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 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3391 if (PL_lex_starts) {
3397 sv_free(PL_thistoken);
3398 PL_thistoken = newSVpvs("");
3401 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3402 if (PL_lex_casemods == 1 && PL_lex_inpat)
3411 case LEX_INTERPPUSH:
3412 return REPORT(sublex_push());
3414 case LEX_INTERPSTART:
3415 if (PL_bufptr == PL_bufend)
3416 return REPORT(sublex_done());
3417 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3418 "### Interpolated variable\n"); });
3420 PL_lex_dojoin = (*PL_bufptr == '@');
3421 PL_lex_state = LEX_INTERPNORMAL;
3422 if (PL_lex_dojoin) {
3423 start_force(PL_curforce);
3424 NEXTVAL_NEXTTOKE.ival = 0;
3426 start_force(PL_curforce);
3427 force_ident("\"", '$');
3428 start_force(PL_curforce);
3429 NEXTVAL_NEXTTOKE.ival = 0;
3431 start_force(PL_curforce);
3432 NEXTVAL_NEXTTOKE.ival = 0;
3434 start_force(PL_curforce);
3435 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3438 if (PL_lex_starts++) {
3443 sv_free(PL_thistoken);
3444 PL_thistoken = newSVpvs("");
3447 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3448 if (!PL_lex_casemods && PL_lex_inpat)
3455 case LEX_INTERPENDMAYBE:
3456 if (intuit_more(PL_bufptr)) {
3457 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3463 if (PL_lex_dojoin) {
3464 PL_lex_dojoin = FALSE;
3465 PL_lex_state = LEX_INTERPCONCAT;
3469 sv_free(PL_thistoken);
3470 PL_thistoken = newSVpvs("");
3475 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3476 && SvEVALED(PL_lex_repl))
3478 if (PL_bufptr != PL_bufend)
3479 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3483 case LEX_INTERPCONCAT:
3485 if (PL_lex_brackets)
3486 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3488 if (PL_bufptr == PL_bufend)
3489 return REPORT(sublex_done());
3491 if (SvIVX(PL_linestr) == '\'') {
3492 SV *sv = newSVsv(PL_linestr);
3495 else if ( PL_hints & HINT_NEW_RE )
3496 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3497 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3501 s = scan_const(PL_bufptr);
3503 PL_lex_state = LEX_INTERPCASEMOD;
3505 PL_lex_state = LEX_INTERPSTART;
3508 if (s != PL_bufptr) {
3509 start_force(PL_curforce);
3511 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3513 NEXTVAL_NEXTTOKE = yylval;
3516 if (PL_lex_starts++) {
3520 sv_free(PL_thistoken);
3521 PL_thistoken = newSVpvs("");
3524 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3525 if (!PL_lex_casemods && PL_lex_inpat)
3538 PL_lex_state = LEX_NORMAL;
3539 s = scan_formline(PL_bufptr);
3540 if (!PL_lex_formbrack)
3546 PL_oldoldbufptr = PL_oldbufptr;
3552 sv_free(PL_thistoken);
3555 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3559 if (isIDFIRST_lazy_if(s,UTF))
3561 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3562 Perl_croak(aTHX_ "Unrecognized character \\x%02X in column %d", *s & 255, (int) len + 1);
3565 goto fake_eof; /* emulate EOF on ^D or ^Z */
3574 if (PL_lex_brackets) {
3575 yyerror((const char *)
3577 ? "Format not terminated"
3578 : "Missing right curly or square bracket"));
3580 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3581 "### Tokener got EOF\n");
3585 if (s++ < PL_bufend)
3586 goto retry; /* ignore stray nulls */
3589 if (!PL_in_eval && !PL_preambled) {
3590 PL_preambled = TRUE;
3595 sv_setpv(PL_linestr,incl_perldb());
3596 if (SvCUR(PL_linestr))
3597 sv_catpvs(PL_linestr,";");
3599 while(AvFILLp(PL_preambleav) >= 0) {
3600 SV *tmpsv = av_shift(PL_preambleav);
3601 sv_catsv(PL_linestr, tmpsv);
3602 sv_catpvs(PL_linestr, ";");
3605 sv_free((SV*)PL_preambleav);
3606 PL_preambleav = NULL;
3608 if (PL_minus_n || PL_minus_p) {
3609 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3611 sv_catpvs(PL_linestr,"chomp;");
3614 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3615 || *PL_splitstr == '"')
3616 && strchr(PL_splitstr + 1, *PL_splitstr))
3617 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3619 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3620 bytes can be used as quoting characters. :-) */
3621 const char *splits = PL_splitstr;
3622 sv_catpvs(PL_linestr, "our @F=split(q\0");
3625 if (*splits == '\\')
3626 sv_catpvn(PL_linestr, splits, 1);
3627 sv_catpvn(PL_linestr, splits, 1);
3628 } while (*splits++);
3629 /* This loop will embed the trailing NUL of
3630 PL_linestr as the last thing it does before
3632 sv_catpvs(PL_linestr, ");");
3636 sv_catpvs(PL_linestr,"our @F=split(' ');");
3640 sv_catpvs(PL_linestr,"use feature ':5.10';");
3641 sv_catpvs(PL_linestr, "\n");
3642 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3643 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3644 PL_last_lop = PL_last_uni = NULL;
3645 if (PERLDB_LINE && PL_curstash != PL_debstash)
3646 update_debugger_info(PL_linestr, NULL, 0);
3650 bof = PL_rsfp ? TRUE : FALSE;
3651 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3654 PL_realtokenstart = -1;
3657 if (PL_preprocess && !PL_in_eval)
3658 (void)PerlProc_pclose(PL_rsfp);
3659 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3660 PerlIO_clearerr(PL_rsfp);
3662 (void)PerlIO_close(PL_rsfp);
3664 PL_doextract = FALSE;
3666 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3671 sv_setpv(PL_linestr,
3674 ? ";}continue{print;}" : ";}"));
3675 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3676 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3677 PL_last_lop = PL_last_uni = NULL;
3678 PL_minus_n = PL_minus_p = 0;
3681 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3682 PL_last_lop = PL_last_uni = NULL;
3683 sv_setpvn(PL_linestr,"",0);
3684 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3686 /* If it looks like the start of a BOM or raw UTF-16,
3687 * check if it in fact is. */
3693 #ifdef PERLIO_IS_STDIO
3694 # ifdef __GNU_LIBRARY__
3695 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3696 # define FTELL_FOR_PIPE_IS_BROKEN
3700 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3701 # define FTELL_FOR_PIPE_IS_BROKEN
3706 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3707 /* This loses the possibility to detect the bof
3708 * situation on perl -P when the libc5 is being used.
3709 * Workaround? Maybe attach some extra state to PL_rsfp?
3712 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3714 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3717 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3718 s = swallow_bom((U8*)s);
3722 /* Incest with pod. */
3725 sv_catsv(PL_thiswhite, PL_linestr);
3727 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3728 sv_setpvn(PL_linestr, "", 0);
3729 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3730 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3731 PL_last_lop = PL_last_uni = NULL;
3732 PL_doextract = FALSE;
3736 } while (PL_doextract);
3737 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3738 if (PERLDB_LINE && PL_curstash != PL_debstash)
3739 update_debugger_info(PL_linestr, NULL, 0);
3740 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3741 PL_last_lop = PL_last_uni = NULL;
3742 if (CopLINE(PL_curcop) == 1) {
3743 while (s < PL_bufend && isSPACE(*s))
3745 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3749 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3753 if (*s == '#' && *(s+1) == '!')
3755 #ifdef ALTERNATE_SHEBANG
3757 static char const as[] = ALTERNATE_SHEBANG;
3758 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3759 d = s + (sizeof(as) - 1);
3761 #endif /* ALTERNATE_SHEBANG */
3770 while (*d && !isSPACE(*d))
3774 #ifdef ARG_ZERO_IS_SCRIPT
3775 if (ipathend > ipath) {
3777 * HP-UX (at least) sets argv[0] to the script name,
3778 * which makes $^X incorrect. And Digital UNIX and Linux,
3779 * at least, set argv[0] to the basename of the Perl
3780 * interpreter. So, having found "#!", we'll set it right.
3782 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3784 assert(SvPOK(x) || SvGMAGICAL(x));
3785 if (sv_eq(x, CopFILESV(PL_curcop))) {
3786 sv_setpvn(x, ipath, ipathend - ipath);
3792 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3793 const char * const lstart = SvPV_const(x,llen);
3795 bstart += blen - llen;
3796 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3797 sv_setpvn(x, ipath, ipathend - ipath);
3802 TAINT_NOT; /* $^X is always tainted, but that's OK */
3804 #endif /* ARG_ZERO_IS_SCRIPT */
3809 d = instr(s,"perl -");
3811 d = instr(s,"perl");
3813 /* avoid getting into infinite loops when shebang
3814 * line contains "Perl" rather than "perl" */
3816 for (d = ipathend-4; d >= ipath; --d) {
3817 if ((*d == 'p' || *d == 'P')
3818 && !ibcmp(d, "perl", 4))
3828 #ifdef ALTERNATE_SHEBANG
3830 * If the ALTERNATE_SHEBANG on this system starts with a
3831 * character that can be part of a Perl expression, then if
3832 * we see it but not "perl", we're probably looking at the
3833 * start of Perl code, not a request to hand off to some
3834 * other interpreter. Similarly, if "perl" is there, but
3835 * not in the first 'word' of the line, we assume the line
3836 * contains the start of the Perl program.
3838 if (d && *s != '#') {
3839 const char *c = ipath;
3840 while (*c && !strchr("; \t\r\n\f\v#", *c))
3843 d = NULL; /* "perl" not in first word; ignore */
3845 *s = '#'; /* Don't try to parse shebang line */
3847 #endif /* ALTERNATE_SHEBANG */
3848 #ifndef MACOS_TRADITIONAL
3853 !instr(s,"indir") &&
3854 instr(PL_origargv[0],"perl"))
3861 while (s < PL_bufend && isSPACE(*s))
3863 if (s < PL_bufend) {
3864 Newxz(newargv,PL_origargc+3,char*);
3866 while (s < PL_bufend && !isSPACE(*s))
3869 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3872 newargv = PL_origargv;
3875 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3877 Perl_croak(aTHX_ "Can't exec %s", ipath);
3881 while (*d && !isSPACE(*d))
3883 while (SPACE_OR_TAB(*d))
3887 const bool switches_done = PL_doswitches;
3888 const U32 oldpdb = PL_perldb;
3889 const bool oldn = PL_minus_n;
3890 const bool oldp = PL_minus_p;
3893 if (*d == 'M' || *d == 'm' || *d == 'C') {
3894 const char * const m = d;
3895 while (*d && !isSPACE(*d))
3897 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3900 d = moreswitches(d);
3902 if (PL_doswitches && !switches_done) {
3903 int argc = PL_origargc;
3904 char **argv = PL_origargv;
3907 } while (argc && argv[0][0] == '-' && argv[0][1]);
3908 init_argv_symbols(argc,argv);
3910 if ((PERLDB_LINE && !oldpdb) ||
3911 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3912 /* if we have already added "LINE: while (<>) {",
3913 we must not do it again */
3915 sv_setpvn(PL_linestr, "", 0);
3916 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3917 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3918 PL_last_lop = PL_last_uni = NULL;
3919 PL_preambled = FALSE;
3921 (void)gv_fetchfile(PL_origfilename);
3928 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3930 PL_lex_state = LEX_FORMLINE;
3935 #ifdef PERL_STRICT_CR
3936 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3938 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3940 case ' ': case '\t': case '\f': case 013:
3941 #ifdef MACOS_TRADITIONAL
3945 PL_realtokenstart = -1;
3947 PL_thiswhite = newSVpvs("");
3948 sv_catpvn(PL_thiswhite, s, 1);
3955 PL_realtokenstart = -1;
3959 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3960 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3961 /* handle eval qq[#line 1 "foo"\n ...] */
3962 CopLINE_dec(PL_curcop);
3965 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3967 if (!PL_in_eval || PL_rsfp)
3972 while (d < PL_bufend && *d != '\n')
3976 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3977 Perl_croak(aTHX_ "panic: input overflow");
3980 PL_thiswhite = newSVpvn(s, d - s);
3985 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3987 PL_lex_state = LEX_FORMLINE;
3993 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3994 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3997 TOKEN(PEG); /* make sure any #! line is accessible */
4002 /* if (PL_madskills && PL_lex_formbrack) { */
4004 while (d < PL_bufend && *d != '\n')
4008 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4009 Perl_croak(aTHX_ "panic: input overflow");
4010 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4012 PL_thiswhite = newSVpvs("");
4013 if (CopLINE(PL_curcop) == 1) {
4014 sv_setpvn(PL_thiswhite, "", 0);
4017 sv_catpvn(PL_thiswhite, s, d - s);
4031 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4039 while (s < PL_bufend && SPACE_OR_TAB(*s))
4042 if (strnEQ(s,"=>",2)) {
4043 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4044 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4045 OPERATOR('-'); /* unary minus */
4047 PL_last_uni = PL_oldbufptr;
4049 case 'r': ftst = OP_FTEREAD; break;
4050 case 'w': ftst = OP_FTEWRITE; break;
4051 case 'x': ftst = OP_FTEEXEC; break;
4052 case 'o': ftst = OP_FTEOWNED; break;
4053 case 'R': ftst = OP_FTRREAD; break;
4054 case 'W': ftst = OP_FTRWRITE; break;
4055 case 'X': ftst = OP_FTREXEC; break;
4056 case 'O': ftst = OP_FTROWNED; break;
4057 case 'e': ftst = OP_FTIS; break;
4058 case 'z': ftst = OP_FTZERO; break;
4059 case 's': ftst = OP_FTSIZE; break;
4060 case 'f': ftst = OP_FTFILE; break;
4061 case 'd': ftst = OP_FTDIR; break;
4062 case 'l': ftst = OP_FTLINK; break;
4063 case 'p': ftst = OP_FTPIPE; break;
4064 case 'S': ftst = OP_FTSOCK; break;
4065 case 'u': ftst = OP_FTSUID; break;
4066 case 'g': ftst = OP_FTSGID; break;
4067 case 'k': ftst = OP_FTSVTX; break;
4068 case 'b': ftst = OP_FTBLK; break;
4069 case 'c': ftst = OP_FTCHR; break;
4070 case 't': ftst = OP_FTTTY; break;
4071 case 'T': ftst = OP_FTTEXT; break;
4072 case 'B': ftst = OP_FTBINARY; break;
4073 case 'M': case 'A': case 'C':
4074 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4076 case 'M': ftst = OP_FTMTIME; break;
4077 case 'A': ftst = OP_FTATIME; break;
4078 case 'C': ftst = OP_FTCTIME; break;
4086 PL_last_lop_op = (OPCODE)ftst;
4087 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4088 "### Saw file test %c\n", (int)tmp);
4093 /* Assume it was a minus followed by a one-letter named
4094 * subroutine call (or a -bareword), then. */
4095 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4096 "### '-%c' looked like a file test but was not\n",
4103 const char tmp = *s++;
4106 if (PL_expect == XOPERATOR)
4111 else if (*s == '>') {
4114 if (isIDFIRST_lazy_if(s,UTF)) {
4115 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4123 if (PL_expect == XOPERATOR)
4126 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4128 OPERATOR('-'); /* unary minus */
4134 const char tmp = *s++;
4137 if (PL_expect == XOPERATOR)
4142 if (PL_expect == XOPERATOR)
4145 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4152 if (PL_expect != XOPERATOR) {
4153 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4154 PL_expect = XOPERATOR;
4155 force_ident(PL_tokenbuf, '*');
4168 if (PL_expect == XOPERATOR) {
4172 PL_tokenbuf[0] = '%';
4173 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4174 sizeof PL_tokenbuf - 1, FALSE);
4175 if (!PL_tokenbuf[1]) {
4178 PL_pending_ident = '%';
4189 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4196 const char tmp = *s++;
4202 goto just_a_word_zero_gv;
4205 switch (PL_expect) {
4211 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4213 PL_bufptr = s; /* update in case we back off */
4219 PL_expect = XTERMBLOCK;
4222 stuffstart = s - SvPVX(PL_linestr) - 1;
4226 while (isIDFIRST_lazy_if(s,UTF)) {
4229 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4230 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4231 if (tmp < 0) tmp = -tmp;
4245 sv = newSVpvn(s, len);
4247 d = scan_str(d,TRUE,TRUE);
4249 /* MUST advance bufptr here to avoid bogus
4250 "at end of line" context messages from yyerror().
4252 PL_bufptr = s + len;
4253 yyerror("Unterminated attribute parameter in attribute list");
4257 return REPORT(0); /* EOF indicator */
4261 sv_catsv(sv, PL_lex_stuff);
4262 attrs = append_elem(OP_LIST, attrs,
4263 newSVOP(OP_CONST, 0, sv));
4264 SvREFCNT_dec(PL_lex_stuff);
4265 PL_lex_stuff = NULL;
4268 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4270 if (PL_in_my == KEY_our) {
4272 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4274 /* skip to avoid loading attributes.pm */
4276 deprecate(":unique");
4279 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4282 /* NOTE: any CV attrs applied here need to be part of
4283 the CVf_BUILTIN_ATTRS define in cv.h! */
4284 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4286 CvLVALUE_on(PL_compcv);
4288 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4290 CvLOCKED_on(PL_compcv);
4292 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4294 CvMETHOD_on(PL_compcv);
4296 /* After we've set the flags, it could be argued that
4297 we don't need to do the attributes.pm-based setting
4298 process, and shouldn't bother appending recognized
4299 flags. To experiment with that, uncomment the
4300 following "else". (Note that's already been
4301 uncommented. That keeps the above-applied built-in
4302 attributes from being intercepted (and possibly
4303 rejected) by a package's attribute routines, but is
4304 justified by the performance win for the common case
4305 of applying only built-in attributes.) */
4307 attrs = append_elem(OP_LIST, attrs,
4308 newSVOP(OP_CONST, 0,
4312 if (*s == ':' && s[1] != ':')
4315 break; /* require real whitespace or :'s */
4316 /* XXX losing whitespace on sequential attributes here */
4320 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4321 if (*s != ';' && *s != '}' && *s != tmp
4322 && (tmp != '=' || *s != ')')) {
4323 const char q = ((*s == '\'') ? '"' : '\'');
4324 /* If here for an expression, and parsed no attrs, back
4326 if (tmp == '=' && !attrs) {
4330 /* MUST advance bufptr here to avoid bogus "at end of line"
4331 context messages from yyerror().
4334 yyerror( (const char *)
4336 ? Perl_form(aTHX_ "Invalid separator character "
4337 "%c%c%c in attribute list", q, *s, q)
4338 : "Unterminated attribute list" ) );
4346 start_force(PL_curforce);
4347 NEXTVAL_NEXTTOKE.opval = attrs;
4348 CURMAD('_', PL_nextwhite);
4353 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4354 (s - SvPVX(PL_linestr)) - stuffstart);
4362 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4363 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4371 const char tmp = *s++;
4376 const char tmp = *s++;
4384 if (PL_lex_brackets <= 0)
4385 yyerror("Unmatched right square bracket");
4388 if (PL_lex_state == LEX_INTERPNORMAL) {
4389 if (PL_lex_brackets == 0) {
4390 if (*s == '-' && s[1] == '>')
4391 PL_lex_state = LEX_INTERPENDMAYBE;
4392 else if (*s != '[' && *s != '{')
4393 PL_lex_state = LEX_INTERPEND;
4400 if (PL_lex_brackets > 100) {
4401 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4403 switch (PL_expect) {
4405 if (PL_lex_formbrack) {
4409 if (PL_oldoldbufptr == PL_last_lop)
4410 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4412 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4413 OPERATOR(HASHBRACK);
4415 while (s < PL_bufend && SPACE_OR_TAB(*s))
4418 PL_tokenbuf[0] = '\0';
4419 if (d < PL_bufend && *d == '-') {
4420 PL_tokenbuf[0] = '-';
4422 while (d < PL_bufend && SPACE_OR_TAB(*d))
4425 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4426 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4428 while (d < PL_bufend && SPACE_OR_TAB(*d))
4431 const char minus = (PL_tokenbuf[0] == '-');
4432 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4440 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4445 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4450 if (PL_oldoldbufptr == PL_last_lop)
4451 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4453 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4456 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4458 /* This hack is to get the ${} in the message. */
4460 yyerror("syntax error");
4463 OPERATOR(HASHBRACK);
4465 /* This hack serves to disambiguate a pair of curlies
4466 * as being a block or an anon hash. Normally, expectation
4467 * determines that, but in cases where we're not in a
4468 * position to expect anything in particular (like inside
4469 * eval"") we have to resolve the ambiguity. This code
4470 * covers the case where the first term in the curlies is a
4471 * quoted string. Most other cases need to be explicitly
4472 * disambiguated by prepending a "+" before the opening
4473 * curly in order to force resolution as an anon hash.
4475 * XXX should probably propagate the outer expectation
4476 * into eval"" to rely less on this hack, but that could
4477 * potentially break current behavior of eval"".
4481 if (*s == '\'' || *s == '"' || *s == '`') {
4482 /* common case: get past first string, handling escapes */
4483 for (t++; t < PL_bufend && *t != *s;)
4484 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4488 else if (*s == 'q') {
4491 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4494 /* skip q//-like construct */
4496 char open, close, term;
4499 while (t < PL_bufend && isSPACE(*t))
4501 /* check for q => */
4502 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4503 OPERATOR(HASHBRACK);
4507 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4511 for (t++; t < PL_bufend; t++) {
4512 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4514 else if (*t == open)
4518 for (t++; t < PL_bufend; t++) {
4519 if (*t == '\\' && t+1 < PL_bufend)
4521 else if (*t == close && --brackets <= 0)
4523 else if (*t == open)
4530 /* skip plain q word */
4531 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4534 else if (isALNUM_lazy_if(t,UTF)) {
4536 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4539 while (t < PL_bufend && isSPACE(*t))
4541 /* if comma follows first term, call it an anon hash */
4542 /* XXX it could be a comma expression with loop modifiers */
4543 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4544 || (*t == '=' && t[1] == '>')))
4545 OPERATOR(HASHBRACK);
4546 if (PL_expect == XREF)
4549 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4555 yylval.ival = CopLINE(PL_curcop);
4556 if (isSPACE(*s) || *s == '#')
4557 PL_copline = NOLINE; /* invalidate current command line number */
4562 if (PL_lex_brackets <= 0)
4563 yyerror("Unmatched right curly bracket");
4565 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4566 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4567 PL_lex_formbrack = 0;
4568 if (PL_lex_state == LEX_INTERPNORMAL) {
4569 if (PL_lex_brackets == 0) {
4570 if (PL_expect & XFAKEBRACK) {
4571 PL_expect &= XENUMMASK;
4572 PL_lex_state = LEX_INTERPEND;
4577 PL_thiswhite = newSVpvs("");
4578 sv_catpvn(PL_thiswhite,"}",1);
4581 return yylex(); /* ignore fake brackets */
4583 if (*s == '-' && s[1] == '>')
4584 PL_lex_state = LEX_INTERPENDMAYBE;
4585 else if (*s != '[' && *s != '{')
4586 PL_lex_state = LEX_INTERPEND;
4589 if (PL_expect & XFAKEBRACK) {
4590 PL_expect &= XENUMMASK;
4592 return yylex(); /* ignore fake brackets */
4594 start_force(PL_curforce);
4596 curmad('X', newSVpvn(s-1,1));
4597 CURMAD('_', PL_thiswhite);
4602 PL_thistoken = newSVpvs("");
4610 if (PL_expect == XOPERATOR) {
4611 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4612 && isIDFIRST_lazy_if(s,UTF))
4614 CopLINE_dec(PL_curcop);
4615 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4616 CopLINE_inc(PL_curcop);
4621 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4623 PL_expect = XOPERATOR;
4624 force_ident(PL_tokenbuf, '&');
4628 yylval.ival = (OPpENTERSUB_AMPER<<8);
4640 const char tmp = *s++;
4647 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4648 && strchr("+-*/%.^&|<",tmp))
4649 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4650 "Reversed %c= operator",(int)tmp);
4652 if (PL_expect == XSTATE && isALPHA(tmp) &&
4653 (s == PL_linestart+1 || s[-2] == '\n') )
4655 if (PL_in_eval && !PL_rsfp) {
4660 if (strnEQ(s,"=cut",4)) {
4676 PL_thiswhite = newSVpvs("");
4677 sv_catpvn(PL_thiswhite, PL_linestart,
4678 PL_bufend - PL_linestart);
4682 PL_doextract = TRUE;
4686 if (PL_lex_brackets < PL_lex_formbrack) {
4688 #ifdef PERL_STRICT_CR
4689 while (SPACE_OR_TAB(*t))
4691 while (SPACE_OR_TAB(*t) || *t == '\r')
4694 if (*t == '\n' || *t == '#') {
4705 const char tmp = *s++;
4707 /* was this !=~ where !~ was meant?
4708 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4710 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4711 const char *t = s+1;
4713 while (t < PL_bufend && isSPACE(*t))
4716 if (*t == '/' || *t == '?' ||
4717 ((*t == 'm' || *t == 's' || *t == 'y')
4718 && !isALNUM(t[1])) ||
4719 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4720 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4721 "!=~ should be !~");
4731 if (PL_expect != XOPERATOR) {
4732 if (s[1] != '<' && !strchr(s,'>'))
4735 s = scan_heredoc(s);
4737 s = scan_inputsymbol(s);
4738 TERM(sublex_start());
4744 SHop(OP_LEFT_SHIFT);
4758 const char tmp = *s++;
4760 SHop(OP_RIGHT_SHIFT);
4761 else if (tmp == '=')
4770 if (PL_expect == XOPERATOR) {
4771 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4773 deprecate_old(commaless_variable_list);
4774 return REPORT(','); /* grandfather non-comma-format format */
4778 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4779 PL_tokenbuf[0] = '@';
4780 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4781 sizeof PL_tokenbuf - 1, FALSE);
4782 if (PL_expect == XOPERATOR)
4783 no_op("Array length", s);
4784 if (!PL_tokenbuf[1])
4786 PL_expect = XOPERATOR;
4787 PL_pending_ident = '#';
4791 PL_tokenbuf[0] = '$';
4792 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4793 sizeof PL_tokenbuf - 1, FALSE);
4794 if (PL_expect == XOPERATOR)
4796 if (!PL_tokenbuf[1]) {
4798 yyerror("Final $ should be \\$ or $name");
4802 /* This kludge not intended to be bulletproof. */
4803 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4804 yylval.opval = newSVOP(OP_CONST, 0,
4805 newSViv(CopARYBASE_get(&PL_compiling)));
4806 yylval.opval->op_private = OPpCONST_ARYBASE;
4812 const char tmp = *s;
4813 if (PL_lex_state == LEX_NORMAL)
4816 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4817 && intuit_more(s)) {
4819 PL_tokenbuf[0] = '@';
4820 if (ckWARN(WARN_SYNTAX)) {
4823 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4826 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4827 while (t < PL_bufend && *t != ']')
4829 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4830 "Multidimensional syntax %.*s not supported",
4831 (int)((t - PL_bufptr) + 1), PL_bufptr);
4835 else if (*s == '{') {
4837 PL_tokenbuf[0] = '%';
4838 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4839 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4841 char tmpbuf[sizeof PL_tokenbuf];
4844 } while (isSPACE(*t));
4845 if (isIDFIRST_lazy_if(t,UTF)) {
4847 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4851 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4852 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4853 "You need to quote \"%s\"",
4860 PL_expect = XOPERATOR;
4861 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4862 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4863 if (!islop || PL_last_lop_op == OP_GREPSTART)
4864 PL_expect = XOPERATOR;
4865 else if (strchr("$@\"'`q", *s))
4866 PL_expect = XTERM; /* e.g. print $fh "foo" */
4867 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4868 PL_expect = XTERM; /* e.g. print $fh &sub */
4869 else if (isIDFIRST_lazy_if(s,UTF)) {
4870 char tmpbuf[sizeof PL_tokenbuf];
4872 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4873 if ((t2 = keyword(tmpbuf, len, 0))) {
4874 /* binary operators exclude handle interpretations */
4886 PL_expect = XTERM; /* e.g. print $fh length() */
4891 PL_expect = XTERM; /* e.g. print $fh subr() */
4894 else if (isDIGIT(*s))
4895 PL_expect = XTERM; /* e.g. print $fh 3 */
4896 else if (*s == '.' && isDIGIT(s[1]))
4897 PL_expect = XTERM; /* e.g. print $fh .3 */
4898 else if ((*s == '?' || *s == '-' || *s == '+')
4899 && !isSPACE(s[1]) && s[1] != '=')
4900 PL_expect = XTERM; /* e.g. print $fh -1 */
4901 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4903 PL_expect = XTERM; /* e.g. print $fh /.../
4904 XXX except DORDOR operator
4906 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4908 PL_expect = XTERM; /* print $fh <<"EOF" */
4911 PL_pending_ident = '$';
4915 if (PL_expect == XOPERATOR)
4917 PL_tokenbuf[0] = '@';
4918 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4919 if (!PL_tokenbuf[1]) {
4922 if (PL_lex_state == LEX_NORMAL)
4924 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4926 PL_tokenbuf[0] = '%';
4928 /* Warn about @ where they meant $. */
4929 if (*s == '[' || *s == '{') {
4930 if (ckWARN(WARN_SYNTAX)) {
4931 const char *t = s + 1;
4932 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4934 if (*t == '}' || *t == ']') {
4936 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4937 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4938 "Scalar value %.*s better written as $%.*s",
4939 (int)(t-PL_bufptr), PL_bufptr,
4940 (int)(t-PL_bufptr-1), PL_bufptr+1);
4945 PL_pending_ident = '@';
4948 case '/': /* may be division, defined-or, or pattern */
4949 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4953 case '?': /* may either be conditional or pattern */
4954 if(PL_expect == XOPERATOR) {
4962 /* A // operator. */
4972 /* Disable warning on "study /blah/" */
4973 if (PL_oldoldbufptr == PL_last_uni
4974 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4975 || memNE(PL_last_uni, "study", 5)
4976 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4979 s = scan_pat(s,OP_MATCH);
4980 TERM(sublex_start());
4984 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4985 #ifdef PERL_STRICT_CR
4988 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4990 && (s == PL_linestart || s[-1] == '\n') )
4992 PL_lex_formbrack = 0;
4996 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5002 yylval.ival = OPf_SPECIAL;
5008 if (PL_expect != XOPERATOR)
5013 case '0': case '1': case '2': case '3': case '4':
5014 case '5': case '6': case '7': case '8': case '9':
5015 s = scan_num(s, &yylval);
5016 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5017 if (PL_expect == XOPERATOR)
5022 s = scan_str(s,!!PL_madskills,FALSE);
5023 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5024 if (PL_expect == XOPERATOR) {
5025 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5027 deprecate_old(commaless_variable_list);
5028 return REPORT(','); /* grandfather non-comma-format format */
5035 yylval.ival = OP_CONST;
5036 TERM(sublex_start());
5039 s = scan_str(s,!!PL_madskills,FALSE);
5040 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5041 if (PL_expect == XOPERATOR) {
5042 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5044 deprecate_old(commaless_variable_list);
5045 return REPORT(','); /* grandfather non-comma-format format */
5052 yylval.ival = OP_CONST;
5053 /* FIXME. I think that this can be const if char *d is replaced by
5054 more localised variables. */
5055 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5056 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5057 yylval.ival = OP_STRINGIFY;
5061 TERM(sublex_start());
5064 s = scan_str(s,!!PL_madskills,FALSE);
5065 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5066 if (PL_expect == XOPERATOR)
5067 no_op("Backticks",s);
5070 readpipe_override();
5071 TERM(sublex_start());
5075 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5076 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5078 if (PL_expect == XOPERATOR)
5079 no_op("Backslash",s);
5083 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5084 char *start = s + 2;
5085 while (isDIGIT(*start) || *start == '_')
5087 if (*start == '.' && isDIGIT(start[1])) {
5088 s = scan_num(s, &yylval);
5091 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5092 else if (!isALPHA(*start) && (PL_expect == XTERM
5093 || PL_expect == XREF || PL_expect == XSTATE
5094 || PL_expect == XTERMORDORDOR)) {
5095 /* XXX Use gv_fetchpvn rather than stomping on a const string */
5096 const char c = *start;
5099 gv = gv_fetchpv(s, 0, SVt_PVCV);
5102 s = scan_num(s, &yylval);
5109 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5151 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5153 /* Some keywords can be followed by any delimiter, including ':' */
5154 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5155 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5156 (PL_tokenbuf[0] == 'q' &&
5157 strchr("qwxr", PL_tokenbuf[1])))));
5159 /* x::* is just a word, unless x is "CORE" */
5160 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5164 while (d < PL_bufend && isSPACE(*d))
5165 d++; /* no comments skipped here, or s### is misparsed */
5167 /* Is this a label? */
5168 if (!tmp && PL_expect == XSTATE
5169 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5171 yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5176 /* Check for keywords */
5177 tmp = keyword(PL_tokenbuf, len, 0);
5179 /* Is this a word before a => operator? */
5180 if (*d == '=' && d[1] == '>') {
5183 = (OP*)newSVOP(OP_CONST, 0,
5184 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5185 yylval.opval->op_private = OPpCONST_BARE;
5189 if (tmp < 0) { /* second-class keyword? */
5190 GV *ogv = NULL; /* override (winner) */
5191 GV *hgv = NULL; /* hidden (loser) */
5192 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5194 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5197 if (GvIMPORTED_CV(gv))
5199 else if (! CvMETHOD(cv))
5203 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5204 (gv = *gvp) && isGV_with_GP(gv) &&
5205 GvCVu(gv) && GvIMPORTED_CV(gv))
5212 tmp = 0; /* overridden by import or by GLOBAL */
5215 && -tmp==KEY_lock /* XXX generalizable kludge */
5218 tmp = 0; /* any sub overrides "weak" keyword */
5220 else { /* no override */
5222 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5223 Perl_warner(aTHX_ packWARN(WARN_MISC),
5224 "dump() better written as CORE::dump()");
5228 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5229 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5230 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5231 "Ambiguous call resolved as CORE::%s(), %s",
5232 GvENAME(hgv), "qualify as such or use &");
5239 default: /* not a keyword */
5240 /* Trade off - by using this evil construction we can pull the
5241 variable gv into the block labelled keylookup. If not, then
5242 we have to give it function scope so that the goto from the
5243 earlier ':' case doesn't bypass the initialisation. */
5245 just_a_word_zero_gv:
5253 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5256 SV *nextPL_nextwhite = 0;
5260 /* Get the rest if it looks like a package qualifier */
5262 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5264 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5267 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5268 *s == '\'' ? "'" : "::");
5273 if (PL_expect == XOPERATOR) {
5274 if (PL_bufptr == PL_linestart) {
5275 CopLINE_dec(PL_curcop);
5276 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5277 CopLINE_inc(PL_curcop);
5280 no_op("Bareword",s);
5283 /* Look for a subroutine with this name in current package,
5284 unless name is "Foo::", in which case Foo is a bearword
5285 (and a package name). */
5287 if (len > 2 && !PL_madskills &&
5288 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5290 if (ckWARN(WARN_BAREWORD)
5291 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5292 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5293 "Bareword \"%s\" refers to nonexistent package",
5296 PL_tokenbuf[len] = '\0';
5302 /* Mustn't actually add anything to a symbol table.
5303 But also don't want to "initialise" any placeholder
5304 constants that might already be there into full
5305 blown PVGVs with attached PVCV. */
5306 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5307 GV_NOADD_NOINIT, SVt_PVCV);
5312 /* if we saw a global override before, get the right name */
5315 sv = newSVpvs("CORE::GLOBAL::");
5316 sv_catpv(sv,PL_tokenbuf);
5319 /* If len is 0, newSVpv does strlen(), which is correct.
5320 If len is non-zero, then it will be the true length,
5321 and so the scalar will be created correctly. */
5322 sv = newSVpv(PL_tokenbuf,len);
5325 if (PL_madskills && !PL_thistoken) {
5326 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5327 PL_thistoken = newSVpv(start,s - start);
5328 PL_realtokenstart = s - SvPVX(PL_linestr);
5332 /* Presume this is going to be a bareword of some sort. */
5335 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5336 yylval.opval->op_private = OPpCONST_BARE;
5337 /* UTF-8 package name? */
5338 if (UTF && !IN_BYTES &&
5339 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5342 /* And if "Foo::", then that's what it certainly is. */
5347 /* Do the explicit type check so that we don't need to force
5348 the initialisation of the symbol table to have a real GV.
5349 Beware - gv may not really be a PVGV, cv may not really be
5350 a PVCV, (because of the space optimisations that gv_init
5351 understands) But they're true if for this symbol there is
5352 respectively a typeglob and a subroutine.
5354 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5355 /* Real typeglob, so get the real subroutine: */
5357 /* A proxy for a subroutine in this package? */
5358 : SvOK(gv) ? (CV *) gv : NULL)
5361 /* See if it's the indirect object for a list operator. */
5363 if (PL_oldoldbufptr &&
5364 PL_oldoldbufptr < PL_bufptr &&
5365 (PL_oldoldbufptr == PL_last_lop
5366 || PL_oldoldbufptr == PL_last_uni) &&
5367 /* NO SKIPSPACE BEFORE HERE! */
5368 (PL_expect == XREF ||
5369 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5371 bool immediate_paren = *s == '(';
5373 /* (Now we can afford to cross potential line boundary.) */
5374 s = SKIPSPACE2(s,nextPL_nextwhite);
5376 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5379 /* Two barewords in a row may indicate method call. */
5381 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5382 (tmp = intuit_method(s, gv, cv)))
5385 /* If not a declared subroutine, it's an indirect object. */
5386 /* (But it's an indir obj regardless for sort.) */
5387 /* Also, if "_" follows a filetest operator, it's a bareword */
5390 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5392 (PL_last_lop_op != OP_MAPSTART &&
5393 PL_last_lop_op != OP_GREPSTART))))
5394 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5395 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5398 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5403 PL_expect = XOPERATOR;
5406 s = SKIPSPACE2(s,nextPL_nextwhite);
5407 PL_nextwhite = nextPL_nextwhite;
5412 /* Is this a word before a => operator? */
5413 if (*s == '=' && s[1] == '>' && !pkgname) {
5415 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5416 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5417 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5421 /* If followed by a paren, it's certainly a subroutine. */
5426 while (SPACE_OR_TAB(*d))
5428 if (*d == ')' && (sv = gv_const_sv(gv))) {
5435 PL_nextwhite = PL_thiswhite;
5438 start_force(PL_curforce);
5440 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5441 PL_expect = XOPERATOR;
5444 PL_nextwhite = nextPL_nextwhite;
5445 curmad('X', PL_thistoken);
5446 PL_thistoken = newSVpvs("");
5454 /* If followed by var or block, call it a method (unless sub) */
5456 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5457 PL_last_lop = PL_oldbufptr;
5458 PL_last_lop_op = OP_METHOD;
5462 /* If followed by a bareword, see if it looks like indir obj. */
5465 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5466 && (tmp = intuit_method(s, gv, cv)))
5469 /* Not a method, so call it a subroutine (if defined) */
5472 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5473 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5474 "Ambiguous use of -%s resolved as -&%s()",
5475 PL_tokenbuf, PL_tokenbuf);
5476 /* Check for a constant sub */
5477 if ((sv = gv_const_sv(gv))) {
5479 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5480 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5481 yylval.opval->op_private = 0;
5485 /* Resolve to GV now. */
5486 if (SvTYPE(gv) != SVt_PVGV) {
5487 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5488 assert (SvTYPE(gv) == SVt_PVGV);
5489 /* cv must have been some sort of placeholder, so
5490 now needs replacing with a real code reference. */
5494 op_free(yylval.opval);
5495 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5496 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5497 PL_last_lop = PL_oldbufptr;
5498 PL_last_lop_op = OP_ENTERSUB;
5499 /* Is there a prototype? */
5507 const char *proto = SvPV_const((SV*)cv, protolen);
5510 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5512 while (*proto == ';')
5514 if (*proto == '&' && *s == '{') {
5515 sv_setpv(PL_subname,
5518 "__ANON__" : "__ANON__::__ANON__"));
5525 PL_nextwhite = PL_thiswhite;
5528 start_force(PL_curforce);
5529 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5532 PL_nextwhite = nextPL_nextwhite;
5533 curmad('X', PL_thistoken);
5534 PL_thistoken = newSVpvs("");
5541 /* Guess harder when madskills require "best effort". */
5542 if (PL_madskills && (!gv || !GvCVu(gv))) {
5543 int probable_sub = 0;
5544 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5546 else if (isALPHA(*s)) {
5550 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5551 if (!keyword(tmpbuf, tmplen, 0))
5554 while (d < PL_bufend && isSPACE(*d))
5556 if (*d == '=' && d[1] == '>')
5561 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5562 op_free(yylval.opval);
5563 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5564 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5565 PL_last_lop = PL_oldbufptr;
5566 PL_last_lop_op = OP_ENTERSUB;
5567 PL_nextwhite = PL_thiswhite;
5569 start_force(PL_curforce);
5570 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5572 PL_nextwhite = nextPL_nextwhite;
5573 curmad('X', PL_thistoken);
5574 PL_thistoken = newSVpvs("");
5579 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5586 /* Call it a bare word */
5588 if (PL_hints & HINT_STRICT_SUBS)
5589 yylval.opval->op_private |= OPpCONST_STRICT;
5592 if (lastchar != '-') {
5593 if (ckWARN(WARN_RESERVED)) {
5597 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5598 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5605 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5606 && ckWARN_d(WARN_AMBIGUOUS)) {
5607 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5608 "Operator or semicolon missing before %c%s",
5609 lastchar, PL_tokenbuf);
5610 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5611 "Ambiguous use of %c resolved as operator %c",
5612 lastchar, lastchar);
5618 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5619 newSVpv(CopFILE(PL_curcop),0));
5623 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5624 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5627 case KEY___PACKAGE__:
5628 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5630 ? newSVhek(HvNAME_HEK(PL_curstash))
5637 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5638 const char *pname = "main";
5639 if (PL_tokenbuf[2] == 'D')
5640 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5641 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5645 GvIOp(gv) = newIO();
5646 IoIFP(GvIOp(gv)) = PL_rsfp;
5647 #if defined(HAS_FCNTL) && defined(F_SETFD)
5649 const int fd = PerlIO_fileno(PL_rsfp);
5650 fcntl(fd,F_SETFD,fd >= 3);
5653 /* Mark this internal pseudo-handle as clean */
5654 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5656 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5657 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5658 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5660 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5661 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5662 /* if the script was opened in binmode, we need to revert
5663 * it to text mode for compatibility; but only iff it has CRs
5664 * XXX this is a questionable hack at best. */
5665 if (PL_bufend-PL_bufptr > 2
5666 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5669 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5670 loc = PerlIO_tell(PL_rsfp);
5671 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5674 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5676 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5677 #endif /* NETWARE */
5678 #ifdef PERLIO_IS_STDIO /* really? */
5679 # if defined(__BORLANDC__)
5680 /* XXX see note in do_binmode() */
5681 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5685 PerlIO_seek(PL_rsfp, loc, 0);
5689 #ifdef PERLIO_LAYERS
5692 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5693 else if (PL_encoding) {
5700 XPUSHs(PL_encoding);
5702 call_method("name", G_SCALAR);
5706 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5707 Perl_form(aTHX_ ":encoding(%"SVf")",
5716 if (PL_realtokenstart >= 0) {
5717 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5719 PL_endwhite = newSVpvs("");
5720 sv_catsv(PL_endwhite, PL_thiswhite);
5722 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5723 PL_realtokenstart = -1;
5725 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5726 SvCUR(PL_endwhite))) != Nullch) ;
5741 if (PL_expect == XSTATE) {
5748 if (*s == ':' && s[1] == ':') {
5751 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5752 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5753 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5756 else if (tmp == KEY_require || tmp == KEY_do)
5757 /* that's a way to remember we saw "CORE::" */
5770 LOP(OP_ACCEPT,XTERM);
5776 LOP(OP_ATAN2,XTERM);
5782 LOP(OP_BINMODE,XTERM);
5785 LOP(OP_BLESS,XTERM);
5794 /* When 'use switch' is in effect, continue has a dual
5795 life as a control operator. */
5797 if (!FEATURE_IS_ENABLED("switch"))
5800 /* We have to disambiguate the two senses of
5801 "continue". If the next token is a '{' then
5802 treat it as the start of a continue block;
5803 otherwise treat it as a control operator.
5815 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5832 if (!PL_cryptseen) {
5833 PL_cryptseen = TRUE;
5837 LOP(OP_CRYPT,XTERM);
5840 LOP(OP_CHMOD,XTERM);
5843 LOP(OP_CHOWN,XTERM);
5846 LOP(OP_CONNECT,XTERM);
5865 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5866 if (orig_keyword == KEY_do) {
5875 PL_hints |= HINT_BLOCK_SCOPE;
5885 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5886 LOP(OP_DBMOPEN,XTERM);
5892 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5899 yylval.ival = CopLINE(PL_curcop);
5915 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5916 UNIBRACK(OP_ENTEREVAL);
5930 case KEY_endhostent:
5936 case KEY_endservent:
5939 case KEY_endprotoent:
5950 yylval.ival = CopLINE(PL_curcop);
5952 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5955 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5958 if ((PL_bufend - p) >= 3 &&
5959 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5961 else if ((PL_bufend - p) >= 4 &&
5962 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5965 if (isIDFIRST_lazy_if(p,UTF)) {
5966 p = scan_ident(p, PL_bufend,
5967 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5971 Perl_croak(aTHX_ "Missing $ on loop variable");
5973 s = SvPVX(PL_linestr) + soff;
5979 LOP(OP_FORMLINE,XTERM);
5985 LOP(OP_FCNTL,XTERM);
5991 LOP(OP_FLOCK,XTERM);
6000 LOP(OP_GREPSTART, XREF);
6003 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6018 case KEY_getpriority:
6019 LOP(OP_GETPRIORITY,XTERM);
6021 case KEY_getprotobyname:
6024 case KEY_getprotobynumber:
6025 LOP(OP_GPBYNUMBER,XTERM);
6027 case KEY_getprotoent:
6039 case KEY_getpeername:
6040 UNI(OP_GETPEERNAME);
6042 case KEY_gethostbyname:
6045 case KEY_gethostbyaddr:
6046 LOP(OP_GHBYADDR,XTERM);
6048 case KEY_gethostent:
6051 case KEY_getnetbyname:
6054 case KEY_getnetbyaddr:
6055 LOP(OP_GNBYADDR,XTERM);
6060 case KEY_getservbyname:
6061 LOP(OP_GSBYNAME,XTERM);
6063 case KEY_getservbyport:
6064 LOP(OP_GSBYPORT,XTERM);
6066 case KEY_getservent:
6069 case KEY_getsockname:
6070 UNI(OP_GETSOCKNAME);
6072 case KEY_getsockopt:
6073 LOP(OP_GSOCKOPT,XTERM);
6088 yylval.ival = CopLINE(PL_curcop);
6098 yylval.ival = CopLINE(PL_curcop);
6102 LOP(OP_INDEX,XTERM);
6108 LOP(OP_IOCTL,XTERM);
6120 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6152 LOP(OP_LISTEN,XTERM);
6161 s = scan_pat(s,OP_MATCH);
6162 TERM(sublex_start());
6165 LOP(OP_MAPSTART, XREF);
6168 LOP(OP_MKDIR,XTERM);
6171 LOP(OP_MSGCTL,XTERM);
6174 LOP(OP_MSGGET,XTERM);
6177 LOP(OP_MSGRCV,XTERM);
6180 LOP(OP_MSGSND,XTERM);
6185 PL_in_my = (U16)tmp;
6187 if (isIDFIRST_lazy_if(s,UTF)) {
6191 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6192 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6194 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6195 if (!PL_in_my_stash) {
6198 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6202 if (PL_madskills) { /* just add type to declarator token */
6203 sv_catsv(PL_thistoken, PL_nextwhite);
6205 sv_catpvn(PL_thistoken, start, s - start);
6213 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6220 s = tokenize_use(0, s);
6224 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6231 if (isIDFIRST_lazy_if(s,UTF)) {
6233 for (d = s; isALNUM_lazy_if(d,UTF);)
6235 for (t=d; isSPACE(*t);)
6237 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6239 && !(t[0] == '=' && t[1] == '>')
6241 int parms_len = (int)(d-s);
6242 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6243 "Precedence problem: open %.*s should be open(%.*s)",
6244 parms_len, s, parms_len, s);
6250 yylval.ival = OP_OR;
6260 LOP(OP_OPEN_DIR,XTERM);
6263 checkcomma(s,PL_tokenbuf,"filehandle");
6267 checkcomma(s,PL_tokenbuf,"filehandle");
6286 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6290 LOP(OP_PIPE_OP,XTERM);
6293 s = scan_str(s,!!PL_madskills,FALSE);
6296 yylval.ival = OP_CONST;
6297 TERM(sublex_start());
6303 s = scan_str(s,!!PL_madskills,FALSE);
6306 PL_expect = XOPERATOR;
6308 if (SvCUR(PL_lex_stuff)) {
6311 d = SvPV_force(PL_lex_stuff, len);
6313 for (; isSPACE(*d) && len; --len, ++d)
6318 if (!warned && ckWARN(WARN_QW)) {
6319 for (; !isSPACE(*d) && len; --len, ++d) {
6321 Perl_warner(aTHX_ packWARN(WARN_QW),
6322 "Possible attempt to separate words with commas");
6325 else if (*d == '#') {
6326 Perl_warner(aTHX_ packWARN(WARN_QW),
6327 "Possible attempt to put comments in qw() list");
6333 for (; !isSPACE(*d) && len; --len, ++d)
6336 sv = newSVpvn(b, d-b);
6337 if (DO_UTF8(PL_lex_stuff))
6339 words = append_elem(OP_LIST, words,
6340 newSVOP(OP_CONST, 0, tokeq(sv)));
6344 start_force(PL_curforce);
6345 NEXTVAL_NEXTTOKE.opval = words;
6350 SvREFCNT_dec(PL_lex_stuff);
6351 PL_lex_stuff = NULL;
6357 s = scan_str(s,!!PL_madskills,FALSE);
6360 yylval.ival = OP_STRINGIFY;
6361 if (SvIVX(PL_lex_stuff) == '\'')
6362 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6363 TERM(sublex_start());
6366 s = scan_pat(s,OP_QR);
6367 TERM(sublex_start());
6370 s = scan_str(s,!!PL_madskills,FALSE);
6373 readpipe_override();
6374 TERM(sublex_start());
6382 s = force_version(s, FALSE);
6384 else if (*s != 'v' || !isDIGIT(s[1])
6385 || (s = force_version(s, TRUE), *s == 'v'))
6387 *PL_tokenbuf = '\0';
6388 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6389 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6390 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6392 yyerror("<> should be quotes");
6394 if (orig_keyword == KEY_require) {
6402 PL_last_uni = PL_oldbufptr;
6403 PL_last_lop_op = OP_REQUIRE;
6405 return REPORT( (int)REQUIRE );
6411 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6415 LOP(OP_RENAME,XTERM);
6424 LOP(OP_RINDEX,XTERM);
6433 UNIDOR(OP_READLINE);
6436 UNIDOR(OP_BACKTICK);
6445 LOP(OP_REVERSE,XTERM);
6448 UNIDOR(OP_READLINK);
6456 TERM(sublex_start());
6458 TOKEN(1); /* force error */
6461 checkcomma(s,PL_tokenbuf,"filehandle");
6471 LOP(OP_SELECT,XTERM);
6477 LOP(OP_SEMCTL,XTERM);
6480 LOP(OP_SEMGET,XTERM);
6483 LOP(OP_SEMOP,XTERM);
6489 LOP(OP_SETPGRP,XTERM);
6491 case KEY_setpriority:
6492 LOP(OP_SETPRIORITY,XTERM);
6494 case KEY_sethostent:
6500 case KEY_setservent:
6503 case KEY_setprotoent:
6513 LOP(OP_SEEKDIR,XTERM);
6515 case KEY_setsockopt:
6516 LOP(OP_SSOCKOPT,XTERM);
6522 LOP(OP_SHMCTL,XTERM);
6525 LOP(OP_SHMGET,XTERM);
6528 LOP(OP_SHMREAD,XTERM);
6531 LOP(OP_SHMWRITE,XTERM);
6534 LOP(OP_SHUTDOWN,XTERM);
6543 LOP(OP_SOCKET,XTERM);
6545 case KEY_socketpair:
6546 LOP(OP_SOCKPAIR,XTERM);
6549 checkcomma(s,PL_tokenbuf,"subroutine name");
6551 if (*s == ';' || *s == ')') /* probably a close */
6552 Perl_croak(aTHX_ "sort is now a reserved word");
6554 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6558 LOP(OP_SPLIT,XTERM);
6561 LOP(OP_SPRINTF,XTERM);
6564 LOP(OP_SPLICE,XTERM);
6579 LOP(OP_SUBSTR,XTERM);
6585 char tmpbuf[sizeof PL_tokenbuf];
6586 SSize_t tboffset = 0;
6587 expectation attrful;
6588 bool have_name, have_proto;
6589 const int key = tmp;
6594 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6595 SV *subtoken = newSVpvn(tstart, s - tstart);
6599 s = SKIPSPACE2(s,tmpwhite);
6604 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6605 (*s == ':' && s[1] == ':'))
6612 attrful = XATTRBLOCK;
6613 /* remember buffer pos'n for later force_word */
6614 tboffset = s - PL_oldbufptr;
6615 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6618 nametoke = newSVpvn(s, d - s);
6620 if (memchr(tmpbuf, ':', len))
6621 sv_setpvn(PL_subname, tmpbuf, len);
6623 sv_setsv(PL_subname,PL_curstname);
6624 sv_catpvs(PL_subname,"::");
6625 sv_catpvn(PL_subname,tmpbuf,len);
6632 CURMAD('X', nametoke);
6633 CURMAD('_', tmpwhite);
6634 (void) force_word(PL_oldbufptr + tboffset, WORD,
6637 s = SKIPSPACE2(d,tmpwhite);
6644 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6645 PL_expect = XTERMBLOCK;
6646 attrful = XATTRTERM;
6647 sv_setpvn(PL_subname,"?",1);
6651 if (key == KEY_format) {
6653 PL_lex_formbrack = PL_lex_brackets + 1;
6655 PL_thistoken = subtoken;
6659 (void) force_word(PL_oldbufptr + tboffset, WORD,
6665 /* Look for a prototype */
6668 bool bad_proto = FALSE;
6669 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6671 s = scan_str(s,!!PL_madskills,FALSE);
6673 Perl_croak(aTHX_ "Prototype not terminated");
6674 /* strip spaces and check for bad characters */
6675 d = SvPVX(PL_lex_stuff);
6677 for (p = d; *p; ++p) {
6680 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6686 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6687 "Illegal character in prototype for %"SVf" : %s",
6688 SVfARG(PL_subname), d);
6689 SvCUR_set(PL_lex_stuff, tmp);
6694 CURMAD('q', PL_thisopen);
6695 CURMAD('_', tmpwhite);
6696 CURMAD('=', PL_thisstuff);
6697 CURMAD('Q', PL_thisclose);
6698 NEXTVAL_NEXTTOKE.opval =
6699 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6700 PL_lex_stuff = Nullsv;
6703 s = SKIPSPACE2(s,tmpwhite);
6711 if (*s == ':' && s[1] != ':')
6712 PL_expect = attrful;
6713 else if (*s != '{' && key == KEY_sub) {
6715 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6717 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6724 curmad('^', newSVpvs(""));
6725 CURMAD('_', tmpwhite);
6729 PL_thistoken = subtoken;
6732 NEXTVAL_NEXTTOKE.opval =
6733 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6734 PL_lex_stuff = NULL;
6739 sv_setpv(PL_subname,
6741 (PL_curstash ? "__ANON__" : "__ANON__::__ANON__"));
6745 (void) force_word(PL_oldbufptr + tboffset, WORD,
6754 LOP(OP_SYSTEM,XREF);
6757 LOP(OP_SYMLINK,XTERM);
6760 LOP(OP_SYSCALL,XTERM);
6763 LOP(OP_SYSOPEN,XTERM);
6766 LOP(OP_SYSSEEK,XTERM);
6769 LOP(OP_SYSREAD,XTERM);
6772 LOP(OP_SYSWRITE,XTERM);
6776 TERM(sublex_start());
6797 LOP(OP_TRUNCATE,XTERM);
6809 yylval.ival = CopLINE(PL_curcop);
6813 yylval.ival = CopLINE(PL_curcop);
6817 LOP(OP_UNLINK,XTERM);
6823 LOP(OP_UNPACK,XTERM);
6826 LOP(OP_UTIME,XTERM);
6832 LOP(OP_UNSHIFT,XTERM);
6835 s = tokenize_use(1, s);
6845 yylval.ival = CopLINE(PL_curcop);
6849 yylval.ival = CopLINE(PL_curcop);
6853 PL_hints |= HINT_BLOCK_SCOPE;
6860 LOP(OP_WAITPID,XTERM);
6869 ctl_l[0] = toCTRL('L');
6871 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6874 /* Make sure $^L is defined */
6875 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6880 if (PL_expect == XOPERATOR)
6886 yylval.ival = OP_XOR;
6891 TERM(sublex_start());
6896 #pragma segment Main
6900 S_pending_ident(pTHX)
6905 /* pit holds the identifier we read and pending_ident is reset */
6906 char pit = PL_pending_ident;
6907 PL_pending_ident = 0;
6909 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6910 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6911 "### Pending identifier '%s'\n", PL_tokenbuf); });
6913 /* if we're in a my(), we can't allow dynamics here.
6914 $foo'bar has already been turned into $foo::bar, so
6915 just check for colons.
6917 if it's a legal name, the OP is a PADANY.
6920 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6921 if (strchr(PL_tokenbuf,':'))
6922 yyerror(Perl_form(aTHX_ "No package name allowed for "
6923 "variable %s in \"our\"",
6925 tmp = allocmy(PL_tokenbuf);
6928 if (strchr(PL_tokenbuf,':'))
6929 yyerror(Perl_form(aTHX_ PL_no_myglob,
6930 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6932 yylval.opval = newOP(OP_PADANY, 0);
6933 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6939 build the ops for accesses to a my() variable.
6941 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6942 then used in a comparison. This catches most, but not
6943 all cases. For instance, it catches
6944 sort { my($a); $a <=> $b }
6946 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6947 (although why you'd do that is anyone's guess).
6950 if (!strchr(PL_tokenbuf,':')) {
6952 tmp = pad_findmy(PL_tokenbuf);
6953 if (tmp != NOT_IN_PAD) {
6954 /* might be an "our" variable" */
6955 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6956 /* build ops for a bareword */
6957 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6958 HEK * const stashname = HvNAME_HEK(stash);
6959 SV * const sym = newSVhek(stashname);
6960 sv_catpvs(sym, "::");
6961 sv_catpv(sym, PL_tokenbuf+1);
6962 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6963 yylval.opval->op_private = OPpCONST_ENTERED;
6966 ? (GV_ADDMULTI | GV_ADDINEVAL)
6969 ((PL_tokenbuf[0] == '$') ? SVt_PV
6970 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6975 /* if it's a sort block and they're naming $a or $b */
6976 if (PL_last_lop_op == OP_SORT &&
6977 PL_tokenbuf[0] == '$' &&
6978 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6981 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6982 d < PL_bufend && *d != '\n';
6985 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6986 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6992 yylval.opval = newOP(OP_PADANY, 0);
6993 yylval.opval->op_targ = tmp;
6999 Whine if they've said @foo in a doublequoted string,
7000 and @foo isn't a variable we can find in the symbol
7003 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7004 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
7005 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7006 && ckWARN(WARN_AMBIGUOUS)
7007 /* DO NOT warn for @- and @+ */
7008 && !( PL_tokenbuf[2] == '\0' &&
7009 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7012 /* Downgraded from fatal to warning 20000522 mjd */
7013 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7014 "Possible unintended interpolation of %s in string",
7019 /* build ops for a bareword */
7020 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
7021 yylval.opval->op_private = OPpCONST_ENTERED;
7024 /* If the identifier refers to a stash, don't autovivify it.
7025 * Change 24660 had the side effect of causing symbol table
7026 * hashes to always be defined, even if they were freshly
7027 * created and the only reference in the entire program was
7028 * the single statement with the defined %foo::bar:: test.
7029 * It appears that all code in the wild doing this actually
7030 * wants to know whether sub-packages have been loaded, so
7031 * by avoiding auto-vivifying symbol tables, we ensure that
7032 * defined %foo::bar:: continues to be false, and the existing
7033 * tests still give the expected answers, even though what
7034 * they're actually testing has now changed subtly.
7036 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
7038 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7039 ((PL_tokenbuf[0] == '$') ? SVt_PV
7040 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7046 * The following code was generated by perl_keyword.pl.
7050 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7055 case 1: /* 5 tokens of length 1 */
7087 case 2: /* 18 tokens of length 2 */
7233 case 3: /* 29 tokens of length 3 */
7237 if (name[1] == 'N' &&
7300 if (name[1] == 'i' &&
7332 if (name[1] == 'o' &&
7341 if (name[1] == 'e' &&
7350 if (name[1] == 'n' &&
7359 if (name[1] == 'o' &&
7368 if (name[1] == 'a' &&
7377 if (name[1] == 'o' &&
7439 if (name[1] == 'e' &&
7453 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7479 if (name[1] == 'i' &&
7488 if (name[1] == 's' &&
7497 if (name[1] == 'e' &&
7506 if (name[1] == 'o' &&
7518 case 4: /* 41 tokens of length 4 */
7522 if (name[1] == 'O' &&
7532 if (name[1] == 'N' &&
7542 if (name[1] == 'i' &&
7552 if (name[1] == 'h' &&
7562 if (name[1] == 'u' &&
7575 if (name[2] == 'c' &&
7584 if (name[2] == 's' &&
7593 if (name[2] == 'a' &&
7629 if (name[1] == 'o' &&
7642 if (name[2] == 't' &&
7651 if (name[2] == 'o' &&
7660 if (name[2] == 't' &&
7669 if (name[2] == 'e' &&
7682 if (name[1] == 'o' &&
7695 if (name[2] == 'y' &&
7704 if (name[2] == 'l' &&
7720 if (name[2] == 's' &&
7729 if (name[2] == 'n' &&
7738 if (name[2] == 'c' &&
7751 if (name[1] == 'e' &&
7761 if (name[1] == 'p' &&
7774 if (name[2] == 'c' &&
7783 if (name[2] == 'p' &&
7792 if (name[2] == 's' &&
7808 if (name[2] == 'n' &&
7878 if (name[2] == 'r' &&
7887 if (name[2] == 'r' &&
7896 if (name[2] == 'a' &&
7912 if (name[2] == 'l' &&
7974 if (name[2] == 'e' &&
7977 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7990 case 5: /* 39 tokens of length 5 */
7994 if (name[1] == 'E' &&
8005 if (name[1] == 'H' &&
8019 if (name[2] == 'a' &&
8029 if (name[2] == 'a' &&
8046 if (name[2] == 'e' &&
8056 if (name[2] == 'e' &&
8060 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8076 if (name[3] == 'i' &&
8085 if (name[3] == 'o' &&
8121 if (name[2] == 'o' &&
8131 if (name[2] == 'y' &&
8145 if (name[1] == 'l' &&
8159 if (name[2] == 'n' &&
8169 if (name[2] == 'o' &&
8183 if (name[1] == 'i' &&
8188 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8197 if (name[2] == 'd' &&
8207 if (name[2] == 'c' &&
8224 if (name[2] == 'c' &&
8234 if (name[2] == 't' &&
8248 if (name[1] == 'k' &&
8259 if (name[1] == 'r' &&
8273 if (name[2] == 's' &&
8283 if (name[2] == 'd' &&
8300 if (name[2] == 'm' &&
8310 if (name[2] == 'i' &&
8320 if (name[2] == 'e' &&
8330 if (name[2] == 'l' &&
8340 if (name[2] == 'a' &&
8353 if (name[3] == 't' &&
8356 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8362 if (name[3] == 'd' &&
8379 if (name[1] == 'i' &&
8393 if (name[2] == 'a' &&
8406 if (name[3] == 'e' &&
8441 if (name[2] == 'i' &&
8458 if (name[2] == 'i' &&
8468 if (name[2] == 'i' &&
8485 case 6: /* 33 tokens of length 6 */
8489 if (name[1] == 'c' &&
8504 if (name[2] == 'l' &&
8515 if (name[2] == 'r' &&
8530 if (name[1] == 'e' &&
8545 if (name[2] == 's' &&
8550 if(ckWARN_d(WARN_SYNTAX))
8551 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8557 if (name[2] == 'i' &&
8575 if (name[2] == 'l' &&
8586 if (name[2] == 'r' &&
8601 if (name[1] == 'm' &&
8616 if (name[2] == 'n' &&
8627 if (name[2] == 's' &&
8642 if (name[1] == 's' &&
8648 if (name[4] == 't' &&
8657 if (name[4] == 'e' &&
8666 if (name[4] == 'c' &&
8675 if (name[4] == 'n' &&
8691 if (name[1] == 'r' &&
8709 if (name[3] == 'a' &&
8719 if (name[3] == 'u' &&
8733 if (name[2] == 'n' &&
8751 if (name[2] == 'a' &&
8765 if (name[3] == 'e' &&
8778 if (name[4] == 't' &&
8787 if (name[4] == 'e' &&
8809 if (name[4] == 't' &&
8818 if (name[4] == 'e' &&
8834 if (name[2] == 'c' &&
8845 if (name[2] == 'l' &&
8856 if (name[2] == 'b' &&
8867 if (name[2] == 's' &&
8890 if (name[4] == 's' &&
8899 if (name[4] == 'n' &&
8912 if (name[3] == 'a' &&
8929 if (name[1] == 'a' &&
8944 case 7: /* 29 tokens of length 7 */
8948 if (name[1] == 'E' &&
8961 if (name[1] == '_' &&
8974 if (name[1] == 'i' &&
8981 return -KEY_binmode;
8987 if (name[1] == 'o' &&
8994 return -KEY_connect;
9003 if (name[2] == 'm' &&
9009 return -KEY_dbmopen;
9020 if (name[4] == 'u' &&
9024 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9030 if (name[4] == 'n' &&
9051 if (name[1] == 'o' &&
9064 if (name[1] == 'e' &&
9071 if (name[5] == 'r' &&
9074 return -KEY_getpgrp;
9080 if (name[5] == 'i' &&
9083 return -KEY_getppid;
9096 if (name[1] == 'c' &&
9103 return -KEY_lcfirst;
9109 if (name[1] == 'p' &&
9116 return -KEY_opendir;
9122 if (name[1] == 'a' &&
9140 if (name[3] == 'd' &&
9145 return -KEY_readdir;
9151 if (name[3] == 'u' &&
9162 if (name[3] == 'e' &&
9167 return -KEY_reverse;
9186 if (name[3] == 'k' &&
9191 return -KEY_seekdir;
9197 if (name[3] == 'p' &&
9202 return -KEY_setpgrp;
9212 if (name[2] == 'm' &&
9218 return -KEY_shmread;
9224 if (name[2] == 'r' &&
9230 return -KEY_sprintf;
9239 if (name[3] == 'l' &&
9244 return -KEY_symlink;
9253 if (name[4] == 'a' &&
9257 return -KEY_syscall;
9263 if (name[4] == 'p' &&
9267 return -KEY_sysopen;
9273 if (name[4] == 'e' &&
9277 return -KEY_sysread;
9283 if (name[4] == 'e' &&
9287 return -KEY_sysseek;
9305 if (name[1] == 'e' &&
9312 return -KEY_telldir;
9321 if (name[2] == 'f' &&
9327 return -KEY_ucfirst;
9333 if (name[2] == 's' &&
9339 return -KEY_unshift;
9349 if (name[1] == 'a' &&
9356 return -KEY_waitpid;
9365 case 8: /* 26 tokens of length 8 */
9369 if (name[1] == 'U' &&
9377 return KEY_AUTOLOAD;
9388 if (name[3] == 'A' &&
9394 return KEY___DATA__;
9400 if (name[3] == 'I' &&
9406 return -KEY___FILE__;
9412 if (name[3] == 'I' &&
9418 return -KEY___LINE__;
9434 if (name[2] == 'o' &&
9441 return -KEY_closedir;
9447 if (name[2] == 'n' &&
9454 return -KEY_continue;
9464 if (name[1] == 'b' &&
9472 return -KEY_dbmclose;
9478 if (name[1] == 'n' &&
9484 if (name[4] == 'r' &&
9489 return -KEY_endgrent;
9495 if (name[4] == 'w' &&
9500 return -KEY_endpwent;
9513 if (name[1] == 'o' &&
9521 return -KEY_formline;
9527 if (name[1] == 'e' &&
9538 if (name[6] == 'n' &&
9541 return -KEY_getgrent;
9547 if (name[6] == 'i' &&
9550 return -KEY_getgrgid;
9556 if (name[6] == 'a' &&
9559 return -KEY_getgrnam;
9572 if (name[4] == 'o' &&
9577 return -KEY_getlogin;
9588 if (name[6] == 'n' &&
9591 return -KEY_getpwent;
9597 if (name[6] == 'a' &&
9600 return -KEY_getpwnam;
9606 if (name[6] == 'i' &&
9609 return -KEY_getpwuid;
9629 if (name[1] == 'e' &&
9636 if (name[5] == 'i' &&
9643 return -KEY_readline;
9648 return -KEY_readlink;
9659 if (name[5] == 'i' &&
9663 return -KEY_readpipe;
9684 if (name[4] == 'r' &&
9689 return -KEY_setgrent;
9695 if (name[4] == 'w' &&
9700 return -KEY_setpwent;
9716 if (name[3] == 'w' &&
9722 return -KEY_shmwrite;
9728 if (name[3] == 't' &&
9734 return -KEY_shutdown;
9744 if (name[2] == 's' &&
9751 return -KEY_syswrite;
9761 if (name[1] == 'r' &&
9769 return -KEY_truncate;
9778 case 9: /* 9 tokens of length 9 */
9782 if (name[1] == 'N' &&
9791 return KEY_UNITCHECK;
9797 if (name[1] == 'n' &&
9806 return -KEY_endnetent;
9812 if (name[1] == 'e' &&
9821 return -KEY_getnetent;
9827 if (name[1] == 'o' &&
9836 return -KEY_localtime;
9842 if (name[1] == 'r' &&
9851 return KEY_prototype;
9857 if (name[1] == 'u' &&
9866 return -KEY_quotemeta;
9872 if (name[1] == 'e' &&
9881 return -KEY_rewinddir;
9887 if (name[1] == 'e' &&
9896 return -KEY_setnetent;
9902 if (name[1] == 'a' &&
9911 return -KEY_wantarray;
9920 case 10: /* 9 tokens of length 10 */
9924 if (name[1] == 'n' &&
9930 if (name[4] == 'o' &&
9937 return -KEY_endhostent;
9943 if (name[4] == 'e' &&
9950 return -KEY_endservent;
9963 if (name[1] == 'e' &&
9969 if (name[4] == 'o' &&
9976 return -KEY_gethostent;
9985 if (name[5] == 'r' &&
9991 return -KEY_getservent;
9997 if (name[5] == 'c' &&
10003 return -KEY_getsockopt;
10023 if (name[2] == 't')
10028 if (name[4] == 'o' &&
10035 return -KEY_sethostent;
10044 if (name[5] == 'r' &&
10050 return -KEY_setservent;
10056 if (name[5] == 'c' &&
10062 return -KEY_setsockopt;
10079 if (name[2] == 'c' &&
10088 return -KEY_socketpair;
10101 case 11: /* 8 tokens of length 11 */
10105 if (name[1] == '_' &&
10115 { /* __PACKAGE__ */
10116 return -KEY___PACKAGE__;
10122 if (name[1] == 'n' &&
10132 { /* endprotoent */
10133 return -KEY_endprotoent;
10139 if (name[1] == 'e' &&
10148 if (name[5] == 'e' &&
10154 { /* getpeername */
10155 return -KEY_getpeername;
10164 if (name[6] == 'o' &&
10169 { /* getpriority */
10170 return -KEY_getpriority;
10176 if (name[6] == 't' &&
10181 { /* getprotoent */
10182 return -KEY_getprotoent;
10196 if (name[4] == 'o' &&
10203 { /* getsockname */
10204 return -KEY_getsockname;
10217 if (name[1] == 'e' &&
10225 if (name[6] == 'o' &&
10230 { /* setpriority */
10231 return -KEY_setpriority;
10237 if (name[6] == 't' &&
10242 { /* setprotoent */
10243 return -KEY_setprotoent;
10259 case 12: /* 2 tokens of length 12 */
10260 if (name[0] == 'g' &&
10272 if (name[9] == 'd' &&
10275 { /* getnetbyaddr */
10276 return -KEY_getnetbyaddr;
10282 if (name[9] == 'a' &&
10285 { /* getnetbyname */
10286 return -KEY_getnetbyname;
10298 case 13: /* 4 tokens of length 13 */
10299 if (name[0] == 'g' &&
10306 if (name[4] == 'o' &&
10315 if (name[10] == 'd' &&
10318 { /* gethostbyaddr */
10319 return -KEY_gethostbyaddr;
10325 if (name[10] == 'a' &&
10328 { /* gethostbyname */
10329 return -KEY_gethostbyname;
10342 if (name[4] == 'e' &&
10351 if (name[10] == 'a' &&
10354 { /* getservbyname */
10355 return -KEY_getservbyname;
10361 if (name[10] == 'o' &&
10364 { /* getservbyport */
10365 return -KEY_getservbyport;
10384 case 14: /* 1 tokens of length 14 */
10385 if (name[0] == 'g' &&
10399 { /* getprotobyname */
10400 return -KEY_getprotobyname;
10405 case 16: /* 1 tokens of length 16 */
10406 if (name[0] == 'g' &&
10422 { /* getprotobynumber */
10423 return -KEY_getprotobynumber;
10437 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10441 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10442 if (ckWARN(WARN_SYNTAX)) {
10445 for (w = s+2; *w && level; w++) {
10448 else if (*w == ')')
10451 while (isSPACE(*w))
10453 /* the list of chars below is for end of statements or
10454 * block / parens, boolean operators (&&, ||, //) and branch
10455 * constructs (or, and, if, until, unless, while, err, for).
10456 * Not a very solid hack... */
10457 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10458 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10459 "%s (...) interpreted as function",name);
10462 while (s < PL_bufend && isSPACE(*s))
10466 while (s < PL_bufend && isSPACE(*s))
10468 if (isIDFIRST_lazy_if(s,UTF)) {
10469 const char * const w = s++;
10470 while (isALNUM_lazy_if(s,UTF))
10472 while (s < PL_bufend && isSPACE(*s))
10476 if (keyword(w, s - w, 0))
10479 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10480 if (gv && GvCVu(gv))
10482 Perl_croak(aTHX_ "No comma allowed after %s", what);
10487 /* Either returns sv, or mortalizes sv and returns a new SV*.
10488 Best used as sv=new_constant(..., sv, ...).
10489 If s, pv are NULL, calls subroutine with one argument,
10490 and type is used with error messages only. */
10493 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10497 HV * const table = GvHV(PL_hintgv); /* ^H */
10501 const char *why1 = "", *why2 = "", *why3 = "";
10503 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10506 why2 = (const char *)
10507 (strEQ(key,"charnames")
10508 ? "(possibly a missing \"use charnames ...\")"
10510 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10511 (type ? type: "undef"), why2);
10513 /* This is convoluted and evil ("goto considered harmful")
10514 * but I do not understand the intricacies of all the different
10515 * failure modes of %^H in here. The goal here is to make
10516 * the most probable error message user-friendly. --jhi */
10521 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10522 (type ? type: "undef"), why1, why2, why3);
10524 yyerror(SvPVX_const(msg));
10528 cvp = hv_fetch(table, key, strlen(key), FALSE);
10529 if (!cvp || !SvOK(*cvp)) {
10532 why3 = "} is not defined";
10535 sv_2mortal(sv); /* Parent created it permanently */
10538 pv = sv_2mortal(newSVpvn(s, len));
10540 typesv = sv_2mortal(newSVpv(type, 0));
10542 typesv = &PL_sv_undef;
10544 PUSHSTACKi(PERLSI_OVERLOAD);
10556 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10560 /* Check the eval first */
10561 if (!PL_in_eval && SvTRUE(ERRSV)) {
10562 sv_catpvs(ERRSV, "Propagated");
10563 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10565 res = SvREFCNT_inc_simple(sv);
10569 SvREFCNT_inc_simple_void(res);
10578 why1 = "Call to &{$^H{";
10580 why3 = "}} did not return a defined value";
10588 /* Returns a NUL terminated string, with the length of the string written to
10592 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10595 register char *d = dest;
10596 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10599 Perl_croak(aTHX_ ident_too_long);
10600 if (isALNUM(*s)) /* UTF handled below */
10602 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10607 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10611 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10612 char *t = s + UTF8SKIP(s);
10614 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10618 Perl_croak(aTHX_ ident_too_long);
10619 Copy(s, d, len, char);
10632 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10635 char *bracket = NULL;
10637 register char *d = dest;
10638 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10643 while (isDIGIT(*s)) {
10645 Perl_croak(aTHX_ ident_too_long);
10652 Perl_croak(aTHX_ ident_too_long);
10653 if (isALNUM(*s)) /* UTF handled below */
10655 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10660 else if (*s == ':' && s[1] == ':') {
10664 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10665 char *t = s + UTF8SKIP(s);
10666 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10668 if (d + (t - s) > e)
10669 Perl_croak(aTHX_ ident_too_long);
10670 Copy(s, d, t - s, char);
10681 if (PL_lex_state != LEX_NORMAL)
10682 PL_lex_state = LEX_INTERPENDMAYBE;
10685 if (*s == '$' && s[1] &&
10686 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10699 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10704 if (isSPACE(s[-1])) {
10706 const char ch = *s++;
10707 if (!SPACE_OR_TAB(ch)) {
10713 if (isIDFIRST_lazy_if(d,UTF)) {
10717 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10718 end += UTF8SKIP(end);
10719 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10720 end += UTF8SKIP(end);
10722 Copy(s, d, end - s, char);
10727 while ((isALNUM(*s) || *s == ':') && d < e)
10730 Perl_croak(aTHX_ ident_too_long);
10733 while (s < send && SPACE_OR_TAB(*s))
10735 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10736 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10737 const char * const brack =
10739 ((*s == '[') ? "[...]" : "{...}");
10740 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10741 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10742 funny, dest, brack, funny, dest, brack);
10745 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10749 /* Handle extended ${^Foo} variables
10750 * 1999-02-27 mjd-perl-patch@plover.com */
10751 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10755 while (isALNUM(*s) && d < e) {
10759 Perl_croak(aTHX_ ident_too_long);
10764 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10765 PL_lex_state = LEX_INTERPEND;
10768 if (PL_lex_state == LEX_NORMAL) {
10769 if (ckWARN(WARN_AMBIGUOUS) &&
10770 (keyword(dest, d - dest, 0)
10771 || get_cvn_flags(dest, d - dest, 0)))
10775 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10776 "Ambiguous use of %c{%s} resolved to %c%s",
10777 funny, dest, funny, dest);
10782 s = bracket; /* let the parser handle it */
10786 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10787 PL_lex_state = LEX_INTERPEND;
10792 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10794 PERL_UNUSED_CONTEXT;
10798 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10799 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10800 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10801 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10802 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10808 S_scan_pat(pTHX_ char *start, I32 type)
10812 char *s = scan_str(start,!!PL_madskills,FALSE);
10813 const char * const valid_flags =
10814 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10821 const char * const delimiter = skipspace(start);
10825 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10826 : "Search pattern not terminated" ));
10829 pm = (PMOP*)newPMOP(type, 0);
10830 if (PL_multi_open == '?') {
10831 /* This is the only point in the code that sets PMf_ONCE: */
10832 pm->op_pmflags |= PMf_ONCE;
10834 /* Hence it's safe to do this bit of PMOP book-keeping here, which
10835 allows us to restrict the list needed by reset to just the ??
10837 assert(type != OP_TRANS);
10839 MAGIC *mg = mg_find((SV*)PL_curstash, PERL_MAGIC_symtab);
10842 mg = sv_magicext((SV*)PL_curstash, 0, PERL_MAGIC_symtab, 0, 0,
10845 elements = mg->mg_len / sizeof(PMOP**);
10846 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
10847 ((PMOP**)mg->mg_ptr) [elements++] = pm;
10848 mg->mg_len = elements * sizeof(PMOP**);
10849 PmopSTASH_set(pm,PL_curstash);
10855 while (*s && strchr(valid_flags, *s))
10856 pmflag(&pm->op_pmflags,*s++);
10858 if (PL_madskills && modstart != s) {
10859 SV* tmptoken = newSVpvn(modstart, s - modstart);
10860 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10863 /* issue a warning if /c is specified,but /g is not */
10864 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10865 && ckWARN(WARN_REGEXP))
10867 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
10868 "Use of /c modifier is meaningless without /g" );
10871 PL_lex_op = (OP*)pm;
10872 yylval.ival = OP_MATCH;
10877 S_scan_subst(pTHX_ char *start)
10888 yylval.ival = OP_NULL;
10890 s = scan_str(start,!!PL_madskills,FALSE);
10893 Perl_croak(aTHX_ "Substitution pattern not terminated");
10895 if (s[-1] == PL_multi_open)
10898 if (PL_madskills) {
10899 CURMAD('q', PL_thisopen);
10900 CURMAD('_', PL_thiswhite);
10901 CURMAD('E', PL_thisstuff);
10902 CURMAD('Q', PL_thisclose);
10903 PL_realtokenstart = s - SvPVX(PL_linestr);
10907 first_start = PL_multi_start;
10908 s = scan_str(s,!!PL_madskills,FALSE);
10910 if (PL_lex_stuff) {
10911 SvREFCNT_dec(PL_lex_stuff);
10912 PL_lex_stuff = NULL;
10914 Perl_croak(aTHX_ "Substitution replacement not terminated");
10916 PL_multi_start = first_start; /* so whole substitution is taken together */
10918 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10921 if (PL_madskills) {
10922 CURMAD('z', PL_thisopen);
10923 CURMAD('R', PL_thisstuff);
10924 CURMAD('Z', PL_thisclose);
10930 if (*s == EXEC_PAT_MOD) {
10934 else if (strchr(S_PAT_MODS, *s))
10935 pmflag(&pm->op_pmflags,*s++);
10941 if (PL_madskills) {
10943 curmad('m', newSVpvn(modstart, s - modstart));
10944 append_madprops(PL_thismad, (OP*)pm, 0);
10948 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10949 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10953 SV * const repl = newSVpvs("");
10955 PL_sublex_info.super_bufptr = s;
10956 PL_sublex_info.super_bufend = PL_bufend;
10958 pm->op_pmflags |= PMf_EVAL;
10960 sv_catpv(repl, (const char *)(es ? "eval " : "do "));
10961 sv_catpvs(repl, "{");
10962 sv_catsv(repl, PL_lex_repl);
10963 if (strchr(SvPVX(PL_lex_repl), '#'))
10964 sv_catpvs(repl, "\n");
10965 sv_catpvs(repl, "}");
10967 SvREFCNT_dec(PL_lex_repl);
10968 PL_lex_repl = repl;
10971 PL_lex_op = (OP*)pm;
10972 yylval.ival = OP_SUBST;
10977 S_scan_trans(pTHX_ char *start)
10990 yylval.ival = OP_NULL;
10992 s = scan_str(start,!!PL_madskills,FALSE);
10994 Perl_croak(aTHX_ "Transliteration pattern not terminated");
10996 if (s[-1] == PL_multi_open)
10999 if (PL_madskills) {
11000 CURMAD('q', PL_thisopen);
11001 CURMAD('_', PL_thiswhite);
11002 CURMAD('E', PL_thisstuff);
11003 CURMAD('Q', PL_thisclose);
11004 PL_realtokenstart = s - SvPVX(PL_linestr);
11008 s = scan_str(s,!!PL_madskills,FALSE);
11010 if (PL_lex_stuff) {
11011 SvREFCNT_dec(PL_lex_stuff);
11012 PL_lex_stuff = NULL;
11014 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11016 if (PL_madskills) {
11017 CURMAD('z', PL_thisopen);
11018 CURMAD('R', PL_thisstuff);
11019 CURMAD('Z', PL_thisclose);
11022 complement = del = squash = 0;
11029 complement = OPpTRANS_COMPLEMENT;
11032 del = OPpTRANS_DELETE;
11035 squash = OPpTRANS_SQUASH;
11044 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11045 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11046 o->op_private &= ~OPpTRANS_ALL;
11047 o->op_private |= del|squash|complement|
11048 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11049 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11052 yylval.ival = OP_TRANS;
11055 if (PL_madskills) {
11057 curmad('m', newSVpvn(modstart, s - modstart));
11058 append_madprops(PL_thismad, o, 0);
11067 S_scan_heredoc(pTHX_ register char *s)
11071 I32 op_type = OP_SCALAR;
11075 const char *found_newline;
11079 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11081 I32 stuffstart = s - SvPVX(PL_linestr);
11084 PL_realtokenstart = -1;
11089 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11093 while (SPACE_OR_TAB(*peek))
11095 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11098 s = delimcpy(d, e, s, PL_bufend, term, &len);
11108 if (!isALNUM_lazy_if(s,UTF))
11109 deprecate_old("bare << to mean <<\"\"");
11110 for (; isALNUM_lazy_if(s,UTF); s++) {
11115 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11116 Perl_croak(aTHX_ "Delimiter for here document is too long");
11119 len = d - PL_tokenbuf;
11122 if (PL_madskills) {
11123 tstart = PL_tokenbuf + !outer;
11124 PL_thisclose = newSVpvn(tstart, len - !outer);
11125 tstart = SvPVX(PL_linestr) + stuffstart;
11126 PL_thisopen = newSVpvn(tstart, s - tstart);
11127 stuffstart = s - SvPVX(PL_linestr);
11130 #ifndef PERL_STRICT_CR
11131 d = strchr(s, '\r');
11133 char * const olds = s;
11135 while (s < PL_bufend) {
11141 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11150 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11157 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11158 herewas = newSVpvn(s,PL_bufend-s);
11162 herewas = newSVpvn(s-1,found_newline-s+1);
11165 herewas = newSVpvn(s,found_newline-s);
11169 if (PL_madskills) {
11170 tstart = SvPVX(PL_linestr) + stuffstart;
11172 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11174 PL_thisstuff = newSVpvn(tstart, s - tstart);
11177 s += SvCUR(herewas);
11180 stuffstart = s - SvPVX(PL_linestr);
11186 tmpstr = newSV_type(SVt_PVIV);
11187 SvGROW(tmpstr, 80);
11188 if (term == '\'') {
11189 op_type = OP_CONST;
11190 SvIV_set(tmpstr, -1);
11192 else if (term == '`') {
11193 op_type = OP_BACKTICK;
11194 SvIV_set(tmpstr, '\\');
11198 PL_multi_start = CopLINE(PL_curcop);
11199 PL_multi_open = PL_multi_close = '<';
11200 term = *PL_tokenbuf;
11201 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11202 char * const bufptr = PL_sublex_info.super_bufptr;
11203 char * const bufend = PL_sublex_info.super_bufend;
11204 char * const olds = s - SvCUR(herewas);
11205 s = strchr(bufptr, '\n');
11209 while (s < bufend &&
11210 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11212 CopLINE_inc(PL_curcop);
11215 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11216 missingterm(PL_tokenbuf);
11218 sv_setpvn(herewas,bufptr,d-bufptr+1);
11219 sv_setpvn(tmpstr,d+1,s-d);
11221 sv_catpvn(herewas,s,bufend-s);
11222 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11229 while (s < PL_bufend &&
11230 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11232 CopLINE_inc(PL_curcop);
11234 if (s >= PL_bufend) {
11235 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11236 missingterm(PL_tokenbuf);
11238 sv_setpvn(tmpstr,d+1,s-d);
11240 if (PL_madskills) {
11242 sv_catpvn(PL_thisstuff, d + 1, s - d);
11244 PL_thisstuff = newSVpvn(d + 1, s - d);
11245 stuffstart = s - SvPVX(PL_linestr);
11249 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11251 sv_catpvn(herewas,s,PL_bufend-s);
11252 sv_setsv(PL_linestr,herewas);
11253 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11254 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11255 PL_last_lop = PL_last_uni = NULL;
11258 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11259 while (s >= PL_bufend) { /* multiple line string? */
11261 if (PL_madskills) {
11262 tstart = SvPVX(PL_linestr) + stuffstart;
11264 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11266 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11270 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11271 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11272 missingterm(PL_tokenbuf);
11275 stuffstart = s - SvPVX(PL_linestr);
11277 CopLINE_inc(PL_curcop);
11278 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11279 PL_last_lop = PL_last_uni = NULL;
11280 #ifndef PERL_STRICT_CR
11281 if (PL_bufend - PL_linestart >= 2) {
11282 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11283 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11285 PL_bufend[-2] = '\n';
11287 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11289 else if (PL_bufend[-1] == '\r')
11290 PL_bufend[-1] = '\n';
11292 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11293 PL_bufend[-1] = '\n';
11295 if (PERLDB_LINE && PL_curstash != PL_debstash)
11296 update_debugger_info(PL_linestr, NULL, 0);
11297 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11298 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11299 *(SvPVX(PL_linestr) + off ) = ' ';
11300 sv_catsv(PL_linestr,herewas);
11301 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11302 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11306 sv_catsv(tmpstr,PL_linestr);
11311 PL_multi_end = CopLINE(PL_curcop);
11312 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11313 SvPV_shrink_to_cur(tmpstr);
11315 SvREFCNT_dec(herewas);
11317 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11319 else if (PL_encoding)
11320 sv_recode_to_utf8(tmpstr, PL_encoding);
11322 PL_lex_stuff = tmpstr;
11323 yylval.ival = op_type;
11327 /* scan_inputsymbol
11328 takes: current position in input buffer
11329 returns: new position in input buffer
11330 side-effects: yylval and lex_op are set.
11335 <FH> read from filehandle
11336 <pkg::FH> read from package qualified filehandle
11337 <pkg'FH> read from package qualified filehandle
11338 <$fh> read from filehandle in $fh
11339 <*.h> filename glob
11344 S_scan_inputsymbol(pTHX_ char *start)
11347 register char *s = start; /* current position in buffer */
11351 char *d = PL_tokenbuf; /* start of temp holding space */
11352 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11354 end = strchr(s, '\n');
11357 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11359 /* die if we didn't have space for the contents of the <>,
11360 or if it didn't end, or if we see a newline
11363 if (len >= (I32)sizeof PL_tokenbuf)
11364 Perl_croak(aTHX_ "Excessively long <> operator");
11366 Perl_croak(aTHX_ "Unterminated <> operator");
11371 Remember, only scalar variables are interpreted as filehandles by
11372 this code. Anything more complex (e.g., <$fh{$num}>) will be
11373 treated as a glob() call.
11374 This code makes use of the fact that except for the $ at the front,
11375 a scalar variable and a filehandle look the same.
11377 if (*d == '$' && d[1]) d++;
11379 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11380 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11383 /* If we've tried to read what we allow filehandles to look like, and
11384 there's still text left, then it must be a glob() and not a getline.
11385 Use scan_str to pull out the stuff between the <> and treat it
11386 as nothing more than a string.
11389 if (d - PL_tokenbuf != len) {
11390 yylval.ival = OP_GLOB;
11391 s = scan_str(start,!!PL_madskills,FALSE);
11393 Perl_croak(aTHX_ "Glob not terminated");
11397 bool readline_overriden = FALSE;
11400 /* we're in a filehandle read situation */
11403 /* turn <> into <ARGV> */
11405 Copy("ARGV",d,5,char);
11407 /* Check whether readline() is overriden */
11408 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11410 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11412 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11413 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11414 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11415 readline_overriden = TRUE;
11417 /* if <$fh>, create the ops to turn the variable into a
11421 /* try to find it in the pad for this block, otherwise find
11422 add symbol table ops
11424 const PADOFFSET tmp = pad_findmy(d);
11425 if (tmp != NOT_IN_PAD) {
11426 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11427 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11428 HEK * const stashname = HvNAME_HEK(stash);
11429 SV * const sym = sv_2mortal(newSVhek(stashname));
11430 sv_catpvs(sym, "::");
11431 sv_catpv(sym, d+1);
11436 OP * const o = newOP(OP_PADSV, 0);
11438 PL_lex_op = readline_overriden
11439 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11440 append_elem(OP_LIST, o,
11441 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11442 : (OP*)newUNOP(OP_READLINE, 0, o);
11451 ? (GV_ADDMULTI | GV_ADDINEVAL)
11454 PL_lex_op = readline_overriden
11455 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11456 append_elem(OP_LIST,
11457 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11458 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11459 : (OP*)newUNOP(OP_READLINE, 0,
11460 newUNOP(OP_RV2SV, 0,
11461 newGVOP(OP_GV, 0, gv)));
11463 if (!readline_overriden)
11464 PL_lex_op->op_flags |= OPf_SPECIAL;
11465 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11466 yylval.ival = OP_NULL;
11469 /* If it's none of the above, it must be a literal filehandle
11470 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11472 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11473 PL_lex_op = readline_overriden
11474 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11475 append_elem(OP_LIST,
11476 newGVOP(OP_GV, 0, gv),
11477 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11478 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11479 yylval.ival = OP_NULL;
11488 takes: start position in buffer
11489 keep_quoted preserve \ on the embedded delimiter(s)
11490 keep_delims preserve the delimiters around the string
11491 returns: position to continue reading from buffer
11492 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11493 updates the read buffer.
11495 This subroutine pulls a string out of the input. It is called for:
11496 q single quotes q(literal text)
11497 ' single quotes 'literal text'
11498 qq double quotes qq(interpolate $here please)
11499 " double quotes "interpolate $here please"
11500 qx backticks qx(/bin/ls -l)
11501 ` backticks `/bin/ls -l`
11502 qw quote words @EXPORT_OK = qw( func() $spam )
11503 m// regexp match m/this/
11504 s/// regexp substitute s/this/that/
11505 tr/// string transliterate tr/this/that/
11506 y/// string transliterate y/this/that/
11507 ($*@) sub prototypes sub foo ($)
11508 (stuff) sub attr parameters sub foo : attr(stuff)
11509 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11511 In most of these cases (all but <>, patterns and transliterate)
11512 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11513 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11514 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11517 It skips whitespace before the string starts, and treats the first
11518 character as the delimiter. If the delimiter is one of ([{< then
11519 the corresponding "close" character )]}> is used as the closing
11520 delimiter. It allows quoting of delimiters, and if the string has
11521 balanced delimiters ([{<>}]) it allows nesting.
11523 On success, the SV with the resulting string is put into lex_stuff or,
11524 if that is already non-NULL, into lex_repl. The second case occurs only
11525 when parsing the RHS of the special constructs s/// and tr/// (y///).
11526 For convenience, the terminating delimiter character is stuffed into
11531 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11534 SV *sv; /* scalar value: string */
11535 const char *tmps; /* temp string, used for delimiter matching */
11536 register char *s = start; /* current position in the buffer */
11537 register char term; /* terminating character */
11538 register char *to; /* current position in the sv's data */
11539 I32 brackets = 1; /* bracket nesting level */
11540 bool has_utf8 = FALSE; /* is there any utf8 content? */
11541 I32 termcode; /* terminating char. code */
11542 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11543 STRLEN termlen; /* length of terminating string */
11544 int last_off = 0; /* last position for nesting bracket */
11550 /* skip space before the delimiter */
11556 if (PL_realtokenstart >= 0) {
11557 stuffstart = PL_realtokenstart;
11558 PL_realtokenstart = -1;
11561 stuffstart = start - SvPVX(PL_linestr);
11563 /* mark where we are, in case we need to report errors */
11566 /* after skipping whitespace, the next character is the terminator */
11569 termcode = termstr[0] = term;
11573 termcode = utf8_to_uvchr((U8*)s, &termlen);
11574 Copy(s, termstr, termlen, U8);
11575 if (!UTF8_IS_INVARIANT(term))
11579 /* mark where we are */
11580 PL_multi_start = CopLINE(PL_curcop);
11581 PL_multi_open = term;
11583 /* find corresponding closing delimiter */
11584 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11585 termcode = termstr[0] = term = tmps[5];
11587 PL_multi_close = term;
11589 /* create a new SV to hold the contents. 79 is the SV's initial length.
11590 What a random number. */
11591 sv = newSV_type(SVt_PVIV);
11593 SvIV_set(sv, termcode);
11594 (void)SvPOK_only(sv); /* validate pointer */
11596 /* move past delimiter and try to read a complete string */
11598 sv_catpvn(sv, s, termlen);
11601 tstart = SvPVX(PL_linestr) + stuffstart;
11602 if (!PL_thisopen && !keep_delims) {
11603 PL_thisopen = newSVpvn(tstart, s - tstart);
11604 stuffstart = s - SvPVX(PL_linestr);
11608 if (PL_encoding && !UTF) {
11612 int offset = s - SvPVX_const(PL_linestr);
11613 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11614 &offset, (char*)termstr, termlen);
11615 const char * const ns = SvPVX_const(PL_linestr) + offset;
11616 char * const svlast = SvEND(sv) - 1;
11618 for (; s < ns; s++) {
11619 if (*s == '\n' && !PL_rsfp)
11620 CopLINE_inc(PL_curcop);
11623 goto read_more_line;
11625 /* handle quoted delimiters */
11626 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11628 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11630 if ((svlast-1 - t) % 2) {
11631 if (!keep_quoted) {
11632 *(svlast-1) = term;
11634 SvCUR_set(sv, SvCUR(sv) - 1);
11639 if (PL_multi_open == PL_multi_close) {
11645 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11646 /* At here, all closes are "was quoted" one,
11647 so we don't check PL_multi_close. */
11649 if (!keep_quoted && *(t+1) == PL_multi_open)
11654 else if (*t == PL_multi_open)
11662 SvCUR_set(sv, w - SvPVX_const(sv));
11664 last_off = w - SvPVX(sv);
11665 if (--brackets <= 0)
11670 if (!keep_delims) {
11671 SvCUR_set(sv, SvCUR(sv) - 1);
11677 /* extend sv if need be */
11678 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11679 /* set 'to' to the next character in the sv's string */
11680 to = SvPVX(sv)+SvCUR(sv);
11682 /* if open delimiter is the close delimiter read unbridle */
11683 if (PL_multi_open == PL_multi_close) {
11684 for (; s < PL_bufend; s++,to++) {
11685 /* embedded newlines increment the current line number */
11686 if (*s == '\n' && !PL_rsfp)
11687 CopLINE_inc(PL_curcop);
11688 /* handle quoted delimiters */
11689 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11690 if (!keep_quoted && s[1] == term)
11692 /* any other quotes are simply copied straight through */
11696 /* terminate when run out of buffer (the for() condition), or
11697 have found the terminator */
11698 else if (*s == term) {
11701 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11704 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11710 /* if the terminator isn't the same as the start character (e.g.,
11711 matched brackets), we have to allow more in the quoting, and
11712 be prepared for nested brackets.
11715 /* read until we run out of string, or we find the terminator */
11716 for (; s < PL_bufend; s++,to++) {
11717 /* embedded newlines increment the line count */
11718 if (*s == '\n' && !PL_rsfp)
11719 CopLINE_inc(PL_curcop);
11720 /* backslashes can escape the open or closing characters */
11721 if (*s == '\\' && s+1 < PL_bufend) {
11722 if (!keep_quoted &&
11723 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11728 /* allow nested opens and closes */
11729 else if (*s == PL_multi_close && --brackets <= 0)
11731 else if (*s == PL_multi_open)
11733 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11738 /* terminate the copied string and update the sv's end-of-string */
11740 SvCUR_set(sv, to - SvPVX_const(sv));
11743 * this next chunk reads more into the buffer if we're not done yet
11747 break; /* handle case where we are done yet :-) */
11749 #ifndef PERL_STRICT_CR
11750 if (to - SvPVX_const(sv) >= 2) {
11751 if ((to[-2] == '\r' && to[-1] == '\n') ||
11752 (to[-2] == '\n' && to[-1] == '\r'))
11756 SvCUR_set(sv, to - SvPVX_const(sv));
11758 else if (to[-1] == '\r')
11761 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11766 /* if we're out of file, or a read fails, bail and reset the current
11767 line marker so we can report where the unterminated string began
11770 if (PL_madskills) {
11771 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11773 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11775 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11779 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11781 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11787 /* we read a line, so increment our line counter */
11788 CopLINE_inc(PL_curcop);
11790 /* update debugger info */
11791 if (PERLDB_LINE && PL_curstash != PL_debstash)
11792 update_debugger_info(PL_linestr, NULL, 0);
11794 /* having changed the buffer, we must update PL_bufend */
11795 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11796 PL_last_lop = PL_last_uni = NULL;
11799 /* at this point, we have successfully read the delimited string */
11801 if (!PL_encoding || UTF) {
11803 if (PL_madskills) {
11804 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11805 const int len = s - tstart;
11807 sv_catpvn(PL_thisstuff, tstart, len);
11809 PL_thisstuff = newSVpvn(tstart, len);
11810 if (!PL_thisclose && !keep_delims)
11811 PL_thisclose = newSVpvn(s,termlen);
11816 sv_catpvn(sv, s, termlen);
11821 if (PL_madskills) {
11822 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11823 const int len = s - tstart - termlen;
11825 sv_catpvn(PL_thisstuff, tstart, len);
11827 PL_thisstuff = newSVpvn(tstart, len);
11828 if (!PL_thisclose && !keep_delims)
11829 PL_thisclose = newSVpvn(s - termlen,termlen);
11833 if (has_utf8 || PL_encoding)
11836 PL_multi_end = CopLINE(PL_curcop);
11838 /* if we allocated too much space, give some back */
11839 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11840 SvLEN_set(sv, SvCUR(sv) + 1);
11841 SvPV_renew(sv, SvLEN(sv));
11844 /* decide whether this is the first or second quoted string we've read
11857 takes: pointer to position in buffer
11858 returns: pointer to new position in buffer
11859 side-effects: builds ops for the constant in yylval.op
11861 Read a number in any of the formats that Perl accepts:
11863 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11864 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11867 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11869 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11872 If it reads a number without a decimal point or an exponent, it will
11873 try converting the number to an integer and see if it can do so
11874 without loss of precision.
11878 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11881 register const char *s = start; /* current position in buffer */
11882 register char *d; /* destination in temp buffer */
11883 register char *e; /* end of temp buffer */
11884 NV nv; /* number read, as a double */
11885 SV *sv = NULL; /* place to put the converted number */
11886 bool floatit; /* boolean: int or float? */
11887 const char *lastub = NULL; /* position of last underbar */
11888 static char const number_too_long[] = "Number too long";
11890 /* We use the first character to decide what type of number this is */
11894 Perl_croak(aTHX_ "panic: scan_num");
11896 /* if it starts with a 0, it could be an octal number, a decimal in
11897 0.13 disguise, or a hexadecimal number, or a binary number. */
11901 u holds the "number so far"
11902 shift the power of 2 of the base
11903 (hex == 4, octal == 3, binary == 1)
11904 overflowed was the number more than we can hold?
11906 Shift is used when we add a digit. It also serves as an "are
11907 we in octal/hex/binary?" indicator to disallow hex characters
11908 when in octal mode.
11913 bool overflowed = FALSE;
11914 bool just_zero = TRUE; /* just plain 0 or binary number? */
11915 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11916 static const char* const bases[5] =
11917 { "", "binary", "", "octal", "hexadecimal" };
11918 static const char* const Bases[5] =
11919 { "", "Binary", "", "Octal", "Hexadecimal" };
11920 static const char* const maxima[5] =
11922 "0b11111111111111111111111111111111",
11926 const char *base, *Base, *max;
11928 /* check for hex */
11933 } else if (s[1] == 'b') {
11938 /* check for a decimal in disguise */
11939 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11941 /* so it must be octal */
11948 if (ckWARN(WARN_SYNTAX))
11949 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11950 "Misplaced _ in number");
11954 base = bases[shift];
11955 Base = Bases[shift];
11956 max = maxima[shift];
11958 /* read the rest of the number */
11960 /* x is used in the overflow test,
11961 b is the digit we're adding on. */
11966 /* if we don't mention it, we're done */
11970 /* _ are ignored -- but warned about if consecutive */
11972 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11973 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11974 "Misplaced _ in number");
11978 /* 8 and 9 are not octal */
11979 case '8': case '9':
11981 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11985 case '2': case '3': case '4':
11986 case '5': case '6': case '7':
11988 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
11991 case '0': case '1':
11992 b = *s++ & 15; /* ASCII digit -> value of digit */
11996 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
11997 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
11998 /* make sure they said 0x */
12001 b = (*s++ & 7) + 9;
12003 /* Prepare to put the digit we have onto the end
12004 of the number so far. We check for overflows.
12010 x = u << shift; /* make room for the digit */
12012 if ((x >> shift) != u
12013 && !(PL_hints & HINT_NEW_BINARY)) {
12016 if (ckWARN_d(WARN_OVERFLOW))
12017 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12018 "Integer overflow in %s number",
12021 u = x | b; /* add the digit to the end */
12024 n *= nvshift[shift];
12025 /* If an NV has not enough bits in its
12026 * mantissa to represent an UV this summing of
12027 * small low-order numbers is a waste of time
12028 * (because the NV cannot preserve the
12029 * low-order bits anyway): we could just
12030 * remember when did we overflow and in the
12031 * end just multiply n by the right
12039 /* if we get here, we had success: make a scalar value from
12044 /* final misplaced underbar check */
12045 if (s[-1] == '_') {
12046 if (ckWARN(WARN_SYNTAX))
12047 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12052 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12053 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12054 "%s number > %s non-portable",
12060 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12061 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12062 "%s number > %s non-portable",
12067 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12068 sv = new_constant(start, s - start, "integer",
12070 else if (PL_hints & HINT_NEW_BINARY)
12071 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
12076 handle decimal numbers.
12077 we're also sent here when we read a 0 as the first digit
12079 case '1': case '2': case '3': case '4': case '5':
12080 case '6': case '7': case '8': case '9': case '.':
12083 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12086 /* read next group of digits and _ and copy into d */
12087 while (isDIGIT(*s) || *s == '_') {
12088 /* skip underscores, checking for misplaced ones
12092 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12093 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12094 "Misplaced _ in number");
12098 /* check for end of fixed-length buffer */
12100 Perl_croak(aTHX_ number_too_long);
12101 /* if we're ok, copy the character */
12106 /* final misplaced underbar check */
12107 if (lastub && s == lastub + 1) {
12108 if (ckWARN(WARN_SYNTAX))
12109 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12112 /* read a decimal portion if there is one. avoid
12113 3..5 being interpreted as the number 3. followed
12116 if (*s == '.' && s[1] != '.') {
12121 if (ckWARN(WARN_SYNTAX))
12122 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12123 "Misplaced _ in number");
12127 /* copy, ignoring underbars, until we run out of digits.
12129 for (; isDIGIT(*s) || *s == '_'; s++) {
12130 /* fixed length buffer check */
12132 Perl_croak(aTHX_ number_too_long);
12134 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12135 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12136 "Misplaced _ in number");
12142 /* fractional part ending in underbar? */
12143 if (s[-1] == '_') {
12144 if (ckWARN(WARN_SYNTAX))
12145 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12146 "Misplaced _ in number");
12148 if (*s == '.' && isDIGIT(s[1])) {
12149 /* oops, it's really a v-string, but without the "v" */
12155 /* read exponent part, if present */
12156 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12160 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12161 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12163 /* stray preinitial _ */
12165 if (ckWARN(WARN_SYNTAX))
12166 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12167 "Misplaced _ in number");
12171 /* allow positive or negative exponent */
12172 if (*s == '+' || *s == '-')
12175 /* stray initial _ */
12177 if (ckWARN(WARN_SYNTAX))
12178 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12179 "Misplaced _ in number");
12183 /* read digits of exponent */
12184 while (isDIGIT(*s) || *s == '_') {
12187 Perl_croak(aTHX_ number_too_long);
12191 if (((lastub && s == lastub + 1) ||
12192 (!isDIGIT(s[1]) && s[1] != '_'))
12193 && ckWARN(WARN_SYNTAX))
12194 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12195 "Misplaced _ in number");
12202 /* make an sv from the string */
12206 We try to do an integer conversion first if no characters
12207 indicating "float" have been found.
12212 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12214 if (flags == IS_NUMBER_IN_UV) {
12216 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12219 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12220 if (uv <= (UV) IV_MIN)
12221 sv_setiv(sv, -(IV)uv);
12228 /* terminate the string */
12230 nv = Atof(PL_tokenbuf);
12234 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
12235 (PL_hints & HINT_NEW_INTEGER) )
12236 sv = new_constant(PL_tokenbuf,
12239 (floatit ? "float" : "integer"),
12243 /* if it starts with a v, it could be a v-string */
12246 sv = newSV(5); /* preallocate storage space */
12247 s = scan_vstring(s, PL_bufend, sv);
12251 /* make the op for the constant and return */
12254 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12256 lvalp->opval = NULL;
12262 S_scan_formline(pTHX_ register char *s)
12265 register char *eol;
12267 SV * const stuff = newSVpvs("");
12268 bool needargs = FALSE;
12269 bool eofmt = FALSE;
12271 char *tokenstart = s;
12274 if (PL_madskills) {
12275 savewhite = PL_thiswhite;
12280 while (!needargs) {
12283 #ifdef PERL_STRICT_CR
12284 while (SPACE_OR_TAB(*t))
12287 while (SPACE_OR_TAB(*t) || *t == '\r')
12290 if (*t == '\n' || t == PL_bufend) {
12295 if (PL_in_eval && !PL_rsfp) {
12296 eol = (char *) memchr(s,'\n',PL_bufend-s);
12301 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12303 for (t = s; t < eol; t++) {
12304 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12306 goto enough; /* ~~ must be first line in formline */
12308 if (*t == '@' || *t == '^')
12312 sv_catpvn(stuff, s, eol-s);
12313 #ifndef PERL_STRICT_CR
12314 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12315 char *end = SvPVX(stuff) + SvCUR(stuff);
12318 SvCUR_set(stuff, SvCUR(stuff) - 1);
12328 if (PL_madskills) {
12330 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12332 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12335 s = filter_gets(PL_linestr, PL_rsfp, 0);
12337 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12339 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12341 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12342 PL_last_lop = PL_last_uni = NULL;
12351 if (SvCUR(stuff)) {
12354 PL_lex_state = LEX_NORMAL;
12355 start_force(PL_curforce);
12356 NEXTVAL_NEXTTOKE.ival = 0;
12360 PL_lex_state = LEX_FORMLINE;
12362 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12364 else if (PL_encoding)
12365 sv_recode_to_utf8(stuff, PL_encoding);
12367 start_force(PL_curforce);
12368 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12370 start_force(PL_curforce);
12371 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12375 SvREFCNT_dec(stuff);
12377 PL_lex_formbrack = 0;
12381 if (PL_madskills) {
12383 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12385 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12386 PL_thiswhite = savewhite;
12393 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12396 const I32 oldsavestack_ix = PL_savestack_ix;
12397 CV* const outsidecv = PL_compcv;
12400 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12402 SAVEI32(PL_subline);
12403 save_item(PL_subname);
12404 SAVESPTR(PL_compcv);
12406 PL_compcv = (CV*)newSV_type(is_format ? SVt_PVFM : SVt_PVCV);
12407 CvFLAGS(PL_compcv) |= flags;
12409 PL_subline = CopLINE(PL_curcop);
12410 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12411 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12412 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12414 return oldsavestack_ix;
12418 #pragma segment Perl_yylex
12421 Perl_yywarn(pTHX_ const char *s)
12424 PL_in_eval |= EVAL_WARNONLY;
12426 PL_in_eval &= ~EVAL_WARNONLY;
12431 Perl_yyerror(pTHX_ const char *s)
12434 const char *where = NULL;
12435 const char *context = NULL;
12438 int yychar = PL_parser->yychar;
12440 if (!yychar || (yychar == ';' && !PL_rsfp))
12442 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12443 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12444 PL_oldbufptr != PL_bufptr) {
12447 The code below is removed for NetWare because it abends/crashes on NetWare
12448 when the script has error such as not having the closing quotes like:
12449 if ($var eq "value)
12450 Checking of white spaces is anyway done in NetWare code.
12453 while (isSPACE(*PL_oldoldbufptr))
12456 context = PL_oldoldbufptr;
12457 contlen = PL_bufptr - PL_oldoldbufptr;
12459 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12460 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12463 The code below is removed for NetWare because it abends/crashes on NetWare
12464 when the script has error such as not having the closing quotes like:
12465 if ($var eq "value)
12466 Checking of white spaces is anyway done in NetWare code.
12469 while (isSPACE(*PL_oldbufptr))
12472 context = PL_oldbufptr;
12473 contlen = PL_bufptr - PL_oldbufptr;
12475 else if (yychar > 255)
12476 where = "next token ???";
12477 else if (yychar == -2) { /* YYEMPTY */
12478 if (PL_lex_state == LEX_NORMAL ||
12479 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12480 where = "at end of line";
12481 else if (PL_lex_inpat)
12482 where = "within pattern";
12484 where = "within string";
12487 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12489 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12490 else if (isPRINT_LC(yychar)) {
12491 const unsigned char string = (unsigned char) yychar;
12492 sv_catpvn(where_sv, &string, 1);
12495 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12496 where = SvPVX_const(where_sv);
12498 msg = sv_2mortal(newSVpv(s, 0));
12499 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12500 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12502 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12504 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12505 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12506 Perl_sv_catpvf(aTHX_ msg,
12507 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12508 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12511 if (PL_in_eval & EVAL_WARNONLY) {
12512 if (ckWARN_d(WARN_SYNTAX))
12513 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12517 if (PL_error_count >= 10) {
12518 if (PL_in_eval && SvCUR(ERRSV))
12519 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12520 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12522 Perl_croak(aTHX_ "%s has too many errors.\n",
12523 OutCopFILE(PL_curcop));
12526 PL_in_my_stash = NULL;
12530 #pragma segment Main
12534 S_swallow_bom(pTHX_ U8 *s)
12537 const STRLEN slen = SvCUR(PL_linestr);
12540 if (s[1] == 0xFE) {
12541 /* UTF-16 little-endian? (or UTF32-LE?) */
12542 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12543 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12544 #ifndef PERL_NO_UTF16_FILTER
12545 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12548 if (PL_bufend > (char*)s) {
12552 filter_add(utf16rev_textfilter, NULL);
12553 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12554 utf16_to_utf8_reversed(s, news,
12555 PL_bufend - (char*)s - 1,
12557 sv_setpvn(PL_linestr, (const char*)news, newlen);
12559 s = (U8*)SvPVX(PL_linestr);
12560 Copy(news, s, newlen, U8);
12564 SvUTF8_on(PL_linestr);
12565 s = (U8*)SvPVX(PL_linestr);
12567 /* FIXME - is this a general bug fix? */
12570 PL_bufend = SvPVX(PL_linestr) + newlen;
12573 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12578 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12579 #ifndef PERL_NO_UTF16_FILTER
12580 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12583 if (PL_bufend > (char *)s) {
12587 filter_add(utf16_textfilter, NULL);
12588 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12589 utf16_to_utf8(s, news,
12590 PL_bufend - (char*)s,
12592 sv_setpvn(PL_linestr, (const char*)news, newlen);
12594 SvUTF8_on(PL_linestr);
12595 s = (U8*)SvPVX(PL_linestr);
12596 PL_bufend = SvPVX(PL_linestr) + newlen;
12599 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12604 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12605 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12606 s += 3; /* UTF-8 */
12612 if (s[2] == 0xFE && s[3] == 0xFF) {
12613 /* UTF-32 big-endian */
12614 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12617 else if (s[2] == 0 && s[3] != 0) {
12620 * are a good indicator of UTF-16BE. */
12621 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12627 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12628 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12629 s += 4; /* UTF-8 */
12635 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12638 * are a good indicator of UTF-16LE. */
12639 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12647 #ifndef PERL_NO_UTF16_FILTER
12649 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12652 const STRLEN old = SvCUR(sv);
12653 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12654 DEBUG_P(PerlIO_printf(Perl_debug_log,
12655 "utf16_textfilter(%p): %d %d (%d)\n",
12656 FPTR2DPTR(void *, utf16_textfilter),
12657 idx, maxlen, (int) count));
12661 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12662 Copy(SvPVX_const(sv), tmps, old, char);
12663 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12664 SvCUR(sv) - old, &newlen);
12665 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12667 DEBUG_P({sv_dump(sv);});
12672 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12675 const STRLEN old = SvCUR(sv);
12676 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12677 DEBUG_P(PerlIO_printf(Perl_debug_log,
12678 "utf16rev_textfilter(%p): %d %d (%d)\n",
12679 FPTR2DPTR(void *, utf16rev_textfilter),
12680 idx, maxlen, (int) count));
12684 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12685 Copy(SvPVX_const(sv), tmps, old, char);
12686 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12687 SvCUR(sv) - old, &newlen);
12688 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12690 DEBUG_P({ sv_dump(sv); });
12696 Returns a pointer to the next character after the parsed
12697 vstring, as well as updating the passed in sv.
12699 Function must be called like
12702 s = scan_vstring(s,e,sv);
12704 where s and e are the start and end of the string.
12705 The sv should already be large enough to store the vstring
12706 passed in, for performance reasons.
12711 Perl_scan_vstring(pTHX_ const char *s, const char *e, SV *sv)
12714 const char *pos = s;
12715 const char *start = s;
12716 if (*pos == 'v') pos++; /* get past 'v' */
12717 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12719 if ( *pos != '.') {
12720 /* this may not be a v-string if followed by => */
12721 const char *next = pos;
12722 while (next < e && isSPACE(*next))
12724 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12725 /* return string not v-string */
12726 sv_setpvn(sv,(char *)s,pos-s);
12727 return (char *)pos;
12731 if (!isALPHA(*pos)) {
12732 U8 tmpbuf[UTF8_MAXBYTES+1];
12735 s++; /* get past 'v' */
12737 sv_setpvn(sv, "", 0);
12740 /* this is atoi() that tolerates underscores */
12743 const char *end = pos;
12745 while (--end >= s) {
12747 const UV orev = rev;
12748 rev += (*end - '0') * mult;
12750 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12751 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12752 "Integer overflow in decimal number");
12756 if (rev > 0x7FFFFFFF)
12757 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12759 /* Append native character for the rev point */
12760 tmpend = uvchr_to_utf8(tmpbuf, rev);
12761 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12762 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12764 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12770 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12774 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12782 * c-indentation-style: bsd
12783 * c-basic-offset: 4
12784 * indent-tabs-mode: t
12787 * ex: set ts=8 sts=4 sw=4 noet: