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))));
2980 * The intent of this yylex wrapper is to minimize the changes to the
2981 * tokener when we aren't interested in collecting madprops. It remains
2982 * to be seen how successful this strategy will be...
2989 char *s = PL_bufptr;
2991 /* make sure PL_thiswhite is initialized */
2995 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2996 if (PL_pending_ident)
2997 return S_pending_ident(aTHX);
2999 /* previous token ate up our whitespace? */
3000 if (!PL_lasttoke && PL_nextwhite) {
3001 PL_thiswhite = PL_nextwhite;
3005 /* isolate the token, and figure out where it is without whitespace */
3006 PL_realtokenstart = -1;
3010 assert(PL_curforce < 0);
3012 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3013 if (!PL_thistoken) {
3014 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3015 PL_thistoken = newSVpvs("");
3017 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3018 PL_thistoken = newSVpvn(tstart, s - tstart);
3021 if (PL_thismad) /* install head */
3022 CURMAD('X', PL_thistoken);
3025 /* last whitespace of a sublex? */
3026 if (optype == ')' && PL_endwhite) {
3027 CURMAD('X', PL_endwhite);
3032 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3033 if (!PL_thiswhite && !PL_endwhite && !optype) {
3034 sv_free(PL_thistoken);
3039 /* put off final whitespace till peg */
3040 if (optype == ';' && !PL_rsfp) {
3041 PL_nextwhite = PL_thiswhite;
3044 else if (PL_thisopen) {
3045 CURMAD('q', PL_thisopen);
3047 sv_free(PL_thistoken);
3051 /* Store actual token text as madprop X */
3052 CURMAD('X', PL_thistoken);
3056 /* add preceding whitespace as madprop _ */
3057 CURMAD('_', PL_thiswhite);
3061 /* add quoted material as madprop = */
3062 CURMAD('=', PL_thisstuff);
3066 /* add terminating quote as madprop Q */
3067 CURMAD('Q', PL_thisclose);
3071 /* special processing based on optype */
3075 /* opval doesn't need a TOKEN since it can already store mp */
3086 append_madprops(PL_thismad, yylval.opval, 0);
3094 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3103 /* remember any fake bracket that lexer is about to discard */
3104 if (PL_lex_brackets == 1 &&
3105 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3108 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3111 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3112 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3115 break; /* don't bother looking for trailing comment */
3124 /* attach a trailing comment to its statement instead of next token */
3128 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3130 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3132 if (*s == '\n' || *s == '#') {
3133 while (s < PL_bufend && *s != '\n')
3137 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3138 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3155 /* Create new token struct. Note: opvals return early above. */
3156 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3163 S_tokenize_use(pTHX_ int is_use, char *s) {
3165 if (PL_expect != XSTATE)
3166 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3167 is_use ? "use" : "no"));
3169 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3170 s = force_version(s, TRUE);
3171 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3172 start_force(PL_curforce);
3173 NEXTVAL_NEXTTOKE.opval = NULL;
3176 else if (*s == 'v') {
3177 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3178 s = force_version(s, FALSE);
3182 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3183 s = force_version(s, FALSE);
3185 yylval.ival = is_use;
3189 static const char* const exp_name[] =
3190 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3191 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3198 Works out what to call the token just pulled out of the input
3199 stream. The yacc parser takes care of taking the ops we return and
3200 stitching them into a tree.
3206 if read an identifier
3207 if we're in a my declaration
3208 croak if they tried to say my($foo::bar)
3209 build the ops for a my() declaration
3210 if it's an access to a my() variable
3211 are we in a sort block?
3212 croak if my($a); $a <=> $b
3213 build ops for access to a my() variable
3214 if in a dq string, and they've said @foo and we can't find @foo
3216 build ops for a bareword
3217 if we already built the token before, use it.
3222 #pragma segment Perl_yylex
3228 register char *s = PL_bufptr;
3233 /* orig_keyword, gvp, and gv are initialized here because
3234 * jump to the label just_a_word_zero can bypass their
3235 * initialization later. */
3236 I32 orig_keyword = 0;
3241 SV* tmp = newSVpvs("");
3242 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3243 (IV)CopLINE(PL_curcop),
3244 lex_state_names[PL_lex_state],
3245 exp_name[PL_expect],
3246 pv_display(tmp, s, strlen(s), 0, 60));
3249 /* check if there's an identifier for us to look at */
3250 if (PL_pending_ident)
3251 return REPORT(S_pending_ident(aTHX));
3253 /* no identifier pending identification */
3255 switch (PL_lex_state) {
3257 case LEX_NORMAL: /* Some compilers will produce faster */
3258 case LEX_INTERPNORMAL: /* code if we comment these out. */
3262 /* when we've already built the next token, just pull it out of the queue */
3266 yylval = PL_nexttoke[PL_lasttoke].next_val;
3268 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3269 PL_nexttoke[PL_lasttoke].next_mad = 0;
3270 if (PL_thismad && PL_thismad->mad_key == '_') {
3271 PL_thiswhite = (SV*)PL_thismad->mad_val;
3272 PL_thismad->mad_val = 0;
3273 mad_free(PL_thismad);
3278 PL_lex_state = PL_lex_defer;
3279 PL_expect = PL_lex_expect;
3280 PL_lex_defer = LEX_NORMAL;
3281 if (!PL_nexttoke[PL_lasttoke].next_type)
3286 yylval = PL_nextval[PL_nexttoke];
3288 PL_lex_state = PL_lex_defer;
3289 PL_expect = PL_lex_expect;
3290 PL_lex_defer = LEX_NORMAL;
3294 /* FIXME - can these be merged? */
3295 return(PL_nexttoke[PL_lasttoke].next_type);
3297 return REPORT(PL_nexttype[PL_nexttoke]);
3300 /* interpolated case modifiers like \L \U, including \Q and \E.
3301 when we get here, PL_bufptr is at the \
3303 case LEX_INTERPCASEMOD:
3305 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3306 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3308 /* handle \E or end of string */
3309 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3311 if (PL_lex_casemods) {
3312 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3313 PL_lex_casestack[PL_lex_casemods] = '\0';
3315 if (PL_bufptr != PL_bufend
3316 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3318 PL_lex_state = LEX_INTERPCONCAT;
3321 PL_thistoken = newSVpvs("\\E");
3327 while (PL_bufptr != PL_bufend &&
3328 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3330 PL_thiswhite = newSVpvs("");
3331 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3335 if (PL_bufptr != PL_bufend)
3338 PL_lex_state = LEX_INTERPCONCAT;
3342 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3343 "### Saw case modifier\n"); });
3345 if (s[1] == '\\' && s[2] == 'E') {
3348 PL_thiswhite = newSVpvs("");
3349 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3352 PL_lex_state = LEX_INTERPCONCAT;
3357 if (!PL_madskills) /* when just compiling don't need correct */
3358 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3359 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3360 if ((*s == 'L' || *s == 'U') &&
3361 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3362 PL_lex_casestack[--PL_lex_casemods] = '\0';
3365 if (PL_lex_casemods > 10)
3366 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3367 PL_lex_casestack[PL_lex_casemods++] = *s;
3368 PL_lex_casestack[PL_lex_casemods] = '\0';
3369 PL_lex_state = LEX_INTERPCONCAT;
3370 start_force(PL_curforce);
3371 NEXTVAL_NEXTTOKE.ival = 0;
3373 start_force(PL_curforce);
3375 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3377 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3379 NEXTVAL_NEXTTOKE.ival = OP_LC;
3381 NEXTVAL_NEXTTOKE.ival = OP_UC;
3383 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3385 Perl_croak(aTHX_ "panic: yylex");
3387 SV* const tmpsv = newSVpvs("");
3388 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3394 if (PL_lex_starts) {
3400 sv_free(PL_thistoken);
3401 PL_thistoken = newSVpvs("");
3404 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3405 if (PL_lex_casemods == 1 && PL_lex_inpat)
3414 case LEX_INTERPPUSH:
3415 return REPORT(sublex_push());
3417 case LEX_INTERPSTART:
3418 if (PL_bufptr == PL_bufend)
3419 return REPORT(sublex_done());
3420 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3421 "### Interpolated variable\n"); });
3423 PL_lex_dojoin = (*PL_bufptr == '@');
3424 PL_lex_state = LEX_INTERPNORMAL;
3425 if (PL_lex_dojoin) {
3426 start_force(PL_curforce);
3427 NEXTVAL_NEXTTOKE.ival = 0;
3429 start_force(PL_curforce);
3430 force_ident("\"", '$');
3431 start_force(PL_curforce);
3432 NEXTVAL_NEXTTOKE.ival = 0;
3434 start_force(PL_curforce);
3435 NEXTVAL_NEXTTOKE.ival = 0;
3437 start_force(PL_curforce);
3438 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3441 if (PL_lex_starts++) {
3446 sv_free(PL_thistoken);
3447 PL_thistoken = newSVpvs("");
3450 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3451 if (!PL_lex_casemods && PL_lex_inpat)
3458 case LEX_INTERPENDMAYBE:
3459 if (intuit_more(PL_bufptr)) {
3460 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3466 if (PL_lex_dojoin) {
3467 PL_lex_dojoin = FALSE;
3468 PL_lex_state = LEX_INTERPCONCAT;
3472 sv_free(PL_thistoken);
3473 PL_thistoken = newSVpvs("");
3478 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3479 && SvEVALED(PL_lex_repl))
3481 if (PL_bufptr != PL_bufend)
3482 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3486 case LEX_INTERPCONCAT:
3488 if (PL_lex_brackets)
3489 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3491 if (PL_bufptr == PL_bufend)
3492 return REPORT(sublex_done());
3494 if (SvIVX(PL_linestr) == '\'') {
3495 SV *sv = newSVsv(PL_linestr);
3498 else if ( PL_hints & HINT_NEW_RE )
3499 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3500 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3504 s = scan_const(PL_bufptr);
3506 PL_lex_state = LEX_INTERPCASEMOD;
3508 PL_lex_state = LEX_INTERPSTART;
3511 if (s != PL_bufptr) {
3512 start_force(PL_curforce);
3514 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3516 NEXTVAL_NEXTTOKE = yylval;
3519 if (PL_lex_starts++) {
3523 sv_free(PL_thistoken);
3524 PL_thistoken = newSVpvs("");
3527 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3528 if (!PL_lex_casemods && PL_lex_inpat)
3541 PL_lex_state = LEX_NORMAL;
3542 s = scan_formline(PL_bufptr);
3543 if (!PL_lex_formbrack)
3549 PL_oldoldbufptr = PL_oldbufptr;
3555 sv_free(PL_thistoken);
3558 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3562 if (isIDFIRST_lazy_if(s,UTF))
3564 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3565 Perl_croak(aTHX_ "Unrecognized character \\x%02X in column %d", *s & 255, (int) len + 1);
3568 goto fake_eof; /* emulate EOF on ^D or ^Z */
3577 if (PL_lex_brackets) {
3578 yyerror((const char *)
3580 ? "Format not terminated"
3581 : "Missing right curly or square bracket"));
3583 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3584 "### Tokener got EOF\n");
3588 if (s++ < PL_bufend)
3589 goto retry; /* ignore stray nulls */
3592 if (!PL_in_eval && !PL_preambled) {
3593 PL_preambled = TRUE;
3598 sv_setpv(PL_linestr,incl_perldb());
3599 if (SvCUR(PL_linestr))
3600 sv_catpvs(PL_linestr,";");
3602 while(AvFILLp(PL_preambleav) >= 0) {
3603 SV *tmpsv = av_shift(PL_preambleav);
3604 sv_catsv(PL_linestr, tmpsv);
3605 sv_catpvs(PL_linestr, ";");
3608 sv_free((SV*)PL_preambleav);
3609 PL_preambleav = NULL;
3611 if (PL_minus_n || PL_minus_p) {
3612 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3614 sv_catpvs(PL_linestr,"chomp;");
3617 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3618 || *PL_splitstr == '"')
3619 && strchr(PL_splitstr + 1, *PL_splitstr))
3620 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3622 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3623 bytes can be used as quoting characters. :-) */
3624 const char *splits = PL_splitstr;
3625 sv_catpvs(PL_linestr, "our @F=split(q\0");
3628 if (*splits == '\\')
3629 sv_catpvn(PL_linestr, splits, 1);
3630 sv_catpvn(PL_linestr, splits, 1);
3631 } while (*splits++);
3632 /* This loop will embed the trailing NUL of
3633 PL_linestr as the last thing it does before
3635 sv_catpvs(PL_linestr, ");");
3639 sv_catpvs(PL_linestr,"our @F=split(' ');");
3643 sv_catpvs(PL_linestr,"use feature ':5.10';");
3644 sv_catpvs(PL_linestr, "\n");
3645 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3646 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3647 PL_last_lop = PL_last_uni = NULL;
3648 if (PERLDB_LINE && PL_curstash != PL_debstash)
3649 update_debugger_info(PL_linestr, NULL, 0);
3653 bof = PL_rsfp ? TRUE : FALSE;
3654 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3657 PL_realtokenstart = -1;
3660 if (PL_preprocess && !PL_in_eval)
3661 (void)PerlProc_pclose(PL_rsfp);
3662 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3663 PerlIO_clearerr(PL_rsfp);
3665 (void)PerlIO_close(PL_rsfp);
3667 PL_doextract = FALSE;
3669 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3674 sv_setpv(PL_linestr,
3677 ? ";}continue{print;}" : ";}"));
3678 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3679 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3680 PL_last_lop = PL_last_uni = NULL;
3681 PL_minus_n = PL_minus_p = 0;
3684 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3685 PL_last_lop = PL_last_uni = NULL;
3686 sv_setpvn(PL_linestr,"",0);
3687 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3689 /* If it looks like the start of a BOM or raw UTF-16,
3690 * check if it in fact is. */
3696 #ifdef PERLIO_IS_STDIO
3697 # ifdef __GNU_LIBRARY__
3698 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3699 # define FTELL_FOR_PIPE_IS_BROKEN
3703 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3704 # define FTELL_FOR_PIPE_IS_BROKEN
3709 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3710 /* This loses the possibility to detect the bof
3711 * situation on perl -P when the libc5 is being used.
3712 * Workaround? Maybe attach some extra state to PL_rsfp?
3715 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3717 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3720 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3721 s = swallow_bom((U8*)s);
3725 /* Incest with pod. */
3728 sv_catsv(PL_thiswhite, PL_linestr);
3730 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3731 sv_setpvn(PL_linestr, "", 0);
3732 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3733 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3734 PL_last_lop = PL_last_uni = NULL;
3735 PL_doextract = FALSE;
3739 } while (PL_doextract);
3740 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3741 if (PERLDB_LINE && PL_curstash != PL_debstash)
3742 update_debugger_info(PL_linestr, NULL, 0);
3743 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3744 PL_last_lop = PL_last_uni = NULL;
3745 if (CopLINE(PL_curcop) == 1) {
3746 while (s < PL_bufend && isSPACE(*s))
3748 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3752 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3756 if (*s == '#' && *(s+1) == '!')
3758 #ifdef ALTERNATE_SHEBANG
3760 static char const as[] = ALTERNATE_SHEBANG;
3761 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3762 d = s + (sizeof(as) - 1);
3764 #endif /* ALTERNATE_SHEBANG */
3773 while (*d && !isSPACE(*d))
3777 #ifdef ARG_ZERO_IS_SCRIPT
3778 if (ipathend > ipath) {
3780 * HP-UX (at least) sets argv[0] to the script name,
3781 * which makes $^X incorrect. And Digital UNIX and Linux,
3782 * at least, set argv[0] to the basename of the Perl
3783 * interpreter. So, having found "#!", we'll set it right.
3785 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3787 assert(SvPOK(x) || SvGMAGICAL(x));
3788 if (sv_eq(x, CopFILESV(PL_curcop))) {
3789 sv_setpvn(x, ipath, ipathend - ipath);
3795 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3796 const char * const lstart = SvPV_const(x,llen);
3798 bstart += blen - llen;
3799 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3800 sv_setpvn(x, ipath, ipathend - ipath);
3805 TAINT_NOT; /* $^X is always tainted, but that's OK */
3807 #endif /* ARG_ZERO_IS_SCRIPT */
3812 d = instr(s,"perl -");
3814 d = instr(s,"perl");
3816 /* avoid getting into infinite loops when shebang
3817 * line contains "Perl" rather than "perl" */
3819 for (d = ipathend-4; d >= ipath; --d) {
3820 if ((*d == 'p' || *d == 'P')
3821 && !ibcmp(d, "perl", 4))
3831 #ifdef ALTERNATE_SHEBANG
3833 * If the ALTERNATE_SHEBANG on this system starts with a
3834 * character that can be part of a Perl expression, then if
3835 * we see it but not "perl", we're probably looking at the
3836 * start of Perl code, not a request to hand off to some
3837 * other interpreter. Similarly, if "perl" is there, but
3838 * not in the first 'word' of the line, we assume the line
3839 * contains the start of the Perl program.
3841 if (d && *s != '#') {
3842 const char *c = ipath;
3843 while (*c && !strchr("; \t\r\n\f\v#", *c))
3846 d = NULL; /* "perl" not in first word; ignore */
3848 *s = '#'; /* Don't try to parse shebang line */
3850 #endif /* ALTERNATE_SHEBANG */
3851 #ifndef MACOS_TRADITIONAL
3856 !instr(s,"indir") &&
3857 instr(PL_origargv[0],"perl"))
3864 while (s < PL_bufend && isSPACE(*s))
3866 if (s < PL_bufend) {
3867 Newxz(newargv,PL_origargc+3,char*);
3869 while (s < PL_bufend && !isSPACE(*s))
3872 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3875 newargv = PL_origargv;
3878 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3880 Perl_croak(aTHX_ "Can't exec %s", ipath);
3884 while (*d && !isSPACE(*d))
3886 while (SPACE_OR_TAB(*d))
3890 const bool switches_done = PL_doswitches;
3891 const U32 oldpdb = PL_perldb;
3892 const bool oldn = PL_minus_n;
3893 const bool oldp = PL_minus_p;
3896 if (*d == 'M' || *d == 'm' || *d == 'C') {
3897 const char * const m = d;
3898 while (*d && !isSPACE(*d))
3900 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3903 d = moreswitches(d);
3905 if (PL_doswitches && !switches_done) {
3906 int argc = PL_origargc;
3907 char **argv = PL_origargv;
3910 } while (argc && argv[0][0] == '-' && argv[0][1]);
3911 init_argv_symbols(argc,argv);
3913 if ((PERLDB_LINE && !oldpdb) ||
3914 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3915 /* if we have already added "LINE: while (<>) {",
3916 we must not do it again */
3918 sv_setpvn(PL_linestr, "", 0);
3919 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3920 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3921 PL_last_lop = PL_last_uni = NULL;
3922 PL_preambled = FALSE;
3924 (void)gv_fetchfile(PL_origfilename);
3931 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3933 PL_lex_state = LEX_FORMLINE;
3938 #ifdef PERL_STRICT_CR
3939 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3941 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3943 case ' ': case '\t': case '\f': case 013:
3944 #ifdef MACOS_TRADITIONAL
3948 PL_realtokenstart = -1;
3950 PL_thiswhite = newSVpvs("");
3951 sv_catpvn(PL_thiswhite, s, 1);
3958 PL_realtokenstart = -1;
3962 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3963 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3964 /* handle eval qq[#line 1 "foo"\n ...] */
3965 CopLINE_dec(PL_curcop);
3968 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3970 if (!PL_in_eval || PL_rsfp)
3975 while (d < PL_bufend && *d != '\n')
3979 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3980 Perl_croak(aTHX_ "panic: input overflow");
3983 PL_thiswhite = newSVpvn(s, d - s);
3988 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3990 PL_lex_state = LEX_FORMLINE;
3996 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3997 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4000 TOKEN(PEG); /* make sure any #! line is accessible */
4005 /* if (PL_madskills && PL_lex_formbrack) { */
4007 while (d < PL_bufend && *d != '\n')
4011 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4012 Perl_croak(aTHX_ "panic: input overflow");
4013 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4015 PL_thiswhite = newSVpvs("");
4016 if (CopLINE(PL_curcop) == 1) {
4017 sv_setpvn(PL_thiswhite, "", 0);
4020 sv_catpvn(PL_thiswhite, s, d - s);
4034 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4042 while (s < PL_bufend && SPACE_OR_TAB(*s))
4045 if (strnEQ(s,"=>",2)) {
4046 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4047 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4048 OPERATOR('-'); /* unary minus */
4050 PL_last_uni = PL_oldbufptr;
4052 case 'r': ftst = OP_FTEREAD; break;
4053 case 'w': ftst = OP_FTEWRITE; break;
4054 case 'x': ftst = OP_FTEEXEC; break;
4055 case 'o': ftst = OP_FTEOWNED; break;
4056 case 'R': ftst = OP_FTRREAD; break;
4057 case 'W': ftst = OP_FTRWRITE; break;
4058 case 'X': ftst = OP_FTREXEC; break;
4059 case 'O': ftst = OP_FTROWNED; break;
4060 case 'e': ftst = OP_FTIS; break;
4061 case 'z': ftst = OP_FTZERO; break;
4062 case 's': ftst = OP_FTSIZE; break;
4063 case 'f': ftst = OP_FTFILE; break;
4064 case 'd': ftst = OP_FTDIR; break;
4065 case 'l': ftst = OP_FTLINK; break;
4066 case 'p': ftst = OP_FTPIPE; break;
4067 case 'S': ftst = OP_FTSOCK; break;
4068 case 'u': ftst = OP_FTSUID; break;
4069 case 'g': ftst = OP_FTSGID; break;
4070 case 'k': ftst = OP_FTSVTX; break;
4071 case 'b': ftst = OP_FTBLK; break;
4072 case 'c': ftst = OP_FTCHR; break;
4073 case 't': ftst = OP_FTTTY; break;
4074 case 'T': ftst = OP_FTTEXT; break;
4075 case 'B': ftst = OP_FTBINARY; break;
4076 case 'M': case 'A': case 'C':
4077 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4079 case 'M': ftst = OP_FTMTIME; break;
4080 case 'A': ftst = OP_FTATIME; break;
4081 case 'C': ftst = OP_FTCTIME; break;
4089 PL_last_lop_op = (OPCODE)ftst;
4090 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4091 "### Saw file test %c\n", (int)tmp);
4096 /* Assume it was a minus followed by a one-letter named
4097 * subroutine call (or a -bareword), then. */
4098 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4099 "### '-%c' looked like a file test but was not\n",
4106 const char tmp = *s++;
4109 if (PL_expect == XOPERATOR)
4114 else if (*s == '>') {
4117 if (isIDFIRST_lazy_if(s,UTF)) {
4118 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4126 if (PL_expect == XOPERATOR)
4129 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4131 OPERATOR('-'); /* unary minus */
4137 const char tmp = *s++;
4140 if (PL_expect == XOPERATOR)
4145 if (PL_expect == XOPERATOR)
4148 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4155 if (PL_expect != XOPERATOR) {
4156 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4157 PL_expect = XOPERATOR;
4158 force_ident(PL_tokenbuf, '*');
4171 if (PL_expect == XOPERATOR) {
4175 PL_tokenbuf[0] = '%';
4176 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4177 sizeof PL_tokenbuf - 1, FALSE);
4178 if (!PL_tokenbuf[1]) {
4181 PL_pending_ident = '%';
4192 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4199 const char tmp = *s++;
4205 goto just_a_word_zero_gv;
4208 switch (PL_expect) {
4214 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4216 PL_bufptr = s; /* update in case we back off */
4222 PL_expect = XTERMBLOCK;
4225 stuffstart = s - SvPVX(PL_linestr) - 1;
4229 while (isIDFIRST_lazy_if(s,UTF)) {
4232 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4233 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4234 if (tmp < 0) tmp = -tmp;
4249 sv = newSVpvn(s, len);
4251 d = scan_str(d,TRUE,TRUE);
4253 /* MUST advance bufptr here to avoid bogus
4254 "at end of line" context messages from yyerror().
4256 PL_bufptr = s + len;
4257 yyerror("Unterminated attribute parameter in attribute list");
4261 return REPORT(0); /* EOF indicator */
4265 sv_catsv(sv, PL_lex_stuff);
4266 attrs = append_elem(OP_LIST, attrs,
4267 newSVOP(OP_CONST, 0, sv));
4268 SvREFCNT_dec(PL_lex_stuff);
4269 PL_lex_stuff = NULL;
4272 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4274 if (PL_in_my == KEY_our) {
4276 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4278 /* skip to avoid loading attributes.pm */
4280 deprecate(":unique");
4283 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4286 /* NOTE: any CV attrs applied here need to be part of
4287 the CVf_BUILTIN_ATTRS define in cv.h! */
4288 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4290 CvLVALUE_on(PL_compcv);
4292 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4294 CvLOCKED_on(PL_compcv);
4296 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4298 CvMETHOD_on(PL_compcv);
4300 /* After we've set the flags, it could be argued that
4301 we don't need to do the attributes.pm-based setting
4302 process, and shouldn't bother appending recognized
4303 flags. To experiment with that, uncomment the
4304 following "else". (Note that's already been
4305 uncommented. That keeps the above-applied built-in
4306 attributes from being intercepted (and possibly
4307 rejected) by a package's attribute routines, but is
4308 justified by the performance win for the common case
4309 of applying only built-in attributes.) */
4311 attrs = append_elem(OP_LIST, attrs,
4312 newSVOP(OP_CONST, 0,
4316 if (*s == ':' && s[1] != ':')
4319 break; /* require real whitespace or :'s */
4320 /* XXX losing whitespace on sequential attributes here */
4324 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4325 if (*s != ';' && *s != '}' && *s != tmp
4326 && (tmp != '=' || *s != ')')) {
4327 const char q = ((*s == '\'') ? '"' : '\'');
4328 /* If here for an expression, and parsed no attrs, back
4330 if (tmp == '=' && !attrs) {
4334 /* MUST advance bufptr here to avoid bogus "at end of line"
4335 context messages from yyerror().
4338 yyerror( (const char *)
4340 ? Perl_form(aTHX_ "Invalid separator character "
4341 "%c%c%c in attribute list", q, *s, q)
4342 : "Unterminated attribute list" ) );
4350 start_force(PL_curforce);
4351 NEXTVAL_NEXTTOKE.opval = attrs;
4352 CURMAD('_', PL_nextwhite);
4357 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4358 (s - SvPVX(PL_linestr)) - stuffstart);
4366 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4367 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4375 const char tmp = *s++;
4380 const char tmp = *s++;
4388 if (PL_lex_brackets <= 0)
4389 yyerror("Unmatched right square bracket");
4392 if (PL_lex_state == LEX_INTERPNORMAL) {
4393 if (PL_lex_brackets == 0) {
4394 if (*s == '-' && s[1] == '>')
4395 PL_lex_state = LEX_INTERPENDMAYBE;
4396 else if (*s != '[' && *s != '{')
4397 PL_lex_state = LEX_INTERPEND;
4404 if (PL_lex_brackets > 100) {
4405 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4407 switch (PL_expect) {
4409 if (PL_lex_formbrack) {
4413 if (PL_oldoldbufptr == PL_last_lop)
4414 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4416 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4417 OPERATOR(HASHBRACK);
4419 while (s < PL_bufend && SPACE_OR_TAB(*s))
4422 PL_tokenbuf[0] = '\0';
4423 if (d < PL_bufend && *d == '-') {
4424 PL_tokenbuf[0] = '-';
4426 while (d < PL_bufend && SPACE_OR_TAB(*d))
4429 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4430 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4432 while (d < PL_bufend && SPACE_OR_TAB(*d))
4435 const char minus = (PL_tokenbuf[0] == '-');
4436 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4444 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4449 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4454 if (PL_oldoldbufptr == PL_last_lop)
4455 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4457 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4460 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4462 /* This hack is to get the ${} in the message. */
4464 yyerror("syntax error");
4467 OPERATOR(HASHBRACK);
4469 /* This hack serves to disambiguate a pair of curlies
4470 * as being a block or an anon hash. Normally, expectation
4471 * determines that, but in cases where we're not in a
4472 * position to expect anything in particular (like inside
4473 * eval"") we have to resolve the ambiguity. This code
4474 * covers the case where the first term in the curlies is a
4475 * quoted string. Most other cases need to be explicitly
4476 * disambiguated by prepending a "+" before the opening
4477 * curly in order to force resolution as an anon hash.
4479 * XXX should probably propagate the outer expectation
4480 * into eval"" to rely less on this hack, but that could
4481 * potentially break current behavior of eval"".
4485 if (*s == '\'' || *s == '"' || *s == '`') {
4486 /* common case: get past first string, handling escapes */
4487 for (t++; t < PL_bufend && *t != *s;)
4488 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4492 else if (*s == 'q') {
4495 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4498 /* skip q//-like construct */
4500 char open, close, term;
4503 while (t < PL_bufend && isSPACE(*t))
4505 /* check for q => */
4506 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4507 OPERATOR(HASHBRACK);
4511 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4515 for (t++; t < PL_bufend; t++) {
4516 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4518 else if (*t == open)
4522 for (t++; t < PL_bufend; t++) {
4523 if (*t == '\\' && t+1 < PL_bufend)
4525 else if (*t == close && --brackets <= 0)
4527 else if (*t == open)
4534 /* skip plain q word */
4535 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4538 else if (isALNUM_lazy_if(t,UTF)) {
4540 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4543 while (t < PL_bufend && isSPACE(*t))
4545 /* if comma follows first term, call it an anon hash */
4546 /* XXX it could be a comma expression with loop modifiers */
4547 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4548 || (*t == '=' && t[1] == '>')))
4549 OPERATOR(HASHBRACK);
4550 if (PL_expect == XREF)
4553 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4559 yylval.ival = CopLINE(PL_curcop);
4560 if (isSPACE(*s) || *s == '#')
4561 PL_copline = NOLINE; /* invalidate current command line number */
4566 if (PL_lex_brackets <= 0)
4567 yyerror("Unmatched right curly bracket");
4569 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4570 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4571 PL_lex_formbrack = 0;
4572 if (PL_lex_state == LEX_INTERPNORMAL) {
4573 if (PL_lex_brackets == 0) {
4574 if (PL_expect & XFAKEBRACK) {
4575 PL_expect &= XENUMMASK;
4576 PL_lex_state = LEX_INTERPEND;
4581 PL_thiswhite = newSVpvs("");
4582 sv_catpvn(PL_thiswhite,"}",1);
4585 return yylex(); /* ignore fake brackets */
4587 if (*s == '-' && s[1] == '>')
4588 PL_lex_state = LEX_INTERPENDMAYBE;
4589 else if (*s != '[' && *s != '{')
4590 PL_lex_state = LEX_INTERPEND;
4593 if (PL_expect & XFAKEBRACK) {
4594 PL_expect &= XENUMMASK;
4596 return yylex(); /* ignore fake brackets */
4598 start_force(PL_curforce);
4600 curmad('X', newSVpvn(s-1,1));
4601 CURMAD('_', PL_thiswhite);
4606 PL_thistoken = newSVpvs("");
4614 if (PL_expect == XOPERATOR) {
4615 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4616 && isIDFIRST_lazy_if(s,UTF))
4618 CopLINE_dec(PL_curcop);
4619 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4620 CopLINE_inc(PL_curcop);
4625 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4627 PL_expect = XOPERATOR;
4628 force_ident(PL_tokenbuf, '&');
4632 yylval.ival = (OPpENTERSUB_AMPER<<8);
4644 const char tmp = *s++;
4651 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4652 && strchr("+-*/%.^&|<",tmp))
4653 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4654 "Reversed %c= operator",(int)tmp);
4656 if (PL_expect == XSTATE && isALPHA(tmp) &&
4657 (s == PL_linestart+1 || s[-2] == '\n') )
4659 if (PL_in_eval && !PL_rsfp) {
4664 if (strnEQ(s,"=cut",4)) {
4680 PL_thiswhite = newSVpvs("");
4681 sv_catpvn(PL_thiswhite, PL_linestart,
4682 PL_bufend - PL_linestart);
4686 PL_doextract = TRUE;
4690 if (PL_lex_brackets < PL_lex_formbrack) {
4692 #ifdef PERL_STRICT_CR
4693 while (SPACE_OR_TAB(*t))
4695 while (SPACE_OR_TAB(*t) || *t == '\r')
4698 if (*t == '\n' || *t == '#') {
4709 const char tmp = *s++;
4711 /* was this !=~ where !~ was meant?
4712 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4714 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4715 const char *t = s+1;
4717 while (t < PL_bufend && isSPACE(*t))
4720 if (*t == '/' || *t == '?' ||
4721 ((*t == 'm' || *t == 's' || *t == 'y')
4722 && !isALNUM(t[1])) ||
4723 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4724 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4725 "!=~ should be !~");
4735 if (PL_expect != XOPERATOR) {
4736 if (s[1] != '<' && !strchr(s,'>'))
4739 s = scan_heredoc(s);
4741 s = scan_inputsymbol(s);
4742 TERM(sublex_start());
4748 SHop(OP_LEFT_SHIFT);
4762 const char tmp = *s++;
4764 SHop(OP_RIGHT_SHIFT);
4765 else if (tmp == '=')
4774 if (PL_expect == XOPERATOR) {
4775 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4777 deprecate_old(commaless_variable_list);
4778 return REPORT(','); /* grandfather non-comma-format format */
4782 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4783 PL_tokenbuf[0] = '@';
4784 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4785 sizeof PL_tokenbuf - 1, FALSE);
4786 if (PL_expect == XOPERATOR)
4787 no_op("Array length", s);
4788 if (!PL_tokenbuf[1])
4790 PL_expect = XOPERATOR;
4791 PL_pending_ident = '#';
4795 PL_tokenbuf[0] = '$';
4796 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4797 sizeof PL_tokenbuf - 1, FALSE);
4798 if (PL_expect == XOPERATOR)
4800 if (!PL_tokenbuf[1]) {
4802 yyerror("Final $ should be \\$ or $name");
4806 /* This kludge not intended to be bulletproof. */
4807 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4808 yylval.opval = newSVOP(OP_CONST, 0,
4809 newSViv(CopARYBASE_get(&PL_compiling)));
4810 yylval.opval->op_private = OPpCONST_ARYBASE;
4816 const char tmp = *s;
4817 if (PL_lex_state == LEX_NORMAL)
4820 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4821 && intuit_more(s)) {
4823 PL_tokenbuf[0] = '@';
4824 if (ckWARN(WARN_SYNTAX)) {
4827 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4830 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4831 while (t < PL_bufend && *t != ']')
4833 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4834 "Multidimensional syntax %.*s not supported",
4835 (int)((t - PL_bufptr) + 1), PL_bufptr);
4839 else if (*s == '{') {
4841 PL_tokenbuf[0] = '%';
4842 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4843 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4845 char tmpbuf[sizeof PL_tokenbuf];
4848 } while (isSPACE(*t));
4849 if (isIDFIRST_lazy_if(t,UTF)) {
4851 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4855 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4856 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4857 "You need to quote \"%s\"",
4864 PL_expect = XOPERATOR;
4865 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4866 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4867 if (!islop || PL_last_lop_op == OP_GREPSTART)
4868 PL_expect = XOPERATOR;
4869 else if (strchr("$@\"'`q", *s))
4870 PL_expect = XTERM; /* e.g. print $fh "foo" */
4871 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4872 PL_expect = XTERM; /* e.g. print $fh &sub */
4873 else if (isIDFIRST_lazy_if(s,UTF)) {
4874 char tmpbuf[sizeof PL_tokenbuf];
4876 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4877 if ((t2 = keyword(tmpbuf, len, 0))) {
4878 /* binary operators exclude handle interpretations */
4890 PL_expect = XTERM; /* e.g. print $fh length() */
4895 PL_expect = XTERM; /* e.g. print $fh subr() */
4898 else if (isDIGIT(*s))
4899 PL_expect = XTERM; /* e.g. print $fh 3 */
4900 else if (*s == '.' && isDIGIT(s[1]))
4901 PL_expect = XTERM; /* e.g. print $fh .3 */
4902 else if ((*s == '?' || *s == '-' || *s == '+')
4903 && !isSPACE(s[1]) && s[1] != '=')
4904 PL_expect = XTERM; /* e.g. print $fh -1 */
4905 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4907 PL_expect = XTERM; /* e.g. print $fh /.../
4908 XXX except DORDOR operator
4910 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4912 PL_expect = XTERM; /* print $fh <<"EOF" */
4915 PL_pending_ident = '$';
4919 if (PL_expect == XOPERATOR)
4921 PL_tokenbuf[0] = '@';
4922 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4923 if (!PL_tokenbuf[1]) {
4926 if (PL_lex_state == LEX_NORMAL)
4928 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4930 PL_tokenbuf[0] = '%';
4932 /* Warn about @ where they meant $. */
4933 if (*s == '[' || *s == '{') {
4934 if (ckWARN(WARN_SYNTAX)) {
4935 const char *t = s + 1;
4936 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4938 if (*t == '}' || *t == ']') {
4940 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4941 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4942 "Scalar value %.*s better written as $%.*s",
4943 (int)(t-PL_bufptr), PL_bufptr,
4944 (int)(t-PL_bufptr-1), PL_bufptr+1);
4949 PL_pending_ident = '@';
4952 case '/': /* may be division, defined-or, or pattern */
4953 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4957 case '?': /* may either be conditional or pattern */
4958 if(PL_expect == XOPERATOR) {
4966 /* A // operator. */
4976 /* Disable warning on "study /blah/" */
4977 if (PL_oldoldbufptr == PL_last_uni
4978 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4979 || memNE(PL_last_uni, "study", 5)
4980 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4983 s = scan_pat(s,OP_MATCH);
4984 TERM(sublex_start());
4988 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4989 #ifdef PERL_STRICT_CR
4992 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4994 && (s == PL_linestart || s[-1] == '\n') )
4996 PL_lex_formbrack = 0;
5000 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5006 yylval.ival = OPf_SPECIAL;
5012 if (PL_expect != XOPERATOR)
5017 case '0': case '1': case '2': case '3': case '4':
5018 case '5': case '6': case '7': case '8': case '9':
5019 s = scan_num(s, &yylval);
5020 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5021 if (PL_expect == XOPERATOR)
5026 s = scan_str(s,!!PL_madskills,FALSE);
5027 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5028 if (PL_expect == XOPERATOR) {
5029 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5031 deprecate_old(commaless_variable_list);
5032 return REPORT(','); /* grandfather non-comma-format format */
5039 yylval.ival = OP_CONST;
5040 TERM(sublex_start());
5043 s = scan_str(s,!!PL_madskills,FALSE);
5044 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5045 if (PL_expect == XOPERATOR) {
5046 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5048 deprecate_old(commaless_variable_list);
5049 return REPORT(','); /* grandfather non-comma-format format */
5056 yylval.ival = OP_CONST;
5057 /* FIXME. I think that this can be const if char *d is replaced by
5058 more localised variables. */
5059 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5060 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5061 yylval.ival = OP_STRINGIFY;
5065 TERM(sublex_start());
5068 s = scan_str(s,!!PL_madskills,FALSE);
5069 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5070 if (PL_expect == XOPERATOR)
5071 no_op("Backticks",s);
5074 readpipe_override();
5075 TERM(sublex_start());
5079 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5080 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5082 if (PL_expect == XOPERATOR)
5083 no_op("Backslash",s);
5087 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5088 char *start = s + 2;
5089 while (isDIGIT(*start) || *start == '_')
5091 if (*start == '.' && isDIGIT(start[1])) {
5092 s = scan_num(s, &yylval);
5095 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5096 else if (!isALPHA(*start) && (PL_expect == XTERM
5097 || PL_expect == XREF || PL_expect == XSTATE
5098 || PL_expect == XTERMORDORDOR)) {
5099 /* XXX Use gv_fetchpvn rather than stomping on a const string */
5100 const char c = *start;
5103 gv = gv_fetchpv(s, 0, SVt_PVCV);
5106 s = scan_num(s, &yylval);
5113 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5155 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5157 /* Some keywords can be followed by any delimiter, including ':' */
5158 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5159 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5160 (PL_tokenbuf[0] == 'q' &&
5161 strchr("qwxr", PL_tokenbuf[1])))));
5163 /* x::* is just a word, unless x is "CORE" */
5164 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5168 while (d < PL_bufend && isSPACE(*d))
5169 d++; /* no comments skipped here, or s### is misparsed */
5171 /* Is this a label? */
5172 if (!tmp && PL_expect == XSTATE
5173 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5175 yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5180 /* Check for keywords */
5181 tmp = keyword(PL_tokenbuf, len, 0);
5183 /* Is this a word before a => operator? */
5184 if (*d == '=' && d[1] == '>') {
5187 = (OP*)newSVOP(OP_CONST, 0,
5188 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5189 yylval.opval->op_private = OPpCONST_BARE;
5193 if (tmp < 0) { /* second-class keyword? */
5194 GV *ogv = NULL; /* override (winner) */
5195 GV *hgv = NULL; /* hidden (loser) */
5196 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5198 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5201 if (GvIMPORTED_CV(gv))
5203 else if (! CvMETHOD(cv))
5207 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5208 (gv = *gvp) && isGV_with_GP(gv) &&
5209 GvCVu(gv) && GvIMPORTED_CV(gv))
5216 tmp = 0; /* overridden by import or by GLOBAL */
5219 && -tmp==KEY_lock /* XXX generalizable kludge */
5222 tmp = 0; /* any sub overrides "weak" keyword */
5224 else { /* no override */
5226 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5227 Perl_warner(aTHX_ packWARN(WARN_MISC),
5228 "dump() better written as CORE::dump()");
5232 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5233 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5234 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5235 "Ambiguous call resolved as CORE::%s(), %s",
5236 GvENAME(hgv), "qualify as such or use &");
5243 default: /* not a keyword */
5244 /* Trade off - by using this evil construction we can pull the
5245 variable gv into the block labelled keylookup. If not, then
5246 we have to give it function scope so that the goto from the
5247 earlier ':' case doesn't bypass the initialisation. */
5249 just_a_word_zero_gv:
5257 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5260 SV *nextPL_nextwhite = 0;
5264 /* Get the rest if it looks like a package qualifier */
5266 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5268 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5271 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5272 *s == '\'' ? "'" : "::");
5277 if (PL_expect == XOPERATOR) {
5278 if (PL_bufptr == PL_linestart) {
5279 CopLINE_dec(PL_curcop);
5280 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5281 CopLINE_inc(PL_curcop);
5284 no_op("Bareword",s);
5287 /* Look for a subroutine with this name in current package,
5288 unless name is "Foo::", in which case Foo is a bearword
5289 (and a package name). */
5291 if (len > 2 && !PL_madskills &&
5292 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5294 if (ckWARN(WARN_BAREWORD)
5295 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5296 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5297 "Bareword \"%s\" refers to nonexistent package",
5300 PL_tokenbuf[len] = '\0';
5306 /* Mustn't actually add anything to a symbol table.
5307 But also don't want to "initialise" any placeholder
5308 constants that might already be there into full
5309 blown PVGVs with attached PVCV. */
5310 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5311 GV_NOADD_NOINIT, SVt_PVCV);
5316 /* if we saw a global override before, get the right name */
5319 sv = newSVpvs("CORE::GLOBAL::");
5320 sv_catpv(sv,PL_tokenbuf);
5323 /* If len is 0, newSVpv does strlen(), which is correct.
5324 If len is non-zero, then it will be the true length,
5325 and so the scalar will be created correctly. */
5326 sv = newSVpv(PL_tokenbuf,len);
5329 if (PL_madskills && !PL_thistoken) {
5330 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5331 PL_thistoken = newSVpv(start,s - start);
5332 PL_realtokenstart = s - SvPVX(PL_linestr);
5336 /* Presume this is going to be a bareword of some sort. */
5339 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5340 yylval.opval->op_private = OPpCONST_BARE;
5341 /* UTF-8 package name? */
5342 if (UTF && !IN_BYTES &&
5343 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5346 /* And if "Foo::", then that's what it certainly is. */
5351 /* Do the explicit type check so that we don't need to force
5352 the initialisation of the symbol table to have a real GV.
5353 Beware - gv may not really be a PVGV, cv may not really be
5354 a PVCV, (because of the space optimisations that gv_init
5355 understands) But they're true if for this symbol there is
5356 respectively a typeglob and a subroutine.
5358 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5359 /* Real typeglob, so get the real subroutine: */
5361 /* A proxy for a subroutine in this package? */
5362 : SvOK(gv) ? (CV *) gv : NULL)
5365 /* See if it's the indirect object for a list operator. */
5367 if (PL_oldoldbufptr &&
5368 PL_oldoldbufptr < PL_bufptr &&
5369 (PL_oldoldbufptr == PL_last_lop
5370 || PL_oldoldbufptr == PL_last_uni) &&
5371 /* NO SKIPSPACE BEFORE HERE! */
5372 (PL_expect == XREF ||
5373 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5375 bool immediate_paren = *s == '(';
5377 /* (Now we can afford to cross potential line boundary.) */
5378 s = SKIPSPACE2(s,nextPL_nextwhite);
5380 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5383 /* Two barewords in a row may indicate method call. */
5385 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5386 (tmp = intuit_method(s, gv, cv)))
5389 /* If not a declared subroutine, it's an indirect object. */
5390 /* (But it's an indir obj regardless for sort.) */
5391 /* Also, if "_" follows a filetest operator, it's a bareword */
5394 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5396 (PL_last_lop_op != OP_MAPSTART &&
5397 PL_last_lop_op != OP_GREPSTART))))
5398 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5399 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5402 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5407 PL_expect = XOPERATOR;
5410 s = SKIPSPACE2(s,nextPL_nextwhite);
5411 PL_nextwhite = nextPL_nextwhite;
5416 /* Is this a word before a => operator? */
5417 if (*s == '=' && s[1] == '>' && !pkgname) {
5419 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5420 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5421 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5425 /* If followed by a paren, it's certainly a subroutine. */
5430 while (SPACE_OR_TAB(*d))
5432 if (*d == ')' && (sv = gv_const_sv(gv))) {
5439 PL_nextwhite = PL_thiswhite;
5442 start_force(PL_curforce);
5444 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5445 PL_expect = XOPERATOR;
5448 PL_nextwhite = nextPL_nextwhite;
5449 curmad('X', PL_thistoken);
5450 PL_thistoken = newSVpvs("");
5458 /* If followed by var or block, call it a method (unless sub) */
5460 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5461 PL_last_lop = PL_oldbufptr;
5462 PL_last_lop_op = OP_METHOD;
5466 /* If followed by a bareword, see if it looks like indir obj. */
5469 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5470 && (tmp = intuit_method(s, gv, cv)))
5473 /* Not a method, so call it a subroutine (if defined) */
5476 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5477 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5478 "Ambiguous use of -%s resolved as -&%s()",
5479 PL_tokenbuf, PL_tokenbuf);
5480 /* Check for a constant sub */
5481 if ((sv = gv_const_sv(gv))) {
5483 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5484 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5485 yylval.opval->op_private = 0;
5489 /* Resolve to GV now. */
5490 if (SvTYPE(gv) != SVt_PVGV) {
5491 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5492 assert (SvTYPE(gv) == SVt_PVGV);
5493 /* cv must have been some sort of placeholder, so
5494 now needs replacing with a real code reference. */
5498 op_free(yylval.opval);
5499 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5500 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5501 PL_last_lop = PL_oldbufptr;
5502 PL_last_lop_op = OP_ENTERSUB;
5503 /* Is there a prototype? */
5511 const char *proto = SvPV_const((SV*)cv, protolen);
5514 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5516 while (*proto == ';')
5518 if (*proto == '&' && *s == '{') {
5519 sv_setpv(PL_subname,
5522 "__ANON__" : "__ANON__::__ANON__"));
5529 PL_nextwhite = PL_thiswhite;
5532 start_force(PL_curforce);
5533 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5536 PL_nextwhite = nextPL_nextwhite;
5537 curmad('X', PL_thistoken);
5538 PL_thistoken = newSVpvs("");
5545 /* Guess harder when madskills require "best effort". */
5546 if (PL_madskills && (!gv || !GvCVu(gv))) {
5547 int probable_sub = 0;
5548 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5550 else if (isALPHA(*s)) {
5554 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5555 if (!keyword(tmpbuf, tmplen, 0))
5558 while (d < PL_bufend && isSPACE(*d))
5560 if (*d == '=' && d[1] == '>')
5565 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5566 op_free(yylval.opval);
5567 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5568 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5569 PL_last_lop = PL_oldbufptr;
5570 PL_last_lop_op = OP_ENTERSUB;
5571 PL_nextwhite = PL_thiswhite;
5573 start_force(PL_curforce);
5574 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5576 PL_nextwhite = nextPL_nextwhite;
5577 curmad('X', PL_thistoken);
5578 PL_thistoken = newSVpvs("");
5583 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5590 /* Call it a bare word */
5592 if (PL_hints & HINT_STRICT_SUBS)
5593 yylval.opval->op_private |= OPpCONST_STRICT;
5596 if (lastchar != '-') {
5597 if (ckWARN(WARN_RESERVED)) {
5601 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5602 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5609 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5610 && ckWARN_d(WARN_AMBIGUOUS)) {
5611 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5612 "Operator or semicolon missing before %c%s",
5613 lastchar, PL_tokenbuf);
5614 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5615 "Ambiguous use of %c resolved as operator %c",
5616 lastchar, lastchar);
5622 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5623 newSVpv(CopFILE(PL_curcop),0));
5627 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5628 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5631 case KEY___PACKAGE__:
5632 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5634 ? newSVhek(HvNAME_HEK(PL_curstash))
5641 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5642 const char *pname = "main";
5643 if (PL_tokenbuf[2] == 'D')
5644 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5645 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5649 GvIOp(gv) = newIO();
5650 IoIFP(GvIOp(gv)) = PL_rsfp;
5651 #if defined(HAS_FCNTL) && defined(F_SETFD)
5653 const int fd = PerlIO_fileno(PL_rsfp);
5654 fcntl(fd,F_SETFD,fd >= 3);
5657 /* Mark this internal pseudo-handle as clean */
5658 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5660 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5661 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5662 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5664 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5665 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5666 /* if the script was opened in binmode, we need to revert
5667 * it to text mode for compatibility; but only iff it has CRs
5668 * XXX this is a questionable hack at best. */
5669 if (PL_bufend-PL_bufptr > 2
5670 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5673 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5674 loc = PerlIO_tell(PL_rsfp);
5675 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5678 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5680 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5681 #endif /* NETWARE */
5682 #ifdef PERLIO_IS_STDIO /* really? */
5683 # if defined(__BORLANDC__)
5684 /* XXX see note in do_binmode() */
5685 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5689 PerlIO_seek(PL_rsfp, loc, 0);
5693 #ifdef PERLIO_LAYERS
5696 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5697 else if (PL_encoding) {
5704 XPUSHs(PL_encoding);
5706 call_method("name", G_SCALAR);
5710 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5711 Perl_form(aTHX_ ":encoding(%"SVf")",
5720 if (PL_realtokenstart >= 0) {
5721 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5723 PL_endwhite = newSVpvs("");
5724 sv_catsv(PL_endwhite, PL_thiswhite);
5726 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5727 PL_realtokenstart = -1;
5729 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5730 SvCUR(PL_endwhite))) != Nullch) ;
5745 if (PL_expect == XSTATE) {
5752 if (*s == ':' && s[1] == ':') {
5755 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5756 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5757 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5760 else if (tmp == KEY_require || tmp == KEY_do)
5761 /* that's a way to remember we saw "CORE::" */
5774 LOP(OP_ACCEPT,XTERM);
5780 LOP(OP_ATAN2,XTERM);
5786 LOP(OP_BINMODE,XTERM);
5789 LOP(OP_BLESS,XTERM);
5798 /* When 'use switch' is in effect, continue has a dual
5799 life as a control operator. */
5801 if (!FEATURE_IS_ENABLED("switch"))
5804 /* We have to disambiguate the two senses of
5805 "continue". If the next token is a '{' then
5806 treat it as the start of a continue block;
5807 otherwise treat it as a control operator.
5819 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5836 if (!PL_cryptseen) {
5837 PL_cryptseen = TRUE;
5841 LOP(OP_CRYPT,XTERM);
5844 LOP(OP_CHMOD,XTERM);
5847 LOP(OP_CHOWN,XTERM);
5850 LOP(OP_CONNECT,XTERM);
5869 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5870 if (orig_keyword == KEY_do) {
5879 PL_hints |= HINT_BLOCK_SCOPE;
5889 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5890 LOP(OP_DBMOPEN,XTERM);
5896 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5903 yylval.ival = CopLINE(PL_curcop);
5919 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5920 UNIBRACK(OP_ENTEREVAL);
5938 case KEY_endhostent:
5944 case KEY_endservent:
5947 case KEY_endprotoent:
5958 yylval.ival = CopLINE(PL_curcop);
5960 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5963 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5966 if ((PL_bufend - p) >= 3 &&
5967 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5969 else if ((PL_bufend - p) >= 4 &&
5970 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5973 if (isIDFIRST_lazy_if(p,UTF)) {
5974 p = scan_ident(p, PL_bufend,
5975 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5979 Perl_croak(aTHX_ "Missing $ on loop variable");
5981 s = SvPVX(PL_linestr) + soff;
5987 LOP(OP_FORMLINE,XTERM);
5993 LOP(OP_FCNTL,XTERM);
5999 LOP(OP_FLOCK,XTERM);
6008 LOP(OP_GREPSTART, XREF);
6011 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6026 case KEY_getpriority:
6027 LOP(OP_GETPRIORITY,XTERM);
6029 case KEY_getprotobyname:
6032 case KEY_getprotobynumber:
6033 LOP(OP_GPBYNUMBER,XTERM);
6035 case KEY_getprotoent:
6047 case KEY_getpeername:
6048 UNI(OP_GETPEERNAME);
6050 case KEY_gethostbyname:
6053 case KEY_gethostbyaddr:
6054 LOP(OP_GHBYADDR,XTERM);
6056 case KEY_gethostent:
6059 case KEY_getnetbyname:
6062 case KEY_getnetbyaddr:
6063 LOP(OP_GNBYADDR,XTERM);
6068 case KEY_getservbyname:
6069 LOP(OP_GSBYNAME,XTERM);
6071 case KEY_getservbyport:
6072 LOP(OP_GSBYPORT,XTERM);
6074 case KEY_getservent:
6077 case KEY_getsockname:
6078 UNI(OP_GETSOCKNAME);
6080 case KEY_getsockopt:
6081 LOP(OP_GSOCKOPT,XTERM);
6096 yylval.ival = CopLINE(PL_curcop);
6107 yylval.ival = CopLINE(PL_curcop);
6111 LOP(OP_INDEX,XTERM);
6117 LOP(OP_IOCTL,XTERM);
6129 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6161 LOP(OP_LISTEN,XTERM);
6170 s = scan_pat(s,OP_MATCH);
6171 TERM(sublex_start());
6174 LOP(OP_MAPSTART, XREF);
6177 LOP(OP_MKDIR,XTERM);
6180 LOP(OP_MSGCTL,XTERM);
6183 LOP(OP_MSGGET,XTERM);
6186 LOP(OP_MSGRCV,XTERM);
6189 LOP(OP_MSGSND,XTERM);
6194 PL_in_my = (U16)tmp;
6196 if (isIDFIRST_lazy_if(s,UTF)) {
6200 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6201 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6203 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6204 if (!PL_in_my_stash) {
6207 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6211 if (PL_madskills) { /* just add type to declarator token */
6212 sv_catsv(PL_thistoken, PL_nextwhite);
6214 sv_catpvn(PL_thistoken, start, s - start);
6222 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6229 s = tokenize_use(0, s);
6233 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6240 if (isIDFIRST_lazy_if(s,UTF)) {
6242 for (d = s; isALNUM_lazy_if(d,UTF);)
6244 for (t=d; isSPACE(*t);)
6246 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6248 && !(t[0] == '=' && t[1] == '>')
6250 int parms_len = (int)(d-s);
6251 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6252 "Precedence problem: open %.*s should be open(%.*s)",
6253 parms_len, s, parms_len, s);
6259 yylval.ival = OP_OR;
6269 LOP(OP_OPEN_DIR,XTERM);
6272 checkcomma(s,PL_tokenbuf,"filehandle");
6276 checkcomma(s,PL_tokenbuf,"filehandle");
6295 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6299 LOP(OP_PIPE_OP,XTERM);
6302 s = scan_str(s,!!PL_madskills,FALSE);
6305 yylval.ival = OP_CONST;
6306 TERM(sublex_start());
6312 s = scan_str(s,!!PL_madskills,FALSE);
6315 PL_expect = XOPERATOR;
6317 if (SvCUR(PL_lex_stuff)) {
6320 d = SvPV_force(PL_lex_stuff, len);
6322 for (; isSPACE(*d) && len; --len, ++d)
6327 if (!warned && ckWARN(WARN_QW)) {
6328 for (; !isSPACE(*d) && len; --len, ++d) {
6330 Perl_warner(aTHX_ packWARN(WARN_QW),
6331 "Possible attempt to separate words with commas");
6334 else if (*d == '#') {
6335 Perl_warner(aTHX_ packWARN(WARN_QW),
6336 "Possible attempt to put comments in qw() list");
6342 for (; !isSPACE(*d) && len; --len, ++d)
6345 sv = newSVpvn(b, d-b);
6346 if (DO_UTF8(PL_lex_stuff))
6348 words = append_elem(OP_LIST, words,
6349 newSVOP(OP_CONST, 0, tokeq(sv)));
6353 start_force(PL_curforce);
6354 NEXTVAL_NEXTTOKE.opval = words;
6359 SvREFCNT_dec(PL_lex_stuff);
6360 PL_lex_stuff = NULL;
6366 s = scan_str(s,!!PL_madskills,FALSE);
6369 yylval.ival = OP_STRINGIFY;
6370 if (SvIVX(PL_lex_stuff) == '\'')
6371 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6372 TERM(sublex_start());
6375 s = scan_pat(s,OP_QR);
6376 TERM(sublex_start());
6379 s = scan_str(s,!!PL_madskills,FALSE);
6382 readpipe_override();
6383 TERM(sublex_start());
6391 s = force_version(s, FALSE);
6393 else if (*s != 'v' || !isDIGIT(s[1])
6394 || (s = force_version(s, TRUE), *s == 'v'))
6396 *PL_tokenbuf = '\0';
6397 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6398 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6399 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6401 yyerror("<> should be quotes");
6403 if (orig_keyword == KEY_require) {
6411 PL_last_uni = PL_oldbufptr;
6412 PL_last_lop_op = OP_REQUIRE;
6414 return REPORT( (int)REQUIRE );
6420 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6424 LOP(OP_RENAME,XTERM);
6433 LOP(OP_RINDEX,XTERM);
6443 UNIDOR(OP_READLINE);
6447 UNIDOR(OP_BACKTICK);
6456 LOP(OP_REVERSE,XTERM);
6459 UNIDOR(OP_READLINK);
6467 TERM(sublex_start());
6469 TOKEN(1); /* force error */
6472 checkcomma(s,PL_tokenbuf,"filehandle");
6482 LOP(OP_SELECT,XTERM);
6488 LOP(OP_SEMCTL,XTERM);
6491 LOP(OP_SEMGET,XTERM);
6494 LOP(OP_SEMOP,XTERM);
6500 LOP(OP_SETPGRP,XTERM);
6502 case KEY_setpriority:
6503 LOP(OP_SETPRIORITY,XTERM);
6505 case KEY_sethostent:
6511 case KEY_setservent:
6514 case KEY_setprotoent:
6524 LOP(OP_SEEKDIR,XTERM);
6526 case KEY_setsockopt:
6527 LOP(OP_SSOCKOPT,XTERM);
6533 LOP(OP_SHMCTL,XTERM);
6536 LOP(OP_SHMGET,XTERM);
6539 LOP(OP_SHMREAD,XTERM);
6542 LOP(OP_SHMWRITE,XTERM);
6545 LOP(OP_SHUTDOWN,XTERM);
6554 LOP(OP_SOCKET,XTERM);
6556 case KEY_socketpair:
6557 LOP(OP_SOCKPAIR,XTERM);
6560 checkcomma(s,PL_tokenbuf,"subroutine name");
6562 if (*s == ';' || *s == ')') /* probably a close */
6563 Perl_croak(aTHX_ "sort is now a reserved word");
6565 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6569 LOP(OP_SPLIT,XTERM);
6572 LOP(OP_SPRINTF,XTERM);
6575 LOP(OP_SPLICE,XTERM);
6590 LOP(OP_SUBSTR,XTERM);
6596 char tmpbuf[sizeof PL_tokenbuf];
6597 SSize_t tboffset = 0;
6598 expectation attrful;
6599 bool have_name, have_proto;
6600 const int key = tmp;
6605 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6606 SV *subtoken = newSVpvn(tstart, s - tstart);
6610 s = SKIPSPACE2(s,tmpwhite);
6615 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6616 (*s == ':' && s[1] == ':'))
6623 attrful = XATTRBLOCK;
6624 /* remember buffer pos'n for later force_word */
6625 tboffset = s - PL_oldbufptr;
6626 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6629 nametoke = newSVpvn(s, d - s);
6631 if (memchr(tmpbuf, ':', len))
6632 sv_setpvn(PL_subname, tmpbuf, len);
6634 sv_setsv(PL_subname,PL_curstname);
6635 sv_catpvs(PL_subname,"::");
6636 sv_catpvn(PL_subname,tmpbuf,len);
6643 CURMAD('X', nametoke);
6644 CURMAD('_', tmpwhite);
6645 (void) force_word(PL_oldbufptr + tboffset, WORD,
6648 s = SKIPSPACE2(d,tmpwhite);
6655 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6656 PL_expect = XTERMBLOCK;
6657 attrful = XATTRTERM;
6658 sv_setpvn(PL_subname,"?",1);
6662 if (key == KEY_format) {
6664 PL_lex_formbrack = PL_lex_brackets + 1;
6666 PL_thistoken = subtoken;
6670 (void) force_word(PL_oldbufptr + tboffset, WORD,
6676 /* Look for a prototype */
6679 bool bad_proto = FALSE;
6680 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6682 s = scan_str(s,!!PL_madskills,FALSE);
6684 Perl_croak(aTHX_ "Prototype not terminated");
6685 /* strip spaces and check for bad characters */
6686 d = SvPVX(PL_lex_stuff);
6688 for (p = d; *p; ++p) {
6691 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6697 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6698 "Illegal character in prototype for %"SVf" : %s",
6699 SVfARG(PL_subname), d);
6700 SvCUR_set(PL_lex_stuff, tmp);
6705 CURMAD('q', PL_thisopen);
6706 CURMAD('_', tmpwhite);
6707 CURMAD('=', PL_thisstuff);
6708 CURMAD('Q', PL_thisclose);
6709 NEXTVAL_NEXTTOKE.opval =
6710 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6711 PL_lex_stuff = Nullsv;
6714 s = SKIPSPACE2(s,tmpwhite);
6722 if (*s == ':' && s[1] != ':')
6723 PL_expect = attrful;
6724 else if (*s != '{' && key == KEY_sub) {
6726 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6728 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6735 curmad('^', newSVpvs(""));
6736 CURMAD('_', tmpwhite);
6740 PL_thistoken = subtoken;
6743 NEXTVAL_NEXTTOKE.opval =
6744 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6745 PL_lex_stuff = NULL;
6750 sv_setpv(PL_subname,
6752 (PL_curstash ? "__ANON__" : "__ANON__::__ANON__"));
6756 (void) force_word(PL_oldbufptr + tboffset, WORD,
6766 LOP(OP_SYSTEM,XREF);
6769 LOP(OP_SYMLINK,XTERM);
6772 LOP(OP_SYSCALL,XTERM);
6775 LOP(OP_SYSOPEN,XTERM);
6778 LOP(OP_SYSSEEK,XTERM);
6781 LOP(OP_SYSREAD,XTERM);
6784 LOP(OP_SYSWRITE,XTERM);
6788 TERM(sublex_start());
6809 LOP(OP_TRUNCATE,XTERM);
6821 yylval.ival = CopLINE(PL_curcop);
6825 yylval.ival = CopLINE(PL_curcop);
6829 LOP(OP_UNLINK,XTERM);
6835 LOP(OP_UNPACK,XTERM);
6838 LOP(OP_UTIME,XTERM);
6844 LOP(OP_UNSHIFT,XTERM);
6847 s = tokenize_use(1, s);
6857 yylval.ival = CopLINE(PL_curcop);
6861 yylval.ival = CopLINE(PL_curcop);
6865 PL_hints |= HINT_BLOCK_SCOPE;
6872 LOP(OP_WAITPID,XTERM);
6881 ctl_l[0] = toCTRL('L');
6883 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6886 /* Make sure $^L is defined */
6887 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6892 if (PL_expect == XOPERATOR)
6898 yylval.ival = OP_XOR;
6903 TERM(sublex_start());
6908 #pragma segment Main
6912 S_pending_ident(pTHX)
6917 /* pit holds the identifier we read and pending_ident is reset */
6918 char pit = PL_pending_ident;
6919 PL_pending_ident = 0;
6921 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6922 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6923 "### Pending identifier '%s'\n", PL_tokenbuf); });
6925 /* if we're in a my(), we can't allow dynamics here.
6926 $foo'bar has already been turned into $foo::bar, so
6927 just check for colons.
6929 if it's a legal name, the OP is a PADANY.
6932 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6933 if (strchr(PL_tokenbuf,':'))
6934 yyerror(Perl_form(aTHX_ "No package name allowed for "
6935 "variable %s in \"our\"",
6937 tmp = allocmy(PL_tokenbuf);
6940 if (strchr(PL_tokenbuf,':'))
6941 yyerror(Perl_form(aTHX_ PL_no_myglob,
6942 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6944 yylval.opval = newOP(OP_PADANY, 0);
6945 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6951 build the ops for accesses to a my() variable.
6953 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6954 then used in a comparison. This catches most, but not
6955 all cases. For instance, it catches
6956 sort { my($a); $a <=> $b }
6958 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6959 (although why you'd do that is anyone's guess).
6962 if (!strchr(PL_tokenbuf,':')) {
6964 tmp = pad_findmy(PL_tokenbuf);
6965 if (tmp != NOT_IN_PAD) {
6966 /* might be an "our" variable" */
6967 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6968 /* build ops for a bareword */
6969 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6970 HEK * const stashname = HvNAME_HEK(stash);
6971 SV * const sym = newSVhek(stashname);
6972 sv_catpvs(sym, "::");
6973 sv_catpv(sym, PL_tokenbuf+1);
6974 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6975 yylval.opval->op_private = OPpCONST_ENTERED;
6978 ? (GV_ADDMULTI | GV_ADDINEVAL)
6981 ((PL_tokenbuf[0] == '$') ? SVt_PV
6982 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6987 /* if it's a sort block and they're naming $a or $b */
6988 if (PL_last_lop_op == OP_SORT &&
6989 PL_tokenbuf[0] == '$' &&
6990 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6993 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6994 d < PL_bufend && *d != '\n';
6997 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6998 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7004 yylval.opval = newOP(OP_PADANY, 0);
7005 yylval.opval->op_targ = tmp;
7011 Whine if they've said @foo in a doublequoted string,
7012 and @foo isn't a variable we can find in the symbol
7015 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7016 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
7017 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7018 && ckWARN(WARN_AMBIGUOUS)
7019 /* DO NOT warn for @- and @+ */
7020 && !( PL_tokenbuf[2] == '\0' &&
7021 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7024 /* Downgraded from fatal to warning 20000522 mjd */
7025 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7026 "Possible unintended interpolation of %s in string",
7031 /* build ops for a bareword */
7032 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
7033 yylval.opval->op_private = OPpCONST_ENTERED;
7036 /* If the identifier refers to a stash, don't autovivify it.
7037 * Change 24660 had the side effect of causing symbol table
7038 * hashes to always be defined, even if they were freshly
7039 * created and the only reference in the entire program was
7040 * the single statement with the defined %foo::bar:: test.
7041 * It appears that all code in the wild doing this actually
7042 * wants to know whether sub-packages have been loaded, so
7043 * by avoiding auto-vivifying symbol tables, we ensure that
7044 * defined %foo::bar:: continues to be false, and the existing
7045 * tests still give the expected answers, even though what
7046 * they're actually testing has now changed subtly.
7048 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
7050 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7051 ((PL_tokenbuf[0] == '$') ? SVt_PV
7052 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7058 * The following code was generated by perl_keyword.pl.
7062 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7067 case 1: /* 5 tokens of length 1 */
7099 case 2: /* 18 tokens of length 2 */
7245 case 3: /* 29 tokens of length 3 */
7249 if (name[1] == 'N' &&
7312 if (name[1] == 'i' &&
7334 return (all_keywords || FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7352 if (name[1] == 'o' &&
7361 if (name[1] == 'e' &&
7370 if (name[1] == 'n' &&
7379 if (name[1] == 'o' &&
7388 if (name[1] == 'a' &&
7397 if (name[1] == 'o' &&
7459 if (name[1] == 'e' &&
7473 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7499 if (name[1] == 'i' &&
7508 if (name[1] == 's' &&
7517 if (name[1] == 'e' &&
7526 if (name[1] == 'o' &&
7538 case 4: /* 41 tokens of length 4 */
7542 if (name[1] == 'O' &&
7552 if (name[1] == 'N' &&
7562 if (name[1] == 'i' &&
7572 if (name[1] == 'h' &&
7582 if (name[1] == 'u' &&
7595 if (name[2] == 'c' &&
7604 if (name[2] == 's' &&
7613 if (name[2] == 'a' &&
7649 if (name[1] == 'o' &&
7662 if (name[2] == 't' &&
7671 if (name[2] == 'o' &&
7680 if (name[2] == 't' &&
7689 if (name[2] == 'e' &&
7702 if (name[1] == 'o' &&
7715 if (name[2] == 'y' &&
7724 if (name[2] == 'l' &&
7740 if (name[2] == 's' &&
7749 if (name[2] == 'n' &&
7758 if (name[2] == 'c' &&
7771 if (name[1] == 'e' &&
7781 if (name[1] == 'p' &&
7794 if (name[2] == 'c' &&
7803 if (name[2] == 'p' &&
7812 if (name[2] == 's' &&
7828 if (name[2] == 'n' &&
7898 if (name[2] == 'r' &&
7907 if (name[2] == 'r' &&
7916 if (name[2] == 'a' &&
7932 if (name[2] == 'l' &&
7994 if (name[2] == 'e' &&
7997 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8010 case 5: /* 39 tokens of length 5 */
8014 if (name[1] == 'E' &&
8025 if (name[1] == 'H' &&
8039 if (name[2] == 'a' &&
8049 if (name[2] == 'a' &&
8066 if (name[2] == 'e' &&
8076 if (name[2] == 'e' &&
8080 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8096 if (name[3] == 'i' &&
8105 if (name[3] == 'o' &&
8141 if (name[2] == 'o' &&
8151 if (name[2] == 'y' &&
8165 if (name[1] == 'l' &&
8179 if (name[2] == 'n' &&
8189 if (name[2] == 'o' &&
8203 if (name[1] == 'i' &&
8208 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8217 if (name[2] == 'd' &&
8227 if (name[2] == 'c' &&
8244 if (name[2] == 'c' &&
8254 if (name[2] == 't' &&
8268 if (name[1] == 'k' &&
8279 if (name[1] == 'r' &&
8293 if (name[2] == 's' &&
8303 if (name[2] == 'd' &&
8320 if (name[2] == 'm' &&
8330 if (name[2] == 'i' &&
8340 if (name[2] == 'e' &&
8350 if (name[2] == 'l' &&
8360 if (name[2] == 'a' &&
8373 if (name[3] == 't' &&
8376 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8382 if (name[3] == 'd' &&
8399 if (name[1] == 'i' &&
8413 if (name[2] == 'a' &&
8426 if (name[3] == 'e' &&
8461 if (name[2] == 'i' &&
8478 if (name[2] == 'i' &&
8488 if (name[2] == 'i' &&
8505 case 6: /* 33 tokens of length 6 */
8509 if (name[1] == 'c' &&
8524 if (name[2] == 'l' &&
8535 if (name[2] == 'r' &&
8550 if (name[1] == 'e' &&
8565 if (name[2] == 's' &&
8570 if(ckWARN_d(WARN_SYNTAX))
8571 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8577 if (name[2] == 'i' &&
8595 if (name[2] == 'l' &&
8606 if (name[2] == 'r' &&
8621 if (name[1] == 'm' &&
8636 if (name[2] == 'n' &&
8647 if (name[2] == 's' &&
8662 if (name[1] == 's' &&
8668 if (name[4] == 't' &&
8677 if (name[4] == 'e' &&
8686 if (name[4] == 'c' &&
8695 if (name[4] == 'n' &&
8711 if (name[1] == 'r' &&
8729 if (name[3] == 'a' &&
8739 if (name[3] == 'u' &&
8753 if (name[2] == 'n' &&
8771 if (name[2] == 'a' &&
8785 if (name[3] == 'e' &&
8798 if (name[4] == 't' &&
8807 if (name[4] == 'e' &&
8829 if (name[4] == 't' &&
8838 if (name[4] == 'e' &&
8854 if (name[2] == 'c' &&
8865 if (name[2] == 'l' &&
8876 if (name[2] == 'b' &&
8887 if (name[2] == 's' &&
8910 if (name[4] == 's' &&
8919 if (name[4] == 'n' &&
8932 if (name[3] == 'a' &&
8949 if (name[1] == 'a' &&
8964 case 7: /* 29 tokens of length 7 */
8968 if (name[1] == 'E' &&
8981 if (name[1] == '_' &&
8994 if (name[1] == 'i' &&
9001 return -KEY_binmode;
9007 if (name[1] == 'o' &&
9014 return -KEY_connect;
9023 if (name[2] == 'm' &&
9029 return -KEY_dbmopen;
9040 if (name[4] == 'u' &&
9044 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9050 if (name[4] == 'n' &&
9071 if (name[1] == 'o' &&
9084 if (name[1] == 'e' &&
9091 if (name[5] == 'r' &&
9094 return -KEY_getpgrp;
9100 if (name[5] == 'i' &&
9103 return -KEY_getppid;
9116 if (name[1] == 'c' &&
9123 return -KEY_lcfirst;
9129 if (name[1] == 'p' &&
9136 return -KEY_opendir;
9142 if (name[1] == 'a' &&
9160 if (name[3] == 'd' &&
9165 return -KEY_readdir;
9171 if (name[3] == 'u' &&
9182 if (name[3] == 'e' &&
9187 return -KEY_reverse;
9206 if (name[3] == 'k' &&
9211 return -KEY_seekdir;
9217 if (name[3] == 'p' &&
9222 return -KEY_setpgrp;
9232 if (name[2] == 'm' &&
9238 return -KEY_shmread;
9244 if (name[2] == 'r' &&
9250 return -KEY_sprintf;
9259 if (name[3] == 'l' &&
9264 return -KEY_symlink;
9273 if (name[4] == 'a' &&
9277 return -KEY_syscall;
9283 if (name[4] == 'p' &&
9287 return -KEY_sysopen;
9293 if (name[4] == 'e' &&
9297 return -KEY_sysread;
9303 if (name[4] == 'e' &&
9307 return -KEY_sysseek;
9325 if (name[1] == 'e' &&
9332 return -KEY_telldir;
9341 if (name[2] == 'f' &&
9347 return -KEY_ucfirst;
9353 if (name[2] == 's' &&
9359 return -KEY_unshift;
9369 if (name[1] == 'a' &&
9376 return -KEY_waitpid;
9385 case 8: /* 26 tokens of length 8 */
9389 if (name[1] == 'U' &&
9397 return KEY_AUTOLOAD;
9408 if (name[3] == 'A' &&
9414 return KEY___DATA__;
9420 if (name[3] == 'I' &&
9426 return -KEY___FILE__;
9432 if (name[3] == 'I' &&
9438 return -KEY___LINE__;
9454 if (name[2] == 'o' &&
9461 return -KEY_closedir;
9467 if (name[2] == 'n' &&
9474 return -KEY_continue;
9484 if (name[1] == 'b' &&
9492 return -KEY_dbmclose;
9498 if (name[1] == 'n' &&
9504 if (name[4] == 'r' &&
9509 return -KEY_endgrent;
9515 if (name[4] == 'w' &&
9520 return -KEY_endpwent;
9533 if (name[1] == 'o' &&
9541 return -KEY_formline;
9547 if (name[1] == 'e' &&
9558 if (name[6] == 'n' &&
9561 return -KEY_getgrent;
9567 if (name[6] == 'i' &&
9570 return -KEY_getgrgid;
9576 if (name[6] == 'a' &&
9579 return -KEY_getgrnam;
9592 if (name[4] == 'o' &&
9597 return -KEY_getlogin;
9608 if (name[6] == 'n' &&
9611 return -KEY_getpwent;
9617 if (name[6] == 'a' &&
9620 return -KEY_getpwnam;
9626 if (name[6] == 'i' &&
9629 return -KEY_getpwuid;
9649 if (name[1] == 'e' &&
9656 if (name[5] == 'i' &&
9663 return -KEY_readline;
9668 return -KEY_readlink;
9679 if (name[5] == 'i' &&
9683 return -KEY_readpipe;
9704 if (name[4] == 'r' &&
9709 return -KEY_setgrent;
9715 if (name[4] == 'w' &&
9720 return -KEY_setpwent;
9736 if (name[3] == 'w' &&
9742 return -KEY_shmwrite;
9748 if (name[3] == 't' &&
9754 return -KEY_shutdown;
9764 if (name[2] == 's' &&
9771 return -KEY_syswrite;
9781 if (name[1] == 'r' &&
9789 return -KEY_truncate;
9798 case 9: /* 9 tokens of length 9 */
9802 if (name[1] == 'N' &&
9811 return KEY_UNITCHECK;
9817 if (name[1] == 'n' &&
9826 return -KEY_endnetent;
9832 if (name[1] == 'e' &&
9841 return -KEY_getnetent;
9847 if (name[1] == 'o' &&
9856 return -KEY_localtime;
9862 if (name[1] == 'r' &&
9871 return KEY_prototype;
9877 if (name[1] == 'u' &&
9886 return -KEY_quotemeta;
9892 if (name[1] == 'e' &&
9901 return -KEY_rewinddir;
9907 if (name[1] == 'e' &&
9916 return -KEY_setnetent;
9922 if (name[1] == 'a' &&
9931 return -KEY_wantarray;
9940 case 10: /* 9 tokens of length 10 */
9944 if (name[1] == 'n' &&
9950 if (name[4] == 'o' &&
9957 return -KEY_endhostent;
9963 if (name[4] == 'e' &&
9970 return -KEY_endservent;
9983 if (name[1] == 'e' &&
9989 if (name[4] == 'o' &&
9996 return -KEY_gethostent;
10005 if (name[5] == 'r' &&
10011 return -KEY_getservent;
10017 if (name[5] == 'c' &&
10023 return -KEY_getsockopt;
10043 if (name[2] == 't')
10048 if (name[4] == 'o' &&
10055 return -KEY_sethostent;
10064 if (name[5] == 'r' &&
10070 return -KEY_setservent;
10076 if (name[5] == 'c' &&
10082 return -KEY_setsockopt;
10099 if (name[2] == 'c' &&
10108 return -KEY_socketpair;
10121 case 11: /* 8 tokens of length 11 */
10125 if (name[1] == '_' &&
10135 { /* __PACKAGE__ */
10136 return -KEY___PACKAGE__;
10142 if (name[1] == 'n' &&
10152 { /* endprotoent */
10153 return -KEY_endprotoent;
10159 if (name[1] == 'e' &&
10168 if (name[5] == 'e' &&
10174 { /* getpeername */
10175 return -KEY_getpeername;
10184 if (name[6] == 'o' &&
10189 { /* getpriority */
10190 return -KEY_getpriority;
10196 if (name[6] == 't' &&
10201 { /* getprotoent */
10202 return -KEY_getprotoent;
10216 if (name[4] == 'o' &&
10223 { /* getsockname */
10224 return -KEY_getsockname;
10237 if (name[1] == 'e' &&
10245 if (name[6] == 'o' &&
10250 { /* setpriority */
10251 return -KEY_setpriority;
10257 if (name[6] == 't' &&
10262 { /* setprotoent */
10263 return -KEY_setprotoent;
10279 case 12: /* 2 tokens of length 12 */
10280 if (name[0] == 'g' &&
10292 if (name[9] == 'd' &&
10295 { /* getnetbyaddr */
10296 return -KEY_getnetbyaddr;
10302 if (name[9] == 'a' &&
10305 { /* getnetbyname */
10306 return -KEY_getnetbyname;
10318 case 13: /* 4 tokens of length 13 */
10319 if (name[0] == 'g' &&
10326 if (name[4] == 'o' &&
10335 if (name[10] == 'd' &&
10338 { /* gethostbyaddr */
10339 return -KEY_gethostbyaddr;
10345 if (name[10] == 'a' &&
10348 { /* gethostbyname */
10349 return -KEY_gethostbyname;
10362 if (name[4] == 'e' &&
10371 if (name[10] == 'a' &&
10374 { /* getservbyname */
10375 return -KEY_getservbyname;
10381 if (name[10] == 'o' &&
10384 { /* getservbyport */
10385 return -KEY_getservbyport;
10404 case 14: /* 1 tokens of length 14 */
10405 if (name[0] == 'g' &&
10419 { /* getprotobyname */
10420 return -KEY_getprotobyname;
10425 case 16: /* 1 tokens of length 16 */
10426 if (name[0] == 'g' &&
10442 { /* getprotobynumber */
10443 return -KEY_getprotobynumber;
10457 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10461 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10462 if (ckWARN(WARN_SYNTAX)) {
10465 for (w = s+2; *w && level; w++) {
10468 else if (*w == ')')
10471 while (isSPACE(*w))
10473 /* the list of chars below is for end of statements or
10474 * block / parens, boolean operators (&&, ||, //) and branch
10475 * constructs (or, and, if, until, unless, while, err, for).
10476 * Not a very solid hack... */
10477 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10478 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10479 "%s (...) interpreted as function",name);
10482 while (s < PL_bufend && isSPACE(*s))
10486 while (s < PL_bufend && isSPACE(*s))
10488 if (isIDFIRST_lazy_if(s,UTF)) {
10489 const char * const w = s++;
10490 while (isALNUM_lazy_if(s,UTF))
10492 while (s < PL_bufend && isSPACE(*s))
10496 if (keyword(w, s - w, 0))
10499 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10500 if (gv && GvCVu(gv))
10502 Perl_croak(aTHX_ "No comma allowed after %s", what);
10507 /* Either returns sv, or mortalizes sv and returns a new SV*.
10508 Best used as sv=new_constant(..., sv, ...).
10509 If s, pv are NULL, calls subroutine with one argument,
10510 and type is used with error messages only. */
10513 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10517 HV * const table = GvHV(PL_hintgv); /* ^H */
10521 const char *why1 = "", *why2 = "", *why3 = "";
10523 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10526 why2 = (const char *)
10527 (strEQ(key,"charnames")
10528 ? "(possibly a missing \"use charnames ...\")"
10530 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10531 (type ? type: "undef"), why2);
10533 /* This is convoluted and evil ("goto considered harmful")
10534 * but I do not understand the intricacies of all the different
10535 * failure modes of %^H in here. The goal here is to make
10536 * the most probable error message user-friendly. --jhi */
10541 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10542 (type ? type: "undef"), why1, why2, why3);
10544 yyerror(SvPVX_const(msg));
10548 cvp = hv_fetch(table, key, strlen(key), FALSE);
10549 if (!cvp || !SvOK(*cvp)) {
10552 why3 = "} is not defined";
10555 sv_2mortal(sv); /* Parent created it permanently */
10558 pv = sv_2mortal(newSVpvn(s, len));
10560 typesv = sv_2mortal(newSVpv(type, 0));
10562 typesv = &PL_sv_undef;
10564 PUSHSTACKi(PERLSI_OVERLOAD);
10576 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10580 /* Check the eval first */
10581 if (!PL_in_eval && SvTRUE(ERRSV)) {
10582 sv_catpvs(ERRSV, "Propagated");
10583 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10585 res = SvREFCNT_inc_simple(sv);
10589 SvREFCNT_inc_simple_void(res);
10598 why1 = "Call to &{$^H{";
10600 why3 = "}} did not return a defined value";
10608 /* Returns a NUL terminated string, with the length of the string written to
10612 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10615 register char *d = dest;
10616 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10619 Perl_croak(aTHX_ ident_too_long);
10620 if (isALNUM(*s)) /* UTF handled below */
10622 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10627 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10631 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10632 char *t = s + UTF8SKIP(s);
10634 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10638 Perl_croak(aTHX_ ident_too_long);
10639 Copy(s, d, len, char);
10652 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10655 char *bracket = NULL;
10657 register char *d = dest;
10658 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10663 while (isDIGIT(*s)) {
10665 Perl_croak(aTHX_ ident_too_long);
10672 Perl_croak(aTHX_ ident_too_long);
10673 if (isALNUM(*s)) /* UTF handled below */
10675 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10680 else if (*s == ':' && s[1] == ':') {
10684 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10685 char *t = s + UTF8SKIP(s);
10686 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10688 if (d + (t - s) > e)
10689 Perl_croak(aTHX_ ident_too_long);
10690 Copy(s, d, t - s, char);
10701 if (PL_lex_state != LEX_NORMAL)
10702 PL_lex_state = LEX_INTERPENDMAYBE;
10705 if (*s == '$' && s[1] &&
10706 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10719 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10724 if (isSPACE(s[-1])) {
10726 const char ch = *s++;
10727 if (!SPACE_OR_TAB(ch)) {
10733 if (isIDFIRST_lazy_if(d,UTF)) {
10737 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10738 end += UTF8SKIP(end);
10739 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10740 end += UTF8SKIP(end);
10742 Copy(s, d, end - s, char);
10747 while ((isALNUM(*s) || *s == ':') && d < e)
10750 Perl_croak(aTHX_ ident_too_long);
10753 while (s < send && SPACE_OR_TAB(*s))
10755 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10756 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10757 const char * const brack =
10759 ((*s == '[') ? "[...]" : "{...}");
10760 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10761 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10762 funny, dest, brack, funny, dest, brack);
10765 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10769 /* Handle extended ${^Foo} variables
10770 * 1999-02-27 mjd-perl-patch@plover.com */
10771 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10775 while (isALNUM(*s) && d < e) {
10779 Perl_croak(aTHX_ ident_too_long);
10784 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10785 PL_lex_state = LEX_INTERPEND;
10788 if (PL_lex_state == LEX_NORMAL) {
10789 if (ckWARN(WARN_AMBIGUOUS) &&
10790 (keyword(dest, d - dest, 0)
10791 || get_cvn_flags(dest, d - dest, 0)))
10795 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10796 "Ambiguous use of %c{%s} resolved to %c%s",
10797 funny, dest, funny, dest);
10802 s = bracket; /* let the parser handle it */
10806 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10807 PL_lex_state = LEX_INTERPEND;
10812 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10814 PERL_UNUSED_CONTEXT;
10818 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10819 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10820 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10821 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10822 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10828 S_scan_pat(pTHX_ char *start, I32 type)
10832 char *s = scan_str(start,!!PL_madskills,FALSE);
10833 const char * const valid_flags =
10834 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10841 const char * const delimiter = skipspace(start);
10845 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10846 : "Search pattern not terminated" ));
10849 pm = (PMOP*)newPMOP(type, 0);
10850 if (PL_multi_open == '?') {
10851 /* This is the only point in the code that sets PMf_ONCE: */
10852 pm->op_pmflags |= PMf_ONCE;
10854 /* Hence it's safe to do this bit of PMOP book-keeping here, which
10855 allows us to restrict the list needed by reset to just the ??
10857 assert(type != OP_TRANS);
10859 MAGIC *mg = mg_find((SV*)PL_curstash, PERL_MAGIC_symtab);
10862 mg = sv_magicext((SV*)PL_curstash, 0, PERL_MAGIC_symtab, 0, 0,
10865 elements = mg->mg_len / sizeof(PMOP**);
10866 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
10867 ((PMOP**)mg->mg_ptr) [elements++] = pm;
10868 mg->mg_len = elements * sizeof(PMOP**);
10869 PmopSTASH_set(pm,PL_curstash);
10875 while (*s && strchr(valid_flags, *s))
10876 pmflag(&pm->op_pmflags,*s++);
10878 if (PL_madskills && modstart != s) {
10879 SV* tmptoken = newSVpvn(modstart, s - modstart);
10880 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10883 /* issue a warning if /c is specified,but /g is not */
10884 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10885 && ckWARN(WARN_REGEXP))
10887 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
10888 "Use of /c modifier is meaningless without /g" );
10891 PL_lex_op = (OP*)pm;
10892 yylval.ival = OP_MATCH;
10897 S_scan_subst(pTHX_ char *start)
10908 yylval.ival = OP_NULL;
10910 s = scan_str(start,!!PL_madskills,FALSE);
10913 Perl_croak(aTHX_ "Substitution pattern not terminated");
10915 if (s[-1] == PL_multi_open)
10918 if (PL_madskills) {
10919 CURMAD('q', PL_thisopen);
10920 CURMAD('_', PL_thiswhite);
10921 CURMAD('E', PL_thisstuff);
10922 CURMAD('Q', PL_thisclose);
10923 PL_realtokenstart = s - SvPVX(PL_linestr);
10927 first_start = PL_multi_start;
10928 s = scan_str(s,!!PL_madskills,FALSE);
10930 if (PL_lex_stuff) {
10931 SvREFCNT_dec(PL_lex_stuff);
10932 PL_lex_stuff = NULL;
10934 Perl_croak(aTHX_ "Substitution replacement not terminated");
10936 PL_multi_start = first_start; /* so whole substitution is taken together */
10938 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10941 if (PL_madskills) {
10942 CURMAD('z', PL_thisopen);
10943 CURMAD('R', PL_thisstuff);
10944 CURMAD('Z', PL_thisclose);
10950 if (*s == EXEC_PAT_MOD) {
10954 else if (strchr(S_PAT_MODS, *s))
10955 pmflag(&pm->op_pmflags,*s++);
10961 if (PL_madskills) {
10963 curmad('m', newSVpvn(modstart, s - modstart));
10964 append_madprops(PL_thismad, (OP*)pm, 0);
10968 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10969 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10973 SV * const repl = newSVpvs("");
10975 PL_sublex_info.super_bufptr = s;
10976 PL_sublex_info.super_bufend = PL_bufend;
10978 pm->op_pmflags |= PMf_EVAL;
10980 sv_catpv(repl, (const char *)(es ? "eval " : "do "));
10981 sv_catpvs(repl, "{");
10982 sv_catsv(repl, PL_lex_repl);
10983 if (strchr(SvPVX(PL_lex_repl), '#'))
10984 sv_catpvs(repl, "\n");
10985 sv_catpvs(repl, "}");
10987 SvREFCNT_dec(PL_lex_repl);
10988 PL_lex_repl = repl;
10991 PL_lex_op = (OP*)pm;
10992 yylval.ival = OP_SUBST;
10997 S_scan_trans(pTHX_ char *start)
11010 yylval.ival = OP_NULL;
11012 s = scan_str(start,!!PL_madskills,FALSE);
11014 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11016 if (s[-1] == PL_multi_open)
11019 if (PL_madskills) {
11020 CURMAD('q', PL_thisopen);
11021 CURMAD('_', PL_thiswhite);
11022 CURMAD('E', PL_thisstuff);
11023 CURMAD('Q', PL_thisclose);
11024 PL_realtokenstart = s - SvPVX(PL_linestr);
11028 s = scan_str(s,!!PL_madskills,FALSE);
11030 if (PL_lex_stuff) {
11031 SvREFCNT_dec(PL_lex_stuff);
11032 PL_lex_stuff = NULL;
11034 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11036 if (PL_madskills) {
11037 CURMAD('z', PL_thisopen);
11038 CURMAD('R', PL_thisstuff);
11039 CURMAD('Z', PL_thisclose);
11042 complement = del = squash = 0;
11049 complement = OPpTRANS_COMPLEMENT;
11052 del = OPpTRANS_DELETE;
11055 squash = OPpTRANS_SQUASH;
11064 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11065 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11066 o->op_private &= ~OPpTRANS_ALL;
11067 o->op_private |= del|squash|complement|
11068 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11069 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11072 yylval.ival = OP_TRANS;
11075 if (PL_madskills) {
11077 curmad('m', newSVpvn(modstart, s - modstart));
11078 append_madprops(PL_thismad, o, 0);
11087 S_scan_heredoc(pTHX_ register char *s)
11091 I32 op_type = OP_SCALAR;
11095 const char *found_newline;
11099 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11101 I32 stuffstart = s - SvPVX(PL_linestr);
11104 PL_realtokenstart = -1;
11109 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11113 while (SPACE_OR_TAB(*peek))
11115 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11118 s = delimcpy(d, e, s, PL_bufend, term, &len);
11128 if (!isALNUM_lazy_if(s,UTF))
11129 deprecate_old("bare << to mean <<\"\"");
11130 for (; isALNUM_lazy_if(s,UTF); s++) {
11135 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11136 Perl_croak(aTHX_ "Delimiter for here document is too long");
11139 len = d - PL_tokenbuf;
11142 if (PL_madskills) {
11143 tstart = PL_tokenbuf + !outer;
11144 PL_thisclose = newSVpvn(tstart, len - !outer);
11145 tstart = SvPVX(PL_linestr) + stuffstart;
11146 PL_thisopen = newSVpvn(tstart, s - tstart);
11147 stuffstart = s - SvPVX(PL_linestr);
11150 #ifndef PERL_STRICT_CR
11151 d = strchr(s, '\r');
11153 char * const olds = s;
11155 while (s < PL_bufend) {
11161 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11170 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11177 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11178 herewas = newSVpvn(s,PL_bufend-s);
11182 herewas = newSVpvn(s-1,found_newline-s+1);
11185 herewas = newSVpvn(s,found_newline-s);
11189 if (PL_madskills) {
11190 tstart = SvPVX(PL_linestr) + stuffstart;
11192 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11194 PL_thisstuff = newSVpvn(tstart, s - tstart);
11197 s += SvCUR(herewas);
11200 stuffstart = s - SvPVX(PL_linestr);
11206 tmpstr = newSV_type(SVt_PVIV);
11207 SvGROW(tmpstr, 80);
11208 if (term == '\'') {
11209 op_type = OP_CONST;
11210 SvIV_set(tmpstr, -1);
11212 else if (term == '`') {
11213 op_type = OP_BACKTICK;
11214 SvIV_set(tmpstr, '\\');
11218 PL_multi_start = CopLINE(PL_curcop);
11219 PL_multi_open = PL_multi_close = '<';
11220 term = *PL_tokenbuf;
11221 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11222 char * const bufptr = PL_sublex_info.super_bufptr;
11223 char * const bufend = PL_sublex_info.super_bufend;
11224 char * const olds = s - SvCUR(herewas);
11225 s = strchr(bufptr, '\n');
11229 while (s < bufend &&
11230 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11232 CopLINE_inc(PL_curcop);
11235 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11236 missingterm(PL_tokenbuf);
11238 sv_setpvn(herewas,bufptr,d-bufptr+1);
11239 sv_setpvn(tmpstr,d+1,s-d);
11241 sv_catpvn(herewas,s,bufend-s);
11242 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11249 while (s < PL_bufend &&
11250 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11252 CopLINE_inc(PL_curcop);
11254 if (s >= PL_bufend) {
11255 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11256 missingterm(PL_tokenbuf);
11258 sv_setpvn(tmpstr,d+1,s-d);
11260 if (PL_madskills) {
11262 sv_catpvn(PL_thisstuff, d + 1, s - d);
11264 PL_thisstuff = newSVpvn(d + 1, s - d);
11265 stuffstart = s - SvPVX(PL_linestr);
11269 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11271 sv_catpvn(herewas,s,PL_bufend-s);
11272 sv_setsv(PL_linestr,herewas);
11273 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11274 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11275 PL_last_lop = PL_last_uni = NULL;
11278 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11279 while (s >= PL_bufend) { /* multiple line string? */
11281 if (PL_madskills) {
11282 tstart = SvPVX(PL_linestr) + stuffstart;
11284 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11286 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11290 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11291 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11292 missingterm(PL_tokenbuf);
11295 stuffstart = s - SvPVX(PL_linestr);
11297 CopLINE_inc(PL_curcop);
11298 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11299 PL_last_lop = PL_last_uni = NULL;
11300 #ifndef PERL_STRICT_CR
11301 if (PL_bufend - PL_linestart >= 2) {
11302 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11303 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11305 PL_bufend[-2] = '\n';
11307 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11309 else if (PL_bufend[-1] == '\r')
11310 PL_bufend[-1] = '\n';
11312 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11313 PL_bufend[-1] = '\n';
11315 if (PERLDB_LINE && PL_curstash != PL_debstash)
11316 update_debugger_info(PL_linestr, NULL, 0);
11317 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11318 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11319 *(SvPVX(PL_linestr) + off ) = ' ';
11320 sv_catsv(PL_linestr,herewas);
11321 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11322 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11326 sv_catsv(tmpstr,PL_linestr);
11331 PL_multi_end = CopLINE(PL_curcop);
11332 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11333 SvPV_shrink_to_cur(tmpstr);
11335 SvREFCNT_dec(herewas);
11337 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11339 else if (PL_encoding)
11340 sv_recode_to_utf8(tmpstr, PL_encoding);
11342 PL_lex_stuff = tmpstr;
11343 yylval.ival = op_type;
11347 /* scan_inputsymbol
11348 takes: current position in input buffer
11349 returns: new position in input buffer
11350 side-effects: yylval and lex_op are set.
11355 <FH> read from filehandle
11356 <pkg::FH> read from package qualified filehandle
11357 <pkg'FH> read from package qualified filehandle
11358 <$fh> read from filehandle in $fh
11359 <*.h> filename glob
11364 S_scan_inputsymbol(pTHX_ char *start)
11367 register char *s = start; /* current position in buffer */
11371 char *d = PL_tokenbuf; /* start of temp holding space */
11372 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11374 end = strchr(s, '\n');
11377 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11379 /* die if we didn't have space for the contents of the <>,
11380 or if it didn't end, or if we see a newline
11383 if (len >= (I32)sizeof PL_tokenbuf)
11384 Perl_croak(aTHX_ "Excessively long <> operator");
11386 Perl_croak(aTHX_ "Unterminated <> operator");
11391 Remember, only scalar variables are interpreted as filehandles by
11392 this code. Anything more complex (e.g., <$fh{$num}>) will be
11393 treated as a glob() call.
11394 This code makes use of the fact that except for the $ at the front,
11395 a scalar variable and a filehandle look the same.
11397 if (*d == '$' && d[1]) d++;
11399 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11400 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11403 /* If we've tried to read what we allow filehandles to look like, and
11404 there's still text left, then it must be a glob() and not a getline.
11405 Use scan_str to pull out the stuff between the <> and treat it
11406 as nothing more than a string.
11409 if (d - PL_tokenbuf != len) {
11410 yylval.ival = OP_GLOB;
11412 s = scan_str(start,!!PL_madskills,FALSE);
11414 Perl_croak(aTHX_ "Glob not terminated");
11418 bool readline_overriden = FALSE;
11421 /* we're in a filehandle read situation */
11424 /* turn <> into <ARGV> */
11426 Copy("ARGV",d,5,char);
11428 /* Check whether readline() is overriden */
11429 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11431 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11433 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11434 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11435 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11436 readline_overriden = TRUE;
11438 /* if <$fh>, create the ops to turn the variable into a
11442 /* try to find it in the pad for this block, otherwise find
11443 add symbol table ops
11445 const PADOFFSET tmp = pad_findmy(d);
11446 if (tmp != NOT_IN_PAD) {
11447 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11448 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11449 HEK * const stashname = HvNAME_HEK(stash);
11450 SV * const sym = sv_2mortal(newSVhek(stashname));
11451 sv_catpvs(sym, "::");
11452 sv_catpv(sym, d+1);
11457 OP * const o = newOP(OP_PADSV, 0);
11459 PL_lex_op = readline_overriden
11460 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11461 append_elem(OP_LIST, o,
11462 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11463 : (OP*)newUNOP(OP_READLINE, 0, o);
11472 ? (GV_ADDMULTI | GV_ADDINEVAL)
11475 PL_lex_op = readline_overriden
11476 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11477 append_elem(OP_LIST,
11478 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11479 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11480 : (OP*)newUNOP(OP_READLINE, 0,
11481 newUNOP(OP_RV2SV, 0,
11482 newGVOP(OP_GV, 0, gv)));
11484 if (!readline_overriden)
11485 PL_lex_op->op_flags |= OPf_SPECIAL;
11486 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11487 yylval.ival = OP_NULL;
11490 /* If it's none of the above, it must be a literal filehandle
11491 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11493 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11494 PL_lex_op = readline_overriden
11495 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11496 append_elem(OP_LIST,
11497 newGVOP(OP_GV, 0, gv),
11498 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11499 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11500 yylval.ival = OP_NULL;
11509 takes: start position in buffer
11510 keep_quoted preserve \ on the embedded delimiter(s)
11511 keep_delims preserve the delimiters around the string
11512 returns: position to continue reading from buffer
11513 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11514 updates the read buffer.
11516 This subroutine pulls a string out of the input. It is called for:
11517 q single quotes q(literal text)
11518 ' single quotes 'literal text'
11519 qq double quotes qq(interpolate $here please)
11520 " double quotes "interpolate $here please"
11521 qx backticks qx(/bin/ls -l)
11522 ` backticks `/bin/ls -l`
11523 qw quote words @EXPORT_OK = qw( func() $spam )
11524 m// regexp match m/this/
11525 s/// regexp substitute s/this/that/
11526 tr/// string transliterate tr/this/that/
11527 y/// string transliterate y/this/that/
11528 ($*@) sub prototypes sub foo ($)
11529 (stuff) sub attr parameters sub foo : attr(stuff)
11530 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11532 In most of these cases (all but <>, patterns and transliterate)
11533 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11534 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11535 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11538 It skips whitespace before the string starts, and treats the first
11539 character as the delimiter. If the delimiter is one of ([{< then
11540 the corresponding "close" character )]}> is used as the closing
11541 delimiter. It allows quoting of delimiters, and if the string has
11542 balanced delimiters ([{<>}]) it allows nesting.
11544 On success, the SV with the resulting string is put into lex_stuff or,
11545 if that is already non-NULL, into lex_repl. The second case occurs only
11546 when parsing the RHS of the special constructs s/// and tr/// (y///).
11547 For convenience, the terminating delimiter character is stuffed into
11552 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11555 SV *sv; /* scalar value: string */
11556 const char *tmps; /* temp string, used for delimiter matching */
11557 register char *s = start; /* current position in the buffer */
11558 register char term; /* terminating character */
11559 register char *to; /* current position in the sv's data */
11560 I32 brackets = 1; /* bracket nesting level */
11561 bool has_utf8 = FALSE; /* is there any utf8 content? */
11562 I32 termcode; /* terminating char. code */
11563 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11564 STRLEN termlen; /* length of terminating string */
11565 int last_off = 0; /* last position for nesting bracket */
11571 /* skip space before the delimiter */
11577 if (PL_realtokenstart >= 0) {
11578 stuffstart = PL_realtokenstart;
11579 PL_realtokenstart = -1;
11582 stuffstart = start - SvPVX(PL_linestr);
11584 /* mark where we are, in case we need to report errors */
11587 /* after skipping whitespace, the next character is the terminator */
11590 termcode = termstr[0] = term;
11594 termcode = utf8_to_uvchr((U8*)s, &termlen);
11595 Copy(s, termstr, termlen, U8);
11596 if (!UTF8_IS_INVARIANT(term))
11600 /* mark where we are */
11601 PL_multi_start = CopLINE(PL_curcop);
11602 PL_multi_open = term;
11604 /* find corresponding closing delimiter */
11605 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11606 termcode = termstr[0] = term = tmps[5];
11608 PL_multi_close = term;
11610 /* create a new SV to hold the contents. 79 is the SV's initial length.
11611 What a random number. */
11612 sv = newSV_type(SVt_PVIV);
11614 SvIV_set(sv, termcode);
11615 (void)SvPOK_only(sv); /* validate pointer */
11617 /* move past delimiter and try to read a complete string */
11619 sv_catpvn(sv, s, termlen);
11622 tstart = SvPVX(PL_linestr) + stuffstart;
11623 if (!PL_thisopen && !keep_delims) {
11624 PL_thisopen = newSVpvn(tstart, s - tstart);
11625 stuffstart = s - SvPVX(PL_linestr);
11629 if (PL_encoding && !UTF) {
11633 int offset = s - SvPVX_const(PL_linestr);
11634 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11635 &offset, (char*)termstr, termlen);
11636 const char * const ns = SvPVX_const(PL_linestr) + offset;
11637 char * const svlast = SvEND(sv) - 1;
11639 for (; s < ns; s++) {
11640 if (*s == '\n' && !PL_rsfp)
11641 CopLINE_inc(PL_curcop);
11644 goto read_more_line;
11646 /* handle quoted delimiters */
11647 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11649 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11651 if ((svlast-1 - t) % 2) {
11652 if (!keep_quoted) {
11653 *(svlast-1) = term;
11655 SvCUR_set(sv, SvCUR(sv) - 1);
11660 if (PL_multi_open == PL_multi_close) {
11666 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11667 /* At here, all closes are "was quoted" one,
11668 so we don't check PL_multi_close. */
11670 if (!keep_quoted && *(t+1) == PL_multi_open)
11675 else if (*t == PL_multi_open)
11683 SvCUR_set(sv, w - SvPVX_const(sv));
11685 last_off = w - SvPVX(sv);
11686 if (--brackets <= 0)
11691 if (!keep_delims) {
11692 SvCUR_set(sv, SvCUR(sv) - 1);
11698 /* extend sv if need be */
11699 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11700 /* set 'to' to the next character in the sv's string */
11701 to = SvPVX(sv)+SvCUR(sv);
11703 /* if open delimiter is the close delimiter read unbridle */
11704 if (PL_multi_open == PL_multi_close) {
11705 for (; s < PL_bufend; s++,to++) {
11706 /* embedded newlines increment the current line number */
11707 if (*s == '\n' && !PL_rsfp)
11708 CopLINE_inc(PL_curcop);
11709 /* handle quoted delimiters */
11710 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11711 if (!keep_quoted && s[1] == term)
11713 /* any other quotes are simply copied straight through */
11717 /* terminate when run out of buffer (the for() condition), or
11718 have found the terminator */
11719 else if (*s == term) {
11722 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11725 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11731 /* if the terminator isn't the same as the start character (e.g.,
11732 matched brackets), we have to allow more in the quoting, and
11733 be prepared for nested brackets.
11736 /* read until we run out of string, or we find the terminator */
11737 for (; s < PL_bufend; s++,to++) {
11738 /* embedded newlines increment the line count */
11739 if (*s == '\n' && !PL_rsfp)
11740 CopLINE_inc(PL_curcop);
11741 /* backslashes can escape the open or closing characters */
11742 if (*s == '\\' && s+1 < PL_bufend) {
11743 if (!keep_quoted &&
11744 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11749 /* allow nested opens and closes */
11750 else if (*s == PL_multi_close && --brackets <= 0)
11752 else if (*s == PL_multi_open)
11754 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11759 /* terminate the copied string and update the sv's end-of-string */
11761 SvCUR_set(sv, to - SvPVX_const(sv));
11764 * this next chunk reads more into the buffer if we're not done yet
11768 break; /* handle case where we are done yet :-) */
11770 #ifndef PERL_STRICT_CR
11771 if (to - SvPVX_const(sv) >= 2) {
11772 if ((to[-2] == '\r' && to[-1] == '\n') ||
11773 (to[-2] == '\n' && to[-1] == '\r'))
11777 SvCUR_set(sv, to - SvPVX_const(sv));
11779 else if (to[-1] == '\r')
11782 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11787 /* if we're out of file, or a read fails, bail and reset the current
11788 line marker so we can report where the unterminated string began
11791 if (PL_madskills) {
11792 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11794 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11796 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11800 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11802 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11808 /* we read a line, so increment our line counter */
11809 CopLINE_inc(PL_curcop);
11811 /* update debugger info */
11812 if (PERLDB_LINE && PL_curstash != PL_debstash)
11813 update_debugger_info(PL_linestr, NULL, 0);
11815 /* having changed the buffer, we must update PL_bufend */
11816 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11817 PL_last_lop = PL_last_uni = NULL;
11820 /* at this point, we have successfully read the delimited string */
11822 if (!PL_encoding || UTF) {
11824 if (PL_madskills) {
11825 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11826 const int len = s - tstart;
11828 sv_catpvn(PL_thisstuff, tstart, len);
11830 PL_thisstuff = newSVpvn(tstart, len);
11831 if (!PL_thisclose && !keep_delims)
11832 PL_thisclose = newSVpvn(s,termlen);
11837 sv_catpvn(sv, s, termlen);
11842 if (PL_madskills) {
11843 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11844 const int len = s - tstart - termlen;
11846 sv_catpvn(PL_thisstuff, tstart, len);
11848 PL_thisstuff = newSVpvn(tstart, len);
11849 if (!PL_thisclose && !keep_delims)
11850 PL_thisclose = newSVpvn(s - termlen,termlen);
11854 if (has_utf8 || PL_encoding)
11857 PL_multi_end = CopLINE(PL_curcop);
11859 /* if we allocated too much space, give some back */
11860 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11861 SvLEN_set(sv, SvCUR(sv) + 1);
11862 SvPV_renew(sv, SvLEN(sv));
11865 /* decide whether this is the first or second quoted string we've read
11878 takes: pointer to position in buffer
11879 returns: pointer to new position in buffer
11880 side-effects: builds ops for the constant in yylval.op
11882 Read a number in any of the formats that Perl accepts:
11884 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11885 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11888 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11890 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11893 If it reads a number without a decimal point or an exponent, it will
11894 try converting the number to an integer and see if it can do so
11895 without loss of precision.
11899 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11902 register const char *s = start; /* current position in buffer */
11903 register char *d; /* destination in temp buffer */
11904 register char *e; /* end of temp buffer */
11905 NV nv; /* number read, as a double */
11906 SV *sv = NULL; /* place to put the converted number */
11907 bool floatit; /* boolean: int or float? */
11908 const char *lastub = NULL; /* position of last underbar */
11909 static char const number_too_long[] = "Number too long";
11911 /* We use the first character to decide what type of number this is */
11915 Perl_croak(aTHX_ "panic: scan_num");
11917 /* if it starts with a 0, it could be an octal number, a decimal in
11918 0.13 disguise, or a hexadecimal number, or a binary number. */
11922 u holds the "number so far"
11923 shift the power of 2 of the base
11924 (hex == 4, octal == 3, binary == 1)
11925 overflowed was the number more than we can hold?
11927 Shift is used when we add a digit. It also serves as an "are
11928 we in octal/hex/binary?" indicator to disallow hex characters
11929 when in octal mode.
11934 bool overflowed = FALSE;
11935 bool just_zero = TRUE; /* just plain 0 or binary number? */
11936 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11937 static const char* const bases[5] =
11938 { "", "binary", "", "octal", "hexadecimal" };
11939 static const char* const Bases[5] =
11940 { "", "Binary", "", "Octal", "Hexadecimal" };
11941 static const char* const maxima[5] =
11943 "0b11111111111111111111111111111111",
11947 const char *base, *Base, *max;
11949 /* check for hex */
11954 } else if (s[1] == 'b') {
11959 /* check for a decimal in disguise */
11960 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11962 /* so it must be octal */
11969 if (ckWARN(WARN_SYNTAX))
11970 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11971 "Misplaced _ in number");
11975 base = bases[shift];
11976 Base = Bases[shift];
11977 max = maxima[shift];
11979 /* read the rest of the number */
11981 /* x is used in the overflow test,
11982 b is the digit we're adding on. */
11987 /* if we don't mention it, we're done */
11991 /* _ are ignored -- but warned about if consecutive */
11993 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11994 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11995 "Misplaced _ in number");
11999 /* 8 and 9 are not octal */
12000 case '8': case '9':
12002 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12006 case '2': case '3': case '4':
12007 case '5': case '6': case '7':
12009 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12012 case '0': case '1':
12013 b = *s++ & 15; /* ASCII digit -> value of digit */
12017 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12018 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12019 /* make sure they said 0x */
12022 b = (*s++ & 7) + 9;
12024 /* Prepare to put the digit we have onto the end
12025 of the number so far. We check for overflows.
12031 x = u << shift; /* make room for the digit */
12033 if ((x >> shift) != u
12034 && !(PL_hints & HINT_NEW_BINARY)) {
12037 if (ckWARN_d(WARN_OVERFLOW))
12038 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12039 "Integer overflow in %s number",
12042 u = x | b; /* add the digit to the end */
12045 n *= nvshift[shift];
12046 /* If an NV has not enough bits in its
12047 * mantissa to represent an UV this summing of
12048 * small low-order numbers is a waste of time
12049 * (because the NV cannot preserve the
12050 * low-order bits anyway): we could just
12051 * remember when did we overflow and in the
12052 * end just multiply n by the right
12060 /* if we get here, we had success: make a scalar value from
12065 /* final misplaced underbar check */
12066 if (s[-1] == '_') {
12067 if (ckWARN(WARN_SYNTAX))
12068 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12073 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12074 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12075 "%s number > %s non-portable",
12081 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12082 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12083 "%s number > %s non-portable",
12088 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12089 sv = new_constant(start, s - start, "integer",
12091 else if (PL_hints & HINT_NEW_BINARY)
12092 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
12097 handle decimal numbers.
12098 we're also sent here when we read a 0 as the first digit
12100 case '1': case '2': case '3': case '4': case '5':
12101 case '6': case '7': case '8': case '9': case '.':
12104 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12107 /* read next group of digits and _ and copy into d */
12108 while (isDIGIT(*s) || *s == '_') {
12109 /* skip underscores, checking for misplaced ones
12113 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12114 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12115 "Misplaced _ in number");
12119 /* check for end of fixed-length buffer */
12121 Perl_croak(aTHX_ number_too_long);
12122 /* if we're ok, copy the character */
12127 /* final misplaced underbar check */
12128 if (lastub && s == lastub + 1) {
12129 if (ckWARN(WARN_SYNTAX))
12130 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12133 /* read a decimal portion if there is one. avoid
12134 3..5 being interpreted as the number 3. followed
12137 if (*s == '.' && s[1] != '.') {
12142 if (ckWARN(WARN_SYNTAX))
12143 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12144 "Misplaced _ in number");
12148 /* copy, ignoring underbars, until we run out of digits.
12150 for (; isDIGIT(*s) || *s == '_'; s++) {
12151 /* fixed length buffer check */
12153 Perl_croak(aTHX_ number_too_long);
12155 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12156 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12157 "Misplaced _ in number");
12163 /* fractional part ending in underbar? */
12164 if (s[-1] == '_') {
12165 if (ckWARN(WARN_SYNTAX))
12166 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12167 "Misplaced _ in number");
12169 if (*s == '.' && isDIGIT(s[1])) {
12170 /* oops, it's really a v-string, but without the "v" */
12176 /* read exponent part, if present */
12177 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12181 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12182 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12184 /* stray preinitial _ */
12186 if (ckWARN(WARN_SYNTAX))
12187 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12188 "Misplaced _ in number");
12192 /* allow positive or negative exponent */
12193 if (*s == '+' || *s == '-')
12196 /* stray initial _ */
12198 if (ckWARN(WARN_SYNTAX))
12199 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12200 "Misplaced _ in number");
12204 /* read digits of exponent */
12205 while (isDIGIT(*s) || *s == '_') {
12208 Perl_croak(aTHX_ number_too_long);
12212 if (((lastub && s == lastub + 1) ||
12213 (!isDIGIT(s[1]) && s[1] != '_'))
12214 && ckWARN(WARN_SYNTAX))
12215 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12216 "Misplaced _ in number");
12223 /* make an sv from the string */
12227 We try to do an integer conversion first if no characters
12228 indicating "float" have been found.
12233 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12235 if (flags == IS_NUMBER_IN_UV) {
12237 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12240 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12241 if (uv <= (UV) IV_MIN)
12242 sv_setiv(sv, -(IV)uv);
12249 /* terminate the string */
12251 nv = Atof(PL_tokenbuf);
12255 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
12256 (PL_hints & HINT_NEW_INTEGER) )
12257 sv = new_constant(PL_tokenbuf,
12260 (floatit ? "float" : "integer"),
12264 /* if it starts with a v, it could be a v-string */
12267 sv = newSV(5); /* preallocate storage space */
12268 s = scan_vstring(s, PL_bufend, sv);
12272 /* make the op for the constant and return */
12275 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12277 lvalp->opval = NULL;
12283 S_scan_formline(pTHX_ register char *s)
12286 register char *eol;
12288 SV * const stuff = newSVpvs("");
12289 bool needargs = FALSE;
12290 bool eofmt = FALSE;
12292 char *tokenstart = s;
12295 if (PL_madskills) {
12296 savewhite = PL_thiswhite;
12301 while (!needargs) {
12304 #ifdef PERL_STRICT_CR
12305 while (SPACE_OR_TAB(*t))
12308 while (SPACE_OR_TAB(*t) || *t == '\r')
12311 if (*t == '\n' || t == PL_bufend) {
12316 if (PL_in_eval && !PL_rsfp) {
12317 eol = (char *) memchr(s,'\n',PL_bufend-s);
12322 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12324 for (t = s; t < eol; t++) {
12325 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12327 goto enough; /* ~~ must be first line in formline */
12329 if (*t == '@' || *t == '^')
12333 sv_catpvn(stuff, s, eol-s);
12334 #ifndef PERL_STRICT_CR
12335 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12336 char *end = SvPVX(stuff) + SvCUR(stuff);
12339 SvCUR_set(stuff, SvCUR(stuff) - 1);
12349 if (PL_madskills) {
12351 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12353 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12356 s = filter_gets(PL_linestr, PL_rsfp, 0);
12358 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12360 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12362 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12363 PL_last_lop = PL_last_uni = NULL;
12372 if (SvCUR(stuff)) {
12375 PL_lex_state = LEX_NORMAL;
12376 start_force(PL_curforce);
12377 NEXTVAL_NEXTTOKE.ival = 0;
12381 PL_lex_state = LEX_FORMLINE;
12383 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12385 else if (PL_encoding)
12386 sv_recode_to_utf8(stuff, PL_encoding);
12388 start_force(PL_curforce);
12389 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12391 start_force(PL_curforce);
12392 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12396 SvREFCNT_dec(stuff);
12398 PL_lex_formbrack = 0;
12402 if (PL_madskills) {
12404 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12406 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12407 PL_thiswhite = savewhite;
12419 PL_cshlen = strlen(PL_cshname);
12421 #if defined(USE_ITHREADS)
12422 PERL_UNUSED_CONTEXT;
12428 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12431 const I32 oldsavestack_ix = PL_savestack_ix;
12432 CV* const outsidecv = PL_compcv;
12435 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12437 SAVEI32(PL_subline);
12438 save_item(PL_subname);
12439 SAVESPTR(PL_compcv);
12441 PL_compcv = (CV*)newSV_type(is_format ? SVt_PVFM : SVt_PVCV);
12442 CvFLAGS(PL_compcv) |= flags;
12444 PL_subline = CopLINE(PL_curcop);
12445 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12446 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12447 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12449 return oldsavestack_ix;
12453 #pragma segment Perl_yylex
12456 Perl_yywarn(pTHX_ const char *s)
12459 PL_in_eval |= EVAL_WARNONLY;
12461 PL_in_eval &= ~EVAL_WARNONLY;
12466 Perl_yyerror(pTHX_ const char *s)
12469 const char *where = NULL;
12470 const char *context = NULL;
12473 int yychar = PL_parser->yychar;
12475 if (!yychar || (yychar == ';' && !PL_rsfp))
12477 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12478 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12479 PL_oldbufptr != PL_bufptr) {
12482 The code below is removed for NetWare because it abends/crashes on NetWare
12483 when the script has error such as not having the closing quotes like:
12484 if ($var eq "value)
12485 Checking of white spaces is anyway done in NetWare code.
12488 while (isSPACE(*PL_oldoldbufptr))
12491 context = PL_oldoldbufptr;
12492 contlen = PL_bufptr - PL_oldoldbufptr;
12494 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12495 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12498 The code below is removed for NetWare because it abends/crashes on NetWare
12499 when the script has error such as not having the closing quotes like:
12500 if ($var eq "value)
12501 Checking of white spaces is anyway done in NetWare code.
12504 while (isSPACE(*PL_oldbufptr))
12507 context = PL_oldbufptr;
12508 contlen = PL_bufptr - PL_oldbufptr;
12510 else if (yychar > 255)
12511 where = "next token ???";
12512 else if (yychar == -2) { /* YYEMPTY */
12513 if (PL_lex_state == LEX_NORMAL ||
12514 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12515 where = "at end of line";
12516 else if (PL_lex_inpat)
12517 where = "within pattern";
12519 where = "within string";
12522 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12524 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12525 else if (isPRINT_LC(yychar))
12526 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12528 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12529 where = SvPVX_const(where_sv);
12531 msg = sv_2mortal(newSVpv(s, 0));
12532 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12533 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12535 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12537 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12538 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12539 Perl_sv_catpvf(aTHX_ msg,
12540 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12541 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12544 if (PL_in_eval & EVAL_WARNONLY) {
12545 if (ckWARN_d(WARN_SYNTAX))
12546 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12550 if (PL_error_count >= 10) {
12551 if (PL_in_eval && SvCUR(ERRSV))
12552 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12553 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12555 Perl_croak(aTHX_ "%s has too many errors.\n",
12556 OutCopFILE(PL_curcop));
12559 PL_in_my_stash = NULL;
12563 #pragma segment Main
12567 S_swallow_bom(pTHX_ U8 *s)
12570 const STRLEN slen = SvCUR(PL_linestr);
12573 if (s[1] == 0xFE) {
12574 /* UTF-16 little-endian? (or UTF32-LE?) */
12575 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12576 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12577 #ifndef PERL_NO_UTF16_FILTER
12578 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12581 if (PL_bufend > (char*)s) {
12585 filter_add(utf16rev_textfilter, NULL);
12586 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12587 utf16_to_utf8_reversed(s, news,
12588 PL_bufend - (char*)s - 1,
12590 sv_setpvn(PL_linestr, (const char*)news, newlen);
12592 s = (U8*)SvPVX(PL_linestr);
12593 Copy(news, s, newlen, U8);
12597 SvUTF8_on(PL_linestr);
12598 s = (U8*)SvPVX(PL_linestr);
12600 /* FIXME - is this a general bug fix? */
12603 PL_bufend = SvPVX(PL_linestr) + newlen;
12606 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12611 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12612 #ifndef PERL_NO_UTF16_FILTER
12613 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12616 if (PL_bufend > (char *)s) {
12620 filter_add(utf16_textfilter, NULL);
12621 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12622 utf16_to_utf8(s, news,
12623 PL_bufend - (char*)s,
12625 sv_setpvn(PL_linestr, (const char*)news, newlen);
12627 SvUTF8_on(PL_linestr);
12628 s = (U8*)SvPVX(PL_linestr);
12629 PL_bufend = SvPVX(PL_linestr) + newlen;
12632 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12637 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12638 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12639 s += 3; /* UTF-8 */
12645 if (s[2] == 0xFE && s[3] == 0xFF) {
12646 /* UTF-32 big-endian */
12647 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12650 else if (s[2] == 0 && s[3] != 0) {
12653 * are a good indicator of UTF-16BE. */
12654 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12660 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12661 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12662 s += 4; /* UTF-8 */
12668 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12671 * are a good indicator of UTF-16LE. */
12672 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12680 #ifndef PERL_NO_UTF16_FILTER
12682 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12685 const STRLEN old = SvCUR(sv);
12686 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12687 DEBUG_P(PerlIO_printf(Perl_debug_log,
12688 "utf16_textfilter(%p): %d %d (%d)\n",
12689 FPTR2DPTR(void *, utf16_textfilter),
12690 idx, maxlen, (int) count));
12694 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12695 Copy(SvPVX_const(sv), tmps, old, char);
12696 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12697 SvCUR(sv) - old, &newlen);
12698 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12700 DEBUG_P({sv_dump(sv);});
12705 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12708 const STRLEN old = SvCUR(sv);
12709 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12710 DEBUG_P(PerlIO_printf(Perl_debug_log,
12711 "utf16rev_textfilter(%p): %d %d (%d)\n",
12712 FPTR2DPTR(void *, utf16rev_textfilter),
12713 idx, maxlen, (int) count));
12717 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12718 Copy(SvPVX_const(sv), tmps, old, char);
12719 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12720 SvCUR(sv) - old, &newlen);
12721 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12723 DEBUG_P({ sv_dump(sv); });
12729 Returns a pointer to the next character after the parsed
12730 vstring, as well as updating the passed in sv.
12732 Function must be called like
12735 s = scan_vstring(s,e,sv);
12737 where s and e are the start and end of the string.
12738 The sv should already be large enough to store the vstring
12739 passed in, for performance reasons.
12744 Perl_scan_vstring(pTHX_ const char *s, const char *e, SV *sv)
12747 const char *pos = s;
12748 const char *start = s;
12749 if (*pos == 'v') pos++; /* get past 'v' */
12750 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12752 if ( *pos != '.') {
12753 /* this may not be a v-string if followed by => */
12754 const char *next = pos;
12755 while (next < e && isSPACE(*next))
12757 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12758 /* return string not v-string */
12759 sv_setpvn(sv,(char *)s,pos-s);
12760 return (char *)pos;
12764 if (!isALPHA(*pos)) {
12765 U8 tmpbuf[UTF8_MAXBYTES+1];
12768 s++; /* get past 'v' */
12770 sv_setpvn(sv, "", 0);
12773 /* this is atoi() that tolerates underscores */
12776 const char *end = pos;
12778 while (--end >= s) {
12780 const UV orev = rev;
12781 rev += (*end - '0') * mult;
12783 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12784 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12785 "Integer overflow in decimal number");
12789 if (rev > 0x7FFFFFFF)
12790 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12792 /* Append native character for the rev point */
12793 tmpend = uvchr_to_utf8(tmpbuf, rev);
12794 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12795 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12797 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12803 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12807 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12815 * c-indentation-style: bsd
12816 * c-basic-offset: 4
12817 * indent-tabs-mode: t
12820 * ex: set ts=8 sts=4 sw=4 noet: