_casemods);
1632 SAVEI32(PL_lex_starts);
1633 SAVEI32(PL_lex_state);
1634 SAVEVPTR(PL_lex_inpat);
1635 SAVEI32(PL_lex_inwhat);
1636 SAVECOPLINE(PL_curcop);
1637 SAVEPPTR(PL_bufptr);
1638 SAVEPPTR(PL_bufend);
1639 SAVEPPTR(PL_oldbufptr);
1640 SAVEPPTR(PL_oldoldbufptr);
1641 SAVEPPTR(PL_last_lop);
1642 SAVEPPTR(PL_last_uni);
1643 SAVEPPTR(PL_linestart);
1644 SAVESPTR(PL_linestr);
1645 SAVEGENERICPV(PL_lex_brackstack);
1646 SAVEGENERICPV(PL_lex_casestack);
1648 PL_linestr = PL_lex_stuff;
1649 PL_lex_stuff = NULL;
1651 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1652 = SvPVX(PL_linestr);
1653 PL_bufend += SvCUR(PL_linestr);
1654 PL_last_lop = PL_last_uni = NULL;
1655 SAVEFREESV(PL_linestr);
1657 PL_lex_dojoin = FALSE;
1658 PL_lex_brackets = 0;
1659 Newx(PL_lex_brackstack, 120, char);
1660 Newx(PL_lex_casestack, 12, char);
1661 PL_lex_casemods = 0;
1662 *PL_lex_casestack = '\0';
1664 PL_lex_state = LEX_INTERPCONCAT;
1665 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1667 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1668 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1669 PL_lex_inpat = PL_sublex_info.sub_op;
1671 PL_lex_inpat = NULL;
1678 * Restores lexer state after a S_sublex_push.
1685 if (!PL_lex_starts++) {
1686 SV * const sv = newSVpvs("");
1687 if (SvUTF8(PL_linestr))
1689 PL_expect = XOPERATOR;
1690 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1694 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1695 PL_lex_state = LEX_INTERPCASEMOD;
1699 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1700 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1701 PL_linestr = PL_lex_repl;
1703 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1704 PL_bufend += SvCUR(PL_linestr);
1705 PL_last_lop = PL_last_uni = NULL;
1706 SAVEFREESV(PL_linestr);
1707 PL_lex_dojoin = FALSE;
1708 PL_lex_brackets = 0;
1709 PL_lex_casemods = 0;
1710 *PL_lex_casestack = '\0';
1712 if (SvEVALED(PL_lex_repl)) {
1713 PL_lex_state = LEX_INTERPNORMAL;
1715 /* we don't clear PL_lex_repl here, so that we can check later
1716 whether this is an evalled subst; that means we rely on the
1717 logic to ensure sublex_done() is called again only via the
1718 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1721 PL_lex_state = LEX_INTERPCONCAT;
1731 PL_endwhite = newSVpvs("");
1732 sv_catsv(PL_endwhite, PL_thiswhite);
1736 sv_setpvn(PL_thistoken,"",0);
1738 PL_realtokenstart = -1;
1742 PL_bufend = SvPVX(PL_linestr);
1743 PL_bufend += SvCUR(PL_linestr);
1744 PL_expect = XOPERATOR;
1745 PL_sublex_info.sub_inwhat = 0;
1753 Extracts a pattern, double-quoted string, or transliteration. This
1756 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1757 processing a pattern (PL_lex_inpat is true), a transliteration
1758 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1760 Returns a pointer to the character scanned up to. If this is
1761 advanced from the start pointer supplied (i.e. if anything was
1762 successfully parsed), will leave an OP for the substring scanned
1763 in yylval. Caller must intuit reason for not parsing further
1764 by looking at the next characters herself.
1768 double-quoted style: \r and \n
1769 regexp special ones: \D \s
1772 case and quoting: \U \Q \E
1773 stops on @ and $, but not for $ as tail anchor
1775 In transliterations:
1776 characters are VERY literal, except for - not at the start or end
1777 of the string, which indicates a range. If the range is in bytes,
1778 scan_const expands the range to the full set of intermediate
1779 characters. If the range is in utf8, the hyphen is replaced with
1780 a certain range mark which will be handled by pmtrans() in op.c.
1782 In double-quoted strings:
1784 double-quoted style: \r and \n
1786 deprecated backrefs: \1 (in substitution replacements)
1787 case and quoting: \U \Q \E
1790 scan_const does *not* construct ops to handle interpolated strings.
1791 It stops processing as soon as it finds an embedded $ or @ variable
1792 and leaves it to the caller to work out what's going on.
1794 embedded arrays (whether in pattern or not) could be:
1795 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1797 $ in double-quoted strings must be the symbol of an embedded scalar.
1799 $ in pattern could be $foo or could be tail anchor. Assumption:
1800 it's a tail anchor if $ is the last thing in the string, or if it's
1801 followed by one of "()| \r\n\t"
1803 \1 (backreferences) are turned into $1
1805 The structure of the code is
1806 while (there's a character to process) {
1807 handle transliteration ranges
1808 skip regexp comments /(?#comment)/ and codes /(?{code})/
1809 skip #-initiated comments in //x patterns
1810 check for embedded arrays
1811 check for embedded scalars
1813 leave intact backslashes from leaveit (below)
1814 deprecate \1 in substitution replacements
1815 handle string-changing backslashes \l \U \Q \E, etc.
1816 switch (what was escaped) {
1817 handle \- in a transliteration (becomes a literal -)
1818 handle \132 (octal characters)
1819 handle \x15 and \x{1234} (hex characters)
1820 handle \N{name} (named characters)
1821 handle \cV (control characters)
1822 handle printf-style backslashes (\f, \r, \n, etc)
1824 } (end if backslash)
1825 } (end while character to read)
1830 S_scan_const(pTHX_ char *start)
1833 register char *send = PL_bufend; /* end of the constant */
1834 SV *sv = newSV(send - start); /* sv for the constant */
1835 register char *s = start; /* start of the constant */
1836 register char *d = SvPVX(sv); /* destination for copies */
1837 bool dorange = FALSE; /* are we in a translit range? */
1838 bool didrange = FALSE; /* did we just finish a range? */
1839 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1840 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1843 UV literal_endpoint = 0;
1844 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1847 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1848 /* If we are doing a trans and we know we want UTF8 set expectation */
1849 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1850 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1854 while (s < send || dorange) {
1855 /* get transliterations out of the way (they're most literal) */
1856 if (PL_lex_inwhat == OP_TRANS) {
1857 /* expand a range A-Z to the full set of characters. AIE! */
1859 I32 i; /* current expanded character */
1860 I32 min; /* first character in range */
1861 I32 max; /* last character in range */
1872 char * const c = (char*)utf8_hop((U8*)d, -1);
1876 *c = (char)UTF_TO_NATIVE(0xff);
1877 /* mark the range as done, and continue */
1883 i = d - SvPVX_const(sv); /* remember current offset */
1886 SvLEN(sv) + (has_utf8 ?
1887 (512 - UTF_CONTINUATION_MARK +
1890 /* How many two-byte within 0..255: 128 in UTF-8,
1891 * 96 in UTF-8-mod. */
1893 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1895 d = SvPVX(sv) + i; /* refresh d after realloc */
1899 for (j = 0; j <= 1; j++) {
1900 char * const c = (char*)utf8_hop((U8*)d, -1);
1901 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1907 max = (U8)0xff; /* only to \xff */
1908 uvmax = uv; /* \x{100} to uvmax */
1910 d = c; /* eat endpoint chars */
1915 d -= 2; /* eat the first char and the - */
1916 min = (U8)*d; /* first char in range */
1917 max = (U8)d[1]; /* last char in range */
1924 "Invalid range \"%c-%c\" in transliteration operator",
1925 (char)min, (char)max);
1929 if (literal_endpoint == 2 &&
1930 ((isLOWER(min) && isLOWER(max)) ||
1931 (isUPPER(min) && isUPPER(max)))) {
1933 for (i = min; i <= max; i++)
1935 *d++ = NATIVE_TO_NEED(has_utf8,i);
1937 for (i = min; i <= max; i++)
1939 *d++ = NATIVE_TO_NEED(has_utf8,i);
1944 for (i = min; i <= max; i++)
1947 const U8 ch = (U8)NATIVE_TO_UTF(i);
1948 if (UNI_IS_INVARIANT(ch))
1951 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
1952 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
1961 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
1963 *d++ = (char)UTF_TO_NATIVE(0xff);
1965 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
1969 /* mark the range as done, and continue */
1973 literal_endpoint = 0;
1978 /* range begins (ignore - as first or last char) */
1979 else if (*s == '-' && s+1 < send && s != start) {
1981 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
1988 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
1998 literal_endpoint = 0;
1999 native_range = TRUE;
2004 /* if we get here, we're not doing a transliteration */
2006 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2007 except for the last char, which will be done separately. */
2008 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2010 while (s+1 < send && *s != ')')
2011 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2013 else if (s[2] == '{' /* This should match regcomp.c */
2014 || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))
2017 char *regparse = s + (s[2] == '{' ? 3 : 4);
2020 while (count && (c = *regparse)) {
2021 if (c == '\\' && regparse[1])
2029 if (*regparse != ')')
2030 regparse--; /* Leave one char for continuation. */
2031 while (s < regparse)
2032 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2036 /* likewise skip #-initiated comments in //x patterns */
2037 else if (*s == '#' && PL_lex_inpat &&
2038 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2039 while (s+1 < send && *s != '\n')
2040 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2043 /* check for embedded arrays
2044 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2046 else if (*s == '@' && s[1]) {
2047 if (isALNUM_lazy_if(s+1,UTF))
2049 if (strchr(":'{$", s[1]))
2051 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2052 break; /* in regexp, neither @+ nor @- are interpolated */
2055 /* check for embedded scalars. only stop if we're sure it's a
2058 else if (*s == '$') {
2059 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2061 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2062 break; /* in regexp, $ might be tail anchor */
2065 /* End of else if chain - OP_TRANS rejoin rest */
2068 if (*s == '\\' && s+1 < send) {
2071 /* deprecate \1 in strings and substitution replacements */
2072 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2073 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2075 if (ckWARN(WARN_SYNTAX))
2076 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2081 /* string-change backslash escapes */
2082 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2086 /* skip any other backslash escapes in a pattern */
2087 else if (PL_lex_inpat) {
2088 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2089 goto default_action;
2092 /* if we get here, it's either a quoted -, or a digit */
2095 /* quoted - in transliterations */
2097 if (PL_lex_inwhat == OP_TRANS) {
2104 if ((isALPHA(*s) || isDIGIT(*s)) &&
2106 Perl_warner(aTHX_ packWARN(WARN_MISC),
2107 "Unrecognized escape \\%c passed through",
2109 /* default action is to copy the quoted character */
2110 goto default_action;
2113 /* \132 indicates an octal constant */
2114 case '0': case '1': case '2': case '3':
2115 case '4': case '5': case '6': case '7':
2119 uv = grok_oct(s, &len, &flags, NULL);
2122 goto NUM_ESCAPE_INSERT;
2124 /* \x24 indicates a hex constant */
2128 char* const e = strchr(s, '}');
2129 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2130 PERL_SCAN_DISALLOW_PREFIX;
2135 yyerror("Missing right brace on \\x{}");
2139 uv = grok_hex(s, &len, &flags, NULL);
2145 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2146 uv = grok_hex(s, &len, &flags, NULL);
2152 /* Insert oct or hex escaped character.
2153 * There will always enough room in sv since such
2154 * escapes will be longer than any UTF-8 sequence
2155 * they can end up as. */
2157 /* We need to map to chars to ASCII before doing the tests
2160 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2161 if (!has_utf8 && uv > 255) {
2162 /* Might need to recode whatever we have
2163 * accumulated so far if it contains any
2166 * (Can't we keep track of that and avoid
2167 * this rescan? --jhi)
2171 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2172 if (!NATIVE_IS_INVARIANT(*c)) {
2177 const STRLEN offset = d - SvPVX_const(sv);
2179 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2183 while (src >= (const U8 *)SvPVX_const(sv)) {
2184 if (!NATIVE_IS_INVARIANT(*src)) {
2185 const U8 ch = NATIVE_TO_ASCII(*src);
2186 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2187 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2197 if (has_utf8 || uv > 255) {
2198 d = (char*)uvchr_to_utf8((U8*)d, uv);
2200 if (PL_lex_inwhat == OP_TRANS &&
2201 PL_sublex_info.sub_op) {
2202 PL_sublex_info.sub_op->op_private |=
2203 (PL_lex_repl ? OPpTRANS_FROM_UTF
2207 if (uv > 255 && !dorange)
2208 native_range = FALSE;
2220 /* \N{LATIN SMALL LETTER A} is a named character */
2224 char* e = strchr(s, '}');
2231 yyerror("Missing right brace on \\N{}");
2235 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2237 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2238 PERL_SCAN_DISALLOW_PREFIX;
2241 uv = grok_hex(s, &len, &flags, NULL);
2242 if ( e > s && len != (STRLEN)(e - s) ) {
2246 goto NUM_ESCAPE_INSERT;
2248 res = newSVpvn(s + 1, e - s - 1);
2249 type = newSVpvn(s - 2,e - s + 3);
2250 res = new_constant( NULL, 0, "charnames",
2251 res, NULL, SvPVX(type) );
2254 sv_utf8_upgrade(res);
2255 str = SvPV_const(res,len);
2256 #ifdef EBCDIC_NEVER_MIND
2257 /* charnames uses pack U and that has been
2258 * recently changed to do the below uni->native
2259 * mapping, so this would be redundant (and wrong,
2260 * the code point would be doubly converted).
2261 * But leave this in just in case the pack U change
2262 * gets revoked, but the semantics is still
2263 * desireable for charnames. --jhi */
2265 UV uv = utf8_to_uvchr((const U8*)str, 0);
2268 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2270 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2271 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2272 str = SvPV_const(res, len);
2276 if (!has_utf8 && SvUTF8(res)) {
2277 const char * const ostart = SvPVX_const(sv);
2278 SvCUR_set(sv, d - ostart);
2281 sv_utf8_upgrade(sv);
2282 /* this just broke our allocation above... */
2283 SvGROW(sv, (STRLEN)(send - start));
2284 d = SvPVX(sv) + SvCUR(sv);
2287 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2288 const char * const odest = SvPVX_const(sv);
2290 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2291 d = SvPVX(sv) + (d - odest);
2295 native_range = FALSE; /* \N{} is guessed to be Unicode */
2297 Copy(str, d, len, char);
2304 yyerror("Missing braces on \\N{}");
2307 /* \c is a control character */
2316 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2319 yyerror("Missing control char name in \\c");
2323 /* printf-style backslashes, formfeeds, newlines, etc */
2325 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2328 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2331 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2334 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2337 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2340 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2343 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2349 } /* end if (backslash) */
2356 /* If we started with encoded form, or already know we want it
2357 and then encode the next character */
2358 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2360 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2361 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2364 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2365 const STRLEN off = d - SvPVX_const(sv);
2366 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2368 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2371 if (uv > 255 && !dorange)
2372 native_range = FALSE;
2376 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2378 } /* while loop to process each character */
2380 /* terminate the string and set up the sv */
2382 SvCUR_set(sv, d - SvPVX_const(sv));
2383 if (SvCUR(sv) >= SvLEN(sv))
2384 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2387 if (PL_encoding && !has_utf8) {
2388 sv_recode_to_utf8(sv, PL_encoding);
2394 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2395 PL_sublex_info.sub_op->op_private |=
2396 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2400 /* shrink the sv if we allocated more than we used */
2401 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2402 SvPV_shrink_to_cur(sv);
2405 /* return the substring (via yylval) only if we parsed anything */
2406 if (s > PL_bufptr) {
2407 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2408 sv = new_constant(start, s - start,
2409 (const char *)(PL_lex_inpat ? "qr" : "q"),
2412 (( PL_lex_inwhat == OP_TRANS
2414 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2417 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2424 * Returns TRUE if there's more to the expression (e.g., a subscript),
2427 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2429 * ->[ and ->{ return TRUE
2430 * { and [ outside a pattern are always subscripts, so return TRUE
2431 * if we're outside a pattern and it's not { or [, then return FALSE
2432 * if we're in a pattern and the first char is a {
2433 * {4,5} (any digits around the comma) returns FALSE
2434 * if we're in a pattern and the first char is a [
2436 * [SOMETHING] has a funky algorithm to decide whether it's a
2437 * character class or not. It has to deal with things like
2438 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2439 * anything else returns TRUE
2442 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2445 S_intuit_more(pTHX_ register char *s)
2448 if (PL_lex_brackets)
2450 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2452 if (*s != '{' && *s != '[')
2457 /* In a pattern, so maybe we have {n,m}. */
2474 /* On the other hand, maybe we have a character class */
2477 if (*s == ']' || *s == '^')
2480 /* this is terrifying, and it works */
2481 int weight = 2; /* let's weigh the evidence */
2483 unsigned char un_char = 255, last_un_char;
2484 const char * const send = strchr(s,']');
2485 char tmpbuf[sizeof PL_tokenbuf * 4];
2487 if (!send) /* has to be an expression */
2490 Zero(seen,256,char);
2493 else if (isDIGIT(*s)) {
2495 if (isDIGIT(s[1]) && s[2] == ']')
2501 for (; s < send; s++) {
2502 last_un_char = un_char;
2503 un_char = (unsigned char)*s;
2508 weight -= seen[un_char] * 10;
2509 if (isALNUM_lazy_if(s+1,UTF)) {
2511 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2512 len = (int)strlen(tmpbuf);
2513 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2518 else if (*s == '$' && s[1] &&
2519 strchr("[#!%*<>()-=",s[1])) {
2520 if (/*{*/ strchr("])} =",s[2]))
2529 if (strchr("wds]",s[1]))
2531 else if (seen[(U8)'\''] || seen[(U8)'"'])
2533 else if (strchr("rnftbxcav",s[1]))
2535 else if (isDIGIT(s[1])) {
2537 while (s[1] && isDIGIT(s[1]))
2547 if (strchr("aA01! ",last_un_char))
2549 if (strchr("zZ79~",s[1]))
2551 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2552 weight -= 5; /* cope with negative subscript */
2555 if (!isALNUM(last_un_char)
2556 && !(last_un_char == '$' || last_un_char == '@'
2557 || last_un_char == '&')
2558 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2563 if (keyword(tmpbuf, d - tmpbuf, 0))
2566 if (un_char == last_un_char + 1)
2568 weight -= seen[un_char];
2573 if (weight >= 0) /* probably a character class */
2583 * Does all the checking to disambiguate
2585 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2586 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2588 * First argument is the stuff after the first token, e.g. "bar".
2590 * Not a method if bar is a filehandle.
2591 * Not a method if foo is a subroutine prototyped to take a filehandle.
2592 * Not a method if it's really "Foo $bar"
2593 * Method if it's "foo $bar"
2594 * Not a method if it's really "print foo $bar"
2595 * Method if it's really "foo package::" (interpreted as package->foo)
2596 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2597 * Not a method if bar is a filehandle or package, but is quoted with
2602 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2605 char *s = start + (*start == '$');
2606 char tmpbuf[sizeof PL_tokenbuf];
2614 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2618 const char *proto = SvPVX_const(cv);
2629 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2630 /* start is the beginning of the possible filehandle/object,
2631 * and s is the end of it
2632 * tmpbuf is a copy of it
2635 if (*start == '$') {
2636 if (gv || PL_last_lop_op == OP_PRINT || isUPPER(*PL_tokenbuf))
2639 len = start - SvPVX(PL_linestr);
2643 start = SvPVX(PL_linestr) + len;
2647 return *s == '(' ? FUNCMETH : METHOD;
2649 if (!keyword(tmpbuf, len, 0)) {
2650 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2654 soff = s - SvPVX(PL_linestr);
2658 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2659 if (indirgv && GvCVu(indirgv))
2661 /* filehandle or package name makes it a method */
2662 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2664 soff = s - SvPVX(PL_linestr);
2667 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2668 return 0; /* no assumptions -- "=>" quotes bearword */
2670 start_force(PL_curforce);
2671 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2672 newSVpvn(tmpbuf,len));
2673 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2675 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2680 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2682 return *s == '(' ? FUNCMETH : METHOD;
2690 * Return a string of Perl code to load the debugger. If PERL5DB
2691 * is set, it will return the contents of that, otherwise a
2692 * compile-time require of perl5db.pl.
2700 const char * const pdb = PerlEnv_getenv("PERL5DB");
2704 SETERRNO(0,SS_NORMAL);
2705 return "BEGIN { require 'perl5db.pl' }";
2711 /* Encoded script support. filter_add() effectively inserts a
2712 * 'pre-processing' function into the current source input stream.
2713 * Note that the filter function only applies to the current source file
2714 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2716 * The datasv parameter (which may be NULL) can be used to pass
2717 * private data to this instance of the filter. The filter function
2718 * can recover the SV using the FILTER_DATA macro and use it to
2719 * store private buffers and state information.
2721 * The supplied datasv parameter is upgraded to a PVIO type
2722 * and the IoDIRP/IoANY field is used to store the function pointer,
2723 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2724 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2725 * private use must be set using malloc'd pointers.
2729 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2735 if (!PL_rsfp_filters)
2736 PL_rsfp_filters = newAV();
2739 SvUPGRADE(datasv, SVt_PVIO);
2740 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2741 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2742 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2743 FPTR2DPTR(void *, IoANY(datasv)),
2744 SvPV_nolen(datasv)));
2745 av_unshift(PL_rsfp_filters, 1);
2746 av_store(PL_rsfp_filters, 0, datasv) ;
2751 /* Delete most recently added instance of this filter function. */
2753 Perl_filter_del(pTHX_ filter_t funcp)
2759 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2760 FPTR2DPTR(void*, funcp)));
2762 if (!PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2764 /* if filter is on top of stack (usual case) just pop it off */
2765 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2766 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2767 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2768 IoANY(datasv) = (void *)NULL;
2769 sv_free(av_pop(PL_rsfp_filters));
2773 /* we need to search for the correct entry and clear it */
2774 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2778 /* Invoke the idxth filter function for the current rsfp. */
2779 /* maxlen 0 = read one text line */
2781 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2786 /* This API is bad. It should have been using unsigned int for maxlen.
2787 Not sure if we want to change the API, but if not we should sanity
2788 check the value here. */
2789 const unsigned int correct_length
2798 if (!PL_rsfp_filters)
2800 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2801 /* Provide a default input filter to make life easy. */
2802 /* Note that we append to the line. This is handy. */
2803 DEBUG_P(PerlIO_printf(Perl_debug_log,
2804 "filter_read %d: from rsfp\n", idx));
2805 if (correct_length) {
2808 const int old_len = SvCUR(buf_sv);
2810 /* ensure buf_sv is large enough */
2811 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2812 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2813 correct_length)) <= 0) {
2814 if (PerlIO_error(PL_rsfp))
2815 return -1; /* error */
2817 return 0 ; /* end of file */
2819 SvCUR_set(buf_sv, old_len + len) ;
2822 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2823 if (PerlIO_error(PL_rsfp))
2824 return -1; /* error */
2826 return 0 ; /* end of file */
2829 return SvCUR(buf_sv);
2831 /* Skip this filter slot if filter has been deleted */
2832 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2833 DEBUG_P(PerlIO_printf(Perl_debug_log,
2834 "filter_read %d: skipped (filter deleted)\n",
2836 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2838 /* Get function pointer hidden within datasv */
2839 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2840 DEBUG_P(PerlIO_printf(Perl_debug_log,
2841 "filter_read %d: via function %p (%s)\n",
2842 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2843 /* Call function. The function is expected to */
2844 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2845 /* Return: <0:error, =0:eof, >0:not eof */
2846 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2850 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2853 #ifdef PERL_CR_FILTER
2854 if (!PL_rsfp_filters) {
2855 filter_add(S_cr_textfilter,NULL);
2858 if (PL_rsfp_filters) {
2860 SvCUR_set(sv, 0); /* start with empty line */
2861 if (FILTER_READ(0, sv, 0) > 0)
2862 return ( SvPVX(sv) ) ;
2867 return (sv_gets(sv, fp, append));
2871 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2876 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2880 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2881 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2883 return GvHV(gv); /* Foo:: */
2886 /* use constant CLASS => 'MyClass' */
2887 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2888 if (gv && GvCV(gv)) {
2889 SV * const sv = cv_const_sv(GvCV(gv));
2891 pkgname = SvPV_nolen_const(sv);
2894 return gv_stashpv(pkgname, 0);
2898 * S_readpipe_override
2899 * Check whether readpipe() is overriden, and generates the appropriate
2900 * optree, provided sublex_start() is called afterwards.
2903 S_readpipe_override(pTHX)
2906 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2907 yylval.ival = OP_BACKTICK;
2909 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2911 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2912 && (gv_readpipe = *gvp) != (GV*)&PL_sv_undef
2913 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2915 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2916 append_elem(OP_LIST,
2917 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2918 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2928 * The intent of this yylex wrapper is to minimize the changes to the
2929 * tokener when we aren't interested in collecting madprops. It remains
2930 * to be seen how successful this strategy will be...
2937 char *s = PL_bufptr;
2939 /* make sure PL_thiswhite is initialized */
2943 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2944 if (PL_pending_ident)
2945 return S_pending_ident(aTHX);
2947 /* previous token ate up our whitespace? */
2948 if (!PL_lasttoke && PL_nextwhite) {
2949 PL_thiswhite = PL_nextwhite;
2953 /* isolate the token, and figure out where it is without whitespace */
2954 PL_realtokenstart = -1;
2958 assert(PL_curforce < 0);
2960 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
2961 if (!PL_thistoken) {
2962 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
2963 PL_thistoken = newSVpvs("");
2965 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
2966 PL_thistoken = newSVpvn(tstart, s - tstart);
2969 if (PL_thismad) /* install head */
2970 CURMAD('X', PL_thistoken);
2973 /* last whitespace of a sublex? */
2974 if (optype == ')' && PL_endwhite) {
2975 CURMAD('X', PL_endwhite);
2980 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
2981 if (!PL_thiswhite && !PL_endwhite && !optype) {
2982 sv_free(PL_thistoken);
2987 /* put off final whitespace till peg */
2988 if (optype == ';' && !PL_rsfp) {
2989 PL_nextwhite = PL_thiswhite;
2992 else if (PL_thisopen) {
2993 CURMAD('q', PL_thisopen);
2995 sv_free(PL_thistoken);
2999 /* Store actual token text as madprop X */
3000 CURMAD('X', PL_thistoken);
3004 /* add preceding whitespace as madprop _ */
3005 CURMAD('_', PL_thiswhite);
3009 /* add quoted material as madprop = */
3010 CURMAD('=', PL_thisstuff);
3014 /* add terminating quote as madprop Q */
3015 CURMAD('Q', PL_thisclose);
3019 /* special processing based on optype */
3023 /* opval doesn't need a TOKEN since it can already store mp */
3034 append_madprops(PL_thismad, yylval.opval, 0);
3042 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3051 /* remember any fake bracket that lexer is about to discard */
3052 if (PL_lex_brackets == 1 &&
3053 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3056 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3059 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3060 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3063 break; /* don't bother looking for trailing comment */
3072 /* attach a trailing comment to its statement instead of next token */
3076 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3078 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3080 if (*s == '\n' || *s == '#') {
3081 while (s < PL_bufend && *s != '\n')
3085 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3086 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3103 /* Create new token struct. Note: opvals return early above. */
3104 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3111 S_tokenize_use(pTHX_ int is_use, char *s) {
3113 if (PL_expect != XSTATE)
3114 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3115 is_use ? "use" : "no"));
3117 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3118 s = force_version(s, TRUE);
3119 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3120 start_force(PL_curforce);
3121 NEXTVAL_NEXTTOKE.opval = NULL;
3124 else if (*s == 'v') {
3125 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3126 s = force_version(s, FALSE);
3130 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3131 s = force_version(s, FALSE);
3133 yylval.ival = is_use;
3137 static const char* const exp_name[] =
3138 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3139 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3146 Works out what to call the token just pulled out of the input
3147 stream. The yacc parser takes care of taking the ops we return and
3148 stitching them into a tree.
3154 if read an identifier
3155 if we're in a my declaration
3156 croak if they tried to say my($foo::bar)
3157 build the ops for a my() declaration
3158 if it's an access to a my() variable
3159 are we in a sort block?
3160 croak if my($a); $a <=> $b
3161 build ops for access to a my() variable
3162 if in a dq string, and they've said @foo and we can't find @foo
3164 build ops for a bareword
3165 if we already built the token before, use it.
3170 #pragma segment Perl_yylex
3176 register char *s = PL_bufptr;
3181 /* orig_keyword, gvp, and gv are initialized here because
3182 * jump to the label just_a_word_zero can bypass their
3183 * initialization later. */
3184 I32 orig_keyword = 0;
3189 SV* tmp = newSVpvs("");
3190 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3191 (IV)CopLINE(PL_curcop),
3192 lex_state_names[PL_lex_state],
3193 exp_name[PL_expect],
3194 pv_display(tmp, s, strlen(s), 0, 60));
3197 /* check if there's an identifier for us to look at */
3198 if (PL_pending_ident)
3199 return REPORT(S_pending_ident(aTHX));
3201 /* no identifier pending identification */
3203 switch (PL_lex_state) {
3205 case LEX_NORMAL: /* Some compilers will produce faster */
3206 case LEX_INTERPNORMAL: /* code if we comment these out. */
3210 /* when we've already built the next token, just pull it out of the queue */
3214 yylval = PL_nexttoke[PL_lasttoke].next_val;
3216 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3217 PL_nexttoke[PL_lasttoke].next_mad = 0;
3218 if (PL_thismad && PL_thismad->mad_key == '_') {
3219 PL_thiswhite = (SV*)PL_thismad->mad_val;
3220 PL_thismad->mad_val = 0;
3221 mad_free(PL_thismad);
3226 PL_lex_state = PL_lex_defer;
3227 PL_expect = PL_lex_expect;
3228 PL_lex_defer = LEX_NORMAL;
3229 if (!PL_nexttoke[PL_lasttoke].next_type)
3234 yylval = PL_nextval[PL_nexttoke];
3236 PL_lex_state = PL_lex_defer;
3237 PL_expect = PL_lex_expect;
3238 PL_lex_defer = LEX_NORMAL;
3242 /* FIXME - can these be merged? */
3243 return(PL_nexttoke[PL_lasttoke].next_type);
3245 return REPORT(PL_nexttype[PL_nexttoke]);
3248 /* interpolated case modifiers like \L \U, including \Q and \E.
3249 when we get here, PL_bufptr is at the \
3251 case LEX_INTERPCASEMOD:
3253 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3254 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3256 /* handle \E or end of string */
3257 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3259 if (PL_lex_casemods) {
3260 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3261 PL_lex_casestack[PL_lex_casemods] = '\0';
3263 if (PL_bufptr != PL_bufend
3264 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3266 PL_lex_state = LEX_INTERPCONCAT;
3269 PL_thistoken = newSVpvs("\\E");
3275 while (PL_bufptr != PL_bufend &&
3276 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3278 PL_thiswhite = newSVpvs("");
3279 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3283 if (PL_bufptr != PL_bufend)
3286 PL_lex_state = LEX_INTERPCONCAT;
3290 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3291 "### Saw case modifier\n"); });
3293 if (s[1] == '\\' && s[2] == 'E') {
3296 PL_thiswhite = newSVpvs("");
3297 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3300 PL_lex_state = LEX_INTERPCONCAT;
3305 if (!PL_madskills) /* when just compiling don't need correct */
3306 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3307 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3308 if ((*s == 'L' || *s == 'U') &&
3309 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3310 PL_lex_casestack[--PL_lex_casemods] = '\0';
3313 if (PL_lex_casemods > 10)
3314 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3315 PL_lex_casestack[PL_lex_casemods++] = *s;
3316 PL_lex_casestack[PL_lex_casemods] = '\0';
3317 PL_lex_state = LEX_INTERPCONCAT;
3318 start_force(PL_curforce);
3319 NEXTVAL_NEXTTOKE.ival = 0;
3321 start_force(PL_curforce);
3323 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3325 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3327 NEXTVAL_NEXTTOKE.ival = OP_LC;
3329 NEXTVAL_NEXTTOKE.ival = OP_UC;
3331 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3333 Perl_croak(aTHX_ "panic: yylex");
3335 SV* const tmpsv = newSVpvs("");
3336 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3342 if (PL_lex_starts) {
3348 sv_free(PL_thistoken);
3349 PL_thistoken = newSVpvs("");
3352 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3353 if (PL_lex_casemods == 1 && PL_lex_inpat)
3362 case LEX_INTERPPUSH:
3363 return REPORT(sublex_push());
3365 case LEX_INTERPSTART:
3366 if (PL_bufptr == PL_bufend)
3367 return REPORT(sublex_done());
3368 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3369 "### Interpolated variable\n"); });
3371 PL_lex_dojoin = (*PL_bufptr == '@');
3372 PL_lex_state = LEX_INTERPNORMAL;
3373 if (PL_lex_dojoin) {
3374 start_force(PL_curforce);
3375 NEXTVAL_NEXTTOKE.ival = 0;
3377 start_force(PL_curforce);
3378 force_ident("\"", '$');
3379 start_force(PL_curforce);
3380 NEXTVAL_NEXTTOKE.ival = 0;
3382 start_force(PL_curforce);
3383 NEXTVAL_NEXTTOKE.ival = 0;
3385 start_force(PL_curforce);
3386 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3389 if (PL_lex_starts++) {
3394 sv_free(PL_thistoken);
3395 PL_thistoken = newSVpvs("");
3398 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3399 if (!PL_lex_casemods && PL_lex_inpat)
3406 case LEX_INTERPENDMAYBE:
3407 if (intuit_more(PL_bufptr)) {
3408 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3414 if (PL_lex_dojoin) {
3415 PL_lex_dojoin = FALSE;
3416 PL_lex_state = LEX_INTERPCONCAT;
3420 sv_free(PL_thistoken);
3421 PL_thistoken = newSVpvs("");
3426 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3427 && SvEVALED(PL_lex_repl))
3429 if (PL_bufptr != PL_bufend)
3430 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3434 case LEX_INTERPCONCAT:
3436 if (PL_lex_brackets)
3437 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3439 if (PL_bufptr == PL_bufend)
3440 return REPORT(sublex_done());
3442 if (SvIVX(PL_linestr) == '\'') {
3443 SV *sv = newSVsv(PL_linestr);
3446 else if ( PL_hints & HINT_NEW_RE )
3447 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3448 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3452 s = scan_const(PL_bufptr);
3454 PL_lex_state = LEX_INTERPCASEMOD;
3456 PL_lex_state = LEX_INTERPSTART;
3459 if (s != PL_bufptr) {
3460 start_force(PL_curforce);
3462 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3464 NEXTVAL_NEXTTOKE = yylval;
3467 if (PL_lex_starts++) {
3471 sv_free(PL_thistoken);
3472 PL_thistoken = newSVpvs("");
3475 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3476 if (!PL_lex_casemods && PL_lex_inpat)
3489 PL_lex_state = LEX_NORMAL;
3490 s = scan_formline(PL_bufptr);
3491 if (!PL_lex_formbrack)
3497 PL_oldoldbufptr = PL_oldbufptr;
3503 sv_free(PL_thistoken);
3506 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3510 if (isIDFIRST_lazy_if(s,UTF))
3512 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3515 goto fake_eof; /* emulate EOF on ^D or ^Z */
3524 if (PL_lex_brackets) {
3525 yyerror((const char *)
3527 ? "Format not terminated"
3528 : "Missing right curly or square bracket"));
3530 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3531 "### Tokener got EOF\n");
3535 if (s++ < PL_bufend)
3536 goto retry; /* ignore stray nulls */
3539 if (!PL_in_eval && !PL_preambled) {
3540 PL_preambled = TRUE;
3545 sv_setpv(PL_linestr,incl_perldb());
3546 if (SvCUR(PL_linestr))
3547 sv_catpvs(PL_linestr,";");
3549 while(AvFILLp(PL_preambleav) >= 0) {
3550 SV *tmpsv = av_shift(PL_preambleav);
3551 sv_catsv(PL_linestr, tmpsv);
3552 sv_catpvs(PL_linestr, ";");
3555 sv_free((SV*)PL_preambleav);
3556 PL_preambleav = NULL;
3558 if (PL_minus_n || PL_minus_p) {
3559 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3561 sv_catpvs(PL_linestr,"chomp;");
3564 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3565 || *PL_splitstr == '"')
3566 && strchr(PL_splitstr + 1, *PL_splitstr))
3567 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3569 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3570 bytes can be used as quoting characters. :-) */
3571 const char *splits = PL_splitstr;
3572 sv_catpvs(PL_linestr, "our @F=split(q\0");
3575 if (*splits == '\\')
3576 sv_catpvn(PL_linestr, splits, 1);
3577 sv_catpvn(PL_linestr, splits, 1);
3578 } while (*splits++);
3579 /* This loop will embed the trailing NUL of
3580 PL_linestr as the last thing it does before
3582 sv_catpvs(PL_linestr, ");");
3586 sv_catpvs(PL_linestr,"our @F=split(' ');");
3590 sv_catpvs(PL_linestr,"use feature ':5.10';");
3591 sv_catpvs(PL_linestr, "\n");
3592 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3593 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3594 PL_last_lop = PL_last_uni = NULL;
3595 if (PERLDB_LINE && PL_curstash != PL_debstash)
3596 update_debugger_info(PL_linestr, NULL, 0);
3600 bof = PL_rsfp ? TRUE : FALSE;
3601 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3604 PL_realtokenstart = -1;
3607 if (PL_preprocess && !PL_in_eval)
3608 (void)PerlProc_pclose(PL_rsfp);
3609 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3610 PerlIO_clearerr(PL_rsfp);
3612 (void)PerlIO_close(PL_rsfp);
3614 PL_doextract = FALSE;
3616 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3621 sv_setpv(PL_linestr,
3624 ? ";}continue{print;}" : ";}"));
3625 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3626 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3627 PL_last_lop = PL_last_uni = NULL;
3628 PL_minus_n = PL_minus_p = 0;
3631 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3632 PL_last_lop = PL_last_uni = NULL;
3633 sv_setpvn(PL_linestr,"",0);
3634 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3636 /* If it looks like the start of a BOM or raw UTF-16,
3637 * check if it in fact is. */
3643 #ifdef PERLIO_IS_STDIO
3644 # ifdef __GNU_LIBRARY__
3645 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3646 # define FTELL_FOR_PIPE_IS_BROKEN
3650 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3651 # define FTELL_FOR_PIPE_IS_BROKEN
3656 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3657 /* This loses the possibility to detect the bof
3658 * situation on perl -P when the libc5 is being used.
3659 * Workaround? Maybe attach some extra state to PL_rsfp?
3662 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3664 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3667 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3668 s = swallow_bom((U8*)s);
3672 /* Incest with pod. */
3675 sv_catsv(PL_thiswhite, PL_linestr);
3677 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3678 sv_setpvn(PL_linestr, "", 0);
3679 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3680 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3681 PL_last_lop = PL_last_uni = NULL;
3682 PL_doextract = FALSE;
3686 } while (PL_doextract);
3687 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3688 if (PERLDB_LINE && PL_curstash != PL_debstash)
3689 update_debugger_info(PL_linestr, NULL, 0);
3690 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3691 PL_last_lop = PL_last_uni = NULL;
3692 if (CopLINE(PL_curcop) == 1) {
3693 while (s < PL_bufend && isSPACE(*s))
3695 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3699 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3703 if (*s == '#' && *(s+1) == '!')
3705 #ifdef ALTERNATE_SHEBANG
3707 static char const as[] = ALTERNATE_SHEBANG;
3708 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3709 d = s + (sizeof(as) - 1);
3711 #endif /* ALTERNATE_SHEBANG */
3720 while (*d && !isSPACE(*d))
3724 #ifdef ARG_ZERO_IS_SCRIPT
3725 if (ipathend > ipath) {
3727 * HP-UX (at least) sets argv[0] to the script name,
3728 * which makes $^X incorrect. And Digital UNIX and Linux,
3729 * at least, set argv[0] to the basename of the Perl
3730 * interpreter. So, having found "#!", we'll set it right.
3732 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3734 assert(SvPOK(x) || SvGMAGICAL(x));
3735 if (sv_eq(x, CopFILESV(PL_curcop))) {
3736 sv_setpvn(x, ipath, ipathend - ipath);
3742 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3743 const char * const lstart = SvPV_const(x,llen);
3745 bstart += blen - llen;
3746 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3747 sv_setpvn(x, ipath, ipathend - ipath);
3752 TAINT_NOT; /* $^X is always tainted, but that's OK */
3754 #endif /* ARG_ZERO_IS_SCRIPT */
3759 d = instr(s,"perl -");
3761 d = instr(s,"perl");
3763 /* avoid getting into infinite loops when shebang
3764 * line contains "Perl" rather than "perl" */
3766 for (d = ipathend-4; d >= ipath; --d) {
3767 if ((*d == 'p' || *d == 'P')
3768 && !ibcmp(d, "perl", 4))
3778 #ifdef ALTERNATE_SHEBANG
3780 * If the ALTERNATE_SHEBANG on this system starts with a
3781 * character that can be part of a Perl expression, then if
3782 * we see it but not "perl", we're probably looking at the
3783 * start of Perl code, not a request to hand off to some
3784 * other interpreter. Similarly, if "perl" is there, but
3785 * not in the first 'word' of the line, we assume the line
3786 * contains the start of the Perl program.
3788 if (d && *s != '#') {
3789 const char *c = ipath;
3790 while (*c && !strchr("; \t\r\n\f\v#", *c))
3793 d = NULL; /* "perl" not in first word; ignore */
3795 *s = '#'; /* Don't try to parse shebang line */
3797 #endif /* ALTERNATE_SHEBANG */
3798 #ifndef MACOS_TRADITIONAL
3803 !instr(s,"indir") &&
3804 instr(PL_origargv[0],"perl"))
3811 while (s < PL_bufend && isSPACE(*s))
3813 if (s < PL_bufend) {
3814 Newxz(newargv,PL_origargc+3,char*);
3816 while (s < PL_bufend && !isSPACE(*s))
3819 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3822 newargv = PL_origargv;
3825 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3827 Perl_croak(aTHX_ "Can't exec %s", ipath);
3831 while (*d && !isSPACE(*d))
3833 while (SPACE_OR_TAB(*d))
3837 const bool switches_done = PL_doswitches;
3838 const U32 oldpdb = PL_perldb;
3839 const bool oldn = PL_minus_n;
3840 const bool oldp = PL_minus_p;
3843 if (*d == 'M' || *d == 'm' || *d == 'C') {
3844 const char * const m = d;
3845 while (*d && !isSPACE(*d))
3847 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3850 d = moreswitches(d);
3852 if (PL_doswitches && !switches_done) {
3853 int argc = PL_origargc;
3854 char **argv = PL_origargv;
3857 } while (argc && argv[0][0] == '-' && argv[0][1]);
3858 init_argv_symbols(argc,argv);
3860 if ((PERLDB_LINE && !oldpdb) ||
3861 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3862 /* if we have already added "LINE: while (<>) {",
3863 we must not do it again */
3865 sv_setpvn(PL_linestr, "", 0);
3866 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3867 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3868 PL_last_lop = PL_last_uni = NULL;
3869 PL_preambled = FALSE;
3871 (void)gv_fetchfile(PL_origfilename);
3878 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3880 PL_lex_state = LEX_FORMLINE;
3885 #ifdef PERL_STRICT_CR
3886 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3888 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3890 case ' ': case '\t': case '\f': case 013:
3891 #ifdef MACOS_TRADITIONAL
3895 PL_realtokenstart = -1;
3904 PL_realtokenstart = -1;
3908 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3909 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3910 /* handle eval qq[#line 1 "foo"\n ...] */
3911 CopLINE_dec(PL_curcop);
3914 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3916 if (!PL_in_eval || PL_rsfp)
3921 while (d < PL_bufend && *d != '\n')
3925 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3926 Perl_croak(aTHX_ "panic: input overflow");
3929 PL_thiswhite = newSVpvn(s, d - s);
3934 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3936 PL_lex_state = LEX_FORMLINE;
3942 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3943 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3946 TOKEN(PEG); /* make sure any #! line is accessible */
3951 /* if (PL_madskills && PL_lex_formbrack) { */
3953 while (d < PL_bufend && *d != '\n')
3957 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3958 Perl_croak(aTHX_ "panic: input overflow");
3959 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
3961 PL_thiswhite = newSVpvs("");
3962 if (CopLINE(PL_curcop) == 1) {
3963 sv_setpvn(PL_thiswhite, "", 0);
3966 sv_catpvn(PL_thiswhite, s, d - s);
3980 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
3988 while (s < PL_bufend && SPACE_OR_TAB(*s))
3991 if (strnEQ(s,"=>",2)) {
3992 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
3993 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
3994 OPERATOR('-'); /* unary minus */
3996 PL_last_uni = PL_oldbufptr;
3998 case 'r': ftst = OP_FTEREAD; break;
3999 case 'w': ftst = OP_FTEWRITE; break;
4000 case 'x': ftst = OP_FTEEXEC; break;
4001 case 'o': ftst = OP_FTEOWNED; break;
4002 case 'R': ftst = OP_FTRREAD; break;
4003 case 'W': ftst = OP_FTRWRITE; break;
4004 case 'X': ftst = OP_FTREXEC; break;
4005 case 'O': ftst = OP_FTROWNED; break;
4006 case 'e': ftst = OP_FTIS; break;
4007 case 'z': ftst = OP_FTZERO; break;
4008 case 's': ftst = OP_FTSIZE; break;
4009 case 'f': ftst = OP_FTFILE; break;
4010 case 'd': ftst = OP_FTDIR; break;
4011 case 'l': ftst = OP_FTLINK; break;
4012 case 'p': ftst = OP_FTPIPE; break;
4013 case 'S': ftst = OP_FTSOCK; break;
4014 case 'u': ftst = OP_FTSUID; break;
4015 case 'g': ftst = OP_FTSGID; break;
4016 case 'k': ftst = OP_FTSVTX; break;
4017 case 'b': ftst = OP_FTBLK; break;
4018 case 'c': ftst = OP_FTCHR; break;
4019 case 't': ftst = OP_FTTTY; break;
4020 case 'T': ftst = OP_FTTEXT; break;
4021 case 'B': ftst = OP_FTBINARY; break;
4022 case 'M': case 'A': case 'C':
4023 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4025 case 'M': ftst = OP_FTMTIME; break;
4026 case 'A': ftst = OP_FTATIME; break;
4027 case 'C': ftst = OP_FTCTIME; break;
4035 PL_last_lop_op = (OPCODE)ftst;
4036 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4037 "### Saw file test %c\n", (int)tmp);
4042 /* Assume it was a minus followed by a one-letter named
4043 * subroutine call (or a -bareword), then. */
4044 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4045 "### '-%c' looked like a file test but was not\n",
4052 const char tmp = *s++;
4055 if (PL_expect == XOPERATOR)
4060 else if (*s == '>') {
4063 if (isIDFIRST_lazy_if(s,UTF)) {
4064 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4072 if (PL_expect == XOPERATOR)
4075 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4077 OPERATOR('-'); /* unary minus */
4083 const char tmp = *s++;
4086 if (PL_expect == XOPERATOR)
4091 if (PL_expect == XOPERATOR)
4094 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4101 if (PL_expect != XOPERATOR) {
4102 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4103 PL_expect = XOPERATOR;
4104 force_ident(PL_tokenbuf, '*');
4117 if (PL_expect == XOPERATOR) {
4121 PL_tokenbuf[0] = '%';
4122 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, TRUE);
4123 if (!PL_tokenbuf[1]) {
4126 PL_pending_ident = '%';
4137 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4144 const char tmp = *s++;
4150 goto just_a_word_zero_gv;
4153 switch (PL_expect) {
4159 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4161 PL_bufptr = s; /* update in case we back off */
4167 PL_expect = XTERMBLOCK;
4170 stuffstart = s - SvPVX(PL_linestr) - 1;
4174 while (isIDFIRST_lazy_if(s,UTF)) {
4177 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4178 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4179 if (tmp < 0) tmp = -tmp;
4194 sv = newSVpvn(s, len);
4196 d = scan_str(d,TRUE,TRUE);
4198 /* MUST advance bufptr here to avoid bogus
4199 "at end of line" context messages from yyerror().
4201 PL_bufptr = s + len;
4202 yyerror("Unterminated attribute parameter in attribute list");
4206 return REPORT(0); /* EOF indicator */
4210 sv_catsv(sv, PL_lex_stuff);
4211 attrs = append_elem(OP_LIST, attrs,
4212 newSVOP(OP_CONST, 0, sv));
4213 SvREFCNT_dec(PL_lex_stuff);
4214 PL_lex_stuff = NULL;
4217 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4219 if (PL_in_my == KEY_our) {
4221 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4223 /* skip to avoid loading attributes.pm */
4225 deprecate(":unique");
4228 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4231 /* NOTE: any CV attrs applied here need to be part of
4232 the CVf_BUILTIN_ATTRS define in cv.h! */
4233 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4235 CvLVALUE_on(PL_compcv);
4237 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4239 CvLOCKED_on(PL_compcv);
4241 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4243 CvMETHOD_on(PL_compcv);
4245 else if (!PL_in_my && len == 9 && strnEQ(SvPVX(sv), "assertion", len)) {
4247 CvASSERTION_on(PL_compcv);
4249 /* After we've set the flags, it could be argued that
4250 we don't need to do the attributes.pm-based setting
4251 process, and shouldn't bother appending recognized
4252 flags. To experiment with that, uncomment the
4253 following "else". (Note that's already been
4254 uncommented. That keeps the above-applied built-in
4255 attributes from being intercepted (and possibly
4256 rejected) by a package's attribute routines, but is
4257 justified by the performance win for the common case
4258 of applying only built-in attributes.) */
4260 attrs = append_elem(OP_LIST, attrs,
4261 newSVOP(OP_CONST, 0,
4265 if (*s == ':' && s[1] != ':')
4268 break; /* require real whitespace or :'s */
4269 /* XXX losing whitespace on sequential attributes here */
4273 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4274 if (*s != ';' && *s != '}' && *s != tmp
4275 && (tmp != '=' || *s != ')')) {
4276 const char q = ((*s == '\'') ? '"' : '\'');
4277 /* If here for an expression, and parsed no attrs, back
4279 if (tmp == '=' && !attrs) {
4283 /* MUST advance bufptr here to avoid bogus "at end of line"
4284 context messages from yyerror().
4287 yyerror( (const char *)
4289 ? Perl_form(aTHX_ "Invalid separator character "
4290 "%c%c%c in attribute list", q, *s, q)
4291 : "Unterminated attribute list" ) );
4299 start_force(PL_curforce);
4300 NEXTVAL_NEXTTOKE.opval = attrs;
4301 CURMAD('_', PL_nextwhite);
4306 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4307 (s - SvPVX(PL_linestr)) - stuffstart);
4315 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4316 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4324 const char tmp = *s++;
4329 const char tmp = *s++;
4337 if (PL_lex_brackets <= 0)
4338 yyerror("Unmatched right square bracket");
4341 if (PL_lex_state == LEX_INTERPNORMAL) {
4342 if (PL_lex_brackets == 0) {
4343 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4344 PL_lex_state = LEX_INTERPEND;
4351 if (PL_lex_brackets > 100) {
4352 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4354 switch (PL_expect) {
4356 if (PL_lex_formbrack) {
4360 if (PL_oldoldbufptr == PL_last_lop)
4361 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4363 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4364 OPERATOR(HASHBRACK);
4366 while (s < PL_bufend && SPACE_OR_TAB(*s))
4369 PL_tokenbuf[0] = '\0';
4370 if (d < PL_bufend && *d == '-') {
4371 PL_tokenbuf[0] = '-';
4373 while (d < PL_bufend && SPACE_OR_TAB(*d))
4376 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4377 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4379 while (d < PL_bufend && SPACE_OR_TAB(*d))
4382 const char minus = (PL_tokenbuf[0] == '-');
4383 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4391 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4396 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4401 if (PL_oldoldbufptr == PL_last_lop)
4402 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4404 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;