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)
52 #define PL_linestr (PL_parser->linestr)
53 #define PL_expect (PL_parser->expect)
54 #define PL_copline (PL_parser->copline)
55 #define PL_bufptr (PL_parser->bufptr)
56 #define PL_oldbufptr (PL_parser->oldbufptr)
57 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
58 #define PL_linestart (PL_parser->linestart)
59 #define PL_bufend (PL_parser->bufend)
60 #define PL_last_uni (PL_parser->last_uni)
61 #define PL_last_lop (PL_parser->last_lop)
62 #define PL_last_lop_op (PL_parser->last_lop_op)
63 #define PL_lex_state (PL_parser->lex_state)
64 #define PL_rsfp (PL_parser->rsfp)
65 #define PL_rsfp_filters (PL_parser->rsfp_filters)
66 #define PL_in_my (PL_parser->in_my)
67 #define PL_in_my_stash (PL_parser->in_my_stash)
68 #define PL_tokenbuf (PL_parser->tokenbuf)
69 #define PL_multi_end (PL_parser->multi_end)
70 #define PL_error_count (PL_parser->error_count)
73 # define PL_endwhite (PL_parser->endwhite)
74 # define PL_faketokens (PL_parser->faketokens)
75 # define PL_lasttoke (PL_parser->lasttoke)
76 # define PL_nextwhite (PL_parser->nextwhite)
77 # define PL_realtokenstart (PL_parser->realtokenstart)
78 # define PL_skipwhite (PL_parser->skipwhite)
79 # define PL_thisclose (PL_parser->thisclose)
80 # define PL_thismad (PL_parser->thismad)
81 # define PL_thisopen (PL_parser->thisopen)
82 # define PL_thisstuff (PL_parser->thisstuff)
83 # define PL_thistoken (PL_parser->thistoken)
84 # define PL_thiswhite (PL_parser->thiswhite)
85 # define PL_thiswhite (PL_parser->thiswhite)
86 # define PL_nexttoke (PL_parser->nexttoke)
87 # define PL_curforce (PL_parser->curforce)
89 # define PL_nexttoke (PL_parser->nexttoke)
90 # define PL_nexttype (PL_parser->nexttype)
91 # define PL_nextval (PL_parser->nextval)
95 S_pending_ident(pTHX);
97 static const char ident_too_long[] = "Identifier too long";
98 static const char commaless_variable_list[] = "comma-less variable list";
100 #ifndef PERL_NO_UTF16_FILTER
101 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
102 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
106 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
107 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
109 # define CURMAD(slot,sv)
110 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
113 #define XFAKEBRACK 128
114 #define XENUMMASK 127
116 #ifdef USE_UTF8_SCRIPTS
117 # define UTF (!IN_BYTES)
119 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
122 /* In variables named $^X, these are the legal values for X.
123 * 1999-02-27 mjd-perl-patch@plover.com */
124 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
126 /* On MacOS, respect nonbreaking spaces */
127 #ifdef MACOS_TRADITIONAL
128 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
130 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
133 /* LEX_* are values for PL_lex_state, the state of the lexer.
134 * They are arranged oddly so that the guard on the switch statement
135 * can get by with a single comparison (if the compiler is smart enough).
138 /* #define LEX_NOTPARSING 11 is done in perl.h. */
140 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
141 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
142 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
143 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
144 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
146 /* at end of code, eg "$x" followed by: */
147 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
148 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
150 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
151 string or after \E, $foo, etc */
152 #define LEX_INTERPCONST 2 /* NOT USED */
153 #define LEX_FORMLINE 1 /* expecting a format line */
154 #define LEX_KNOWNEXT 0 /* next token known; just return it */
158 static const char* const lex_state_names[] = {
177 #include "keywords.h"
179 /* CLINE is a macro that ensures PL_copline has a sane value */
184 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
187 # define SKIPSPACE0(s) skipspace0(s)
188 # define SKIPSPACE1(s) skipspace1(s)
189 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
190 # define PEEKSPACE(s) skipspace2(s,0)
192 # define SKIPSPACE0(s) skipspace(s)
193 # define SKIPSPACE1(s) skipspace(s)
194 # define SKIPSPACE2(s,tsv) skipspace(s)
195 # define PEEKSPACE(s) skipspace(s)
199 * Convenience functions to return different tokens and prime the
200 * lexer for the next token. They all take an argument.
202 * TOKEN : generic token (used for '(', DOLSHARP, etc)
203 * OPERATOR : generic operator
204 * AOPERATOR : assignment operator
205 * PREBLOCK : beginning the block after an if, while, foreach, ...
206 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
207 * PREREF : *EXPR where EXPR is not a simple identifier
208 * TERM : expression term
209 * LOOPX : loop exiting command (goto, last, dump, etc)
210 * FTST : file test operator
211 * FUN0 : zero-argument function
212 * FUN1 : not used, except for not, which isn't a UNIOP
213 * BOop : bitwise or or xor
215 * SHop : shift operator
216 * PWop : power operator
217 * PMop : pattern-matching operator
218 * Aop : addition-level operator
219 * Mop : multiplication-level operator
220 * Eop : equality-testing operator
221 * Rop : relational operator <= != gt
223 * Also see LOP and lop() below.
226 #ifdef DEBUGGING /* Serve -DT. */
227 # define REPORT(retval) tokereport((I32)retval)
229 # define REPORT(retval) (retval)
232 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
233 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
234 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
235 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
236 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
237 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
238 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
239 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
240 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
241 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
242 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
243 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
244 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
245 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
246 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
247 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
248 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
249 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
250 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
251 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
253 /* This bit of chicanery makes a unary function followed by
254 * a parenthesis into a function with one argument, highest precedence.
255 * The UNIDOR macro is for unary functions that can be followed by the //
256 * operator (such as C<shift // 0>).
258 #define UNI2(f,x) { \
262 PL_last_uni = PL_oldbufptr; \
263 PL_last_lop_op = f; \
265 return REPORT( (int)FUNC1 ); \
267 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
269 #define UNI(f) UNI2(f,XTERM)
270 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
272 #define UNIBRACK(f) { \
275 PL_last_uni = PL_oldbufptr; \
277 return REPORT( (int)FUNC1 ); \
279 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
282 /* grandfather return to old style */
283 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
287 /* how to interpret the yylval associated with the token */
291 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
297 static struct debug_tokens {
299 enum token_type type;
301 } const debug_tokens[] =
303 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
304 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
305 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
306 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
307 { ARROW, TOKENTYPE_NONE, "ARROW" },
308 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
309 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
310 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
311 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
312 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
313 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
314 { DO, TOKENTYPE_NONE, "DO" },
315 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
316 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
317 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
318 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
319 { ELSE, TOKENTYPE_NONE, "ELSE" },
320 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
321 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
322 { FOR, TOKENTYPE_IVAL, "FOR" },
323 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
324 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
325 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
326 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
327 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
328 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
329 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
330 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
331 { IF, TOKENTYPE_IVAL, "IF" },
332 { LABEL, TOKENTYPE_PVAL, "LABEL" },
333 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
334 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
335 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
336 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
337 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
338 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
339 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
340 { MY, TOKENTYPE_IVAL, "MY" },
341 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
342 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
343 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
344 { OROP, TOKENTYPE_IVAL, "OROP" },
345 { OROR, TOKENTYPE_NONE, "OROR" },
346 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
347 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
348 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
349 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
350 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
351 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
352 { PREINC, TOKENTYPE_NONE, "PREINC" },
353 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
354 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
355 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
356 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
357 { SUB, TOKENTYPE_NONE, "SUB" },
358 { THING, TOKENTYPE_OPVAL, "THING" },
359 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
360 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
361 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
362 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
363 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
364 { USE, TOKENTYPE_IVAL, "USE" },
365 { WHEN, TOKENTYPE_IVAL, "WHEN" },
366 { WHILE, TOKENTYPE_IVAL, "WHILE" },
367 { WORD, TOKENTYPE_OPVAL, "WORD" },
368 { 0, TOKENTYPE_NONE, NULL }
371 /* dump the returned token in rv, plus any optional arg in yylval */
374 S_tokereport(pTHX_ I32 rv)
378 const char *name = NULL;
379 enum token_type type = TOKENTYPE_NONE;
380 const struct debug_tokens *p;
381 SV* const report = newSVpvs("<== ");
383 for (p = debug_tokens; p->token; p++) {
384 if (p->token == (int)rv) {
391 Perl_sv_catpv(aTHX_ report, name);
392 else if ((char)rv > ' ' && (char)rv < '~')
393 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
395 sv_catpvs(report, "EOF");
397 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
400 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
403 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
405 case TOKENTYPE_OPNUM:
406 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
407 PL_op_name[yylval.ival]);
410 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
412 case TOKENTYPE_OPVAL:
414 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
415 PL_op_name[yylval.opval->op_type]);
416 if (yylval.opval->op_type == OP_CONST) {
417 Perl_sv_catpvf(aTHX_ report, " %s",
418 SvPEEK(cSVOPx_sv(yylval.opval)));
423 sv_catpvs(report, "(opval=null)");
426 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
432 /* print the buffer with suitable escapes */
435 S_printbuf(pTHX_ const char* fmt, const char* s)
437 SV* const tmp = newSVpvs("");
438 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
447 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
448 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
452 S_ao(pTHX_ int toketype)
455 if (*PL_bufptr == '=') {
457 if (toketype == ANDAND)
458 yylval.ival = OP_ANDASSIGN;
459 else if (toketype == OROR)
460 yylval.ival = OP_ORASSIGN;
461 else if (toketype == DORDOR)
462 yylval.ival = OP_DORASSIGN;
470 * When Perl expects an operator and finds something else, no_op
471 * prints the warning. It always prints "<something> found where
472 * operator expected. It prints "Missing semicolon on previous line?"
473 * if the surprise occurs at the start of the line. "do you need to
474 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
475 * where the compiler doesn't know if foo is a method call or a function.
476 * It prints "Missing operator before end of line" if there's nothing
477 * after the missing operator, or "... before <...>" if there is something
478 * after the missing operator.
482 S_no_op(pTHX_ const char *what, char *s)
485 char * const oldbp = PL_bufptr;
486 const bool is_first = (PL_oldbufptr == PL_linestart);
492 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
493 if (ckWARN_d(WARN_SYNTAX)) {
495 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
496 "\t(Missing semicolon on previous line?)\n");
497 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
499 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
501 if (t < PL_bufptr && isSPACE(*t))
502 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
503 "\t(Do you need to predeclare %.*s?)\n",
504 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
508 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
509 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
517 * Complain about missing quote/regexp/heredoc terminator.
518 * If it's called with NULL then it cauterizes the line buffer.
519 * If we're in a delimited string and the delimiter is a control
520 * character, it's reformatted into a two-char sequence like ^C.
525 S_missingterm(pTHX_ char *s)
531 char * const nl = strrchr(s,'\n');
537 iscntrl(PL_multi_close)
539 PL_multi_close < 32 || PL_multi_close == 127
543 tmpbuf[1] = (char)toCTRL(PL_multi_close);
548 *tmpbuf = (char)PL_multi_close;
552 q = strchr(s,'"') ? '\'' : '"';
553 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
556 #define FEATURE_IS_ENABLED(name) \
557 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
558 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
560 * S_feature_is_enabled
561 * Check whether the named feature is enabled.
564 S_feature_is_enabled(pTHX_ const char *name, STRLEN namelen)
567 HV * const hinthv = GvHV(PL_hintgv);
568 char he_name[32] = "feature_";
569 (void) my_strlcpy(&he_name[8], name, 24);
571 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
579 Perl_deprecate(pTHX_ const char *s)
581 if (ckWARN(WARN_DEPRECATED))
582 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
586 Perl_deprecate_old(pTHX_ const char *s)
588 /* This function should NOT be called for any new deprecated warnings */
589 /* Use Perl_deprecate instead */
591 /* It is here to maintain backward compatibility with the pre-5.8 */
592 /* warnings category hierarchy. The "deprecated" category used to */
593 /* live under the "syntax" category. It is now a top-level category */
594 /* in its own right. */
596 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
597 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
598 "Use of %s is deprecated", s);
602 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
603 * utf16-to-utf8-reversed.
606 #ifdef PERL_CR_FILTER
610 register const char *s = SvPVX_const(sv);
611 register const char * const e = s + SvCUR(sv);
612 /* outer loop optimized to do nothing if there are no CR-LFs */
614 if (*s++ == '\r' && *s == '\n') {
615 /* hit a CR-LF, need to copy the rest */
616 register char *d = s - 1;
619 if (*s == '\r' && s[1] == '\n')
630 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
632 const I32 count = FILTER_READ(idx+1, sv, maxlen);
633 if (count > 0 && !maxlen)
644 * Create a parser object and initialise its parser and lexer fields
646 * rsfp is the opened file handle to read from (if any),
648 * line holds any initial content already read from the file (or in
649 * the case of no file, such as an eval, the whole contents);
651 * new_filter indicates that this is a new file and it shouldn't inherit
652 * the filters from the current parser (ie require).
656 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
659 const char *s = NULL;
661 yy_parser *parser, *oparser;
663 /* create and initialise a parser */
665 Newxz(parser, 1, yy_parser);
666 parser->old_parser = oparser = PL_parser;
669 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
670 parser->ps = parser->stack;
671 parser->stack_size = YYINITDEPTH;
673 parser->stack->state = 0;
674 parser->yyerrstatus = 0;
675 parser->yychar = YYEMPTY; /* Cause a token to be read. */
677 /* on scope exit, free this parser and restore any outer one */
679 parser->saved_curcop = PL_curcop;
681 /* initialise lexer state */
684 parser->curforce = -1;
686 parser->nexttoke = 0;
688 parser->copline = NOLINE;
689 parser->lex_state = LEX_NORMAL;
690 parser->expect = XSTATE;
692 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
693 : (AV*)SvREFCNT_inc(oparser->rsfp_filters);
695 Newx(parser->lex_brackstack, 120, char);
696 Newx(parser->lex_casestack, 12, char);
697 *parser->lex_casestack = '\0';
700 s = SvPV_const(line, len);
706 parser->linestr = newSVpvs("\n;");
707 } else if (SvREADONLY(line) || s[len-1] != ';') {
708 parser->linestr = newSVsv(line);
710 sv_catpvs(parser->linestr, "\n;");
713 SvREFCNT_inc_simple_void_NN(line);
714 parser->linestr = line;
716 parser->oldoldbufptr =
719 parser->linestart = SvPVX(parser->linestr);
720 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
721 parser->last_lop = parser->last_uni = NULL;
725 /* delete a parser object */
728 Perl_parser_free(pTHX_ const yy_parser *parser)
730 PL_curcop = parser->saved_curcop;
731 SvREFCNT_dec(parser->linestr);
733 if (parser->rsfp == PerlIO_stdin())
734 PerlIO_clearerr(parser->rsfp);
735 else if (parser->rsfp && parser->old_parser
736 && parser->rsfp != parser->old_parser->rsfp)
737 PerlIO_close(parser->rsfp);
738 SvREFCNT_dec(parser->rsfp_filters);
740 Safefree(parser->stack);
741 Safefree(parser->lex_brackstack);
742 Safefree(parser->lex_casestack);
743 PL_parser = parser->old_parser;
750 * Finalizer for lexing operations. Must be called when the parser is
751 * done with the lexer.
758 PL_doextract = FALSE;
763 * This subroutine has nothing to do with tilting, whether at windmills
764 * or pinball tables. Its name is short for "increment line". It
765 * increments the current line number in CopLINE(PL_curcop) and checks
766 * to see whether the line starts with a comment of the form
767 * # line 500 "foo.pm"
768 * If so, it sets the current line number and file to the values in the comment.
772 S_incline(pTHX_ const char *s)
779 CopLINE_inc(PL_curcop);
782 while (SPACE_OR_TAB(*s))
784 if (strnEQ(s, "line", 4))
788 if (SPACE_OR_TAB(*s))
792 while (SPACE_OR_TAB(*s))
800 while (SPACE_OR_TAB(*s))
802 if (*s == '"' && (t = strchr(s+1, '"'))) {
812 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
814 if (*e != '\n' && *e != '\0')
815 return; /* false alarm */
818 const STRLEN len = t - s;
820 const char * const cf = CopFILE(PL_curcop);
821 STRLEN tmplen = cf ? strlen(cf) : 0;
822 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
823 /* must copy *{"::_<(eval N)[oldfilename:L]"}
824 * to *{"::_<newfilename"} */
825 /* However, the long form of evals is only turned on by the
826 debugger - usually they're "(eval %lu)" */
830 STRLEN tmplen2 = len;
831 if (tmplen + 2 <= sizeof smallbuf)
834 Newx(tmpbuf, tmplen + 2, char);
837 memcpy(tmpbuf + 2, cf, tmplen);
839 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
844 if (tmplen2 + 2 <= sizeof smallbuf)
847 Newx(tmpbuf2, tmplen2 + 2, char);
849 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
850 /* Either they malloc'd it, or we malloc'd it,
851 so no prefix is present in ours. */
856 memcpy(tmpbuf2 + 2, s, tmplen2);
859 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
861 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
862 /* adjust ${"::_<newfilename"} to store the new file name */
863 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
864 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
865 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
868 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
870 if (tmpbuf != smallbuf) Safefree(tmpbuf);
873 CopFILE_free(PL_curcop);
874 CopFILE_setn(PL_curcop, s, len);
876 CopLINE_set(PL_curcop, atoi(n)-1);
880 /* skip space before PL_thistoken */
883 S_skipspace0(pTHX_ register char *s)
890 PL_thiswhite = newSVpvs("");
891 sv_catsv(PL_thiswhite, PL_skipwhite);
892 sv_free(PL_skipwhite);
895 PL_realtokenstart = s - SvPVX(PL_linestr);
899 /* skip space after PL_thistoken */
902 S_skipspace1(pTHX_ register char *s)
904 const char *start = s;
905 I32 startoff = start - SvPVX(PL_linestr);
910 start = SvPVX(PL_linestr) + startoff;
911 if (!PL_thistoken && PL_realtokenstart >= 0) {
912 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
913 PL_thistoken = newSVpvn(tstart, start - tstart);
915 PL_realtokenstart = -1;
918 PL_nextwhite = newSVpvs("");
919 sv_catsv(PL_nextwhite, PL_skipwhite);
920 sv_free(PL_skipwhite);
927 S_skipspace2(pTHX_ register char *s, SV **svp)
930 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
931 const I32 startoff = s - SvPVX(PL_linestr);
934 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
935 if (!PL_madskills || !svp)
937 start = SvPVX(PL_linestr) + startoff;
938 if (!PL_thistoken && PL_realtokenstart >= 0) {
939 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
940 PL_thistoken = newSVpvn(tstart, start - tstart);
941 PL_realtokenstart = -1;
946 sv_setsv(*svp, PL_skipwhite);
947 sv_free(PL_skipwhite);
956 S_update_debugger_info(pTHX_ SV *orig_sv, const char *buf, STRLEN len)
958 AV *av = CopFILEAVx(PL_curcop);
960 SV * const sv = newSV_type(SVt_PVMG);
962 sv_setsv(sv, orig_sv);
964 sv_setpvn(sv, buf, len);
967 av_store(av, (I32)CopLINE(PL_curcop), sv);
973 * Called to gobble the appropriate amount and type of whitespace.
974 * Skips comments as well.
978 S_skipspace(pTHX_ register char *s)
983 int startoff = s - SvPVX(PL_linestr);
986 sv_free(PL_skipwhite);
991 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
992 while (s < PL_bufend && SPACE_OR_TAB(*s))
1002 SSize_t oldprevlen, oldoldprevlen;
1003 SSize_t oldloplen = 0, oldunilen = 0;
1004 while (s < PL_bufend && isSPACE(*s)) {
1005 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
1010 if (s < PL_bufend && *s == '#') {
1011 while (s < PL_bufend && *s != '\n')
1013 if (s < PL_bufend) {
1015 if (PL_in_eval && !PL_rsfp) {
1022 /* only continue to recharge the buffer if we're at the end
1023 * of the buffer, we're not reading from a source filter, and
1024 * we're in normal lexing mode
1026 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1027 PL_lex_state == LEX_FORMLINE)
1034 /* try to recharge the buffer */
1036 curoff = s - SvPVX(PL_linestr);
1039 if ((s = filter_gets(PL_linestr, PL_rsfp,
1040 (prevlen = SvCUR(PL_linestr)))) == NULL)
1043 if (PL_madskills && curoff != startoff) {
1045 PL_skipwhite = newSVpvs("");
1046 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1050 /* mustn't throw out old stuff yet if madpropping */
1051 SvCUR(PL_linestr) = curoff;
1052 s = SvPVX(PL_linestr) + curoff;
1054 if (curoff && s[-1] == '\n')
1058 /* end of file. Add on the -p or -n magic */
1059 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1062 sv_catpvs(PL_linestr,
1063 ";}continue{print or die qq(-p destination: $!\\n);}");
1065 sv_setpvs(PL_linestr,
1066 ";}continue{print or die qq(-p destination: $!\\n);}");
1068 PL_minus_n = PL_minus_p = 0;
1070 else if (PL_minus_n) {
1072 sv_catpvn(PL_linestr, ";}", 2);
1074 sv_setpvn(PL_linestr, ";}", 2);
1080 sv_catpvn(PL_linestr,";", 1);
1082 sv_setpvn(PL_linestr,";", 1);
1085 /* reset variables for next time we lex */
1086 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1092 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1093 PL_last_lop = PL_last_uni = NULL;
1095 /* Close the filehandle. Could be from -P preprocessor,
1096 * STDIN, or a regular file. If we were reading code from
1097 * STDIN (because the commandline held no -e or filename)
1098 * then we don't close it, we reset it so the code can
1099 * read from STDIN too.
1102 if (PL_preprocess && !PL_in_eval)
1103 (void)PerlProc_pclose(PL_rsfp);
1104 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1105 PerlIO_clearerr(PL_rsfp);
1107 (void)PerlIO_close(PL_rsfp);
1112 /* not at end of file, so we only read another line */
1113 /* make corresponding updates to old pointers, for yyerror() */
1114 oldprevlen = PL_oldbufptr - PL_bufend;
1115 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1117 oldunilen = PL_last_uni - PL_bufend;
1119 oldloplen = PL_last_lop - PL_bufend;
1120 PL_linestart = PL_bufptr = s + prevlen;
1121 PL_bufend = s + SvCUR(PL_linestr);
1123 PL_oldbufptr = s + oldprevlen;
1124 PL_oldoldbufptr = s + oldoldprevlen;
1126 PL_last_uni = s + oldunilen;
1128 PL_last_lop = s + oldloplen;
1131 /* debugger active and we're not compiling the debugger code,
1132 * so store the line into the debugger's array of lines
1134 if (PERLDB_LINE && PL_curstash != PL_debstash)
1135 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1142 PL_skipwhite = newSVpvs("");
1143 curoff = s - SvPVX(PL_linestr);
1144 if (curoff - startoff)
1145 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1154 * Check the unary operators to ensure there's no ambiguity in how they're
1155 * used. An ambiguous piece of code would be:
1157 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1158 * the +5 is its argument.
1168 if (PL_oldoldbufptr != PL_last_uni)
1170 while (isSPACE(*PL_last_uni))
1173 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1175 if ((t = strchr(s, '(')) && t < PL_bufptr)
1178 if (ckWARN_d(WARN_AMBIGUOUS)){
1179 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1180 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1181 (int)(s - PL_last_uni), PL_last_uni);
1186 * LOP : macro to build a list operator. Its behaviour has been replaced
1187 * with a subroutine, S_lop() for which LOP is just another name.
1190 #define LOP(f,x) return lop(f,x,s)
1194 * Build a list operator (or something that might be one). The rules:
1195 * - if we have a next token, then it's a list operator [why?]
1196 * - if the next thing is an opening paren, then it's a function
1197 * - else it's a list operator
1201 S_lop(pTHX_ I32 f, int x, char *s)
1208 PL_last_lop = PL_oldbufptr;
1209 PL_last_lop_op = (OPCODE)f;
1212 return REPORT(LSTOP);
1215 return REPORT(LSTOP);
1218 return REPORT(FUNC);
1221 return REPORT(FUNC);
1223 return REPORT(LSTOP);
1229 * Sets up for an eventual force_next(). start_force(0) basically does
1230 * an unshift, while start_force(-1) does a push. yylex removes items
1235 S_start_force(pTHX_ int where)
1239 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1240 where = PL_lasttoke;
1241 assert(PL_curforce < 0 || PL_curforce == where);
1242 if (PL_curforce != where) {
1243 for (i = PL_lasttoke; i > where; --i) {
1244 PL_nexttoke[i] = PL_nexttoke[i-1];
1248 if (PL_curforce < 0) /* in case of duplicate start_force() */
1249 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1250 PL_curforce = where;
1253 curmad('^', newSVpvs(""));
1254 CURMAD('_', PL_nextwhite);
1259 S_curmad(pTHX_ char slot, SV *sv)
1265 if (PL_curforce < 0)
1266 where = &PL_thismad;
1268 where = &PL_nexttoke[PL_curforce].next_mad;
1271 sv_setpvn(sv, "", 0);
1274 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1276 else if (PL_encoding) {
1277 sv_recode_to_utf8(sv, PL_encoding);
1282 /* keep a slot open for the head of the list? */
1283 if (slot != '_' && *where && (*where)->mad_key == '^') {
1284 (*where)->mad_key = slot;
1285 sv_free((SV*)((*where)->mad_val));
1286 (*where)->mad_val = (void*)sv;
1289 addmad(newMADsv(slot, sv), where, 0);
1292 # define start_force(where) NOOP
1293 # define curmad(slot, sv) NOOP
1298 * When the lexer realizes it knows the next token (for instance,
1299 * it is reordering tokens for the parser) then it can call S_force_next
1300 * to know what token to return the next time the lexer is called. Caller
1301 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1302 * and possibly PL_expect to ensure the lexer handles the token correctly.
1306 S_force_next(pTHX_ I32 type)
1310 if (PL_curforce < 0)
1311 start_force(PL_lasttoke);
1312 PL_nexttoke[PL_curforce].next_type = type;
1313 if (PL_lex_state != LEX_KNOWNEXT)
1314 PL_lex_defer = PL_lex_state;
1315 PL_lex_state = LEX_KNOWNEXT;
1316 PL_lex_expect = PL_expect;
1319 PL_nexttype[PL_nexttoke] = type;
1321 if (PL_lex_state != LEX_KNOWNEXT) {
1322 PL_lex_defer = PL_lex_state;
1323 PL_lex_expect = PL_expect;
1324 PL_lex_state = LEX_KNOWNEXT;
1330 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1333 SV * const sv = newSVpvn(start,len);
1334 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1341 * When the lexer knows the next thing is a word (for instance, it has
1342 * just seen -> and it knows that the next char is a word char, then
1343 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1347 * char *start : buffer position (must be within PL_linestr)
1348 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1349 * int check_keyword : if true, Perl checks to make sure the word isn't
1350 * a keyword (do this if the word is a label, e.g. goto FOO)
1351 * int allow_pack : if true, : characters will also be allowed (require,
1352 * use, etc. do this)
1353 * int allow_initial_tick : used by the "sub" lexer only.
1357 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1363 start = SKIPSPACE1(start);
1365 if (isIDFIRST_lazy_if(s,UTF) ||
1366 (allow_pack && *s == ':') ||
1367 (allow_initial_tick && *s == '\'') )
1369 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1370 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1372 start_force(PL_curforce);
1374 curmad('X', newSVpvn(start,s-start));
1375 if (token == METHOD) {
1380 PL_expect = XOPERATOR;
1384 curmad('g', newSVpvs( "forced" ));
1385 NEXTVAL_NEXTTOKE.opval
1386 = (OP*)newSVOP(OP_CONST,0,
1387 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1388 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1396 * Called when the lexer wants $foo *foo &foo etc, but the program
1397 * text only contains the "foo" portion. The first argument is a pointer
1398 * to the "foo", and the second argument is the type symbol to prefix.
1399 * Forces the next token to be a "WORD".
1400 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1404 S_force_ident(pTHX_ register const char *s, int kind)
1408 const STRLEN len = strlen(s);
1409 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1410 start_force(PL_curforce);
1411 NEXTVAL_NEXTTOKE.opval = o;
1414 o->op_private = OPpCONST_ENTERED;
1415 /* XXX see note in pp_entereval() for why we forgo typo
1416 warnings if the symbol must be introduced in an eval.
1418 gv_fetchpvn_flags(s, len,
1419 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1421 kind == '$' ? SVt_PV :
1422 kind == '@' ? SVt_PVAV :
1423 kind == '%' ? SVt_PVHV :
1431 Perl_str_to_version(pTHX_ SV *sv)
1436 const char *start = SvPV_const(sv,len);
1437 const char * const end = start + len;
1438 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1439 while (start < end) {
1443 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1448 retval += ((NV)n)/nshift;
1457 * Forces the next token to be a version number.
1458 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1459 * and if "guessing" is TRUE, then no new token is created (and the caller
1460 * must use an alternative parsing method).
1464 S_force_version(pTHX_ char *s, int guessing)
1470 I32 startoff = s - SvPVX(PL_linestr);
1479 while (isDIGIT(*d) || *d == '_' || *d == '.')
1483 start_force(PL_curforce);
1484 curmad('X', newSVpvn(s,d-s));
1487 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1489 s = scan_num(s, &yylval);
1490 version = yylval.opval;
1491 ver = cSVOPx(version)->op_sv;
1492 if (SvPOK(ver) && !SvNIOK(ver)) {
1493 SvUPGRADE(ver, SVt_PVNV);
1494 SvNV_set(ver, str_to_version(ver));
1495 SvNOK_on(ver); /* hint that it is a version */
1498 else if (guessing) {
1501 sv_free(PL_nextwhite); /* let next token collect whitespace */
1503 s = SvPVX(PL_linestr) + startoff;
1511 if (PL_madskills && !version) {
1512 sv_free(PL_nextwhite); /* let next token collect whitespace */
1514 s = SvPVX(PL_linestr) + startoff;
1517 /* NOTE: The parser sees the package name and the VERSION swapped */
1518 start_force(PL_curforce);
1519 NEXTVAL_NEXTTOKE.opval = version;
1527 * Tokenize a quoted string passed in as an SV. It finds the next
1528 * chunk, up to end of string or a backslash. It may make a new
1529 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1534 S_tokeq(pTHX_ SV *sv)
1538 register char *send;
1546 s = SvPV_force(sv, len);
1547 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1550 while (s < send && *s != '\\')
1555 if ( PL_hints & HINT_NEW_STRING ) {
1556 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1562 if (s + 1 < send && (s[1] == '\\'))
1563 s++; /* all that, just for this */
1568 SvCUR_set(sv, d - SvPVX_const(sv));
1570 if ( PL_hints & HINT_NEW_STRING )
1571 return new_constant(NULL, 0, "q", sv, pv, "q");
1576 * Now come three functions related to double-quote context,
1577 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1578 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1579 * interact with PL_lex_state, and create fake ( ... ) argument lists
1580 * to handle functions and concatenation.
1581 * They assume that whoever calls them will be setting up a fake
1582 * join call, because each subthing puts a ',' after it. This lets
1585 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1587 * (I'm not sure whether the spurious commas at the end of lcfirst's
1588 * arguments and join's arguments are created or not).
1593 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1595 * Pattern matching will set PL_lex_op to the pattern-matching op to
1596 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1598 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1600 * Everything else becomes a FUNC.
1602 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1603 * had an OP_CONST or OP_READLINE). This just sets us up for a
1604 * call to S_sublex_push().
1608 S_sublex_start(pTHX)
1611 register const I32 op_type = yylval.ival;
1613 if (op_type == OP_NULL) {
1614 yylval.opval = PL_lex_op;
1618 if (op_type == OP_CONST || op_type == OP_READLINE) {
1619 SV *sv = tokeq(PL_lex_stuff);
1621 if (SvTYPE(sv) == SVt_PVIV) {
1622 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1624 const char * const p = SvPV_const(sv, len);
1625 SV * const nsv = newSVpvn(p, len);
1631 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1632 PL_lex_stuff = NULL;
1633 /* Allow <FH> // "foo" */
1634 if (op_type == OP_READLINE)
1635 PL_expect = XTERMORDORDOR;
1638 else if (op_type == OP_BACKTICK && PL_lex_op) {
1639 /* readpipe() vas overriden */
1640 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1641 yylval.opval = PL_lex_op;
1643 PL_lex_stuff = NULL;
1647 PL_sublex_info.super_state = PL_lex_state;
1648 PL_sublex_info.sub_inwhat = (U16)op_type;
1649 PL_sublex_info.sub_op = PL_lex_op;
1650 PL_lex_state = LEX_INTERPPUSH;
1654 yylval.opval = PL_lex_op;
1664 * Create a new scope to save the lexing state. The scope will be
1665 * ended in S_sublex_done. Returns a '(', starting the function arguments
1666 * to the uc, lc, etc. found before.
1667 * Sets PL_lex_state to LEX_INTERPCONCAT.
1676 PL_lex_state = PL_sublex_info.super_state;
1677 SAVEBOOL(PL_lex_dojoin);
1678 SAVEI32(PL_lex_brackets);
1679 SAVEI32(PL_lex_casemods);
1680 SAVEI32(PL_lex_starts);
1681 SAVEI8(PL_lex_state);
1682 SAVEVPTR(PL_lex_inpat);
1683 SAVEI16(PL_lex_inwhat);
1684 SAVECOPLINE(PL_curcop);
1685 SAVEPPTR(PL_bufptr);
1686 SAVEPPTR(PL_bufend);
1687 SAVEPPTR(PL_oldbufptr);
1688 SAVEPPTR(PL_oldoldbufptr);
1689 SAVEPPTR(PL_last_lop);
1690 SAVEPPTR(PL_last_uni);
1691 SAVEPPTR(PL_linestart);
1692 SAVESPTR(PL_linestr);
1693 SAVEGENERICPV(PL_lex_brackstack);
1694 SAVEGENERICPV(PL_lex_casestack);
1696 PL_linestr = PL_lex_stuff;
1697 PL_lex_stuff = NULL;
1699 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1700 = SvPVX(PL_linestr);
1701 PL_bufend += SvCUR(PL_linestr);
1702 PL_last_lop = PL_last_uni = NULL;
1703 SAVEFREESV(PL_linestr);
1705 PL_lex_dojoin = FALSE;
1706 PL_lex_brackets = 0;
1707 Newx(PL_lex_brackstack, 120, char);
1708 Newx(PL_lex_casestack, 12, char);
1709 PL_lex_casemods = 0;
1710 *PL_lex_casestack = '\0';
1712 PL_lex_state = LEX_INTERPCONCAT;
1713 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1715 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1716 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1717 PL_lex_inpat = PL_sublex_info.sub_op;
1719 PL_lex_inpat = NULL;
1726 * Restores lexer state after a S_sublex_push.
1733 if (!PL_lex_starts++) {
1734 SV * const sv = newSVpvs("");
1735 if (SvUTF8(PL_linestr))
1737 PL_expect = XOPERATOR;
1738 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1742 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1743 PL_lex_state = LEX_INTERPCASEMOD;
1747 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1748 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1749 PL_linestr = PL_lex_repl;
1751 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1752 PL_bufend += SvCUR(PL_linestr);
1753 PL_last_lop = PL_last_uni = NULL;
1754 SAVEFREESV(PL_linestr);
1755 PL_lex_dojoin = FALSE;
1756 PL_lex_brackets = 0;
1757 PL_lex_casemods = 0;
1758 *PL_lex_casestack = '\0';
1760 if (SvEVALED(PL_lex_repl)) {
1761 PL_lex_state = LEX_INTERPNORMAL;
1763 /* we don't clear PL_lex_repl here, so that we can check later
1764 whether this is an evalled subst; that means we rely on the
1765 logic to ensure sublex_done() is called again only via the
1766 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1769 PL_lex_state = LEX_INTERPCONCAT;
1779 PL_endwhite = newSVpvs("");
1780 sv_catsv(PL_endwhite, PL_thiswhite);
1784 sv_setpvn(PL_thistoken,"",0);
1786 PL_realtokenstart = -1;
1790 PL_bufend = SvPVX(PL_linestr);
1791 PL_bufend += SvCUR(PL_linestr);
1792 PL_expect = XOPERATOR;
1793 PL_sublex_info.sub_inwhat = 0;
1801 Extracts a pattern, double-quoted string, or transliteration. This
1804 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1805 processing a pattern (PL_lex_inpat is true), a transliteration
1806 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1808 Returns a pointer to the character scanned up to. If this is
1809 advanced from the start pointer supplied (i.e. if anything was
1810 successfully parsed), will leave an OP for the substring scanned
1811 in yylval. Caller must intuit reason for not parsing further
1812 by looking at the next characters herself.
1816 double-quoted style: \r and \n
1817 regexp special ones: \D \s
1820 case and quoting: \U \Q \E
1821 stops on @ and $, but not for $ as tail anchor
1823 In transliterations:
1824 characters are VERY literal, except for - not at the start or end
1825 of the string, which indicates a range. If the range is in bytes,
1826 scan_const expands the range to the full set of intermediate
1827 characters. If the range is in utf8, the hyphen is replaced with
1828 a certain range mark which will be handled by pmtrans() in op.c.
1830 In double-quoted strings:
1832 double-quoted style: \r and \n
1834 deprecated backrefs: \1 (in substitution replacements)
1835 case and quoting: \U \Q \E
1838 scan_const does *not* construct ops to handle interpolated strings.
1839 It stops processing as soon as it finds an embedded $ or @ variable
1840 and leaves it to the caller to work out what's going on.
1842 embedded arrays (whether in pattern or not) could be:
1843 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1845 $ in double-quoted strings must be the symbol of an embedded scalar.
1847 $ in pattern could be $foo or could be tail anchor. Assumption:
1848 it's a tail anchor if $ is the last thing in the string, or if it's
1849 followed by one of "()| \r\n\t"
1851 \1 (backreferences) are turned into $1
1853 The structure of the code is
1854 while (there's a character to process) {
1855 handle transliteration ranges
1856 skip regexp comments /(?#comment)/ and codes /(?{code})/
1857 skip #-initiated comments in //x patterns
1858 check for embedded arrays
1859 check for embedded scalars
1861 leave intact backslashes from leaveit (below)
1862 deprecate \1 in substitution replacements
1863 handle string-changing backslashes \l \U \Q \E, etc.
1864 switch (what was escaped) {
1865 handle \- in a transliteration (becomes a literal -)
1866 handle \132 (octal characters)
1867 handle \x15 and \x{1234} (hex characters)
1868 handle \N{name} (named characters)
1869 handle \cV (control characters)
1870 handle printf-style backslashes (\f, \r, \n, etc)
1872 } (end if backslash)
1873 } (end while character to read)
1878 S_scan_const(pTHX_ char *start)
1881 register char *send = PL_bufend; /* end of the constant */
1882 SV *sv = newSV(send - start); /* sv for the constant */
1883 register char *s = start; /* start of the constant */
1884 register char *d = SvPVX(sv); /* destination for copies */
1885 bool dorange = FALSE; /* are we in a translit range? */
1886 bool didrange = FALSE; /* did we just finish a range? */
1887 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1888 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1891 UV literal_endpoint = 0;
1892 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1895 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1896 /* If we are doing a trans and we know we want UTF8 set expectation */
1897 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1898 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1902 while (s < send || dorange) {
1903 /* get transliterations out of the way (they're most literal) */
1904 if (PL_lex_inwhat == OP_TRANS) {
1905 /* expand a range A-Z to the full set of characters. AIE! */
1907 I32 i; /* current expanded character */
1908 I32 min; /* first character in range */
1909 I32 max; /* last character in range */
1920 char * const c = (char*)utf8_hop((U8*)d, -1);
1924 *c = (char)UTF_TO_NATIVE(0xff);
1925 /* mark the range as done, and continue */
1931 i = d - SvPVX_const(sv); /* remember current offset */
1934 SvLEN(sv) + (has_utf8 ?
1935 (512 - UTF_CONTINUATION_MARK +
1938 /* How many two-byte within 0..255: 128 in UTF-8,
1939 * 96 in UTF-8-mod. */
1941 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1943 d = SvPVX(sv) + i; /* refresh d after realloc */
1947 for (j = 0; j <= 1; j++) {
1948 char * const c = (char*)utf8_hop((U8*)d, -1);
1949 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1955 max = (U8)0xff; /* only to \xff */
1956 uvmax = uv; /* \x{100} to uvmax */
1958 d = c; /* eat endpoint chars */
1963 d -= 2; /* eat the first char and the - */
1964 min = (U8)*d; /* first char in range */
1965 max = (U8)d[1]; /* last char in range */
1972 "Invalid range \"%c-%c\" in transliteration operator",
1973 (char)min, (char)max);
1977 if (literal_endpoint == 2 &&
1978 ((isLOWER(min) && isLOWER(max)) ||
1979 (isUPPER(min) && isUPPER(max)))) {
1981 for (i = min; i <= max; i++)
1983 *d++ = NATIVE_TO_NEED(has_utf8,i);
1985 for (i = min; i <= max; i++)
1987 *d++ = NATIVE_TO_NEED(has_utf8,i);
1992 for (i = min; i <= max; i++)
1995 const U8 ch = (U8)NATIVE_TO_UTF(i);
1996 if (UNI_IS_INVARIANT(ch))
1999 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2000 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2009 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2011 *d++ = (char)UTF_TO_NATIVE(0xff);
2013 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2017 /* mark the range as done, and continue */
2021 literal_endpoint = 0;
2026 /* range begins (ignore - as first or last char) */
2027 else if (*s == '-' && s+1 < send && s != start) {
2029 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2036 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2046 literal_endpoint = 0;
2047 native_range = TRUE;
2052 /* if we get here, we're not doing a transliteration */
2054 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2055 except for the last char, which will be done separately. */
2056 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2058 while (s+1 < send && *s != ')')
2059 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2061 else if (s[2] == '{' /* This should match regcomp.c */
2062 || (s[2] == '?' && s[3] == '{'))
2065 char *regparse = s + (s[2] == '{' ? 3 : 4);
2068 while (count && (c = *regparse)) {
2069 if (c == '\\' && regparse[1])
2077 if (*regparse != ')')
2078 regparse--; /* Leave one char for continuation. */
2079 while (s < regparse)
2080 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2084 /* likewise skip #-initiated comments in //x patterns */
2085 else if (*s == '#' && PL_lex_inpat &&
2086 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2087 while (s+1 < send && *s != '\n')
2088 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2091 /* check for embedded arrays
2092 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2094 else if (*s == '@' && s[1]) {
2095 if (isALNUM_lazy_if(s+1,UTF))
2097 if (strchr(":'{$", s[1]))
2099 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2100 break; /* in regexp, neither @+ nor @- are interpolated */
2103 /* check for embedded scalars. only stop if we're sure it's a
2106 else if (*s == '$') {
2107 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2109 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2110 break; /* in regexp, $ might be tail anchor */
2113 /* End of else if chain - OP_TRANS rejoin rest */
2116 if (*s == '\\' && s+1 < send) {
2119 /* deprecate \1 in strings and substitution replacements */
2120 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2121 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2123 if (ckWARN(WARN_SYNTAX))
2124 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2129 /* string-change backslash escapes */
2130 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2134 /* skip any other backslash escapes in a pattern */
2135 else if (PL_lex_inpat) {
2136 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2137 goto default_action;
2140 /* if we get here, it's either a quoted -, or a digit */
2143 /* quoted - in transliterations */
2145 if (PL_lex_inwhat == OP_TRANS) {
2152 if ((isALPHA(*s) || isDIGIT(*s)) &&
2154 Perl_warner(aTHX_ packWARN(WARN_MISC),
2155 "Unrecognized escape \\%c passed through",
2157 /* default action is to copy the quoted character */
2158 goto default_action;
2161 /* \132 indicates an octal constant */
2162 case '0': case '1': case '2': case '3':
2163 case '4': case '5': case '6': case '7':
2167 uv = grok_oct(s, &len, &flags, NULL);
2170 goto NUM_ESCAPE_INSERT;
2172 /* \x24 indicates a hex constant */
2176 char* const e = strchr(s, '}');
2177 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2178 PERL_SCAN_DISALLOW_PREFIX;
2183 yyerror("Missing right brace on \\x{}");
2187 uv = grok_hex(s, &len, &flags, NULL);
2193 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2194 uv = grok_hex(s, &len, &flags, NULL);
2200 /* Insert oct or hex escaped character.
2201 * There will always enough room in sv since such
2202 * escapes will be longer than any UTF-8 sequence
2203 * they can end up as. */
2205 /* We need to map to chars to ASCII before doing the tests
2208 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2209 if (!has_utf8 && uv > 255) {
2210 /* Might need to recode whatever we have
2211 * accumulated so far if it contains any
2214 * (Can't we keep track of that and avoid
2215 * this rescan? --jhi)
2219 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2220 if (!NATIVE_IS_INVARIANT(*c)) {
2225 const STRLEN offset = d - SvPVX_const(sv);
2227 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2231 while (src >= (const U8 *)SvPVX_const(sv)) {
2232 if (!NATIVE_IS_INVARIANT(*src)) {
2233 const U8 ch = NATIVE_TO_ASCII(*src);
2234 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2235 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2245 if (has_utf8 || uv > 255) {
2246 d = (char*)uvchr_to_utf8((U8*)d, uv);
2248 if (PL_lex_inwhat == OP_TRANS &&
2249 PL_sublex_info.sub_op) {
2250 PL_sublex_info.sub_op->op_private |=
2251 (PL_lex_repl ? OPpTRANS_FROM_UTF
2255 if (uv > 255 && !dorange)
2256 native_range = FALSE;
2268 /* \N{LATIN SMALL LETTER A} is a named character */
2272 char* e = strchr(s, '}');
2279 yyerror("Missing right brace on \\N{}");
2283 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2285 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2286 PERL_SCAN_DISALLOW_PREFIX;
2289 uv = grok_hex(s, &len, &flags, NULL);
2290 if ( e > s && len != (STRLEN)(e - s) ) {
2294 goto NUM_ESCAPE_INSERT;
2296 res = newSVpvn(s + 1, e - s - 1);
2297 type = newSVpvn(s - 2,e - s + 3);
2298 res = new_constant( NULL, 0, "charnames",
2299 res, NULL, SvPVX(type) );
2302 sv_utf8_upgrade(res);
2303 str = SvPV_const(res,len);
2304 #ifdef EBCDIC_NEVER_MIND
2305 /* charnames uses pack U and that has been
2306 * recently changed to do the below uni->native
2307 * mapping, so this would be redundant (and wrong,
2308 * the code point would be doubly converted).
2309 * But leave this in just in case the pack U change
2310 * gets revoked, but the semantics is still
2311 * desireable for charnames. --jhi */
2313 UV uv = utf8_to_uvchr((const U8*)str, 0);
2316 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2318 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2319 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2320 str = SvPV_const(res, len);
2324 if (!has_utf8 && SvUTF8(res)) {
2325 const char * const ostart = SvPVX_const(sv);
2326 SvCUR_set(sv, d - ostart);
2329 sv_utf8_upgrade(sv);
2330 /* this just broke our allocation above... */
2331 SvGROW(sv, (STRLEN)(send - start));
2332 d = SvPVX(sv) + SvCUR(sv);
2335 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2336 const char * const odest = SvPVX_const(sv);
2338 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2339 d = SvPVX(sv) + (d - odest);
2343 native_range = FALSE; /* \N{} is guessed to be Unicode */
2345 Copy(str, d, len, char);
2352 yyerror("Missing braces on \\N{}");
2355 /* \c is a control character */
2364 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2367 yyerror("Missing control char name in \\c");
2371 /* printf-style backslashes, formfeeds, newlines, etc */
2373 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2376 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2379 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2382 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2385 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2388 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2391 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2397 } /* end if (backslash) */
2404 /* If we started with encoded form, or already know we want it
2405 and then encode the next character */
2406 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2408 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2409 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2412 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2413 const STRLEN off = d - SvPVX_const(sv);
2414 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2416 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2419 if (uv > 255 && !dorange)
2420 native_range = FALSE;
2424 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2426 } /* while loop to process each character */
2428 /* terminate the string and set up the sv */
2430 SvCUR_set(sv, d - SvPVX_const(sv));
2431 if (SvCUR(sv) >= SvLEN(sv))
2432 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2435 if (PL_encoding && !has_utf8) {
2436 sv_recode_to_utf8(sv, PL_encoding);
2442 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2443 PL_sublex_info.sub_op->op_private |=
2444 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2448 /* shrink the sv if we allocated more than we used */
2449 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2450 SvPV_shrink_to_cur(sv);
2453 /* return the substring (via yylval) only if we parsed anything */
2454 if (s > PL_bufptr) {
2455 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2456 sv = new_constant(start, s - start,
2457 (const char *)(PL_lex_inpat ? "qr" : "q"),
2460 (( PL_lex_inwhat == OP_TRANS
2462 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2465 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2472 * Returns TRUE if there's more to the expression (e.g., a subscript),
2475 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2477 * ->[ and ->{ return TRUE
2478 * { and [ outside a pattern are always subscripts, so return TRUE
2479 * if we're outside a pattern and it's not { or [, then return FALSE
2480 * if we're in a pattern and the first char is a {
2481 * {4,5} (any digits around the comma) returns FALSE
2482 * if we're in a pattern and the first char is a [
2484 * [SOMETHING] has a funky algorithm to decide whether it's a
2485 * character class or not. It has to deal with things like
2486 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2487 * anything else returns TRUE
2490 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2493 S_intuit_more(pTHX_ register char *s)
2496 if (PL_lex_brackets)
2498 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2500 if (*s != '{' && *s != '[')
2505 /* In a pattern, so maybe we have {n,m}. */
2522 /* On the other hand, maybe we have a character class */
2525 if (*s == ']' || *s == '^')
2528 /* this is terrifying, and it works */
2529 int weight = 2; /* let's weigh the evidence */
2531 unsigned char un_char = 255, last_un_char;
2532 const char * const send = strchr(s,']');
2533 char tmpbuf[sizeof PL_tokenbuf * 4];
2535 if (!send) /* has to be an expression */
2538 Zero(seen,256,char);
2541 else if (isDIGIT(*s)) {
2543 if (isDIGIT(s[1]) && s[2] == ']')
2549 for (; s < send; s++) {
2550 last_un_char = un_char;
2551 un_char = (unsigned char)*s;
2556 weight -= seen[un_char] * 10;
2557 if (isALNUM_lazy_if(s+1,UTF)) {
2559 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2560 len = (int)strlen(tmpbuf);
2561 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2566 else if (*s == '$' && s[1] &&
2567 strchr("[#!%*<>()-=",s[1])) {
2568 if (/*{*/ strchr("])} =",s[2]))
2577 if (strchr("wds]",s[1]))
2579 else if (seen[(U8)'\''] || seen[(U8)'"'])
2581 else if (strchr("rnftbxcav",s[1]))
2583 else if (isDIGIT(s[1])) {
2585 while (s[1] && isDIGIT(s[1]))
2595 if (strchr("aA01! ",last_un_char))
2597 if (strchr("zZ79~",s[1]))
2599 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2600 weight -= 5; /* cope with negative subscript */
2603 if (!isALNUM(last_un_char)
2604 && !(last_un_char == '$' || last_un_char == '@'
2605 || last_un_char == '&')
2606 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2611 if (keyword(tmpbuf, d - tmpbuf, 0))
2614 if (un_char == last_un_char + 1)
2616 weight -= seen[un_char];
2621 if (weight >= 0) /* probably a character class */
2631 * Does all the checking to disambiguate
2633 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2634 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2636 * First argument is the stuff after the first token, e.g. "bar".
2638 * Not a method if bar is a filehandle.
2639 * Not a method if foo is a subroutine prototyped to take a filehandle.
2640 * Not a method if it's really "Foo $bar"
2641 * Method if it's "foo $bar"
2642 * Not a method if it's really "print foo $bar"
2643 * Method if it's really "foo package::" (interpreted as package->foo)
2644 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2645 * Not a method if bar is a filehandle or package, but is quoted with
2650 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2653 char *s = start + (*start == '$');
2654 char tmpbuf[sizeof PL_tokenbuf];
2662 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2666 const char *proto = SvPVX_const(cv);
2677 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2678 /* start is the beginning of the possible filehandle/object,
2679 * and s is the end of it
2680 * tmpbuf is a copy of it
2683 if (*start == '$') {
2684 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
2685 isUPPER(*PL_tokenbuf))
2688 len = start - SvPVX(PL_linestr);
2692 start = SvPVX(PL_linestr) + len;
2696 return *s == '(' ? FUNCMETH : METHOD;
2698 if (!keyword(tmpbuf, len, 0)) {
2699 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2703 soff = s - SvPVX(PL_linestr);
2707 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2708 if (indirgv && GvCVu(indirgv))
2710 /* filehandle or package name makes it a method */
2711 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2713 soff = s - SvPVX(PL_linestr);
2716 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2717 return 0; /* no assumptions -- "=>" quotes bearword */
2719 start_force(PL_curforce);
2720 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2721 newSVpvn(tmpbuf,len));
2722 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2724 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2729 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2731 return *s == '(' ? FUNCMETH : METHOD;
2739 * Return a string of Perl code to load the debugger. If PERL5DB
2740 * is set, it will return the contents of that, otherwise a
2741 * compile-time require of perl5db.pl.
2749 const char * const pdb = PerlEnv_getenv("PERL5DB");
2753 SETERRNO(0,SS_NORMAL);
2754 return "BEGIN { require 'perl5db.pl' }";
2760 /* Encoded script support. filter_add() effectively inserts a
2761 * 'pre-processing' function into the current source input stream.
2762 * Note that the filter function only applies to the current source file
2763 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2765 * The datasv parameter (which may be NULL) can be used to pass
2766 * private data to this instance of the filter. The filter function
2767 * can recover the SV using the FILTER_DATA macro and use it to
2768 * store private buffers and state information.
2770 * The supplied datasv parameter is upgraded to a PVIO type
2771 * and the IoDIRP/IoANY field is used to store the function pointer,
2772 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2773 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2774 * private use must be set using malloc'd pointers.
2778 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2787 if (!PL_rsfp_filters)
2788 PL_rsfp_filters = newAV();
2791 SvUPGRADE(datasv, SVt_PVIO);
2792 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2793 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2794 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2795 FPTR2DPTR(void *, IoANY(datasv)),
2796 SvPV_nolen(datasv)));
2797 av_unshift(PL_rsfp_filters, 1);
2798 av_store(PL_rsfp_filters, 0, datasv) ;
2803 /* Delete most recently added instance of this filter function. */
2805 Perl_filter_del(pTHX_ filter_t funcp)
2811 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2812 FPTR2DPTR(void*, funcp)));
2814 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2816 /* if filter is on top of stack (usual case) just pop it off */
2817 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2818 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2819 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2820 IoANY(datasv) = (void *)NULL;
2821 sv_free(av_pop(PL_rsfp_filters));
2825 /* we need to search for the correct entry and clear it */
2826 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2830 /* Invoke the idxth filter function for the current rsfp. */
2831 /* maxlen 0 = read one text line */
2833 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2838 /* This API is bad. It should have been using unsigned int for maxlen.
2839 Not sure if we want to change the API, but if not we should sanity
2840 check the value here. */
2841 const unsigned int correct_length
2850 if (!PL_parser || !PL_rsfp_filters)
2852 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2853 /* Provide a default input filter to make life easy. */
2854 /* Note that we append to the line. This is handy. */
2855 DEBUG_P(PerlIO_printf(Perl_debug_log,
2856 "filter_read %d: from rsfp\n", idx));
2857 if (correct_length) {
2860 const int old_len = SvCUR(buf_sv);
2862 /* ensure buf_sv is large enough */
2863 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2864 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2865 correct_length)) <= 0) {
2866 if (PerlIO_error(PL_rsfp))
2867 return -1; /* error */
2869 return 0 ; /* end of file */
2871 SvCUR_set(buf_sv, old_len + len) ;
2874 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2875 if (PerlIO_error(PL_rsfp))
2876 return -1; /* error */
2878 return 0 ; /* end of file */
2881 return SvCUR(buf_sv);
2883 /* Skip this filter slot if filter has been deleted */
2884 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2885 DEBUG_P(PerlIO_printf(Perl_debug_log,
2886 "filter_read %d: skipped (filter deleted)\n",
2888 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2890 /* Get function pointer hidden within datasv */
2891 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2892 DEBUG_P(PerlIO_printf(Perl_debug_log,
2893 "filter_read %d: via function %p (%s)\n",
2894 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2895 /* Call function. The function is expected to */
2896 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2897 /* Return: <0:error, =0:eof, >0:not eof */
2898 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2902 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2905 #ifdef PERL_CR_FILTER
2906 if (!PL_rsfp_filters) {
2907 filter_add(S_cr_textfilter,NULL);
2910 if (PL_rsfp_filters) {
2912 SvCUR_set(sv, 0); /* start with empty line */
2913 if (FILTER_READ(0, sv, 0) > 0)
2914 return ( SvPVX(sv) ) ;
2919 return (sv_gets(sv, fp, append));
2923 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2928 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2932 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2933 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2935 return GvHV(gv); /* Foo:: */
2938 /* use constant CLASS => 'MyClass' */
2939 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2940 if (gv && GvCV(gv)) {
2941 SV * const sv = cv_const_sv(GvCV(gv));
2943 pkgname = SvPV_nolen_const(sv);
2946 return gv_stashpv(pkgname, 0);
2950 * S_readpipe_override
2951 * Check whether readpipe() is overriden, and generates the appropriate
2952 * optree, provided sublex_start() is called afterwards.
2955 S_readpipe_override(pTHX)
2958 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2959 yylval.ival = OP_BACKTICK;
2961 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2963 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2964 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
2965 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2967 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2968 append_elem(OP_LIST,
2969 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2970 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2977 * The intent of this yylex wrapper is to minimize the changes to the
2978 * tokener when we aren't interested in collecting madprops. It remains
2979 * to be seen how successful this strategy will be...
2986 char *s = PL_bufptr;
2988 /* make sure PL_thiswhite is initialized */
2992 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2993 if (PL_pending_ident)
2994 return S_pending_ident(aTHX);
2996 /* previous token ate up our whitespace? */
2997 if (!PL_lasttoke && PL_nextwhite) {
2998 PL_thiswhite = PL_nextwhite;
3002 /* isolate the token, and figure out where it is without whitespace */
3003 PL_realtokenstart = -1;
3007 assert(PL_curforce < 0);
3009 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3010 if (!PL_thistoken) {
3011 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3012 PL_thistoken = newSVpvs("");
3014 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3015 PL_thistoken = newSVpvn(tstart, s - tstart);
3018 if (PL_thismad) /* install head */
3019 CURMAD('X', PL_thistoken);
3022 /* last whitespace of a sublex? */
3023 if (optype == ')' && PL_endwhite) {
3024 CURMAD('X', PL_endwhite);
3029 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3030 if (!PL_thiswhite && !PL_endwhite && !optype) {
3031 sv_free(PL_thistoken);
3036 /* put off final whitespace till peg */
3037 if (optype == ';' && !PL_rsfp) {
3038 PL_nextwhite = PL_thiswhite;
3041 else if (PL_thisopen) {
3042 CURMAD('q', PL_thisopen);
3044 sv_free(PL_thistoken);
3048 /* Store actual token text as madprop X */
3049 CURMAD('X', PL_thistoken);
3053 /* add preceding whitespace as madprop _ */
3054 CURMAD('_', PL_thiswhite);
3058 /* add quoted material as madprop = */
3059 CURMAD('=', PL_thisstuff);
3063 /* add terminating quote as madprop Q */
3064 CURMAD('Q', PL_thisclose);
3068 /* special processing based on optype */
3072 /* opval doesn't need a TOKEN since it can already store mp */
3083 append_madprops(PL_thismad, yylval.opval, 0);
3091 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3100 /* remember any fake bracket that lexer is about to discard */
3101 if (PL_lex_brackets == 1 &&
3102 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3105 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3108 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3109 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3112 break; /* don't bother looking for trailing comment */
3121 /* attach a trailing comment to its statement instead of next token */
3125 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3127 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3129 if (*s == '\n' || *s == '#') {
3130 while (s < PL_bufend && *s != '\n')
3134 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3135 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3152 /* Create new token struct. Note: opvals return early above. */
3153 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3160 S_tokenize_use(pTHX_ int is_use, char *s) {
3162 if (PL_expect != XSTATE)
3163 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3164 is_use ? "use" : "no"));
3166 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3167 s = force_version(s, TRUE);
3168 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3169 start_force(PL_curforce);
3170 NEXTVAL_NEXTTOKE.opval = NULL;
3173 else if (*s == 'v') {
3174 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3175 s = force_version(s, FALSE);
3179 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3180 s = force_version(s, FALSE);
3182 yylval.ival = is_use;
3186 static const char* const exp_name[] =
3187 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3188 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3195 Works out what to call the token just pulled out of the input
3196 stream. The yacc parser takes care of taking the ops we return and
3197 stitching them into a tree.
3203 if read an identifier
3204 if we're in a my declaration
3205 croak if they tried to say my($foo::bar)
3206 build the ops for a my() declaration
3207 if it's an access to a my() variable
3208 are we in a sort block?
3209 croak if my($a); $a <=> $b
3210 build ops for access to a my() variable
3211 if in a dq string, and they've said @foo and we can't find @foo
3213 build ops for a bareword
3214 if we already built the token before, use it.
3219 #pragma segment Perl_yylex
3225 register char *s = PL_bufptr;
3230 /* orig_keyword, gvp, and gv are initialized here because
3231 * jump to the label just_a_word_zero can bypass their
3232 * initialization later. */
3233 I32 orig_keyword = 0;
3238 SV* tmp = newSVpvs("");
3239 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3240 (IV)CopLINE(PL_curcop),
3241 lex_state_names[PL_lex_state],
3242 exp_name[PL_expect],
3243 pv_display(tmp, s, strlen(s), 0, 60));
3246 /* check if there's an identifier for us to look at */
3247 if (PL_pending_ident)
3248 return REPORT(S_pending_ident(aTHX));
3250 /* no identifier pending identification */
3252 switch (PL_lex_state) {
3254 case LEX_NORMAL: /* Some compilers will produce faster */
3255 case LEX_INTERPNORMAL: /* code if we comment these out. */
3259 /* when we've already built the next token, just pull it out of the queue */
3263 yylval = PL_nexttoke[PL_lasttoke].next_val;
3265 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3266 PL_nexttoke[PL_lasttoke].next_mad = 0;
3267 if (PL_thismad && PL_thismad->mad_key == '_') {
3268 PL_thiswhite = (SV*)PL_thismad->mad_val;
3269 PL_thismad->mad_val = 0;
3270 mad_free(PL_thismad);
3275 PL_lex_state = PL_lex_defer;
3276 PL_expect = PL_lex_expect;
3277 PL_lex_defer = LEX_NORMAL;
3278 if (!PL_nexttoke[PL_lasttoke].next_type)
3283 yylval = PL_nextval[PL_nexttoke];
3285 PL_lex_state = PL_lex_defer;
3286 PL_expect = PL_lex_expect;
3287 PL_lex_defer = LEX_NORMAL;
3291 /* FIXME - can these be merged? */
3292 return(PL_nexttoke[PL_lasttoke].next_type);
3294 return REPORT(PL_nexttype[PL_nexttoke]);
3297 /* interpolated case modifiers like \L \U, including \Q and \E.
3298 when we get here, PL_bufptr is at the \
3300 case LEX_INTERPCASEMOD:
3302 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3303 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3305 /* handle \E or end of string */
3306 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3308 if (PL_lex_casemods) {
3309 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3310 PL_lex_casestack[PL_lex_casemods] = '\0';
3312 if (PL_bufptr != PL_bufend
3313 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3315 PL_lex_state = LEX_INTERPCONCAT;
3318 PL_thistoken = newSVpvs("\\E");
3324 while (PL_bufptr != PL_bufend &&
3325 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3327 PL_thiswhite = newSVpvs("");
3328 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3332 if (PL_bufptr != PL_bufend)
3335 PL_lex_state = LEX_INTERPCONCAT;
3339 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3340 "### Saw case modifier\n"); });
3342 if (s[1] == '\\' && s[2] == 'E') {
3345 PL_thiswhite = newSVpvs("");
3346 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3349 PL_lex_state = LEX_INTERPCONCAT;
3354 if (!PL_madskills) /* when just compiling don't need correct */
3355 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3356 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3357 if ((*s == 'L' || *s == 'U') &&
3358 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3359 PL_lex_casestack[--PL_lex_casemods] = '\0';
3362 if (PL_lex_casemods > 10)
3363 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3364 PL_lex_casestack[PL_lex_casemods++] = *s;
3365 PL_lex_casestack[PL_lex_casemods] = '\0';
3366 PL_lex_state = LEX_INTERPCONCAT;
3367 start_force(PL_curforce);
3368 NEXTVAL_NEXTTOKE.ival = 0;
3370 start_force(PL_curforce);
3372 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3374 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3376 NEXTVAL_NEXTTOKE.ival = OP_LC;
3378 NEXTVAL_NEXTTOKE.ival = OP_UC;
3380 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3382 Perl_croak(aTHX_ "panic: yylex");
3384 SV* const tmpsv = newSVpvs("\\ ");
3385 /* replace the space with the character we want to escape
3387 SvPVX(tmpsv)[1] = *s;
3393 if (PL_lex_starts) {
3399 sv_free(PL_thistoken);
3400 PL_thistoken = newSVpvs("");
3403 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3404 if (PL_lex_casemods == 1 && PL_lex_inpat)
3413 case LEX_INTERPPUSH:
3414 return REPORT(sublex_push());
3416 case LEX_INTERPSTART:
3417 if (PL_bufptr == PL_bufend)
3418 return REPORT(sublex_done());
3419 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3420 "### Interpolated variable\n"); });
3422 PL_lex_dojoin = (*PL_bufptr == '@');
3423 PL_lex_state = LEX_INTERPNORMAL;
3424 if (PL_lex_dojoin) {
3425 start_force(PL_curforce);
3426 NEXTVAL_NEXTTOKE.ival = 0;
3428 start_force(PL_curforce);
3429 force_ident("\"", '$');
3430 start_force(PL_curforce);
3431 NEXTVAL_NEXTTOKE.ival = 0;
3433 start_force(PL_curforce);
3434 NEXTVAL_NEXTTOKE.ival = 0;
3436 start_force(PL_curforce);
3437 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3440 if (PL_lex_starts++) {
3445 sv_free(PL_thistoken);
3446 PL_thistoken = newSVpvs("");
3449 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3450 if (!PL_lex_casemods && PL_lex_inpat)
3457 case LEX_INTERPENDMAYBE:
3458 if (intuit_more(PL_bufptr)) {
3459 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3465 if (PL_lex_dojoin) {
3466 PL_lex_dojoin = FALSE;
3467 PL_lex_state = LEX_INTERPCONCAT;
3471 sv_free(PL_thistoken);
3472 PL_thistoken = newSVpvs("");
3477 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3478 && SvEVALED(PL_lex_repl))
3480 if (PL_bufptr != PL_bufend)
3481 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3485 case LEX_INTERPCONCAT:
3487 if (PL_lex_brackets)
3488 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3490 if (PL_bufptr == PL_bufend)
3491 return REPORT(sublex_done());
3493 if (SvIVX(PL_linestr) == '\'') {
3494 SV *sv = newSVsv(PL_linestr);
3497 else if ( PL_hints & HINT_NEW_RE )
3498 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3499 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3503 s = scan_const(PL_bufptr);
3505 PL_lex_state = LEX_INTERPCASEMOD;
3507 PL_lex_state = LEX_INTERPSTART;
3510 if (s != PL_bufptr) {
3511 start_force(PL_curforce);
3513 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3515 NEXTVAL_NEXTTOKE = yylval;
3518 if (PL_lex_starts++) {
3522 sv_free(PL_thistoken);
3523 PL_thistoken = newSVpvs("");
3526 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3527 if (!PL_lex_casemods && PL_lex_inpat)
3540 PL_lex_state = LEX_NORMAL;
3541 s = scan_formline(PL_bufptr);
3542 if (!PL_lex_formbrack)
3548 PL_oldoldbufptr = PL_oldbufptr;
3554 sv_free(PL_thistoken);
3557 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3561 if (isIDFIRST_lazy_if(s,UTF))
3563 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3564 Perl_croak(aTHX_ "Unrecognized character \\x%02X in column %d", *s & 255, (int) len + 1);
3567 goto fake_eof; /* emulate EOF on ^D or ^Z */
3576 if (PL_lex_brackets) {
3577 yyerror((const char *)
3579 ? "Format not terminated"
3580 : "Missing right curly or square bracket"));
3582 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3583 "### Tokener got EOF\n");
3587 if (s++ < PL_bufend)
3588 goto retry; /* ignore stray nulls */
3591 if (!PL_in_eval && !PL_preambled) {
3592 PL_preambled = TRUE;
3597 sv_setpv(PL_linestr,incl_perldb());
3598 if (SvCUR(PL_linestr))
3599 sv_catpvs(PL_linestr,";");
3601 while(AvFILLp(PL_preambleav) >= 0) {
3602 SV *tmpsv = av_shift(PL_preambleav);
3603 sv_catsv(PL_linestr, tmpsv);
3604 sv_catpvs(PL_linestr, ";");
3607 sv_free((SV*)PL_preambleav);
3608 PL_preambleav = NULL;
3610 if (PL_minus_n || PL_minus_p) {
3611 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3613 sv_catpvs(PL_linestr,"chomp;");
3616 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3617 || *PL_splitstr == '"')
3618 && strchr(PL_splitstr + 1, *PL_splitstr))
3619 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3621 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3622 bytes can be used as quoting characters. :-) */
3623 const char *splits = PL_splitstr;
3624 sv_catpvs(PL_linestr, "our @F=split(q\0");
3627 if (*splits == '\\')
3628 sv_catpvn(PL_linestr, splits, 1);
3629 sv_catpvn(PL_linestr, splits, 1);
3630 } while (*splits++);
3631 /* This loop will embed the trailing NUL of
3632 PL_linestr as the last thing it does before
3634 sv_catpvs(PL_linestr, ");");
3638 sv_catpvs(PL_linestr,"our @F=split(' ');");
3642 sv_catpvs(PL_linestr,"use feature ':5.10';");
3643 sv_catpvs(PL_linestr, "\n");
3644 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3645 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3646 PL_last_lop = PL_last_uni = NULL;
3647 if (PERLDB_LINE && PL_curstash != PL_debstash)
3648 update_debugger_info(PL_linestr, NULL, 0);
3652 bof = PL_rsfp ? TRUE : FALSE;
3653 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3656 PL_realtokenstart = -1;
3659 if (PL_preprocess && !PL_in_eval)
3660 (void)PerlProc_pclose(PL_rsfp);
3661 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3662 PerlIO_clearerr(PL_rsfp);
3664 (void)PerlIO_close(PL_rsfp);
3666 PL_doextract = FALSE;
3668 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3673 sv_setpv(PL_linestr,
3676 ? ";}continue{print;}" : ";}"));
3677 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3678 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3679 PL_last_lop = PL_last_uni = NULL;
3680 PL_minus_n = PL_minus_p = 0;
3683 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3684 PL_last_lop = PL_last_uni = NULL;
3685 sv_setpvn(PL_linestr,"",0);
3686 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3688 /* If it looks like the start of a BOM or raw UTF-16,
3689 * check if it in fact is. */
3695 #ifdef PERLIO_IS_STDIO
3696 # ifdef __GNU_LIBRARY__
3697 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3698 # define FTELL_FOR_PIPE_IS_BROKEN
3702 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3703 # define FTELL_FOR_PIPE_IS_BROKEN
3708 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3709 /* This loses the possibility to detect the bof
3710 * situation on perl -P when the libc5 is being used.
3711 * Workaround? Maybe attach some extra state to PL_rsfp?
3714 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3716 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3719 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3720 s = swallow_bom((U8*)s);
3724 /* Incest with pod. */
3727 sv_catsv(PL_thiswhite, PL_linestr);
3729 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3730 sv_setpvn(PL_linestr, "", 0);
3731 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3732 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3733 PL_last_lop = PL_last_uni = NULL;
3734 PL_doextract = FALSE;
3738 } while (PL_doextract);
3739 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3740 if (PERLDB_LINE && PL_curstash != PL_debstash)
3741 update_debugger_info(PL_linestr, NULL, 0);
3742 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3743 PL_last_lop = PL_last_uni = NULL;
3744 if (CopLINE(PL_curcop) == 1) {
3745 while (s < PL_bufend && isSPACE(*s))
3747 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3751 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3755 if (*s == '#' && *(s+1) == '!')
3757 #ifdef ALTERNATE_SHEBANG
3759 static char const as[] = ALTERNATE_SHEBANG;
3760 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3761 d = s + (sizeof(as) - 1);
3763 #endif /* ALTERNATE_SHEBANG */
3772 while (*d && !isSPACE(*d))
3776 #ifdef ARG_ZERO_IS_SCRIPT
3777 if (ipathend > ipath) {
3779 * HP-UX (at least) sets argv[0] to the script name,
3780 * which makes $^X incorrect. And Digital UNIX and Linux,
3781 * at least, set argv[0] to the basename of the Perl
3782 * interpreter. So, having found "#!", we'll set it right.
3784 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3786 assert(SvPOK(x) || SvGMAGICAL(x));
3787 if (sv_eq(x, CopFILESV(PL_curcop))) {
3788 sv_setpvn(x, ipath, ipathend - ipath);
3794 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3795 const char * const lstart = SvPV_const(x,llen);
3797 bstart += blen - llen;
3798 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3799 sv_setpvn(x, ipath, ipathend - ipath);
3804 TAINT_NOT; /* $^X is always tainted, but that's OK */
3806 #endif /* ARG_ZERO_IS_SCRIPT */
3811 d = instr(s,"perl -");
3813 d = instr(s,"perl");
3815 /* avoid getting into infinite loops when shebang
3816 * line contains "Perl" rather than "perl" */
3818 for (d = ipathend-4; d >= ipath; --d) {
3819 if ((*d == 'p' || *d == 'P')
3820 && !ibcmp(d, "perl", 4))
3830 #ifdef ALTERNATE_SHEBANG
3832 * If the ALTERNATE_SHEBANG on this system starts with a
3833 * character that can be part of a Perl expression, then if
3834 * we see it but not "perl", we're probably looking at the
3835 * start of Perl code, not a request to hand off to some
3836 * other interpreter. Similarly, if "perl" is there, but
3837 * not in the first 'word' of the line, we assume the line
3838 * contains the start of the Perl program.
3840 if (d && *s != '#') {
3841 const char *c = ipath;
3842 while (*c && !strchr("; \t\r\n\f\v#", *c))
3845 d = NULL; /* "perl" not in first word; ignore */
3847 *s = '#'; /* Don't try to parse shebang line */
3849 #endif /* ALTERNATE_SHEBANG */
3850 #ifndef MACOS_TRADITIONAL
3855 !instr(s,"indir") &&
3856 instr(PL_origargv[0],"perl"))
3863 while (s < PL_bufend && isSPACE(*s))
3865 if (s < PL_bufend) {
3866 Newxz(newargv,PL_origargc+3,char*);
3868 while (s < PL_bufend && !isSPACE(*s))
3871 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3874 newargv = PL_origargv;
3877 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3879 Perl_croak(aTHX_ "Can't exec %s", ipath);
3883 while (*d && !isSPACE(*d))
3885 while (SPACE_OR_TAB(*d))
3889 const bool switches_done = PL_doswitches;
3890 const U32 oldpdb = PL_perldb;
3891 const bool oldn = PL_minus_n;
3892 const bool oldp = PL_minus_p;
3896 if (*d1 == 'M' || *d1 == 'm' || *d1 == 'C') {
3897 const char * const m = d1;
3898 while (*d1 && !isSPACE(*d1))
3900 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3903 d1 = moreswitches(d1);
3905 if (PL_doswitches && !switches_done) {
3906 int argc = PL_origargc;
3907 char **argv = PL_origargv;
3910 } while (argc && argv[0][0] == '-' && argv[0][1]);
3911 init_argv_symbols(argc,argv);
3913 if ((PERLDB_LINE && !oldpdb) ||
3914 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3915 /* if we have already added "LINE: while (<>) {",
3916 we must not do it again */
3918 sv_setpvn(PL_linestr, "", 0);
3919 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3920 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3921 PL_last_lop = PL_last_uni = NULL;
3922 PL_preambled = FALSE;
3924 (void)gv_fetchfile(PL_origfilename);
3931 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3933 PL_lex_state = LEX_FORMLINE;
3938 #ifdef PERL_STRICT_CR
3939 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3941 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3943 case ' ': case '\t': case '\f': case 013:
3944 #ifdef MACOS_TRADITIONAL
3948 PL_realtokenstart = -1;
3950 PL_thiswhite = newSVpvs("");
3951 sv_catpvn(PL_thiswhite, s, 1);
3958 PL_realtokenstart = -1;
3962 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3963 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3964 /* handle eval qq[#line 1 "foo"\n ...] */
3965 CopLINE_dec(PL_curcop);
3968 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3970 if (!PL_in_eval || PL_rsfp)
3975 while (d < PL_bufend && *d != '\n')
3979 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3980 Perl_croak(aTHX_ "panic: input overflow");
3983 PL_thiswhite = newSVpvn(s, d - s);
3988 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3990 PL_lex_state = LEX_FORMLINE;
3996 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3997 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4000 TOKEN(PEG); /* make sure any #! line is accessible */
4005 /* if (PL_madskills && PL_lex_formbrack) { */
4007 while (d < PL_bufend && *d != '\n')
4011 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4012 Perl_croak(aTHX_ "panic: input overflow");
4013 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4015 PL_thiswhite = newSVpvs("");
4016 if (CopLINE(PL_curcop) == 1) {
4017 sv_setpvn(PL_thiswhite, "", 0);
4020 sv_catpvn(PL_thiswhite, s, d - s);
4034 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4042 while (s < PL_bufend && SPACE_OR_TAB(*s))
4045 if (strnEQ(s,"=>",2)) {
4046 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4047 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4048 OPERATOR('-'); /* unary minus */
4050 PL_last_uni = PL_oldbufptr;
4052 case 'r': ftst = OP_FTEREAD; break;
4053 case 'w': ftst = OP_FTEWRITE; break;
4054 case 'x': ftst = OP_FTEEXEC; break;
4055 case 'o': ftst = OP_FTEOWNED; break;
4056 case 'R': ftst = OP_FTRREAD; break;
4057 case 'W': ftst = OP_FTRWRITE; break;
4058 case 'X': ftst = OP_FTREXEC; break;
4059 case 'O': ftst = OP_FTROWNED; break;
4060 case 'e': ftst = OP_FTIS; break;
4061 case 'z': ftst = OP_FTZERO; break;
4062 case 's': ftst = OP_FTSIZE; break;
4063 case 'f': ftst = OP_FTFILE; break;
4064 case 'd': ftst = OP_FTDIR; break;
4065 case 'l': ftst = OP_FTLINK; break;
4066 case 'p': ftst = OP_FTPIPE; break;
4067 case 'S': ftst = OP_FTSOCK; break;
4068 case 'u': ftst = OP_FTSUID; break;
4069 case 'g': ftst = OP_FTSGID; break;
4070 case 'k': ftst = OP_FTSVTX; break;
4071 case 'b': ftst = OP_FTBLK; break;
4072 case 'c': ftst = OP_FTCHR; break;
4073 case 't': ftst = OP_FTTTY; break;
4074 case 'T': ftst = OP_FTTEXT; break;
4075 case 'B': ftst = OP_FTBINARY; break;
4076 case 'M': case 'A': case 'C':
4077 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4079 case 'M': ftst = OP_FTMTIME; break;
4080 case 'A': ftst = OP_FTATIME; break;
4081 case 'C': ftst = OP_FTCTIME; break;
4089 PL_last_lop_op = (OPCODE)ftst;
4090 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4091 "### Saw file test %c\n", (int)tmp);
4096 /* Assume it was a minus followed by a one-letter named
4097 * subroutine call (or a -bareword), then. */
4098 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4099 "### '-%c' looked like a file test but was not\n",
4106 const char tmp = *s++;
4109 if (PL_expect == XOPERATOR)
4114 else if (*s == '>') {
4117 if (isIDFIRST_lazy_if(s,UTF)) {
4118 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4126 if (PL_expect == XOPERATOR)
4129 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4131 OPERATOR('-'); /* unary minus */
4137 const char tmp = *s++;
4140 if (PL_expect == XOPERATOR)
4145 if (PL_expect == XOPERATOR)
4148 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4155 if (PL_expect != XOPERATOR) {
4156 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4157 PL_expect = XOPERATOR;
4158 force_ident(PL_tokenbuf, '*');
4171 if (PL_expect == XOPERATOR) {
4175 PL_tokenbuf[0] = '%';
4176 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4177 sizeof PL_tokenbuf - 1, FALSE);
4178 if (!PL_tokenbuf[1]) {
4181 PL_pending_ident = '%';
4192 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4199 const char tmp = *s++;
4205 goto just_a_word_zero_gv;
4208 switch (PL_expect) {
4214 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4216 PL_bufptr = s; /* update in case we back off */
4222 PL_expect = XTERMBLOCK;
4225 stuffstart = s - SvPVX(PL_linestr) - 1;
4229 while (isIDFIRST_lazy_if(s,UTF)) {
4232 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4233 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4234 if (tmp < 0) tmp = -tmp;
4248 sv = newSVpvn(s, len);
4250 d = scan_str(d,TRUE,TRUE);
4252 /* MUST advance bufptr here to avoid bogus
4253 "at end of line" context messages from yyerror().
4255 PL_bufptr = s + len;
4256 yyerror("Unterminated attribute parameter in attribute list");
4260 return REPORT(0); /* EOF indicator */
4264 sv_catsv(sv, PL_lex_stuff);
4265 attrs = append_elem(OP_LIST, attrs,
4266 newSVOP(OP_CONST, 0, sv));
4267 SvREFCNT_dec(PL_lex_stuff);
4268 PL_lex_stuff = NULL;
4271 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4273 if (PL_in_my == KEY_our) {
4275 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4277 /* skip to avoid loading attributes.pm */
4279 deprecate(":unique");
4282 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4285 /* NOTE: any CV attrs applied here need to be part of
4286 the CVf_BUILTIN_ATTRS define in cv.h! */
4287 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4289 CvLVALUE_on(PL_compcv);
4291 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4293 CvLOCKED_on(PL_compcv);
4295 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4297 CvMETHOD_on(PL_compcv);
4299 /* After we've set the flags, it could be argued that
4300 we don't need to do the attributes.pm-based setting
4301 process, and shouldn't bother appending recognized
4302 flags. To experiment with that, uncomment the
4303 following "else". (Note that's already been
4304 uncommented. That keeps the above-applied built-in
4305 attributes from being intercepted (and possibly
4306 rejected) by a package's attribute routines, but is
4307 justified by the performance win for the common case
4308 of applying only built-in attributes.) */
4310 attrs = append_elem(OP_LIST, attrs,
4311 newSVOP(OP_CONST, 0,
4315 if (*s == ':' && s[1] != ':')
4318 break; /* require real whitespace or :'s */
4319 /* XXX losing whitespace on sequential attributes here */
4323 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4324 if (*s != ';' && *s != '}' && *s != tmp
4325 && (tmp != '=' || *s != ')')) {
4326 const char q = ((*s == '\'') ? '"' : '\'');
4327 /* If here for an expression, and parsed no attrs, back
4329 if (tmp == '=' && !attrs) {
4333 /* MUST advance bufptr here to avoid bogus "at end of line"
4334 context messages from yyerror().
4337 yyerror( (const char *)
4339 ? Perl_form(aTHX_ "Invalid separator character "
4340 "%c%c%c in attribute list", q, *s, q)
4341 : "Unterminated attribute list" ) );
4349 start_force(PL_curforce);
4350 NEXTVAL_NEXTTOKE.opval = attrs;
4351 CURMAD('_', PL_nextwhite);
4356 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4357 (s - SvPVX(PL_linestr)) - stuffstart);
4365 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4366 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4374 const char tmp = *s++;
4379 const char tmp = *s++;
4387 if (PL_lex_brackets <= 0)
4388 yyerror("Unmatched right square bracket");
4391 if (PL_lex_state == LEX_INTERPNORMAL) {
4392 if (PL_lex_brackets == 0) {
4393 if (*s == '-' && s[1] == '>')
4394 PL_lex_state = LEX_INTERPENDMAYBE;
4395 else if (*s != '[' && *s != '{')
4396 PL_lex_state = LEX_INTERPEND;
4403 if (PL_lex_brackets > 100) {
4404 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4406 switch (PL_expect) {
4408 if (PL_lex_formbrack) {
4412 if (PL_oldoldbufptr == PL_last_lop)
4413 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4415 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4416 OPERATOR(HASHBRACK);
4418 while (s < PL_bufend && SPACE_OR_TAB(*s))
4421 PL_tokenbuf[0] = '\0';
4422 if (d < PL_bufend && *d == '-') {
4423 PL_tokenbuf[0] = '-';
4425 while (d < PL_bufend && SPACE_OR_TAB(*d))
4428 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4429 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4431 while (d < PL_bufend && SPACE_OR_TAB(*d))
4434 const char minus = (PL_tokenbuf[0] == '-');
4435 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4443 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4448 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4453 if (PL_oldoldbufptr == PL_last_lop)
4454 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4456 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4459 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4461 /* This hack is to get the ${} in the message. */
4463 yyerror("syntax error");
4466 OPERATOR(HASHBRACK);
4468 /* This hack serves to disambiguate a pair of curlies
4469 * as being a block or an anon hash. Normally, expectation
4470 * determines that, but in cases where we're not in a
4471 * position to expect anything in particular (like inside
4472 * eval"") we have to resolve the ambiguity. This code
4473 * covers the case where the first term in the curlies is a
4474 * quoted string. Most other cases need to be explicitly
4475 * disambiguated by prepending a "+" before the opening
4476 * curly in order to force resolution as an anon hash.
4478 * XXX should probably propagate the outer expectation
4479 * into eval"" to rely less on this hack, but that could
4480 * potentially break current behavior of eval"".
4484 if (*s == '\'' || *s == '"' || *s == '`') {
4485 /* common case: get past first string, handling escapes */
4486 for (t++; t < PL_bufend && *t != *s;)
4487 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4491 else if (*s == 'q') {
4494 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4497 /* skip q//-like construct */
4499 char open, close, term;
4502 while (t < PL_bufend && isSPACE(*t))
4504 /* check for q => */
4505 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4506 OPERATOR(HASHBRACK);
4510 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4514 for (t++; t < PL_bufend; t++) {
4515 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4517 else if (*t == open)
4521 for (t++; t < PL_bufend; t++) {
4522 if (*t == '\\' && t+1 < PL_bufend)
4524 else if (*t == close && --brackets <= 0)
4526 else if (*t == open)
4533 /* skip plain q word */
4534 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4537 else if (isALNUM_lazy_if(t,UTF)) {
4539 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4542 while (t < PL_bufend && isSPACE(*t))
4544 /* if comma follows first term, call it an anon hash */
4545 /* XXX it could be a comma expression with loop modifiers */
4546 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4547 || (*t == '=' && t[1] == '>')))
4548 OPERATOR(HASHBRACK);
4549 if (PL_expect == XREF)
4552 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4558 yylval.ival = CopLINE(PL_curcop);
4559 if (isSPACE(*s) || *s == '#')
4560 PL_copline = NOLINE; /* invalidate current command line number */
4565 if (PL_lex_brackets <= 0)
4566 yyerror("Unmatched right curly bracket");
4568 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4569 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4570 PL_lex_formbrack = 0;
4571 if (PL_lex_state == LEX_INTERPNORMAL) {
4572 if (PL_lex_brackets == 0) {
4573 if (PL_expect & XFAKEBRACK) {
4574 PL_expect &= XENUMMASK;
4575 PL_lex_state = LEX_INTERPEND;
4580 PL_thiswhite = newSVpvs("");
4581 sv_catpvn(PL_thiswhite,"}",1);
4584 return yylex(); /* ignore fake brackets */
4586 if (*s == '-' && s[1] == '>')
4587 PL_lex_state = LEX_INTERPENDMAYBE;
4588 else if (*s != '[' && *s != '{')
4589 PL_lex_state = LEX_INTERPEND;
4592 if (PL_expect & XFAKEBRACK) {
4593 PL_expect &= XENUMMASK;
4595 return yylex(); /* ignore fake brackets */
4597 start_force(PL_curforce);
4599 curmad('X', newSVpvn(s-1,1));
4600 CURMAD('_', PL_thiswhite);
4605 PL_thistoken = newSVpvs("");
4613 if (PL_expect == XOPERATOR) {
4614 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4615 && isIDFIRST_lazy_if(s,UTF))
4617 CopLINE_dec(PL_curcop);
4618 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4619 CopLINE_inc(PL_curcop);
4624 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4626 PL_expect = XOPERATOR;
4627 force_ident(PL_tokenbuf, '&');
4631 yylval.ival = (OPpENTERSUB_AMPER<<8);
4643 const char tmp = *s++;
4650 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4651 && strchr("+-*/%.^&|<",tmp))
4652 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4653 "Reversed %c= operator",(int)tmp);
4655 if (PL_expect == XSTATE && isALPHA(tmp) &&
4656 (s == PL_linestart+1 || s[-2] == '\n') )
4658 if (PL_in_eval && !PL_rsfp) {
4663 if (strnEQ(s,"=cut",4)) {
4679 PL_thiswhite = newSVpvs("");
4680 sv_catpvn(PL_thiswhite, PL_linestart,
4681 PL_bufend - PL_linestart);
4685 PL_doextract = TRUE;
4689 if (PL_lex_brackets < PL_lex_formbrack) {
4691 #ifdef PERL_STRICT_CR
4692 while (SPACE_OR_TAB(*t))
4694 while (SPACE_OR_TAB(*t) || *t == '\r')
4697 if (*t == '\n' || *t == '#') {
4708 const char tmp = *s++;
4710 /* was this !=~ where !~ was meant?
4711 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4713 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4714 const char *t = s+1;
4716 while (t < PL_bufend && isSPACE(*t))
4719 if (*t == '/' || *t == '?' ||
4720 ((*t == 'm' || *t == 's' || *t == 'y')
4721 && !isALNUM(t[1])) ||
4722 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4723 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4724 "!=~ should be !~");
4734 if (PL_expect != XOPERATOR) {
4735 if (s[1] != '<' && !strchr(s,'>'))
4738 s = scan_heredoc(s);
4740 s = scan_inputsymbol(s);
4741 TERM(sublex_start());
4747 SHop(OP_LEFT_SHIFT);
4761 const char tmp = *s++;
4763 SHop(OP_RIGHT_SHIFT);
4764 else if (tmp == '=')
4773 if (PL_expect == XOPERATOR) {
4774 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4776 deprecate_old(commaless_variable_list);
4777 return REPORT(','); /* grandfather non-comma-format format */
4781 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4782 PL_tokenbuf[0] = '@';
4783 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4784 sizeof PL_tokenbuf - 1, FALSE);
4785 if (PL_expect == XOPERATOR)
4786 no_op("Array length", s);
4787 if (!PL_tokenbuf[1])
4789 PL_expect = XOPERATOR;
4790 PL_pending_ident = '#';
4794 PL_tokenbuf[0] = '$';
4795 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4796 sizeof PL_tokenbuf - 1, FALSE);
4797 if (PL_expect == XOPERATOR)
4799 if (!PL_tokenbuf[1]) {
4801 yyerror("Final $ should be \\$ or $name");
4805 /* This kludge not intended to be bulletproof. */
4806 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4807 yylval.opval = newSVOP(OP_CONST, 0,
4808 newSViv(CopARYBASE_get(&PL_compiling)));
4809 yylval.opval->op_private = OPpCONST_ARYBASE;
4815 const char tmp = *s;
4816 if (PL_lex_state == LEX_NORMAL)
4819 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4820 && intuit_more(s)) {
4822 PL_tokenbuf[0] = '@';
4823 if (ckWARN(WARN_SYNTAX)) {
4826 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4829 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4830 while (t < PL_bufend && *t != ']')
4832 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4833 "Multidimensional syntax %.*s not supported",
4834 (int)((t - PL_bufptr) + 1), PL_bufptr);
4838 else if (*s == '{') {
4840 PL_tokenbuf[0] = '%';
4841 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4842 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4844 char tmpbuf[sizeof PL_tokenbuf];
4847 } while (isSPACE(*t));
4848 if (isIDFIRST_lazy_if(t,UTF)) {
4850 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4854 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4855 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4856 "You need to quote \"%s\"",
4863 PL_expect = XOPERATOR;
4864 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4865 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4866 if (!islop || PL_last_lop_op == OP_GREPSTART)
4867 PL_expect = XOPERATOR;
4868 else if (strchr("$@\"'`q", *s))
4869 PL_expect = XTERM; /* e.g. print $fh "foo" */
4870 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4871 PL_expect = XTERM; /* e.g. print $fh &sub */
4872 else if (isIDFIRST_lazy_if(s,UTF)) {
4873 char tmpbuf[sizeof PL_tokenbuf];
4875 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4876 if ((t2 = keyword(tmpbuf, len, 0))) {
4877 /* binary operators exclude handle interpretations */
4889 PL_expect = XTERM; /* e.g. print $fh length() */
4894 PL_expect = XTERM; /* e.g. print $fh subr() */
4897 else if (isDIGIT(*s))
4898 PL_expect = XTERM; /* e.g. print $fh 3 */
4899 else if (*s == '.' && isDIGIT(s[1]))
4900 PL_expect = XTERM; /* e.g. print $fh .3 */
4901 else if ((*s == '?' || *s == '-' || *s == '+')
4902 && !isSPACE(s[1]) && s[1] != '=')
4903 PL_expect = XTERM; /* e.g. print $fh -1 */
4904 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4906 PL_expect = XTERM; /* e.g. print $fh /.../
4907 XXX except DORDOR operator
4909 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4911 PL_expect = XTERM; /* print $fh <<"EOF" */
4914 PL_pending_ident = '$';
4918 if (PL_expect == XOPERATOR)
4920 PL_tokenbuf[0] = '@';
4921 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4922 if (!PL_tokenbuf[1]) {
4925 if (PL_lex_state == LEX_NORMAL)
4927 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4929 PL_tokenbuf[0] = '%';
4931 /* Warn about @ where they meant $. */
4932 if (*s == '[' || *s == '{') {
4933 if (ckWARN(WARN_SYNTAX)) {
4934 const char *t = s + 1;
4935 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4937 if (*t == '}' || *t == ']') {
4939 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4940 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4941 "Scalar value %.*s better written as $%.*s",
4942 (int)(t-PL_bufptr), PL_bufptr,
4943 (int)(t-PL_bufptr-1), PL_bufptr+1);
4948 PL_pending_ident = '@';
4951 case '/': /* may be division, defined-or, or pattern */
4952 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4956 case '?': /* may either be conditional or pattern */
4957 if(PL_expect == XOPERATOR) {
4965 /* A // operator. */
4975 /* Disable warning on "study /blah/" */
4976 if (PL_oldoldbufptr == PL_last_uni
4977 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4978 || memNE(PL_last_uni, "study", 5)
4979 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4982 s = scan_pat(s,OP_MATCH);
4983 TERM(sublex_start());
4987 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4988 #ifdef PERL_STRICT_CR
4991 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4993 && (s == PL_linestart || s[-1] == '\n') )
4995 PL_lex_formbrack = 0;
4999 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5005 yylval.ival = OPf_SPECIAL;
5011 if (PL_expect != XOPERATOR)
5016 case '0': case '1': case '2': case '3': case '4':
5017 case '5': case '6': case '7': case '8': case '9':
5018 s = scan_num(s, &yylval);
5019 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5020 if (PL_expect == XOPERATOR)
5025 s = scan_str(s,!!PL_madskills,FALSE);
5026 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5027 if (PL_expect == XOPERATOR) {
5028 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5030 deprecate_old(commaless_variable_list);
5031 return REPORT(','); /* grandfather non-comma-format format */
5038 yylval.ival = OP_CONST;
5039 TERM(sublex_start());
5042 s = scan_str(s,!!PL_madskills,FALSE);
5043 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5044 if (PL_expect == XOPERATOR) {
5045 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5047 deprecate_old(commaless_variable_list);
5048 return REPORT(','); /* grandfather non-comma-format format */
5055 yylval.ival = OP_CONST;
5056 /* FIXME. I think that this can be const if char *d is replaced by
5057 more localised variables. */
5058 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5059 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5060 yylval.ival = OP_STRINGIFY;
5064 TERM(sublex_start());
5067 s = scan_str(s,!!PL_madskills,FALSE);
5068 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5069 if (PL_expect == XOPERATOR)
5070 no_op("Backticks",s);
5073 readpipe_override();
5074 TERM(sublex_start());
5078 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5079 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5081 if (PL_expect == XOPERATOR)
5082 no_op("Backslash",s);
5086 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5087 char *start = s + 2;
5088 while (isDIGIT(*start) || *start == '_')
5090 if (*start == '.' && isDIGIT(start[1])) {
5091 s = scan_num(s, &yylval);
5094 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5095 else if (!isALPHA(*start) && (PL_expect == XTERM
5096 || PL_expect == XREF || PL_expect == XSTATE
5097 || PL_expect == XTERMORDORDOR)) {
5098 /* XXX Use gv_fetchpvn rather than stomping on a const string */
5099 const char c = *start;
5102 gv = gv_fetchpv(s, 0, SVt_PVCV);
5105 s = scan_num(s, &yylval);
5112 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5154 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5156 /* Some keywords can be followed by any delimiter, including ':' */
5157 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5158 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5159 (PL_tokenbuf[0] == 'q' &&
5160 strchr("qwxr", PL_tokenbuf[1])))));
5162 /* x::* is just a word, unless x is "CORE" */
5163 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5167 while (d < PL_bufend && isSPACE(*d))
5168 d++; /* no comments skipped here, or s### is misparsed */
5170 /* Is this a label? */
5171 if (!tmp && PL_expect == XSTATE
5172 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5174 yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5179 /* Check for keywords */
5180 tmp = keyword(PL_tokenbuf, len, 0);
5182 /* Is this a word before a => operator? */
5183 if (*d == '=' && d[1] == '>') {
5186 = (OP*)newSVOP(OP_CONST, 0,
5187 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5188 yylval.opval->op_private = OPpCONST_BARE;
5192 if (tmp < 0) { /* second-class keyword? */
5193 GV *ogv = NULL; /* override (winner) */
5194 GV *hgv = NULL; /* hidden (loser) */
5195 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5197 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5200 if (GvIMPORTED_CV(gv))
5202 else if (! CvMETHOD(cv))
5206 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5207 (gv = *gvp) && isGV_with_GP(gv) &&
5208 GvCVu(gv) && GvIMPORTED_CV(gv))
5215 tmp = 0; /* overridden by import or by GLOBAL */
5218 && -tmp==KEY_lock /* XXX generalizable kludge */
5221 tmp = 0; /* any sub overrides "weak" keyword */
5223 else { /* no override */
5225 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5226 Perl_warner(aTHX_ packWARN(WARN_MISC),
5227 "dump() better written as CORE::dump()");
5231 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5232 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5233 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5234 "Ambiguous call resolved as CORE::%s(), %s",
5235 GvENAME(hgv), "qualify as such or use &");
5242 default: /* not a keyword */
5243 /* Trade off - by using this evil construction we can pull the
5244 variable gv into the block labelled keylookup. If not, then
5245 we have to give it function scope so that the goto from the
5246 earlier ':' case doesn't bypass the initialisation. */
5248 just_a_word_zero_gv:
5256 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5259 SV *nextPL_nextwhite = 0;
5263 /* Get the rest if it looks like a package qualifier */
5265 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5267 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5270 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5271 *s == '\'' ? "'" : "::");
5276 if (PL_expect == XOPERATOR) {
5277 if (PL_bufptr == PL_linestart) {
5278 CopLINE_dec(PL_curcop);
5279 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5280 CopLINE_inc(PL_curcop);
5283 no_op("Bareword",s);
5286 /* Look for a subroutine with this name in current package,
5287 unless name is "Foo::", in which case Foo is a bearword
5288 (and a package name). */
5290 if (len > 2 && !PL_madskills &&
5291 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5293 if (ckWARN(WARN_BAREWORD)
5294 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5295 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5296 "Bareword \"%s\" refers to nonexistent package",
5299 PL_tokenbuf[len] = '\0';
5305 /* Mustn't actually add anything to a symbol table.
5306 But also don't want to "initialise" any placeholder
5307 constants that might already be there into full
5308 blown PVGVs with attached PVCV. */
5309 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5310 GV_NOADD_NOINIT, SVt_PVCV);
5315 /* if we saw a global override before, get the right name */
5318 sv = newSVpvs("CORE::GLOBAL::");
5319 sv_catpv(sv,PL_tokenbuf);
5322 /* If len is 0, newSVpv does strlen(), which is correct.
5323 If len is non-zero, then it will be the true length,
5324 and so the scalar will be created correctly. */
5325 sv = newSVpv(PL_tokenbuf,len);
5328 if (PL_madskills && !PL_thistoken) {
5329 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5330 PL_thistoken = newSVpv(start,s - start);
5331 PL_realtokenstart = s - SvPVX(PL_linestr);
5335 /* Presume this is going to be a bareword of some sort. */
5338 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5339 yylval.opval->op_private = OPpCONST_BARE;
5340 /* UTF-8 package name? */
5341 if (UTF && !IN_BYTES &&
5342 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5345 /* And if "Foo::", then that's what it certainly is. */
5350 /* Do the explicit type check so that we don't need to force
5351 the initialisation of the symbol table to have a real GV.
5352 Beware - gv may not really be a PVGV, cv may not really be
5353 a PVCV, (because of the space optimisations that gv_init
5354 understands) But they're true if for this symbol there is
5355 respectively a typeglob and a subroutine.
5357 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5358 /* Real typeglob, so get the real subroutine: */
5360 /* A proxy for a subroutine in this package? */
5361 : SvOK(gv) ? (CV *) gv : NULL)
5364 /* See if it's the indirect object for a list operator. */
5366 if (PL_oldoldbufptr &&
5367 PL_oldoldbufptr < PL_bufptr &&
5368 (PL_oldoldbufptr == PL_last_lop
5369 || PL_oldoldbufptr == PL_last_uni) &&
5370 /* NO SKIPSPACE BEFORE HERE! */
5371 (PL_expect == XREF ||
5372 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5374 bool immediate_paren = *s == '(';
5376 /* (Now we can afford to cross potential line boundary.) */
5377 s = SKIPSPACE2(s,nextPL_nextwhite);
5379 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5382 /* Two barewords in a row may indicate method call. */
5384 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5385 (tmp = intuit_method(s, gv, cv)))
5388 /* If not a declared subroutine, it's an indirect object. */
5389 /* (But it's an indir obj regardless for sort.) */
5390 /* Also, if "_" follows a filetest operator, it's a bareword */
5393 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5395 (PL_last_lop_op != OP_MAPSTART &&
5396 PL_last_lop_op != OP_GREPSTART))))
5397 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5398 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5401 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5406 PL_expect = XOPERATOR;
5409 s = SKIPSPACE2(s,nextPL_nextwhite);
5410 PL_nextwhite = nextPL_nextwhite;
5415 /* Is this a word before a => operator? */
5416 if (*s == '=' && s[1] == '>' && !pkgname) {
5418 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5419 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5420 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5424 /* If followed by a paren, it's certainly a subroutine. */
5429 while (SPACE_OR_TAB(*d))
5431 if (*d == ')' && (sv = gv_const_sv(gv))) {
5438 PL_nextwhite = PL_thiswhite;
5441 start_force(PL_curforce);
5443 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5444 PL_expect = XOPERATOR;
5447 PL_nextwhite = nextPL_nextwhite;
5448 curmad('X', PL_thistoken);
5449 PL_thistoken = newSVpvs("");
5457 /* If followed by var or block, call it a method (unless sub) */
5459 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5460 PL_last_lop = PL_oldbufptr;
5461 PL_last_lop_op = OP_METHOD;
5465 /* If followed by a bareword, see if it looks like indir obj. */
5468 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5469 && (tmp = intuit_method(s, gv, cv)))
5472 /* Not a method, so call it a subroutine (if defined) */
5475 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5476 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5477 "Ambiguous use of -%s resolved as -&%s()",
5478 PL_tokenbuf, PL_tokenbuf);
5479 /* Check for a constant sub */
5480 if ((sv = gv_const_sv(gv))) {
5482 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5483 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5484 yylval.opval->op_private = 0;
5488 /* Resolve to GV now. */
5489 if (SvTYPE(gv) != SVt_PVGV) {
5490 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5491 assert (SvTYPE(gv) == SVt_PVGV);
5492 /* cv must have been some sort of placeholder, so
5493 now needs replacing with a real code reference. */
5497 op_free(yylval.opval);
5498 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5499 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5500 PL_last_lop = PL_oldbufptr;
5501 PL_last_lop_op = OP_ENTERSUB;
5502 /* Is there a prototype? */
5510 const char *proto = SvPV_const((SV*)cv, protolen);
5513 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5515 while (*proto == ';')
5517 if (*proto == '&' && *s == '{') {
5518 sv_setpv(PL_subname,
5521 "__ANON__" : "__ANON__::__ANON__"));
5528 PL_nextwhite = PL_thiswhite;
5531 start_force(PL_curforce);
5532 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5535 PL_nextwhite = nextPL_nextwhite;
5536 curmad('X', PL_thistoken);
5537 PL_thistoken = newSVpvs("");
5544 /* Guess harder when madskills require "best effort". */
5545 if (PL_madskills && (!gv || !GvCVu(gv))) {
5546 int probable_sub = 0;
5547 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5549 else if (isALPHA(*s)) {
5553 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5554 if (!keyword(tmpbuf, tmplen, 0))
5557 while (d < PL_bufend && isSPACE(*d))
5559 if (*d == '=' && d[1] == '>')
5564 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5565 op_free(yylval.opval);
5566 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5567 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5568 PL_last_lop = PL_oldbufptr;
5569 PL_last_lop_op = OP_ENTERSUB;
5570 PL_nextwhite = PL_thiswhite;
5572 start_force(PL_curforce);
5573 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5575 PL_nextwhite = nextPL_nextwhite;
5576 curmad('X', PL_thistoken);
5577 PL_thistoken = newSVpvs("");
5582 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5589 /* Call it a bare word */
5591 if (PL_hints & HINT_STRICT_SUBS)
5592 yylval.opval->op_private |= OPpCONST_STRICT;
5595 if (lastchar != '-') {
5596 if (ckWARN(WARN_RESERVED)) {
5600 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5601 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5608 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5609 && ckWARN_d(WARN_AMBIGUOUS)) {
5610 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5611 "Operator or semicolon missing before %c%s",
5612 lastchar, PL_tokenbuf);
5613 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5614 "Ambiguous use of %c resolved as operator %c",
5615 lastchar, lastchar);
5621 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5622 newSVpv(CopFILE(PL_curcop),0));
5626 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5627 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5630 case KEY___PACKAGE__:
5631 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5633 ? newSVhek(HvNAME_HEK(PL_curstash))
5640 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5641 const char *pname = "main";
5642 if (PL_tokenbuf[2] == 'D')
5643 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5644 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5648 GvIOp(gv) = newIO();
5649 IoIFP(GvIOp(gv)) = PL_rsfp;
5650 #if defined(HAS_FCNTL) && defined(F_SETFD)
5652 const int fd = PerlIO_fileno(PL_rsfp);
5653 fcntl(fd,F_SETFD,fd >= 3);
5656 /* Mark this internal pseudo-handle as clean */
5657 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5659 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5660 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5661 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5663 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5664 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5665 /* if the script was opened in binmode, we need to revert
5666 * it to text mode for compatibility; but only iff it has CRs
5667 * XXX this is a questionable hack at best. */
5668 if (PL_bufend-PL_bufptr > 2
5669 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5672 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5673 loc = PerlIO_tell(PL_rsfp);
5674 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5677 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5679 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5680 #endif /* NETWARE */
5681 #ifdef PERLIO_IS_STDIO /* really? */
5682 # if defined(__BORLANDC__)
5683 /* XXX see note in do_binmode() */
5684 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5688 PerlIO_seek(PL_rsfp, loc, 0);
5692 #ifdef PERLIO_LAYERS
5695 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5696 else if (PL_encoding) {
5703 XPUSHs(PL_encoding);
5705 call_method("name", G_SCALAR);
5709 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5710 Perl_form(aTHX_ ":encoding(%"SVf")",
5719 if (PL_realtokenstart >= 0) {
5720 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5722 PL_endwhite = newSVpvs("");
5723 sv_catsv(PL_endwhite, PL_thiswhite);
5725 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5726 PL_realtokenstart = -1;
5728 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5729 SvCUR(PL_endwhite))) != NULL) ;
5744 if (PL_expect == XSTATE) {
5751 if (*s == ':' && s[1] == ':') {
5754 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5755 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5756 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5759 else if (tmp == KEY_require || tmp == KEY_do)
5760 /* that's a way to remember we saw "CORE::" */
5773 LOP(OP_ACCEPT,XTERM);
5779 LOP(OP_ATAN2,XTERM);
5785 LOP(OP_BINMODE,XTERM);
5788 LOP(OP_BLESS,XTERM);
5797 /* When 'use switch' is in effect, continue has a dual
5798 life as a control operator. */
5800 if (!FEATURE_IS_ENABLED("switch"))
5803 /* We have to disambiguate the two senses of
5804 "continue". If the next token is a '{' then
5805 treat it as the start of a continue block;
5806 otherwise treat it as a control operator.
5818 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5835 if (!PL_cryptseen) {
5836 PL_cryptseen = TRUE;
5840 LOP(OP_CRYPT,XTERM);
5843 LOP(OP_CHMOD,XTERM);
5846 LOP(OP_CHOWN,XTERM);
5849 LOP(OP_CONNECT,XTERM);
5868 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5869 if (orig_keyword == KEY_do) {
5878 PL_hints |= HINT_BLOCK_SCOPE;
5888 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5889 LOP(OP_DBMOPEN,XTERM);
5895 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5902 yylval.ival = CopLINE(PL_curcop);
5918 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5919 UNIBRACK(OP_ENTEREVAL);
5933 case KEY_endhostent:
5939 case KEY_endservent:
5942 case KEY_endprotoent:
5953 yylval.ival = CopLINE(PL_curcop);
5955 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5958 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5961 if ((PL_bufend - p) >= 3 &&
5962 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5964 else if ((PL_bufend - p) >= 4 &&
5965 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5968 if (isIDFIRST_lazy_if(p,UTF)) {
5969 p = scan_ident(p, PL_bufend,
5970 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5974 Perl_croak(aTHX_ "Missing $ on loop variable");
5976 s = SvPVX(PL_linestr) + soff;
5982 LOP(OP_FORMLINE,XTERM);
5988 LOP(OP_FCNTL,XTERM);
5994 LOP(OP_FLOCK,XTERM);
6003 LOP(OP_GREPSTART, XREF);
6006 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6021 case KEY_getpriority:
6022 LOP(OP_GETPRIORITY,XTERM);
6024 case KEY_getprotobyname:
6027 case KEY_getprotobynumber:
6028 LOP(OP_GPBYNUMBER,XTERM);
6030 case KEY_getprotoent:
6042 case KEY_getpeername:
6043 UNI(OP_GETPEERNAME);
6045 case KEY_gethostbyname:
6048 case KEY_gethostbyaddr:
6049 LOP(OP_GHBYADDR,XTERM);
6051 case KEY_gethostent:
6054 case KEY_getnetbyname:
6057 case KEY_getnetbyaddr:
6058 LOP(OP_GNBYADDR,XTERM);
6063 case KEY_getservbyname:
6064 LOP(OP_GSBYNAME,XTERM);
6066 case KEY_getservbyport:
6067 LOP(OP_GSBYPORT,XTERM);
6069 case KEY_getservent:
6072 case KEY_getsockname:
6073 UNI(OP_GETSOCKNAME);
6075 case KEY_getsockopt:
6076 LOP(OP_GSOCKOPT,XTERM);
6091 yylval.ival = CopLINE(PL_curcop);
6101 yylval.ival = CopLINE(PL_curcop);
6105 LOP(OP_INDEX,XTERM);
6111 LOP(OP_IOCTL,XTERM);
6123 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6155 LOP(OP_LISTEN,XTERM);
6164 s = scan_pat(s,OP_MATCH);
6165 TERM(sublex_start());
6168 LOP(OP_MAPSTART, XREF);
6171 LOP(OP_MKDIR,XTERM);
6174 LOP(OP_MSGCTL,XTERM);
6177 LOP(OP_MSGGET,XTERM);
6180 LOP(OP_MSGRCV,XTERM);
6183 LOP(OP_MSGSND,XTERM);
6188 PL_in_my = (U16)tmp;
6190 if (isIDFIRST_lazy_if(s,UTF)) {
6194 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6195 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6197 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6198 if (!PL_in_my_stash) {
6201 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6205 if (PL_madskills) { /* just add type to declarator token */
6206 sv_catsv(PL_thistoken, PL_nextwhite);
6208 sv_catpvn(PL_thistoken, start, s - start);
6216 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6223 s = tokenize_use(0, s);
6227 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6234 if (isIDFIRST_lazy_if(s,UTF)) {
6236 for (d = s; isALNUM_lazy_if(d,UTF);)
6238 for (t=d; isSPACE(*t);)
6240 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6242 && !(t[0] == '=' && t[1] == '>')
6244 int parms_len = (int)(d-s);
6245 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6246 "Precedence problem: open %.*s should be open(%.*s)",
6247 parms_len, s, parms_len, s);
6253 yylval.ival = OP_OR;
6263 LOP(OP_OPEN_DIR,XTERM);
6266 checkcomma(s,PL_tokenbuf,"filehandle");
6270 checkcomma(s,PL_tokenbuf,"filehandle");
6289 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6293 LOP(OP_PIPE_OP,XTERM);
6296 s = scan_str(s,!!PL_madskills,FALSE);
6299 yylval.ival = OP_CONST;
6300 TERM(sublex_start());
6306 s = scan_str(s,!!PL_madskills,FALSE);
6309 PL_expect = XOPERATOR;
6311 if (SvCUR(PL_lex_stuff)) {
6314 d = SvPV_force(PL_lex_stuff, len);
6316 for (; isSPACE(*d) && len; --len, ++d)
6321 if (!warned && ckWARN(WARN_QW)) {
6322 for (; !isSPACE(*d) && len; --len, ++d) {
6324 Perl_warner(aTHX_ packWARN(WARN_QW),
6325 "Possible attempt to separate words with commas");
6328 else if (*d == '#') {
6329 Perl_warner(aTHX_ packWARN(WARN_QW),
6330 "Possible attempt to put comments in qw() list");
6336 for (; !isSPACE(*d) && len; --len, ++d)
6339 sv = newSVpvn(b, d-b);
6340 if (DO_UTF8(PL_lex_stuff))
6342 words = append_elem(OP_LIST, words,
6343 newSVOP(OP_CONST, 0, tokeq(sv)));
6347 start_force(PL_curforce);
6348 NEXTVAL_NEXTTOKE.opval = words;
6353 SvREFCNT_dec(PL_lex_stuff);
6354 PL_lex_stuff = NULL;
6360 s = scan_str(s,!!PL_madskills,FALSE);
6363 yylval.ival = OP_STRINGIFY;
6364 if (SvIVX(PL_lex_stuff) == '\'')
6365 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6366 TERM(sublex_start());
6369 s = scan_pat(s,OP_QR);
6370 TERM(sublex_start());
6373 s = scan_str(s,!!PL_madskills,FALSE);
6376 readpipe_override();
6377 TERM(sublex_start());
6385 s = force_version(s, FALSE);
6387 else if (*s != 'v' || !isDIGIT(s[1])
6388 || (s = force_version(s, TRUE), *s == 'v'))
6390 *PL_tokenbuf = '\0';
6391 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6392 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6393 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6395 yyerror("<> should be quotes");
6397 if (orig_keyword == KEY_require) {
6405 PL_last_uni = PL_oldbufptr;
6406 PL_last_lop_op = OP_REQUIRE;
6408 return REPORT( (int)REQUIRE );
6414 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6418 LOP(OP_RENAME,XTERM);
6427 LOP(OP_RINDEX,XTERM);
6436 UNIDOR(OP_READLINE);
6439 UNIDOR(OP_BACKTICK);
6448 LOP(OP_REVERSE,XTERM);
6451 UNIDOR(OP_READLINK);
6459 TERM(sublex_start());
6461 TOKEN(1); /* force error */
6464 checkcomma(s,PL_tokenbuf,"filehandle");
6474 LOP(OP_SELECT,XTERM);
6480 LOP(OP_SEMCTL,XTERM);
6483 LOP(OP_SEMGET,XTERM);
6486 LOP(OP_SEMOP,XTERM);
6492 LOP(OP_SETPGRP,XTERM);
6494 case KEY_setpriority:
6495 LOP(OP_SETPRIORITY,XTERM);
6497 case KEY_sethostent:
6503 case KEY_setservent:
6506 case KEY_setprotoent:
6516 LOP(OP_SEEKDIR,XTERM);
6518 case KEY_setsockopt:
6519 LOP(OP_SSOCKOPT,XTERM);
6525 LOP(OP_SHMCTL,XTERM);
6528 LOP(OP_SHMGET,XTERM);
6531 LOP(OP_SHMREAD,XTERM);
6534 LOP(OP_SHMWRITE,XTERM);
6537 LOP(OP_SHUTDOWN,XTERM);
6546 LOP(OP_SOCKET,XTERM);
6548 case KEY_socketpair:
6549 LOP(OP_SOCKPAIR,XTERM);
6552 checkcomma(s,PL_tokenbuf,"subroutine name");
6554 if (*s == ';' || *s == ')') /* probably a close */
6555 Perl_croak(aTHX_ "sort is now a reserved word");
6557 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6561 LOP(OP_SPLIT,XTERM);
6564 LOP(OP_SPRINTF,XTERM);
6567 LOP(OP_SPLICE,XTERM);
6582 LOP(OP_SUBSTR,XTERM);
6588 char tmpbuf[sizeof PL_tokenbuf];
6589 SSize_t tboffset = 0;
6590 expectation attrful;
6591 bool have_name, have_proto;
6592 const int key = tmp;
6597 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6598 SV *subtoken = newSVpvn(tstart, s - tstart);
6602 s = SKIPSPACE2(s,tmpwhite);
6607 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6608 (*s == ':' && s[1] == ':'))
6615 attrful = XATTRBLOCK;
6616 /* remember buffer pos'n for later force_word */
6617 tboffset = s - PL_oldbufptr;
6618 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6621 nametoke = newSVpvn(s, d - s);
6623 if (memchr(tmpbuf, ':', len))
6624 sv_setpvn(PL_subname, tmpbuf, len);
6626 sv_setsv(PL_subname,PL_curstname);
6627 sv_catpvs(PL_subname,"::");
6628 sv_catpvn(PL_subname,tmpbuf,len);
6635 CURMAD('X', nametoke);
6636 CURMAD('_', tmpwhite);
6637 (void) force_word(PL_oldbufptr + tboffset, WORD,
6640 s = SKIPSPACE2(d,tmpwhite);
6647 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6648 PL_expect = XTERMBLOCK;
6649 attrful = XATTRTERM;
6650 sv_setpvn(PL_subname,"?",1);
6654 if (key == KEY_format) {
6656 PL_lex_formbrack = PL_lex_brackets + 1;
6658 PL_thistoken = subtoken;
6662 (void) force_word(PL_oldbufptr + tboffset, WORD,
6668 /* Look for a prototype */
6671 bool bad_proto = FALSE;
6672 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6674 s = scan_str(s,!!PL_madskills,FALSE);
6676 Perl_croak(aTHX_ "Prototype not terminated");
6677 /* strip spaces and check for bad characters */
6678 d = SvPVX(PL_lex_stuff);
6680 for (p = d; *p; ++p) {
6683 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6689 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6690 "Illegal character in prototype for %"SVf" : %s",
6691 SVfARG(PL_subname), d);
6692 SvCUR_set(PL_lex_stuff, tmp);
6697 CURMAD('q', PL_thisopen);
6698 CURMAD('_', tmpwhite);
6699 CURMAD('=', PL_thisstuff);
6700 CURMAD('Q', PL_thisclose);
6701 NEXTVAL_NEXTTOKE.opval =
6702 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6703 PL_lex_stuff = NULL;
6706 s = SKIPSPACE2(s,tmpwhite);
6714 if (*s == ':' && s[1] != ':')
6715 PL_expect = attrful;
6716 else if (*s != '{' && key == KEY_sub) {
6718 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6720 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6727 curmad('^', newSVpvs(""));
6728 CURMAD('_', tmpwhite);
6732 PL_thistoken = subtoken;
6735 NEXTVAL_NEXTTOKE.opval =
6736 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6737 PL_lex_stuff = NULL;
6742 sv_setpv(PL_subname,
6744 (PL_curstash ? "__ANON__" : "__ANON__::__ANON__"));
6748 (void) force_word(PL_oldbufptr + tboffset, WORD,
6757 LOP(OP_SYSTEM,XREF);
6760 LOP(OP_SYMLINK,XTERM);
6763 LOP(OP_SYSCALL,XTERM);
6766 LOP(OP_SYSOPEN,XTERM);
6769 LOP(OP_SYSSEEK,XTERM);
6772 LOP(OP_SYSREAD,XTERM);
6775 LOP(OP_SYSWRITE,XTERM);
6779 TERM(sublex_start());
6800 LOP(OP_TRUNCATE,XTERM);
6812 yylval.ival = CopLINE(PL_curcop);
6816 yylval.ival = CopLINE(PL_curcop);
6820 LOP(OP_UNLINK,XTERM);
6826 LOP(OP_UNPACK,XTERM);
6829 LOP(OP_UTIME,XTERM);
6835 LOP(OP_UNSHIFT,XTERM);
6838 s = tokenize_use(1, s);
6848 yylval.ival = CopLINE(PL_curcop);
6852 yylval.ival = CopLINE(PL_curcop);
6856 PL_hints |= HINT_BLOCK_SCOPE;
6863 LOP(OP_WAITPID,XTERM);
6872 ctl_l[0] = toCTRL('L');
6874 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6877 /* Make sure $^L is defined */
6878 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6883 if (PL_expect == XOPERATOR)
6889 yylval.ival = OP_XOR;
6894 TERM(sublex_start());
6899 #pragma segment Main
6903 S_pending_ident(pTHX)
6908 /* pit holds the identifier we read and pending_ident is reset */
6909 char pit = PL_pending_ident;
6910 PL_pending_ident = 0;
6912 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6913 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6914 "### Pending identifier '%s'\n", PL_tokenbuf); });
6916 /* if we're in a my(), we can't allow dynamics here.
6917 $foo'bar has already been turned into $foo::bar, so
6918 just check for colons.
6920 if it's a legal name, the OP is a PADANY.
6923 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6924 if (strchr(PL_tokenbuf,':'))
6925 yyerror(Perl_form(aTHX_ "No package name allowed for "
6926 "variable %s in \"our\"",
6928 tmp = allocmy(PL_tokenbuf);
6931 if (strchr(PL_tokenbuf,':'))
6932 yyerror(Perl_form(aTHX_ PL_no_myglob,
6933 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6935 yylval.opval = newOP(OP_PADANY, 0);
6936 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6942 build the ops for accesses to a my() variable.
6944 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6945 then used in a comparison. This catches most, but not
6946 all cases. For instance, it catches
6947 sort { my($a); $a <=> $b }
6949 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6950 (although why you'd do that is anyone's guess).
6953 if (!strchr(PL_tokenbuf,':')) {
6955 tmp = pad_findmy(PL_tokenbuf);
6956 if (tmp != NOT_IN_PAD) {
6957 /* might be an "our" variable" */
6958 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6959 /* build ops for a bareword */
6960 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6961 HEK * const stashname = HvNAME_HEK(stash);
6962 SV * const sym = newSVhek(stashname);
6963 sv_catpvs(sym, "::");
6964 sv_catpv(sym, PL_tokenbuf+1);
6965 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6966 yylval.opval->op_private = OPpCONST_ENTERED;
6969 ? (GV_ADDMULTI | GV_ADDINEVAL)
6972 ((PL_tokenbuf[0] == '$') ? SVt_PV
6973 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6978 /* if it's a sort block and they're naming $a or $b */
6979 if (PL_last_lop_op == OP_SORT &&
6980 PL_tokenbuf[0] == '$' &&
6981 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6984 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6985 d < PL_bufend && *d != '\n';
6988 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6989 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6995 yylval.opval = newOP(OP_PADANY, 0);
6996 yylval.opval->op_targ = tmp;
7002 Whine if they've said @foo in a doublequoted string,
7003 and @foo isn't a variable we can find in the symbol
7006 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7007 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
7008 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7009 && ckWARN(WARN_AMBIGUOUS)
7010 /* DO NOT warn for @- and @+ */
7011 && !( PL_tokenbuf[2] == '\0' &&
7012 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7015 /* Downgraded from fatal to warning 20000522 mjd */
7016 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7017 "Possible unintended interpolation of %s in string",
7022 /* build ops for a bareword */
7023 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
7024 yylval.opval->op_private = OPpCONST_ENTERED;
7027 /* If the identifier refers to a stash, don't autovivify it.
7028 * Change 24660 had the side effect of causing symbol table
7029 * hashes to always be defined, even if they were freshly
7030 * created and the only reference in the entire program was
7031 * the single statement with the defined %foo::bar:: test.
7032 * It appears that all code in the wild doing this actually
7033 * wants to know whether sub-packages have been loaded, so
7034 * by avoiding auto-vivifying symbol tables, we ensure that
7035 * defined %foo::bar:: continues to be false, and the existing
7036 * tests still give the expected answers, even though what
7037 * they're actually testing has now changed subtly.
7039 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
7041 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7042 ((PL_tokenbuf[0] == '$') ? SVt_PV
7043 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7049 * The following code was generated by perl_keyword.pl.
7053 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7058 case 1: /* 5 tokens of length 1 */
7090 case 2: /* 18 tokens of length 2 */
7236 case 3: /* 29 tokens of length 3 */
7240 if (name[1] == 'N' &&
7303 if (name[1] == 'i' &&
7335 if (name[1] == 'o' &&
7344 if (name[1] == 'e' &&
7353 if (name[1] == 'n' &&
7362 if (name[1] == 'o' &&
7371 if (name[1] == 'a' &&
7380 if (name[1] == 'o' &&
7442 if (name[1] == 'e' &&
7456 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7482 if (name[1] == 'i' &&
7491 if (name[1] == 's' &&
7500 if (name[1] == 'e' &&
7509 if (name[1] == 'o' &&
7521 case 4: /* 41 tokens of length 4 */
7525 if (name[1] == 'O' &&
7535 if (name[1] == 'N' &&
7545 if (name[1] == 'i' &&
7555 if (name[1] == 'h' &&
7565 if (name[1] == 'u' &&
7578 if (name[2] == 'c' &&
7587 if (name[2] == 's' &&
7596 if (name[2] == 'a' &&
7632 if (name[1] == 'o' &&
7645 if (name[2] == 't' &&
7654 if (name[2] == 'o' &&
7663 if (name[2] == 't' &&
7672 if (name[2] == 'e' &&
7685 if (name[1] == 'o' &&
7698 if (name[2] == 'y' &&
7707 if (name[2] == 'l' &&
7723 if (name[2] == 's' &&
7732 if (name[2] == 'n' &&
7741 if (name[2] == 'c' &&
7754 if (name[1] == 'e' &&
7764 if (name[1] == 'p' &&
7777 if (name[2] == 'c' &&
7786 if (name[2] == 'p' &&
7795 if (name[2] == 's' &&
7811 if (name[2] == 'n' &&
7881 if (name[2] == 'r' &&
7890 if (name[2] == 'r' &&
7899 if (name[2] == 'a' &&
7915 if (name[2] == 'l' &&
7977 if (name[2] == 'e' &&
7980 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7993 case 5: /* 39 tokens of length 5 */
7997 if (name[1] == 'E' &&
8008 if (name[1] == 'H' &&
8022 if (name[2] == 'a' &&
8032 if (name[2] == 'a' &&
8049 if (name[2] == 'e' &&
8059 if (name[2] == 'e' &&
8063 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8079 if (name[3] == 'i' &&
8088 if (name[3] == 'o' &&
8124 if (name[2] == 'o' &&
8134 if (name[2] == 'y' &&
8148 if (name[1] == 'l' &&
8162 if (name[2] == 'n' &&
8172 if (name[2] == 'o' &&
8186 if (name[1] == 'i' &&
8191 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8200 if (name[2] == 'd' &&
8210 if (name[2] == 'c' &&
8227 if (name[2] == 'c' &&
8237 if (name[2] == 't' &&
8251 if (name[1] == 'k' &&
8262 if (name[1] == 'r' &&
8276 if (name[2] == 's' &&
8286 if (name[2] == 'd' &&
8303 if (name[2] == 'm' &&
8313 if (name[2] == 'i' &&
8323 if (name[2] == 'e' &&
8333 if (name[2] == 'l' &&
8343 if (name[2] == 'a' &&
8356 if (name[3] == 't' &&
8359 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8365 if (name[3] == 'd' &&
8382 if (name[1] == 'i' &&
8396 if (name[2] == 'a' &&
8409 if (name[3] == 'e' &&
8444 if (name[2] == 'i' &&
8461 if (name[2] == 'i' &&
8471 if (name[2] == 'i' &&
8488 case 6: /* 33 tokens of length 6 */
8492 if (name[1] == 'c' &&
8507 if (name[2] == 'l' &&
8518 if (name[2] == 'r' &&
8533 if (name[1] == 'e' &&
8548 if (name[2] == 's' &&
8553 if(ckWARN_d(WARN_SYNTAX))
8554 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8560 if (name[2] == 'i' &&
8578 if (name[2] == 'l' &&
8589 if (name[2] == 'r' &&
8604 if (name[1] == 'm' &&
8619 if (name[2] == 'n' &&
8630 if (name[2] == 's' &&
8645 if (name[1] == 's' &&
8651 if (name[4] == 't' &&
8660 if (name[4] == 'e' &&
8669 if (name[4] == 'c' &&
8678 if (name[4] == 'n' &&
8694 if (name[1] == 'r' &&
8712 if (name[3] == 'a' &&
8722 if (name[3] == 'u' &&
8736 if (name[2] == 'n' &&
8754 if (name[2] == 'a' &&
8768 if (name[3] == 'e' &&
8781 if (name[4] == 't' &&
8790 if (name[4] == 'e' &&
8812 if (name[4] == 't' &&
8821 if (name[4] == 'e' &&
8837 if (name[2] == 'c' &&
8848 if (name[2] == 'l' &&
8859 if (name[2] == 'b' &&
8870 if (name[2] == 's' &&
8893 if (name[4] == 's' &&
8902 if (name[4] == 'n' &&
8915 if (name[3] == 'a' &&
8932 if (name[1] == 'a' &&
8947 case 7: /* 29 tokens of length 7 */
8951 if (name[1] == 'E' &&
8964 if (name[1] == '_' &&
8977 if (name[1] == 'i' &&
8984 return -KEY_binmode;
8990 if (name[1] == 'o' &&
8997 return -KEY_connect;
9006 if (name[2] == 'm' &&
9012 return -KEY_dbmopen;
9023 if (name[4] == 'u' &&
9027 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9033 if (name[4] == 'n' &&
9054 if (name[1] == 'o' &&
9067 if (name[1] == 'e' &&
9074 if (name[5] == 'r' &&
9077 return -KEY_getpgrp;
9083 if (name[5] == 'i' &&
9086 return -KEY_getppid;
9099 if (name[1] == 'c' &&
9106 return -KEY_lcfirst;
9112 if (name[1] == 'p' &&
9119 return -KEY_opendir;
9125 if (name[1] == 'a' &&
9143 if (name[3] == 'd' &&
9148 return -KEY_readdir;
9154 if (name[3] == 'u' &&
9165 if (name[3] == 'e' &&
9170 return -KEY_reverse;
9189 if (name[3] == 'k' &&
9194 return -KEY_seekdir;
9200 if (name[3] == 'p' &&
9205 return -KEY_setpgrp;
9215 if (name[2] == 'm' &&
9221 return -KEY_shmread;
9227 if (name[2] == 'r' &&
9233 return -KEY_sprintf;
9242 if (name[3] == 'l' &&
9247 return -KEY_symlink;
9256 if (name[4] == 'a' &&
9260 return -KEY_syscall;
9266 if (name[4] == 'p' &&
9270 return -KEY_sysopen;
9276 if (name[4] == 'e' &&
9280 return -KEY_sysread;
9286 if (name[4] == 'e' &&
9290 return -KEY_sysseek;
9308 if (name[1] == 'e' &&
9315 return -KEY_telldir;
9324 if (name[2] == 'f' &&
9330 return -KEY_ucfirst;
9336 if (name[2] == 's' &&
9342 return -KEY_unshift;
9352 if (name[1] == 'a' &&
9359 return -KEY_waitpid;
9368 case 8: /* 26 tokens of length 8 */
9372 if (name[1] == 'U' &&
9380 return KEY_AUTOLOAD;
9391 if (name[3] == 'A' &&
9397 return KEY___DATA__;
9403 if (name[3] == 'I' &&
9409 return -KEY___FILE__;
9415 if (name[3] == 'I' &&
9421 return -KEY___LINE__;
9437 if (name[2] == 'o' &&
9444 return -KEY_closedir;
9450 if (name[2] == 'n' &&
9457 return -KEY_continue;
9467 if (name[1] == 'b' &&
9475 return -KEY_dbmclose;
9481 if (name[1] == 'n' &&
9487 if (name[4] == 'r' &&
9492 return -KEY_endgrent;
9498 if (name[4] == 'w' &&
9503 return -KEY_endpwent;
9516 if (name[1] == 'o' &&
9524 return -KEY_formline;
9530 if (name[1] == 'e' &&
9541 if (name[6] == 'n' &&
9544 return -KEY_getgrent;
9550 if (name[6] == 'i' &&
9553 return -KEY_getgrgid;
9559 if (name[6] == 'a' &&
9562 return -KEY_getgrnam;
9575 if (name[4] == 'o' &&
9580 return -KEY_getlogin;
9591 if (name[6] == 'n' &&
9594 return -KEY_getpwent;
9600 if (name[6] == 'a' &&
9603 return -KEY_getpwnam;
9609 if (name[6] == 'i' &&
9612 return -KEY_getpwuid;
9632 if (name[1] == 'e' &&
9639 if (name[5] == 'i' &&
9646 return -KEY_readline;
9651 return -KEY_readlink;
9662 if (name[5] == 'i' &&
9666 return -KEY_readpipe;
9687 if (name[4] == 'r' &&
9692 return -KEY_setgrent;
9698 if (name[4] == 'w' &&
9703 return -KEY_setpwent;
9719 if (name[3] == 'w' &&
9725 return -KEY_shmwrite;
9731 if (name[3] == 't' &&
9737 return -KEY_shutdown;
9747 if (name[2] == 's' &&
9754 return -KEY_syswrite;
9764 if (name[1] == 'r' &&
9772 return -KEY_truncate;
9781 case 9: /* 9 tokens of length 9 */
9785 if (name[1] == 'N' &&
9794 return KEY_UNITCHECK;
9800 if (name[1] == 'n' &&
9809 return -KEY_endnetent;
9815 if (name[1] == 'e' &&
9824 return -KEY_getnetent;
9830 if (name[1] == 'o' &&
9839 return -KEY_localtime;
9845 if (name[1] == 'r' &&
9854 return KEY_prototype;
9860 if (name[1] == 'u' &&
9869 return -KEY_quotemeta;
9875 if (name[1] == 'e' &&
9884 return -KEY_rewinddir;
9890 if (name[1] == 'e' &&
9899 return -KEY_setnetent;
9905 if (name[1] == 'a' &&
9914 return -KEY_wantarray;
9923 case 10: /* 9 tokens of length 10 */
9927 if (name[1] == 'n' &&
9933 if (name[4] == 'o' &&
9940 return -KEY_endhostent;
9946 if (name[4] == 'e' &&
9953 return -KEY_endservent;
9966 if (name[1] == 'e' &&
9972 if (name[4] == 'o' &&
9979 return -KEY_gethostent;
9988 if (name[5] == 'r' &&
9994 return -KEY_getservent;
10000 if (name[5] == 'c' &&
10006 return -KEY_getsockopt;
10026 if (name[2] == 't')
10031 if (name[4] == 'o' &&
10038 return -KEY_sethostent;
10047 if (name[5] == 'r' &&
10053 return -KEY_setservent;
10059 if (name[5] == 'c' &&
10065 return -KEY_setsockopt;
10082 if (name[2] == 'c' &&
10091 return -KEY_socketpair;
10104 case 11: /* 8 tokens of length 11 */
10108 if (name[1] == '_' &&
10118 { /* __PACKAGE__ */
10119 return -KEY___PACKAGE__;
10125 if (name[1] == 'n' &&
10135 { /* endprotoent */
10136 return -KEY_endprotoent;
10142 if (name[1] == 'e' &&
10151 if (name[5] == 'e' &&
10157 { /* getpeername */
10158 return -KEY_getpeername;
10167 if (name[6] == 'o' &&
10172 { /* getpriority */
10173 return -KEY_getpriority;
10179 if (name[6] == 't' &&
10184 { /* getprotoent */
10185 return -KEY_getprotoent;
10199 if (name[4] == 'o' &&
10206 { /* getsockname */
10207 return -KEY_getsockname;
10220 if (name[1] == 'e' &&
10228 if (name[6] == 'o' &&
10233 { /* setpriority */
10234 return -KEY_setpriority;
10240 if (name[6] == 't' &&
10245 { /* setprotoent */
10246 return -KEY_setprotoent;
10262 case 12: /* 2 tokens of length 12 */
10263 if (name[0] == 'g' &&
10275 if (name[9] == 'd' &&
10278 { /* getnetbyaddr */
10279 return -KEY_getnetbyaddr;
10285 if (name[9] == 'a' &&
10288 { /* getnetbyname */
10289 return -KEY_getnetbyname;
10301 case 13: /* 4 tokens of length 13 */
10302 if (name[0] == 'g' &&
10309 if (name[4] == 'o' &&
10318 if (name[10] == 'd' &&
10321 { /* gethostbyaddr */
10322 return -KEY_gethostbyaddr;
10328 if (name[10] == 'a' &&
10331 { /* gethostbyname */
10332 return -KEY_gethostbyname;
10345 if (name[4] == 'e' &&
10354 if (name[10] == 'a' &&
10357 { /* getservbyname */
10358 return -KEY_getservbyname;
10364 if (name[10] == 'o' &&
10367 { /* getservbyport */
10368 return -KEY_getservbyport;
10387 case 14: /* 1 tokens of length 14 */
10388 if (name[0] == 'g' &&
10402 { /* getprotobyname */
10403 return -KEY_getprotobyname;
10408 case 16: /* 1 tokens of length 16 */
10409 if (name[0] == 'g' &&
10425 { /* getprotobynumber */
10426 return -KEY_getprotobynumber;
10440 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10444 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10445 if (ckWARN(WARN_SYNTAX)) {
10448 for (w = s+2; *w && level; w++) {
10451 else if (*w == ')')
10454 while (isSPACE(*w))
10456 /* the list of chars below is for end of statements or
10457 * block / parens, boolean operators (&&, ||, //) and branch
10458 * constructs (or, and, if, until, unless, while, err, for).
10459 * Not a very solid hack... */
10460 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10461 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10462 "%s (...) interpreted as function",name);
10465 while (s < PL_bufend && isSPACE(*s))
10469 while (s < PL_bufend && isSPACE(*s))
10471 if (isIDFIRST_lazy_if(s,UTF)) {
10472 const char * const w = s++;
10473 while (isALNUM_lazy_if(s,UTF))
10475 while (s < PL_bufend && isSPACE(*s))
10479 if (keyword(w, s - w, 0))
10482 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10483 if (gv && GvCVu(gv))
10485 Perl_croak(aTHX_ "No comma allowed after %s", what);
10490 /* Either returns sv, or mortalizes sv and returns a new SV*.
10491 Best used as sv=new_constant(..., sv, ...).
10492 If s, pv are NULL, calls subroutine with one argument,
10493 and type is used with error messages only. */
10496 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10500 HV * const table = GvHV(PL_hintgv); /* ^H */
10504 const char *why1 = "", *why2 = "", *why3 = "";
10506 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10509 why2 = (const char *)
10510 (strEQ(key,"charnames")
10511 ? "(possibly a missing \"use charnames ...\")"
10513 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10514 (type ? type: "undef"), why2);
10516 /* This is convoluted and evil ("goto considered harmful")
10517 * but I do not understand the intricacies of all the different
10518 * failure modes of %^H in here. The goal here is to make
10519 * the most probable error message user-friendly. --jhi */
10524 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10525 (type ? type: "undef"), why1, why2, why3);
10527 yyerror(SvPVX_const(msg));
10531 cvp = hv_fetch(table, key, strlen(key), FALSE);
10532 if (!cvp || !SvOK(*cvp)) {
10535 why3 = "} is not defined";
10538 sv_2mortal(sv); /* Parent created it permanently */
10541 pv = sv_2mortal(newSVpvn(s, len));
10543 typesv = sv_2mortal(newSVpv(type, 0));
10545 typesv = &PL_sv_undef;
10547 PUSHSTACKi(PERLSI_OVERLOAD);
10559 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10563 /* Check the eval first */
10564 if (!PL_in_eval && SvTRUE(ERRSV)) {
10565 sv_catpvs(ERRSV, "Propagated");
10566 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10568 res = SvREFCNT_inc_simple(sv);
10572 SvREFCNT_inc_simple_void(res);
10581 why1 = "Call to &{$^H{";
10583 why3 = "}} did not return a defined value";
10591 /* Returns a NUL terminated string, with the length of the string written to
10595 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10598 register char *d = dest;
10599 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10602 Perl_croak(aTHX_ ident_too_long);
10603 if (isALNUM(*s)) /* UTF handled below */
10605 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10610 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10614 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10615 char *t = s + UTF8SKIP(s);
10617 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10621 Perl_croak(aTHX_ ident_too_long);
10622 Copy(s, d, len, char);
10635 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10638 char *bracket = NULL;
10640 register char *d = dest;
10641 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10646 while (isDIGIT(*s)) {
10648 Perl_croak(aTHX_ ident_too_long);
10655 Perl_croak(aTHX_ ident_too_long);
10656 if (isALNUM(*s)) /* UTF handled below */
10658 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10663 else if (*s == ':' && s[1] == ':') {
10667 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10668 char *t = s + UTF8SKIP(s);
10669 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10671 if (d + (t - s) > e)
10672 Perl_croak(aTHX_ ident_too_long);
10673 Copy(s, d, t - s, char);
10684 if (PL_lex_state != LEX_NORMAL)
10685 PL_lex_state = LEX_INTERPENDMAYBE;
10688 if (*s == '$' && s[1] &&
10689 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10702 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10707 if (isSPACE(s[-1])) {
10709 const char ch = *s++;
10710 if (!SPACE_OR_TAB(ch)) {
10716 if (isIDFIRST_lazy_if(d,UTF)) {
10720 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10721 end += UTF8SKIP(end);
10722 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10723 end += UTF8SKIP(end);
10725 Copy(s, d, end - s, char);
10730 while ((isALNUM(*s) || *s == ':') && d < e)
10733 Perl_croak(aTHX_ ident_too_long);
10736 while (s < send && SPACE_OR_TAB(*s))
10738 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10739 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10740 const char * const brack =
10742 ((*s == '[') ? "[...]" : "{...}");
10743 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10744 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10745 funny, dest, brack, funny, dest, brack);
10748 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10752 /* Handle extended ${^Foo} variables
10753 * 1999-02-27 mjd-perl-patch@plover.com */
10754 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10758 while (isALNUM(*s) && d < e) {
10762 Perl_croak(aTHX_ ident_too_long);
10767 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10768 PL_lex_state = LEX_INTERPEND;
10771 if (PL_lex_state == LEX_NORMAL) {
10772 if (ckWARN(WARN_AMBIGUOUS) &&
10773 (keyword(dest, d - dest, 0)
10774 || get_cvn_flags(dest, d - dest, 0)))
10778 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10779 "Ambiguous use of %c{%s} resolved to %c%s",
10780 funny, dest, funny, dest);
10785 s = bracket; /* let the parser handle it */
10789 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10790 PL_lex_state = LEX_INTERPEND;
10795 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10797 PERL_UNUSED_CONTEXT;
10801 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10802 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10803 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10804 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10805 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10811 S_scan_pat(pTHX_ char *start, I32 type)
10815 char *s = scan_str(start,!!PL_madskills,FALSE);
10816 const char * const valid_flags =
10817 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10824 const char * const delimiter = skipspace(start);
10828 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10829 : "Search pattern not terminated" ));
10832 pm = (PMOP*)newPMOP(type, 0);
10833 if (PL_multi_open == '?') {
10834 /* This is the only point in the code that sets PMf_ONCE: */
10835 pm->op_pmflags |= PMf_ONCE;
10837 /* Hence it's safe to do this bit of PMOP book-keeping here, which
10838 allows us to restrict the list needed by reset to just the ??
10840 assert(type != OP_TRANS);
10842 MAGIC *mg = mg_find((SV*)PL_curstash, PERL_MAGIC_symtab);
10845 mg = sv_magicext((SV*)PL_curstash, 0, PERL_MAGIC_symtab, 0, 0,
10848 elements = mg->mg_len / sizeof(PMOP**);
10849 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
10850 ((PMOP**)mg->mg_ptr) [elements++] = pm;
10851 mg->mg_len = elements * sizeof(PMOP**);
10852 PmopSTASH_set(pm,PL_curstash);
10858 while (*s && strchr(valid_flags, *s))
10859 pmflag(&pm->op_pmflags,*s++);
10861 if (PL_madskills && modstart != s) {
10862 SV* tmptoken = newSVpvn(modstart, s - modstart);
10863 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10866 /* issue a warning if /c is specified,but /g is not */
10867 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10868 && ckWARN(WARN_REGEXP))
10870 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
10871 "Use of /c modifier is meaningless without /g" );
10874 PL_lex_op = (OP*)pm;
10875 yylval.ival = OP_MATCH;
10880 S_scan_subst(pTHX_ char *start)
10891 yylval.ival = OP_NULL;
10893 s = scan_str(start,!!PL_madskills,FALSE);
10896 Perl_croak(aTHX_ "Substitution pattern not terminated");
10898 if (s[-1] == PL_multi_open)
10901 if (PL_madskills) {
10902 CURMAD('q', PL_thisopen);
10903 CURMAD('_', PL_thiswhite);
10904 CURMAD('E', PL_thisstuff);
10905 CURMAD('Q', PL_thisclose);
10906 PL_realtokenstart = s - SvPVX(PL_linestr);
10910 first_start = PL_multi_start;
10911 s = scan_str(s,!!PL_madskills,FALSE);
10913 if (PL_lex_stuff) {
10914 SvREFCNT_dec(PL_lex_stuff);
10915 PL_lex_stuff = NULL;
10917 Perl_croak(aTHX_ "Substitution replacement not terminated");
10919 PL_multi_start = first_start; /* so whole substitution is taken together */
10921 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10924 if (PL_madskills) {
10925 CURMAD('z', PL_thisopen);
10926 CURMAD('R', PL_thisstuff);
10927 CURMAD('Z', PL_thisclose);
10933 if (*s == EXEC_PAT_MOD) {
10937 else if (strchr(S_PAT_MODS, *s))
10938 pmflag(&pm->op_pmflags,*s++);
10944 if (PL_madskills) {
10946 curmad('m', newSVpvn(modstart, s - modstart));
10947 append_madprops(PL_thismad, (OP*)pm, 0);
10951 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10952 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10956 SV * const repl = newSVpvs("");
10958 PL_sublex_info.super_bufptr = s;
10959 PL_sublex_info.super_bufend = PL_bufend;
10961 pm->op_pmflags |= PMf_EVAL;
10964 sv_catpvs(repl, "eval ");
10966 sv_catpvs(repl, "do ");
10968 sv_catpvs(repl, "{");
10969 sv_catsv(repl, PL_lex_repl);
10970 if (strchr(SvPVX(PL_lex_repl), '#'))
10971 sv_catpvs(repl, "\n");
10972 sv_catpvs(repl, "}");
10974 SvREFCNT_dec(PL_lex_repl);
10975 PL_lex_repl = repl;
10978 PL_lex_op = (OP*)pm;
10979 yylval.ival = OP_SUBST;
10984 S_scan_trans(pTHX_ char *start)
10997 yylval.ival = OP_NULL;
10999 s = scan_str(start,!!PL_madskills,FALSE);
11001 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11003 if (s[-1] == PL_multi_open)
11006 if (PL_madskills) {
11007 CURMAD('q', PL_thisopen);
11008 CURMAD('_', PL_thiswhite);
11009 CURMAD('E', PL_thisstuff);
11010 CURMAD('Q', PL_thisclose);
11011 PL_realtokenstart = s - SvPVX(PL_linestr);
11015 s = scan_str(s,!!PL_madskills,FALSE);
11017 if (PL_lex_stuff) {
11018 SvREFCNT_dec(PL_lex_stuff);
11019 PL_lex_stuff = NULL;
11021 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11023 if (PL_madskills) {
11024 CURMAD('z', PL_thisopen);
11025 CURMAD('R', PL_thisstuff);
11026 CURMAD('Z', PL_thisclose);
11029 complement = del = squash = 0;
11036 complement = OPpTRANS_COMPLEMENT;
11039 del = OPpTRANS_DELETE;
11042 squash = OPpTRANS_SQUASH;
11051 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11052 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11053 o->op_private &= ~OPpTRANS_ALL;
11054 o->op_private |= del|squash|complement|
11055 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11056 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11059 yylval.ival = OP_TRANS;
11062 if (PL_madskills) {
11064 curmad('m', newSVpvn(modstart, s - modstart));
11065 append_madprops(PL_thismad, o, 0);
11074 S_scan_heredoc(pTHX_ register char *s)
11078 I32 op_type = OP_SCALAR;
11082 const char *found_newline;
11086 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11088 I32 stuffstart = s - SvPVX(PL_linestr);
11091 PL_realtokenstart = -1;
11096 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11100 while (SPACE_OR_TAB(*peek))
11102 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11105 s = delimcpy(d, e, s, PL_bufend, term, &len);
11115 if (!isALNUM_lazy_if(s,UTF))
11116 deprecate_old("bare << to mean <<\"\"");
11117 for (; isALNUM_lazy_if(s,UTF); s++) {
11122 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11123 Perl_croak(aTHX_ "Delimiter for here document is too long");
11126 len = d - PL_tokenbuf;
11129 if (PL_madskills) {
11130 tstart = PL_tokenbuf + !outer;
11131 PL_thisclose = newSVpvn(tstart, len - !outer);
11132 tstart = SvPVX(PL_linestr) + stuffstart;
11133 PL_thisopen = newSVpvn(tstart, s - tstart);
11134 stuffstart = s - SvPVX(PL_linestr);
11137 #ifndef PERL_STRICT_CR
11138 d = strchr(s, '\r');
11140 char * const olds = s;
11142 while (s < PL_bufend) {
11148 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11157 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11164 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11165 herewas = newSVpvn(s,PL_bufend-s);
11169 herewas = newSVpvn(s-1,found_newline-s+1);
11172 herewas = newSVpvn(s,found_newline-s);
11176 if (PL_madskills) {
11177 tstart = SvPVX(PL_linestr) + stuffstart;
11179 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11181 PL_thisstuff = newSVpvn(tstart, s - tstart);
11184 s += SvCUR(herewas);
11187 stuffstart = s - SvPVX(PL_linestr);
11193 tmpstr = newSV_type(SVt_PVIV);
11194 SvGROW(tmpstr, 80);
11195 if (term == '\'') {
11196 op_type = OP_CONST;
11197 SvIV_set(tmpstr, -1);
11199 else if (term == '`') {
11200 op_type = OP_BACKTICK;
11201 SvIV_set(tmpstr, '\\');
11205 PL_multi_start = CopLINE(PL_curcop);
11206 PL_multi_open = PL_multi_close = '<';
11207 term = *PL_tokenbuf;
11208 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11209 char * const bufptr = PL_sublex_info.super_bufptr;
11210 char * const bufend = PL_sublex_info.super_bufend;
11211 char * const olds = s - SvCUR(herewas);
11212 s = strchr(bufptr, '\n');
11216 while (s < bufend &&
11217 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11219 CopLINE_inc(PL_curcop);
11222 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11223 missingterm(PL_tokenbuf);
11225 sv_setpvn(herewas,bufptr,d-bufptr+1);
11226 sv_setpvn(tmpstr,d+1,s-d);
11228 sv_catpvn(herewas,s,bufend-s);
11229 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11236 while (s < PL_bufend &&
11237 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11239 CopLINE_inc(PL_curcop);
11241 if (s >= PL_bufend) {
11242 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11243 missingterm(PL_tokenbuf);
11245 sv_setpvn(tmpstr,d+1,s-d);
11247 if (PL_madskills) {
11249 sv_catpvn(PL_thisstuff, d + 1, s - d);
11251 PL_thisstuff = newSVpvn(d + 1, s - d);
11252 stuffstart = s - SvPVX(PL_linestr);
11256 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11258 sv_catpvn(herewas,s,PL_bufend-s);
11259 sv_setsv(PL_linestr,herewas);
11260 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11261 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11262 PL_last_lop = PL_last_uni = NULL;
11265 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11266 while (s >= PL_bufend) { /* multiple line string? */
11268 if (PL_madskills) {
11269 tstart = SvPVX(PL_linestr) + stuffstart;
11271 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11273 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11277 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11278 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11279 missingterm(PL_tokenbuf);
11282 stuffstart = s - SvPVX(PL_linestr);
11284 CopLINE_inc(PL_curcop);
11285 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11286 PL_last_lop = PL_last_uni = NULL;
11287 #ifndef PERL_STRICT_CR
11288 if (PL_bufend - PL_linestart >= 2) {
11289 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11290 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11292 PL_bufend[-2] = '\n';
11294 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11296 else if (PL_bufend[-1] == '\r')
11297 PL_bufend[-1] = '\n';
11299 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11300 PL_bufend[-1] = '\n';
11302 if (PERLDB_LINE && PL_curstash != PL_debstash)
11303 update_debugger_info(PL_linestr, NULL, 0);
11304 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11305 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11306 *(SvPVX(PL_linestr) + off ) = ' ';
11307 sv_catsv(PL_linestr,herewas);
11308 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11309 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11313 sv_catsv(tmpstr,PL_linestr);
11318 PL_multi_end = CopLINE(PL_curcop);
11319 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11320 SvPV_shrink_to_cur(tmpstr);
11322 SvREFCNT_dec(herewas);
11324 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11326 else if (PL_encoding)
11327 sv_recode_to_utf8(tmpstr, PL_encoding);
11329 PL_lex_stuff = tmpstr;
11330 yylval.ival = op_type;
11334 /* scan_inputsymbol
11335 takes: current position in input buffer
11336 returns: new position in input buffer
11337 side-effects: yylval and lex_op are set.
11342 <FH> read from filehandle
11343 <pkg::FH> read from package qualified filehandle
11344 <pkg'FH> read from package qualified filehandle
11345 <$fh> read from filehandle in $fh
11346 <*.h> filename glob
11351 S_scan_inputsymbol(pTHX_ char *start)
11354 register char *s = start; /* current position in buffer */
11358 char *d = PL_tokenbuf; /* start of temp holding space */
11359 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11361 end = strchr(s, '\n');
11364 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11366 /* die if we didn't have space for the contents of the <>,
11367 or if it didn't end, or if we see a newline
11370 if (len >= (I32)sizeof PL_tokenbuf)
11371 Perl_croak(aTHX_ "Excessively long <> operator");
11373 Perl_croak(aTHX_ "Unterminated <> operator");
11378 Remember, only scalar variables are interpreted as filehandles by
11379 this code. Anything more complex (e.g., <$fh{$num}>) will be
11380 treated as a glob() call.
11381 This code makes use of the fact that except for the $ at the front,
11382 a scalar variable and a filehandle look the same.
11384 if (*d == '$' && d[1]) d++;
11386 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11387 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11390 /* If we've tried to read what we allow filehandles to look like, and
11391 there's still text left, then it must be a glob() and not a getline.
11392 Use scan_str to pull out the stuff between the <> and treat it
11393 as nothing more than a string.
11396 if (d - PL_tokenbuf != len) {
11397 yylval.ival = OP_GLOB;
11398 s = scan_str(start,!!PL_madskills,FALSE);
11400 Perl_croak(aTHX_ "Glob not terminated");
11404 bool readline_overriden = FALSE;
11407 /* we're in a filehandle read situation */
11410 /* turn <> into <ARGV> */
11412 Copy("ARGV",d,5,char);
11414 /* Check whether readline() is overriden */
11415 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11417 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11419 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11420 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11421 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11422 readline_overriden = TRUE;
11424 /* if <$fh>, create the ops to turn the variable into a
11428 /* try to find it in the pad for this block, otherwise find
11429 add symbol table ops
11431 const PADOFFSET tmp = pad_findmy(d);
11432 if (tmp != NOT_IN_PAD) {
11433 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11434 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11435 HEK * const stashname = HvNAME_HEK(stash);
11436 SV * const sym = sv_2mortal(newSVhek(stashname));
11437 sv_catpvs(sym, "::");
11438 sv_catpv(sym, d+1);
11443 OP * const o = newOP(OP_PADSV, 0);
11445 PL_lex_op = readline_overriden
11446 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11447 append_elem(OP_LIST, o,
11448 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11449 : (OP*)newUNOP(OP_READLINE, 0, o);
11458 ? (GV_ADDMULTI | GV_ADDINEVAL)
11461 PL_lex_op = readline_overriden
11462 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11463 append_elem(OP_LIST,
11464 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11465 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11466 : (OP*)newUNOP(OP_READLINE, 0,
11467 newUNOP(OP_RV2SV, 0,
11468 newGVOP(OP_GV, 0, gv)));
11470 if (!readline_overriden)
11471 PL_lex_op->op_flags |= OPf_SPECIAL;
11472 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11473 yylval.ival = OP_NULL;
11476 /* If it's none of the above, it must be a literal filehandle
11477 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11479 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11480 PL_lex_op = readline_overriden
11481 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11482 append_elem(OP_LIST,
11483 newGVOP(OP_GV, 0, gv),
11484 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11485 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11486 yylval.ival = OP_NULL;
11495 takes: start position in buffer
11496 keep_quoted preserve \ on the embedded delimiter(s)
11497 keep_delims preserve the delimiters around the string
11498 returns: position to continue reading from buffer
11499 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11500 updates the read buffer.
11502 This subroutine pulls a string out of the input. It is called for:
11503 q single quotes q(literal text)
11504 ' single quotes 'literal text'
11505 qq double quotes qq(interpolate $here please)
11506 " double quotes "interpolate $here please"
11507 qx backticks qx(/bin/ls -l)
11508 ` backticks `/bin/ls -l`
11509 qw quote words @EXPORT_OK = qw( func() $spam )
11510 m// regexp match m/this/
11511 s/// regexp substitute s/this/that/
11512 tr/// string transliterate tr/this/that/
11513 y/// string transliterate y/this/that/
11514 ($*@) sub prototypes sub foo ($)
11515 (stuff) sub attr parameters sub foo : attr(stuff)
11516 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11518 In most of these cases (all but <>, patterns and transliterate)
11519 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11520 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11521 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11524 It skips whitespace before the string starts, and treats the first
11525 character as the delimiter. If the delimiter is one of ([{< then
11526 the corresponding "close" character )]}> is used as the closing
11527 delimiter. It allows quoting of delimiters, and if the string has
11528 balanced delimiters ([{<>}]) it allows nesting.
11530 On success, the SV with the resulting string is put into lex_stuff or,
11531 if that is already non-NULL, into lex_repl. The second case occurs only
11532 when parsing the RHS of the special constructs s/// and tr/// (y///).
11533 For convenience, the terminating delimiter character is stuffed into
11538 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11541 SV *sv; /* scalar value: string */
11542 const char *tmps; /* temp string, used for delimiter matching */
11543 register char *s = start; /* current position in the buffer */
11544 register char term; /* terminating character */
11545 register char *to; /* current position in the sv's data */
11546 I32 brackets = 1; /* bracket nesting level */
11547 bool has_utf8 = FALSE; /* is there any utf8 content? */
11548 I32 termcode; /* terminating char. code */
11549 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11550 STRLEN termlen; /* length of terminating string */
11551 int last_off = 0; /* last position for nesting bracket */
11557 /* skip space before the delimiter */
11563 if (PL_realtokenstart >= 0) {
11564 stuffstart = PL_realtokenstart;
11565 PL_realtokenstart = -1;
11568 stuffstart = start - SvPVX(PL_linestr);
11570 /* mark where we are, in case we need to report errors */
11573 /* after skipping whitespace, the next character is the terminator */
11576 termcode = termstr[0] = term;
11580 termcode = utf8_to_uvchr((U8*)s, &termlen);
11581 Copy(s, termstr, termlen, U8);
11582 if (!UTF8_IS_INVARIANT(term))
11586 /* mark where we are */
11587 PL_multi_start = CopLINE(PL_curcop);
11588 PL_multi_open = term;
11590 /* find corresponding closing delimiter */
11591 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11592 termcode = termstr[0] = term = tmps[5];
11594 PL_multi_close = term;
11596 /* create a new SV to hold the contents. 79 is the SV's initial length.
11597 What a random number. */
11598 sv = newSV_type(SVt_PVIV);
11600 SvIV_set(sv, termcode);
11601 (void)SvPOK_only(sv); /* validate pointer */
11603 /* move past delimiter and try to read a complete string */
11605 sv_catpvn(sv, s, termlen);
11608 tstart = SvPVX(PL_linestr) + stuffstart;
11609 if (!PL_thisopen && !keep_delims) {
11610 PL_thisopen = newSVpvn(tstart, s - tstart);
11611 stuffstart = s - SvPVX(PL_linestr);
11615 if (PL_encoding && !UTF) {
11619 int offset = s - SvPVX_const(PL_linestr);
11620 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11621 &offset, (char*)termstr, termlen);
11622 const char * const ns = SvPVX_const(PL_linestr) + offset;
11623 char * const svlast = SvEND(sv) - 1;
11625 for (; s < ns; s++) {
11626 if (*s == '\n' && !PL_rsfp)
11627 CopLINE_inc(PL_curcop);
11630 goto read_more_line;
11632 /* handle quoted delimiters */
11633 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11635 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11637 if ((svlast-1 - t) % 2) {
11638 if (!keep_quoted) {
11639 *(svlast-1) = term;
11641 SvCUR_set(sv, SvCUR(sv) - 1);
11646 if (PL_multi_open == PL_multi_close) {
11652 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11653 /* At here, all closes are "was quoted" one,
11654 so we don't check PL_multi_close. */
11656 if (!keep_quoted && *(t+1) == PL_multi_open)
11661 else if (*t == PL_multi_open)
11669 SvCUR_set(sv, w - SvPVX_const(sv));
11671 last_off = w - SvPVX(sv);
11672 if (--brackets <= 0)
11677 if (!keep_delims) {
11678 SvCUR_set(sv, SvCUR(sv) - 1);
11684 /* extend sv if need be */
11685 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11686 /* set 'to' to the next character in the sv's string */
11687 to = SvPVX(sv)+SvCUR(sv);
11689 /* if open delimiter is the close delimiter read unbridle */
11690 if (PL_multi_open == PL_multi_close) {
11691 for (; s < PL_bufend; s++,to++) {
11692 /* embedded newlines increment the current line number */
11693 if (*s == '\n' && !PL_rsfp)
11694 CopLINE_inc(PL_curcop);
11695 /* handle quoted delimiters */
11696 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11697 if (!keep_quoted && s[1] == term)
11699 /* any other quotes are simply copied straight through */
11703 /* terminate when run out of buffer (the for() condition), or
11704 have found the terminator */
11705 else if (*s == term) {
11708 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11711 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11717 /* if the terminator isn't the same as the start character (e.g.,
11718 matched brackets), we have to allow more in the quoting, and
11719 be prepared for nested brackets.
11722 /* read until we run out of string, or we find the terminator */
11723 for (; s < PL_bufend; s++,to++) {
11724 /* embedded newlines increment the line count */
11725 if (*s == '\n' && !PL_rsfp)
11726 CopLINE_inc(PL_curcop);
11727 /* backslashes can escape the open or closing characters */
11728 if (*s == '\\' && s+1 < PL_bufend) {
11729 if (!keep_quoted &&
11730 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11735 /* allow nested opens and closes */
11736 else if (*s == PL_multi_close && --brackets <= 0)
11738 else if (*s == PL_multi_open)
11740 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11745 /* terminate the copied string and update the sv's end-of-string */
11747 SvCUR_set(sv, to - SvPVX_const(sv));
11750 * this next chunk reads more into the buffer if we're not done yet
11754 break; /* handle case where we are done yet :-) */
11756 #ifndef PERL_STRICT_CR
11757 if (to - SvPVX_const(sv) >= 2) {
11758 if ((to[-2] == '\r' && to[-1] == '\n') ||
11759 (to[-2] == '\n' && to[-1] == '\r'))
11763 SvCUR_set(sv, to - SvPVX_const(sv));
11765 else if (to[-1] == '\r')
11768 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11773 /* if we're out of file, or a read fails, bail and reset the current
11774 line marker so we can report where the unterminated string began
11777 if (PL_madskills) {
11778 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11780 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11782 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11786 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11788 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11794 /* we read a line, so increment our line counter */
11795 CopLINE_inc(PL_curcop);
11797 /* update debugger info */
11798 if (PERLDB_LINE && PL_curstash != PL_debstash)
11799 update_debugger_info(PL_linestr, NULL, 0);
11801 /* having changed the buffer, we must update PL_bufend */
11802 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11803 PL_last_lop = PL_last_uni = NULL;
11806 /* at this point, we have successfully read the delimited string */
11808 if (!PL_encoding || UTF) {
11810 if (PL_madskills) {
11811 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11812 const int len = s - tstart;
11814 sv_catpvn(PL_thisstuff, tstart, len);
11816 PL_thisstuff = newSVpvn(tstart, len);
11817 if (!PL_thisclose && !keep_delims)
11818 PL_thisclose = newSVpvn(s,termlen);
11823 sv_catpvn(sv, s, termlen);
11828 if (PL_madskills) {
11829 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11830 const int len = s - tstart - termlen;
11832 sv_catpvn(PL_thisstuff, tstart, len);
11834 PL_thisstuff = newSVpvn(tstart, len);
11835 if (!PL_thisclose && !keep_delims)
11836 PL_thisclose = newSVpvn(s - termlen,termlen);
11840 if (has_utf8 || PL_encoding)
11843 PL_multi_end = CopLINE(PL_curcop);
11845 /* if we allocated too much space, give some back */
11846 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11847 SvLEN_set(sv, SvCUR(sv) + 1);
11848 SvPV_renew(sv, SvLEN(sv));
11851 /* decide whether this is the first or second quoted string we've read
11864 takes: pointer to position in buffer
11865 returns: pointer to new position in buffer
11866 side-effects: builds ops for the constant in yylval.op
11868 Read a number in any of the formats that Perl accepts:
11870 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11871 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11874 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11876 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11879 If it reads a number without a decimal point or an exponent, it will
11880 try converting the number to an integer and see if it can do so
11881 without loss of precision.
11885 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11888 register const char *s = start; /* current position in buffer */
11889 register char *d; /* destination in temp buffer */
11890 register char *e; /* end of temp buffer */
11891 NV nv; /* number read, as a double */
11892 SV *sv = NULL; /* place to put the converted number */
11893 bool floatit; /* boolean: int or float? */
11894 const char *lastub = NULL; /* position of last underbar */
11895 static char const number_too_long[] = "Number too long";
11897 /* We use the first character to decide what type of number this is */
11901 Perl_croak(aTHX_ "panic: scan_num");
11903 /* if it starts with a 0, it could be an octal number, a decimal in
11904 0.13 disguise, or a hexadecimal number, or a binary number. */
11908 u holds the "number so far"
11909 shift the power of 2 of the base
11910 (hex == 4, octal == 3, binary == 1)
11911 overflowed was the number more than we can hold?
11913 Shift is used when we add a digit. It also serves as an "are
11914 we in octal/hex/binary?" indicator to disallow hex characters
11915 when in octal mode.
11920 bool overflowed = FALSE;
11921 bool just_zero = TRUE; /* just plain 0 or binary number? */
11922 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11923 static const char* const bases[5] =
11924 { "", "binary", "", "octal", "hexadecimal" };
11925 static const char* const Bases[5] =
11926 { "", "Binary", "", "Octal", "Hexadecimal" };
11927 static const char* const maxima[5] =
11929 "0b11111111111111111111111111111111",
11933 const char *base, *Base, *max;
11935 /* check for hex */
11940 } else if (s[1] == 'b') {
11945 /* check for a decimal in disguise */
11946 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11948 /* so it must be octal */
11955 if (ckWARN(WARN_SYNTAX))
11956 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11957 "Misplaced _ in number");
11961 base = bases[shift];
11962 Base = Bases[shift];
11963 max = maxima[shift];
11965 /* read the rest of the number */
11967 /* x is used in the overflow test,
11968 b is the digit we're adding on. */
11973 /* if we don't mention it, we're done */
11977 /* _ are ignored -- but warned about if consecutive */
11979 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11980 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11981 "Misplaced _ in number");
11985 /* 8 and 9 are not octal */
11986 case '8': case '9':
11988 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11992 case '2': case '3': case '4':
11993 case '5': case '6': case '7':
11995 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11998 case '0': case '1':
11999 b = *s++ & 15; /* ASCII digit -> value of digit */
12003 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12004 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12005 /* make sure they said 0x */
12008 b = (*s++ & 7) + 9;
12010 /* Prepare to put the digit we have onto the end
12011 of the number so far. We check for overflows.
12017 x = u << shift; /* make room for the digit */
12019 if ((x >> shift) != u
12020 && !(PL_hints & HINT_NEW_BINARY)) {
12023 if (ckWARN_d(WARN_OVERFLOW))
12024 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12025 "Integer overflow in %s number",
12028 u = x | b; /* add the digit to the end */
12031 n *= nvshift[shift];
12032 /* If an NV has not enough bits in its
12033 * mantissa to represent an UV this summing of
12034 * small low-order numbers is a waste of time
12035 * (because the NV cannot preserve the
12036 * low-order bits anyway): we could just
12037 * remember when did we overflow and in the
12038 * end just multiply n by the right
12046 /* if we get here, we had success: make a scalar value from
12051 /* final misplaced underbar check */
12052 if (s[-1] == '_') {
12053 if (ckWARN(WARN_SYNTAX))
12054 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12059 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12060 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12061 "%s number > %s non-portable",
12067 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12068 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12069 "%s number > %s non-portable",
12074 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12075 sv = new_constant(start, s - start, "integer",
12077 else if (PL_hints & HINT_NEW_BINARY)
12078 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
12083 handle decimal numbers.
12084 we're also sent here when we read a 0 as the first digit
12086 case '1': case '2': case '3': case '4': case '5':
12087 case '6': case '7': case '8': case '9': case '.':
12090 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12093 /* read next group of digits and _ and copy into d */
12094 while (isDIGIT(*s) || *s == '_') {
12095 /* skip underscores, checking for misplaced ones
12099 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12100 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12101 "Misplaced _ in number");
12105 /* check for end of fixed-length buffer */
12107 Perl_croak(aTHX_ number_too_long);
12108 /* if we're ok, copy the character */
12113 /* final misplaced underbar check */
12114 if (lastub && s == lastub + 1) {
12115 if (ckWARN(WARN_SYNTAX))
12116 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12119 /* read a decimal portion if there is one. avoid
12120 3..5 being interpreted as the number 3. followed
12123 if (*s == '.' && s[1] != '.') {
12128 if (ckWARN(WARN_SYNTAX))
12129 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12130 "Misplaced _ in number");
12134 /* copy, ignoring underbars, until we run out of digits.
12136 for (; isDIGIT(*s) || *s == '_'; s++) {
12137 /* fixed length buffer check */
12139 Perl_croak(aTHX_ number_too_long);
12141 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12142 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12143 "Misplaced _ in number");
12149 /* fractional part ending in underbar? */
12150 if (s[-1] == '_') {
12151 if (ckWARN(WARN_SYNTAX))
12152 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12153 "Misplaced _ in number");
12155 if (*s == '.' && isDIGIT(s[1])) {
12156 /* oops, it's really a v-string, but without the "v" */
12162 /* read exponent part, if present */
12163 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12167 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12168 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12170 /* stray preinitial _ */
12172 if (ckWARN(WARN_SYNTAX))
12173 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12174 "Misplaced _ in number");
12178 /* allow positive or negative exponent */
12179 if (*s == '+' || *s == '-')
12182 /* stray initial _ */
12184 if (ckWARN(WARN_SYNTAX))
12185 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12186 "Misplaced _ in number");
12190 /* read digits of exponent */
12191 while (isDIGIT(*s) || *s == '_') {
12194 Perl_croak(aTHX_ number_too_long);
12198 if (((lastub && s == lastub + 1) ||
12199 (!isDIGIT(s[1]) && s[1] != '_'))
12200 && ckWARN(WARN_SYNTAX))
12201 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12202 "Misplaced _ in number");
12209 /* make an sv from the string */
12213 We try to do an integer conversion first if no characters
12214 indicating "float" have been found.
12219 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12221 if (flags == IS_NUMBER_IN_UV) {
12223 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12226 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12227 if (uv <= (UV) IV_MIN)
12228 sv_setiv(sv, -(IV)uv);
12235 /* terminate the string */
12237 nv = Atof(PL_tokenbuf);
12241 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
12242 (PL_hints & HINT_NEW_INTEGER) )
12243 sv = new_constant(PL_tokenbuf,
12246 (floatit ? "float" : "integer"),
12250 /* if it starts with a v, it could be a v-string */
12253 sv = newSV(5); /* preallocate storage space */
12254 s = scan_vstring(s, PL_bufend, sv);
12258 /* make the op for the constant and return */
12261 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12263 lvalp->opval = NULL;
12269 S_scan_formline(pTHX_ register char *s)
12272 register char *eol;
12274 SV * const stuff = newSVpvs("");
12275 bool needargs = FALSE;
12276 bool eofmt = FALSE;
12278 char *tokenstart = s;
12281 if (PL_madskills) {
12282 savewhite = PL_thiswhite;
12287 while (!needargs) {
12290 #ifdef PERL_STRICT_CR
12291 while (SPACE_OR_TAB(*t))
12294 while (SPACE_OR_TAB(*t) || *t == '\r')
12297 if (*t == '\n' || t == PL_bufend) {
12302 if (PL_in_eval && !PL_rsfp) {
12303 eol = (char *) memchr(s,'\n',PL_bufend-s);
12308 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12310 for (t = s; t < eol; t++) {
12311 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12313 goto enough; /* ~~ must be first line in formline */
12315 if (*t == '@' || *t == '^')
12319 sv_catpvn(stuff, s, eol-s);
12320 #ifndef PERL_STRICT_CR
12321 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12322 char *end = SvPVX(stuff) + SvCUR(stuff);
12325 SvCUR_set(stuff, SvCUR(stuff) - 1);
12335 if (PL_madskills) {
12337 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12339 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12342 s = filter_gets(PL_linestr, PL_rsfp, 0);
12344 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12346 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12348 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12349 PL_last_lop = PL_last_uni = NULL;
12358 if (SvCUR(stuff)) {
12361 PL_lex_state = LEX_NORMAL;
12362 start_force(PL_curforce);
12363 NEXTVAL_NEXTTOKE.ival = 0;
12367 PL_lex_state = LEX_FORMLINE;
12369 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12371 else if (PL_encoding)
12372 sv_recode_to_utf8(stuff, PL_encoding);
12374 start_force(PL_curforce);
12375 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12377 start_force(PL_curforce);
12378 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12382 SvREFCNT_dec(stuff);
12384 PL_lex_formbrack = 0;
12388 if (PL_madskills) {
12390 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12392 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12393 PL_thiswhite = savewhite;
12400 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12403 const I32 oldsavestack_ix = PL_savestack_ix;
12404 CV* const outsidecv = PL_compcv;
12407 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12409 SAVEI32(PL_subline);
12410 save_item(PL_subname);
12411 SAVESPTR(PL_compcv);
12413 PL_compcv = (CV*)newSV_type(is_format ? SVt_PVFM : SVt_PVCV);
12414 CvFLAGS(PL_compcv) |= flags;
12416 PL_subline = CopLINE(PL_curcop);
12417 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12418 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12419 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12421 return oldsavestack_ix;
12425 #pragma segment Perl_yylex
12428 Perl_yywarn(pTHX_ const char *s)
12431 PL_in_eval |= EVAL_WARNONLY;
12433 PL_in_eval &= ~EVAL_WARNONLY;
12438 Perl_yyerror(pTHX_ const char *s)
12441 const char *where = NULL;
12442 const char *context = NULL;
12445 int yychar = PL_parser->yychar;
12447 if (!yychar || (yychar == ';' && !PL_rsfp))
12449 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12450 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12451 PL_oldbufptr != PL_bufptr) {
12454 The code below is removed for NetWare because it abends/crashes on NetWare
12455 when the script has error such as not having the closing quotes like:
12456 if ($var eq "value)
12457 Checking of white spaces is anyway done in NetWare code.
12460 while (isSPACE(*PL_oldoldbufptr))
12463 context = PL_oldoldbufptr;
12464 contlen = PL_bufptr - PL_oldoldbufptr;
12466 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12467 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12470 The code below is removed for NetWare because it abends/crashes on NetWare
12471 when the script has error such as not having the closing quotes like:
12472 if ($var eq "value)
12473 Checking of white spaces is anyway done in NetWare code.
12476 while (isSPACE(*PL_oldbufptr))
12479 context = PL_oldbufptr;
12480 contlen = PL_bufptr - PL_oldbufptr;
12482 else if (yychar > 255)
12483 where = "next token ???";
12484 else if (yychar == -2) { /* YYEMPTY */
12485 if (PL_lex_state == LEX_NORMAL ||
12486 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12487 where = "at end of line";
12488 else if (PL_lex_inpat)
12489 where = "within pattern";
12491 where = "within string";
12494 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12496 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12497 else if (isPRINT_LC(yychar)) {
12498 const unsigned char string = (unsigned char) yychar;
12499 sv_catpvn(where_sv, &string, 1);
12502 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12503 where = SvPVX_const(where_sv);
12505 msg = sv_2mortal(newSVpv(s, 0));
12506 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12507 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12509 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12511 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12512 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12513 Perl_sv_catpvf(aTHX_ msg,
12514 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12515 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12518 if (PL_in_eval & EVAL_WARNONLY) {
12519 if (ckWARN_d(WARN_SYNTAX))
12520 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12524 if (PL_error_count >= 10) {
12525 if (PL_in_eval && SvCUR(ERRSV))
12526 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12527 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12529 Perl_croak(aTHX_ "%s has too many errors.\n",
12530 OutCopFILE(PL_curcop));
12533 PL_in_my_stash = NULL;
12537 #pragma segment Main
12541 S_swallow_bom(pTHX_ U8 *s)
12544 const STRLEN slen = SvCUR(PL_linestr);
12547 if (s[1] == 0xFE) {
12548 /* UTF-16 little-endian? (or UTF32-LE?) */
12549 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12550 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12551 #ifndef PERL_NO_UTF16_FILTER
12552 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12555 if (PL_bufend > (char*)s) {
12559 filter_add(utf16rev_textfilter, NULL);
12560 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12561 utf16_to_utf8_reversed(s, news,
12562 PL_bufend - (char*)s - 1,
12564 sv_setpvn(PL_linestr, (const char*)news, newlen);
12566 s = (U8*)SvPVX(PL_linestr);
12567 Copy(news, s, newlen, U8);
12571 SvUTF8_on(PL_linestr);
12572 s = (U8*)SvPVX(PL_linestr);
12574 /* FIXME - is this a general bug fix? */
12577 PL_bufend = SvPVX(PL_linestr) + newlen;
12580 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12585 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12586 #ifndef PERL_NO_UTF16_FILTER
12587 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12590 if (PL_bufend > (char *)s) {
12594 filter_add(utf16_textfilter, NULL);
12595 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12596 utf16_to_utf8(s, news,
12597 PL_bufend - (char*)s,
12599 sv_setpvn(PL_linestr, (const char*)news, newlen);
12601 SvUTF8_on(PL_linestr);
12602 s = (U8*)SvPVX(PL_linestr);
12603 PL_bufend = SvPVX(PL_linestr) + newlen;
12606 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12611 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12612 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12613 s += 3; /* UTF-8 */
12619 if (s[2] == 0xFE && s[3] == 0xFF) {
12620 /* UTF-32 big-endian */
12621 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12624 else if (s[2] == 0 && s[3] != 0) {
12627 * are a good indicator of UTF-16BE. */
12628 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12634 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12635 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12636 s += 4; /* UTF-8 */
12642 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12645 * are a good indicator of UTF-16LE. */
12646 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12654 #ifndef PERL_NO_UTF16_FILTER
12656 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12659 const STRLEN old = SvCUR(sv);
12660 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12661 DEBUG_P(PerlIO_printf(Perl_debug_log,
12662 "utf16_textfilter(%p): %d %d (%d)\n",
12663 FPTR2DPTR(void *, utf16_textfilter),
12664 idx, maxlen, (int) count));
12668 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12669 Copy(SvPVX_const(sv), tmps, old, char);
12670 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12671 SvCUR(sv) - old, &newlen);
12672 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12674 DEBUG_P({sv_dump(sv);});
12679 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12682 const STRLEN old = SvCUR(sv);
12683 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12684 DEBUG_P(PerlIO_printf(Perl_debug_log,
12685 "utf16rev_textfilter(%p): %d %d (%d)\n",
12686 FPTR2DPTR(void *, utf16rev_textfilter),
12687 idx, maxlen, (int) count));
12691 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12692 Copy(SvPVX_const(sv), tmps, old, char);
12693 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12694 SvCUR(sv) - old, &newlen);
12695 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12697 DEBUG_P({ sv_dump(sv); });
12703 Returns a pointer to the next character after the parsed
12704 vstring, as well as updating the passed in sv.
12706 Function must be called like
12709 s = scan_vstring(s,e,sv);
12711 where s and e are the start and end of the string.
12712 The sv should already be large enough to store the vstring
12713 passed in, for performance reasons.
12718 Perl_scan_vstring(pTHX_ const char *s, const char *e, SV *sv)
12721 const char *pos = s;
12722 const char *start = s;
12723 if (*pos == 'v') pos++; /* get past 'v' */
12724 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12726 if ( *pos != '.') {
12727 /* this may not be a v-string if followed by => */
12728 const char *next = pos;
12729 while (next < e && isSPACE(*next))
12731 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12732 /* return string not v-string */
12733 sv_setpvn(sv,(char *)s,pos-s);
12734 return (char *)pos;
12738 if (!isALPHA(*pos)) {
12739 U8 tmpbuf[UTF8_MAXBYTES+1];
12742 s++; /* get past 'v' */
12744 sv_setpvn(sv, "", 0);
12747 /* this is atoi() that tolerates underscores */
12750 const char *end = pos;
12752 while (--end >= s) {
12754 const UV orev = rev;
12755 rev += (*end - '0') * mult;
12757 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12758 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12759 "Integer overflow in decimal number");
12763 if (rev > 0x7FFFFFFF)
12764 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12766 /* Append native character for the rev point */
12767 tmpend = uvchr_to_utf8(tmpbuf, rev);
12768 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12769 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12771 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12777 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12781 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12789 * c-indentation-style: bsd
12790 * c-basic-offset: 4
12791 * indent-tabs-mode: t
12794 * ex: set ts=8 sts=4 sw=4 noet: