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 new_constant(a,b,c,d,e,f,g) \
27 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
29 #define yylval (PL_parser->yylval)
31 /* YYINITDEPTH -- initial size of the parser's stacks. */
32 #define YYINITDEPTH 200
34 /* XXX temporary backwards compatibility */
35 #define PL_lex_brackets (PL_parser->lex_brackets)
36 #define PL_lex_brackstack (PL_parser->lex_brackstack)
37 #define PL_lex_casemods (PL_parser->lex_casemods)
38 #define PL_lex_casestack (PL_parser->lex_casestack)
39 #define PL_lex_defer (PL_parser->lex_defer)
40 #define PL_lex_dojoin (PL_parser->lex_dojoin)
41 #define PL_lex_expect (PL_parser->lex_expect)
42 #define PL_lex_formbrack (PL_parser->lex_formbrack)
43 #define PL_lex_inpat (PL_parser->lex_inpat)
44 #define PL_lex_inwhat (PL_parser->lex_inwhat)
45 #define PL_lex_op (PL_parser->lex_op)
46 #define PL_lex_repl (PL_parser->lex_repl)
47 #define PL_lex_starts (PL_parser->lex_starts)
48 #define PL_lex_stuff (PL_parser->lex_stuff)
49 #define PL_multi_start (PL_parser->multi_start)
50 #define PL_multi_open (PL_parser->multi_open)
51 #define PL_multi_close (PL_parser->multi_close)
52 #define PL_pending_ident (PL_parser->pending_ident)
53 #define PL_preambled (PL_parser->preambled)
54 #define PL_sublex_info (PL_parser->sublex_info)
55 #define PL_linestr (PL_parser->linestr)
56 #define PL_expect (PL_parser->expect)
57 #define PL_copline (PL_parser->copline)
58 #define PL_bufptr (PL_parser->bufptr)
59 #define PL_oldbufptr (PL_parser->oldbufptr)
60 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
61 #define PL_linestart (PL_parser->linestart)
62 #define PL_bufend (PL_parser->bufend)
63 #define PL_last_uni (PL_parser->last_uni)
64 #define PL_last_lop (PL_parser->last_lop)
65 #define PL_last_lop_op (PL_parser->last_lop_op)
66 #define PL_lex_state (PL_parser->lex_state)
67 #define PL_rsfp (PL_parser->rsfp)
68 #define PL_rsfp_filters (PL_parser->rsfp_filters)
69 #define PL_in_my (PL_parser->in_my)
70 #define PL_in_my_stash (PL_parser->in_my_stash)
71 #define PL_tokenbuf (PL_parser->tokenbuf)
72 #define PL_multi_end (PL_parser->multi_end)
73 #define PL_error_count (PL_parser->error_count)
76 # define PL_endwhite (PL_parser->endwhite)
77 # define PL_faketokens (PL_parser->faketokens)
78 # define PL_lasttoke (PL_parser->lasttoke)
79 # define PL_nextwhite (PL_parser->nextwhite)
80 # define PL_realtokenstart (PL_parser->realtokenstart)
81 # define PL_skipwhite (PL_parser->skipwhite)
82 # define PL_thisclose (PL_parser->thisclose)
83 # define PL_thismad (PL_parser->thismad)
84 # define PL_thisopen (PL_parser->thisopen)
85 # define PL_thisstuff (PL_parser->thisstuff)
86 # define PL_thistoken (PL_parser->thistoken)
87 # define PL_thiswhite (PL_parser->thiswhite)
88 # define PL_thiswhite (PL_parser->thiswhite)
89 # define PL_nexttoke (PL_parser->nexttoke)
90 # define PL_curforce (PL_parser->curforce)
92 # define PL_nexttoke (PL_parser->nexttoke)
93 # define PL_nexttype (PL_parser->nexttype)
94 # define PL_nextval (PL_parser->nextval)
98 S_pending_ident(pTHX);
100 static const char ident_too_long[] = "Identifier too long";
101 static const char commaless_variable_list[] = "comma-less variable list";
103 #ifndef PERL_NO_UTF16_FILTER
104 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
105 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
109 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
110 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
112 # define CURMAD(slot,sv)
113 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
116 #define XFAKEBRACK 128
117 #define XENUMMASK 127
119 #ifdef USE_UTF8_SCRIPTS
120 # define UTF (!IN_BYTES)
122 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
125 /* In variables named $^X, these are the legal values for X.
126 * 1999-02-27 mjd-perl-patch@plover.com */
127 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
129 /* On MacOS, respect nonbreaking spaces */
130 #ifdef MACOS_TRADITIONAL
131 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
133 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
136 /* LEX_* are values for PL_lex_state, the state of the lexer.
137 * They are arranged oddly so that the guard on the switch statement
138 * can get by with a single comparison (if the compiler is smart enough).
141 /* #define LEX_NOTPARSING 11 is done in perl.h. */
143 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
144 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
145 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
146 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
147 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
149 /* at end of code, eg "$x" followed by: */
150 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
151 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
153 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
154 string or after \E, $foo, etc */
155 #define LEX_INTERPCONST 2 /* NOT USED */
156 #define LEX_FORMLINE 1 /* expecting a format line */
157 #define LEX_KNOWNEXT 0 /* next token known; just return it */
161 static const char* const lex_state_names[] = {
180 #include "keywords.h"
182 /* CLINE is a macro that ensures PL_copline has a sane value */
187 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
190 # define SKIPSPACE0(s) skipspace0(s)
191 # define SKIPSPACE1(s) skipspace1(s)
192 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
193 # define PEEKSPACE(s) skipspace2(s,0)
195 # define SKIPSPACE0(s) skipspace(s)
196 # define SKIPSPACE1(s) skipspace(s)
197 # define SKIPSPACE2(s,tsv) skipspace(s)
198 # define PEEKSPACE(s) skipspace(s)
202 * Convenience functions to return different tokens and prime the
203 * lexer for the next token. They all take an argument.
205 * TOKEN : generic token (used for '(', DOLSHARP, etc)
206 * OPERATOR : generic operator
207 * AOPERATOR : assignment operator
208 * PREBLOCK : beginning the block after an if, while, foreach, ...
209 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
210 * PREREF : *EXPR where EXPR is not a simple identifier
211 * TERM : expression term
212 * LOOPX : loop exiting command (goto, last, dump, etc)
213 * FTST : file test operator
214 * FUN0 : zero-argument function
215 * FUN1 : not used, except for not, which isn't a UNIOP
216 * BOop : bitwise or or xor
218 * SHop : shift operator
219 * PWop : power operator
220 * PMop : pattern-matching operator
221 * Aop : addition-level operator
222 * Mop : multiplication-level operator
223 * Eop : equality-testing operator
224 * Rop : relational operator <= != gt
226 * Also see LOP and lop() below.
229 #ifdef DEBUGGING /* Serve -DT. */
230 # define REPORT(retval) tokereport((I32)retval)
232 # define REPORT(retval) (retval)
235 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
236 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
237 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
238 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
239 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
240 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
241 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
242 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
243 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
244 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
245 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
246 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
247 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
248 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
249 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
250 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
251 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
252 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
253 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
254 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
256 /* This bit of chicanery makes a unary function followed by
257 * a parenthesis into a function with one argument, highest precedence.
258 * The UNIDOR macro is for unary functions that can be followed by the //
259 * operator (such as C<shift // 0>).
261 #define UNI2(f,x) { \
265 PL_last_uni = PL_oldbufptr; \
266 PL_last_lop_op = f; \
268 return REPORT( (int)FUNC1 ); \
270 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
272 #define UNI(f) UNI2(f,XTERM)
273 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
275 #define UNIBRACK(f) { \
278 PL_last_uni = PL_oldbufptr; \
280 return REPORT( (int)FUNC1 ); \
282 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
285 /* grandfather return to old style */
286 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
290 /* how to interpret the yylval associated with the token */
294 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
300 static struct debug_tokens {
302 enum token_type type;
304 } const debug_tokens[] =
306 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
307 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
308 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
309 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
310 { ARROW, TOKENTYPE_NONE, "ARROW" },
311 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
312 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
313 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
314 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
315 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
316 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
317 { DO, TOKENTYPE_NONE, "DO" },
318 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
319 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
320 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
321 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
322 { ELSE, TOKENTYPE_NONE, "ELSE" },
323 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
324 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
325 { FOR, TOKENTYPE_IVAL, "FOR" },
326 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
327 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
328 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
329 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
330 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
331 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
332 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
333 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
334 { IF, TOKENTYPE_IVAL, "IF" },
335 { LABEL, TOKENTYPE_PVAL, "LABEL" },
336 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
337 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
338 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
339 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
340 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
341 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
342 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
343 { MY, TOKENTYPE_IVAL, "MY" },
344 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
345 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
346 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
347 { OROP, TOKENTYPE_IVAL, "OROP" },
348 { OROR, TOKENTYPE_NONE, "OROR" },
349 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
350 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
351 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
352 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
353 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
354 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
355 { PREINC, TOKENTYPE_NONE, "PREINC" },
356 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
357 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
358 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
359 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
360 { SUB, TOKENTYPE_NONE, "SUB" },
361 { THING, TOKENTYPE_OPVAL, "THING" },
362 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
363 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
364 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
365 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
366 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
367 { USE, TOKENTYPE_IVAL, "USE" },
368 { WHEN, TOKENTYPE_IVAL, "WHEN" },
369 { WHILE, TOKENTYPE_IVAL, "WHILE" },
370 { WORD, TOKENTYPE_OPVAL, "WORD" },
371 { 0, TOKENTYPE_NONE, NULL }
374 /* dump the returned token in rv, plus any optional arg in yylval */
377 S_tokereport(pTHX_ I32 rv)
381 const char *name = NULL;
382 enum token_type type = TOKENTYPE_NONE;
383 const struct debug_tokens *p;
384 SV* const report = newSVpvs("<== ");
386 for (p = debug_tokens; p->token; p++) {
387 if (p->token == (int)rv) {
394 Perl_sv_catpv(aTHX_ report, name);
395 else if ((char)rv > ' ' && (char)rv < '~')
396 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
398 sv_catpvs(report, "EOF");
400 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
403 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
406 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
408 case TOKENTYPE_OPNUM:
409 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
410 PL_op_name[yylval.ival]);
413 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
415 case TOKENTYPE_OPVAL:
417 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
418 PL_op_name[yylval.opval->op_type]);
419 if (yylval.opval->op_type == OP_CONST) {
420 Perl_sv_catpvf(aTHX_ report, " %s",
421 SvPEEK(cSVOPx_sv(yylval.opval)));
426 sv_catpvs(report, "(opval=null)");
429 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
435 /* print the buffer with suitable escapes */
438 S_printbuf(pTHX_ const char* fmt, const char* s)
440 SV* const tmp = newSVpvs("");
441 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
450 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
451 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
455 S_ao(pTHX_ int toketype)
458 if (*PL_bufptr == '=') {
460 if (toketype == ANDAND)
461 yylval.ival = OP_ANDASSIGN;
462 else if (toketype == OROR)
463 yylval.ival = OP_ORASSIGN;
464 else if (toketype == DORDOR)
465 yylval.ival = OP_DORASSIGN;
473 * When Perl expects an operator and finds something else, no_op
474 * prints the warning. It always prints "<something> found where
475 * operator expected. It prints "Missing semicolon on previous line?"
476 * if the surprise occurs at the start of the line. "do you need to
477 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
478 * where the compiler doesn't know if foo is a method call or a function.
479 * It prints "Missing operator before end of line" if there's nothing
480 * after the missing operator, or "... before <...>" if there is something
481 * after the missing operator.
485 S_no_op(pTHX_ const char *what, char *s)
488 char * const oldbp = PL_bufptr;
489 const bool is_first = (PL_oldbufptr == PL_linestart);
495 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
496 if (ckWARN_d(WARN_SYNTAX)) {
498 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
499 "\t(Missing semicolon on previous line?)\n");
500 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
502 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
504 if (t < PL_bufptr && isSPACE(*t))
505 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
506 "\t(Do you need to predeclare %.*s?)\n",
507 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
511 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
512 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
520 * Complain about missing quote/regexp/heredoc terminator.
521 * If it's called with NULL then it cauterizes the line buffer.
522 * If we're in a delimited string and the delimiter is a control
523 * character, it's reformatted into a two-char sequence like ^C.
528 S_missingterm(pTHX_ char *s)
534 char * const nl = strrchr(s,'\n');
540 iscntrl(PL_multi_close)
542 PL_multi_close < 32 || PL_multi_close == 127
546 tmpbuf[1] = (char)toCTRL(PL_multi_close);
551 *tmpbuf = (char)PL_multi_close;
555 q = strchr(s,'"') ? '\'' : '"';
556 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
559 #define FEATURE_IS_ENABLED(name) \
560 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
561 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
562 /* The longest string we pass in. */
563 #define MAX_FEATURE_LEN (sizeof("switch")-1)
566 * S_feature_is_enabled
567 * Check whether the named feature is enabled.
570 S_feature_is_enabled(pTHX_ const char *name, STRLEN namelen)
573 HV * const hinthv = GvHV(PL_hintgv);
574 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
575 assert(namelen <= MAX_FEATURE_LEN);
576 memcpy(&he_name[8], name, namelen);
578 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
586 Perl_deprecate(pTHX_ const char *s)
588 if (ckWARN(WARN_DEPRECATED))
589 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
593 Perl_deprecate_old(pTHX_ const char *s)
595 /* This function should NOT be called for any new deprecated warnings */
596 /* Use Perl_deprecate instead */
598 /* It is here to maintain backward compatibility with the pre-5.8 */
599 /* warnings category hierarchy. The "deprecated" category used to */
600 /* live under the "syntax" category. It is now a top-level category */
601 /* in its own right. */
603 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
604 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
605 "Use of %s is deprecated", s);
609 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
610 * utf16-to-utf8-reversed.
613 #ifdef PERL_CR_FILTER
617 register const char *s = SvPVX_const(sv);
618 register const char * const e = s + SvCUR(sv);
619 /* outer loop optimized to do nothing if there are no CR-LFs */
621 if (*s++ == '\r' && *s == '\n') {
622 /* hit a CR-LF, need to copy the rest */
623 register char *d = s - 1;
626 if (*s == '\r' && s[1] == '\n')
637 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
639 const I32 count = FILTER_READ(idx+1, sv, maxlen);
640 if (count > 0 && !maxlen)
651 * Create a parser object and initialise its parser and lexer fields
653 * rsfp is the opened file handle to read from (if any),
655 * line holds any initial content already read from the file (or in
656 * the case of no file, such as an eval, the whole contents);
658 * new_filter indicates that this is a new file and it shouldn't inherit
659 * the filters from the current parser (ie require).
663 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
666 const char *s = NULL;
668 yy_parser *parser, *oparser;
670 /* create and initialise a parser */
672 Newxz(parser, 1, yy_parser);
673 parser->old_parser = oparser = PL_parser;
676 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
677 parser->ps = parser->stack;
678 parser->stack_size = YYINITDEPTH;
680 parser->stack->state = 0;
681 parser->yyerrstatus = 0;
682 parser->yychar = YYEMPTY; /* Cause a token to be read. */
684 /* on scope exit, free this parser and restore any outer one */
686 parser->saved_curcop = PL_curcop;
688 /* initialise lexer state */
691 parser->curforce = -1;
693 parser->nexttoke = 0;
695 parser->copline = NOLINE;
696 parser->lex_state = LEX_NORMAL;
697 parser->expect = XSTATE;
699 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
700 : (AV*)SvREFCNT_inc(oparser->rsfp_filters);
702 Newx(parser->lex_brackstack, 120, char);
703 Newx(parser->lex_casestack, 12, char);
704 *parser->lex_casestack = '\0';
707 s = SvPV_const(line, len);
713 parser->linestr = newSVpvs("\n;");
714 } else if (SvREADONLY(line) || s[len-1] != ';') {
715 parser->linestr = newSVsv(line);
717 sv_catpvs(parser->linestr, "\n;");
720 SvREFCNT_inc_simple_void_NN(line);
721 parser->linestr = line;
723 parser->oldoldbufptr =
726 parser->linestart = SvPVX(parser->linestr);
727 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
728 parser->last_lop = parser->last_uni = NULL;
732 /* delete a parser object */
735 Perl_parser_free(pTHX_ const yy_parser *parser)
737 PL_curcop = parser->saved_curcop;
738 SvREFCNT_dec(parser->linestr);
740 if (parser->rsfp == PerlIO_stdin())
741 PerlIO_clearerr(parser->rsfp);
742 else if (parser->rsfp && parser->old_parser
743 && parser->rsfp != parser->old_parser->rsfp)
744 PerlIO_close(parser->rsfp);
745 SvREFCNT_dec(parser->rsfp_filters);
747 Safefree(parser->stack);
748 Safefree(parser->lex_brackstack);
749 Safefree(parser->lex_casestack);
750 PL_parser = parser->old_parser;
757 * Finalizer for lexing operations. Must be called when the parser is
758 * done with the lexer.
765 PL_doextract = FALSE;
770 * This subroutine has nothing to do with tilting, whether at windmills
771 * or pinball tables. Its name is short for "increment line". It
772 * increments the current line number in CopLINE(PL_curcop) and checks
773 * to see whether the line starts with a comment of the form
774 * # line 500 "foo.pm"
775 * If so, it sets the current line number and file to the values in the comment.
779 S_incline(pTHX_ const char *s)
786 CopLINE_inc(PL_curcop);
789 while (SPACE_OR_TAB(*s))
791 if (strnEQ(s, "line", 4))
795 if (SPACE_OR_TAB(*s))
799 while (SPACE_OR_TAB(*s))
807 while (SPACE_OR_TAB(*s))
809 if (*s == '"' && (t = strchr(s+1, '"'))) {
819 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
821 if (*e != '\n' && *e != '\0')
822 return; /* false alarm */
825 const STRLEN len = t - s;
827 SV *const temp_sv = CopFILESV(PL_curcop);
833 tmplen = SvCUR(temp_sv);
839 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
840 /* must copy *{"::_<(eval N)[oldfilename:L]"}
841 * to *{"::_<newfilename"} */
842 /* However, the long form of evals is only turned on by the
843 debugger - usually they're "(eval %lu)" */
847 STRLEN tmplen2 = len;
848 if (tmplen + 2 <= sizeof smallbuf)
851 Newx(tmpbuf, tmplen + 2, char);
854 memcpy(tmpbuf + 2, cf, tmplen);
856 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
861 if (tmplen2 + 2 <= sizeof smallbuf)
864 Newx(tmpbuf2, tmplen2 + 2, char);
866 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
867 /* Either they malloc'd it, or we malloc'd it,
868 so no prefix is present in ours. */
873 memcpy(tmpbuf2 + 2, s, tmplen2);
876 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
878 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
879 /* adjust ${"::_<newfilename"} to store the new file name */
880 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
881 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
882 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
885 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
887 if (tmpbuf != smallbuf) Safefree(tmpbuf);
890 CopFILE_free(PL_curcop);
891 CopFILE_setn(PL_curcop, s, len);
893 CopLINE_set(PL_curcop, atoi(n)-1);
897 /* skip space before PL_thistoken */
900 S_skipspace0(pTHX_ register char *s)
907 PL_thiswhite = newSVpvs("");
908 sv_catsv(PL_thiswhite, PL_skipwhite);
909 sv_free(PL_skipwhite);
912 PL_realtokenstart = s - SvPVX(PL_linestr);
916 /* skip space after PL_thistoken */
919 S_skipspace1(pTHX_ register char *s)
921 const char *start = s;
922 I32 startoff = start - SvPVX(PL_linestr);
927 start = SvPVX(PL_linestr) + startoff;
928 if (!PL_thistoken && PL_realtokenstart >= 0) {
929 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
930 PL_thistoken = newSVpvn(tstart, start - tstart);
932 PL_realtokenstart = -1;
935 PL_nextwhite = newSVpvs("");
936 sv_catsv(PL_nextwhite, PL_skipwhite);
937 sv_free(PL_skipwhite);
944 S_skipspace2(pTHX_ register char *s, SV **svp)
947 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
948 const I32 startoff = s - SvPVX(PL_linestr);
951 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
952 if (!PL_madskills || !svp)
954 start = SvPVX(PL_linestr) + startoff;
955 if (!PL_thistoken && PL_realtokenstart >= 0) {
956 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
957 PL_thistoken = newSVpvn(tstart, start - tstart);
958 PL_realtokenstart = -1;
963 sv_setsv(*svp, PL_skipwhite);
964 sv_free(PL_skipwhite);
973 S_update_debugger_info(pTHX_ SV *orig_sv, const char *buf, STRLEN len)
975 AV *av = CopFILEAVx(PL_curcop);
977 SV * const sv = newSV_type(SVt_PVMG);
979 sv_setsv(sv, orig_sv);
981 sv_setpvn(sv, buf, len);
984 av_store(av, (I32)CopLINE(PL_curcop), sv);
990 * Called to gobble the appropriate amount and type of whitespace.
991 * Skips comments as well.
995 S_skipspace(pTHX_ register char *s)
1000 int startoff = s - SvPVX(PL_linestr);
1003 sv_free(PL_skipwhite);
1008 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1009 while (s < PL_bufend && SPACE_OR_TAB(*s))
1019 SSize_t oldprevlen, oldoldprevlen;
1020 SSize_t oldloplen = 0, oldunilen = 0;
1021 while (s < PL_bufend && isSPACE(*s)) {
1022 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
1027 if (s < PL_bufend && *s == '#') {
1028 while (s < PL_bufend && *s != '\n')
1030 if (s < PL_bufend) {
1032 if (PL_in_eval && !PL_rsfp) {
1039 /* only continue to recharge the buffer if we're at the end
1040 * of the buffer, we're not reading from a source filter, and
1041 * we're in normal lexing mode
1043 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1044 PL_lex_state == LEX_FORMLINE)
1051 /* try to recharge the buffer */
1053 curoff = s - SvPVX(PL_linestr);
1056 if ((s = filter_gets(PL_linestr, PL_rsfp,
1057 (prevlen = SvCUR(PL_linestr)))) == NULL)
1060 if (PL_madskills && curoff != startoff) {
1062 PL_skipwhite = newSVpvs("");
1063 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1067 /* mustn't throw out old stuff yet if madpropping */
1068 SvCUR(PL_linestr) = curoff;
1069 s = SvPVX(PL_linestr) + curoff;
1071 if (curoff && s[-1] == '\n')
1075 /* end of file. Add on the -p or -n magic */
1076 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1079 sv_catpvs(PL_linestr,
1080 ";}continue{print or die qq(-p destination: $!\\n);}");
1082 sv_setpvs(PL_linestr,
1083 ";}continue{print or die qq(-p destination: $!\\n);}");
1085 PL_minus_n = PL_minus_p = 0;
1087 else if (PL_minus_n) {
1089 sv_catpvn(PL_linestr, ";}", 2);
1091 sv_setpvn(PL_linestr, ";}", 2);
1097 sv_catpvn(PL_linestr,";", 1);
1099 sv_setpvn(PL_linestr,";", 1);
1102 /* reset variables for next time we lex */
1103 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1109 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1110 PL_last_lop = PL_last_uni = NULL;
1112 /* Close the filehandle. Could be from -P preprocessor,
1113 * STDIN, or a regular file. If we were reading code from
1114 * STDIN (because the commandline held no -e or filename)
1115 * then we don't close it, we reset it so the code can
1116 * read from STDIN too.
1119 if (PL_preprocess && !PL_in_eval)
1120 (void)PerlProc_pclose(PL_rsfp);
1121 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1122 PerlIO_clearerr(PL_rsfp);
1124 (void)PerlIO_close(PL_rsfp);
1129 /* not at end of file, so we only read another line */
1130 /* make corresponding updates to old pointers, for yyerror() */
1131 oldprevlen = PL_oldbufptr - PL_bufend;
1132 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1134 oldunilen = PL_last_uni - PL_bufend;
1136 oldloplen = PL_last_lop - PL_bufend;
1137 PL_linestart = PL_bufptr = s + prevlen;
1138 PL_bufend = s + SvCUR(PL_linestr);
1140 PL_oldbufptr = s + oldprevlen;
1141 PL_oldoldbufptr = s + oldoldprevlen;
1143 PL_last_uni = s + oldunilen;
1145 PL_last_lop = s + oldloplen;
1148 /* debugger active and we're not compiling the debugger code,
1149 * so store the line into the debugger's array of lines
1151 if (PERLDB_LINE && PL_curstash != PL_debstash)
1152 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1159 PL_skipwhite = newSVpvs("");
1160 curoff = s - SvPVX(PL_linestr);
1161 if (curoff - startoff)
1162 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1171 * Check the unary operators to ensure there's no ambiguity in how they're
1172 * used. An ambiguous piece of code would be:
1174 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1175 * the +5 is its argument.
1185 if (PL_oldoldbufptr != PL_last_uni)
1187 while (isSPACE(*PL_last_uni))
1190 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1192 if ((t = strchr(s, '(')) && t < PL_bufptr)
1195 if (ckWARN_d(WARN_AMBIGUOUS)){
1196 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1197 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1198 (int)(s - PL_last_uni), PL_last_uni);
1203 * LOP : macro to build a list operator. Its behaviour has been replaced
1204 * with a subroutine, S_lop() for which LOP is just another name.
1207 #define LOP(f,x) return lop(f,x,s)
1211 * Build a list operator (or something that might be one). The rules:
1212 * - if we have a next token, then it's a list operator [why?]
1213 * - if the next thing is an opening paren, then it's a function
1214 * - else it's a list operator
1218 S_lop(pTHX_ I32 f, int x, char *s)
1225 PL_last_lop = PL_oldbufptr;
1226 PL_last_lop_op = (OPCODE)f;
1229 return REPORT(LSTOP);
1232 return REPORT(LSTOP);
1235 return REPORT(FUNC);
1238 return REPORT(FUNC);
1240 return REPORT(LSTOP);
1246 * Sets up for an eventual force_next(). start_force(0) basically does
1247 * an unshift, while start_force(-1) does a push. yylex removes items
1252 S_start_force(pTHX_ int where)
1256 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1257 where = PL_lasttoke;
1258 assert(PL_curforce < 0 || PL_curforce == where);
1259 if (PL_curforce != where) {
1260 for (i = PL_lasttoke; i > where; --i) {
1261 PL_nexttoke[i] = PL_nexttoke[i-1];
1265 if (PL_curforce < 0) /* in case of duplicate start_force() */
1266 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1267 PL_curforce = where;
1270 curmad('^', newSVpvs(""));
1271 CURMAD('_', PL_nextwhite);
1276 S_curmad(pTHX_ char slot, SV *sv)
1282 if (PL_curforce < 0)
1283 where = &PL_thismad;
1285 where = &PL_nexttoke[PL_curforce].next_mad;
1288 sv_setpvn(sv, "", 0);
1291 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1293 else if (PL_encoding) {
1294 sv_recode_to_utf8(sv, PL_encoding);
1299 /* keep a slot open for the head of the list? */
1300 if (slot != '_' && *where && (*where)->mad_key == '^') {
1301 (*where)->mad_key = slot;
1302 sv_free((SV*)((*where)->mad_val));
1303 (*where)->mad_val = (void*)sv;
1306 addmad(newMADsv(slot, sv), where, 0);
1309 # define start_force(where) NOOP
1310 # define curmad(slot, sv) NOOP
1315 * When the lexer realizes it knows the next token (for instance,
1316 * it is reordering tokens for the parser) then it can call S_force_next
1317 * to know what token to return the next time the lexer is called. Caller
1318 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1319 * and possibly PL_expect to ensure the lexer handles the token correctly.
1323 S_force_next(pTHX_ I32 type)
1327 if (PL_curforce < 0)
1328 start_force(PL_lasttoke);
1329 PL_nexttoke[PL_curforce].next_type = type;
1330 if (PL_lex_state != LEX_KNOWNEXT)
1331 PL_lex_defer = PL_lex_state;
1332 PL_lex_state = LEX_KNOWNEXT;
1333 PL_lex_expect = PL_expect;
1336 PL_nexttype[PL_nexttoke] = type;
1338 if (PL_lex_state != LEX_KNOWNEXT) {
1339 PL_lex_defer = PL_lex_state;
1340 PL_lex_expect = PL_expect;
1341 PL_lex_state = LEX_KNOWNEXT;
1347 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1350 SV * const sv = newSVpvn(start,len);
1351 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1358 * When the lexer knows the next thing is a word (for instance, it has
1359 * just seen -> and it knows that the next char is a word char, then
1360 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1364 * char *start : buffer position (must be within PL_linestr)
1365 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1366 * int check_keyword : if true, Perl checks to make sure the word isn't
1367 * a keyword (do this if the word is a label, e.g. goto FOO)
1368 * int allow_pack : if true, : characters will also be allowed (require,
1369 * use, etc. do this)
1370 * int allow_initial_tick : used by the "sub" lexer only.
1374 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1380 start = SKIPSPACE1(start);
1382 if (isIDFIRST_lazy_if(s,UTF) ||
1383 (allow_pack && *s == ':') ||
1384 (allow_initial_tick && *s == '\'') )
1386 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1387 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1389 start_force(PL_curforce);
1391 curmad('X', newSVpvn(start,s-start));
1392 if (token == METHOD) {
1397 PL_expect = XOPERATOR;
1401 curmad('g', newSVpvs( "forced" ));
1402 NEXTVAL_NEXTTOKE.opval
1403 = (OP*)newSVOP(OP_CONST,0,
1404 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1405 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1413 * Called when the lexer wants $foo *foo &foo etc, but the program
1414 * text only contains the "foo" portion. The first argument is a pointer
1415 * to the "foo", and the second argument is the type symbol to prefix.
1416 * Forces the next token to be a "WORD".
1417 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1421 S_force_ident(pTHX_ register const char *s, int kind)
1425 const STRLEN len = strlen(s);
1426 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1427 start_force(PL_curforce);
1428 NEXTVAL_NEXTTOKE.opval = o;
1431 o->op_private = OPpCONST_ENTERED;
1432 /* XXX see note in pp_entereval() for why we forgo typo
1433 warnings if the symbol must be introduced in an eval.
1435 gv_fetchpvn_flags(s, len,
1436 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1438 kind == '$' ? SVt_PV :
1439 kind == '@' ? SVt_PVAV :
1440 kind == '%' ? SVt_PVHV :
1448 Perl_str_to_version(pTHX_ SV *sv)
1453 const char *start = SvPV_const(sv,len);
1454 const char * const end = start + len;
1455 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1456 while (start < end) {
1460 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1465 retval += ((NV)n)/nshift;
1474 * Forces the next token to be a version number.
1475 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1476 * and if "guessing" is TRUE, then no new token is created (and the caller
1477 * must use an alternative parsing method).
1481 S_force_version(pTHX_ char *s, int guessing)
1487 I32 startoff = s - SvPVX(PL_linestr);
1496 while (isDIGIT(*d) || *d == '_' || *d == '.')
1500 start_force(PL_curforce);
1501 curmad('X', newSVpvn(s,d-s));
1504 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1506 s = scan_num(s, &yylval);
1507 version = yylval.opval;
1508 ver = cSVOPx(version)->op_sv;
1509 if (SvPOK(ver) && !SvNIOK(ver)) {
1510 SvUPGRADE(ver, SVt_PVNV);
1511 SvNV_set(ver, str_to_version(ver));
1512 SvNOK_on(ver); /* hint that it is a version */
1515 else if (guessing) {
1518 sv_free(PL_nextwhite); /* let next token collect whitespace */
1520 s = SvPVX(PL_linestr) + startoff;
1528 if (PL_madskills && !version) {
1529 sv_free(PL_nextwhite); /* let next token collect whitespace */
1531 s = SvPVX(PL_linestr) + startoff;
1534 /* NOTE: The parser sees the package name and the VERSION swapped */
1535 start_force(PL_curforce);
1536 NEXTVAL_NEXTTOKE.opval = version;
1544 * Tokenize a quoted string passed in as an SV. It finds the next
1545 * chunk, up to end of string or a backslash. It may make a new
1546 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1551 S_tokeq(pTHX_ SV *sv)
1555 register char *send;
1563 s = SvPV_force(sv, len);
1564 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1567 while (s < send && *s != '\\')
1572 if ( PL_hints & HINT_NEW_STRING ) {
1573 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1579 if (s + 1 < send && (s[1] == '\\'))
1580 s++; /* all that, just for this */
1585 SvCUR_set(sv, d - SvPVX_const(sv));
1587 if ( PL_hints & HINT_NEW_STRING )
1588 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
1593 * Now come three functions related to double-quote context,
1594 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1595 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1596 * interact with PL_lex_state, and create fake ( ... ) argument lists
1597 * to handle functions and concatenation.
1598 * They assume that whoever calls them will be setting up a fake
1599 * join call, because each subthing puts a ',' after it. This lets
1602 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1604 * (I'm not sure whether the spurious commas at the end of lcfirst's
1605 * arguments and join's arguments are created or not).
1610 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1612 * Pattern matching will set PL_lex_op to the pattern-matching op to
1613 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1615 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1617 * Everything else becomes a FUNC.
1619 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1620 * had an OP_CONST or OP_READLINE). This just sets us up for a
1621 * call to S_sublex_push().
1625 S_sublex_start(pTHX)
1628 register const I32 op_type = yylval.ival;
1630 if (op_type == OP_NULL) {
1631 yylval.opval = PL_lex_op;
1635 if (op_type == OP_CONST || op_type == OP_READLINE) {
1636 SV *sv = tokeq(PL_lex_stuff);
1638 if (SvTYPE(sv) == SVt_PVIV) {
1639 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1641 const char * const p = SvPV_const(sv, len);
1642 SV * const nsv = newSVpvn(p, len);
1648 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1649 PL_lex_stuff = NULL;
1650 /* Allow <FH> // "foo" */
1651 if (op_type == OP_READLINE)
1652 PL_expect = XTERMORDORDOR;
1655 else if (op_type == OP_BACKTICK && PL_lex_op) {
1656 /* readpipe() vas overriden */
1657 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1658 yylval.opval = PL_lex_op;
1660 PL_lex_stuff = NULL;
1664 PL_sublex_info.super_state = PL_lex_state;
1665 PL_sublex_info.sub_inwhat = (U16)op_type;
1666 PL_sublex_info.sub_op = PL_lex_op;
1667 PL_lex_state = LEX_INTERPPUSH;
1671 yylval.opval = PL_lex_op;
1681 * Create a new scope to save the lexing state. The scope will be
1682 * ended in S_sublex_done. Returns a '(', starting the function arguments
1683 * to the uc, lc, etc. found before.
1684 * Sets PL_lex_state to LEX_INTERPCONCAT.
1693 PL_lex_state = PL_sublex_info.super_state;
1694 SAVEBOOL(PL_lex_dojoin);
1695 SAVEI32(PL_lex_brackets);
1696 SAVEI32(PL_lex_casemods);
1697 SAVEI32(PL_lex_starts);
1698 SAVEI8(PL_lex_state);
1699 SAVEVPTR(PL_lex_inpat);
1700 SAVEI16(PL_lex_inwhat);
1701 SAVECOPLINE(PL_curcop);
1702 SAVEPPTR(PL_bufptr);
1703 SAVEPPTR(PL_bufend);
1704 SAVEPPTR(PL_oldbufptr);
1705 SAVEPPTR(PL_oldoldbufptr);
1706 SAVEPPTR(PL_last_lop);
1707 SAVEPPTR(PL_last_uni);
1708 SAVEPPTR(PL_linestart);
1709 SAVESPTR(PL_linestr);
1710 SAVEGENERICPV(PL_lex_brackstack);
1711 SAVEGENERICPV(PL_lex_casestack);
1713 PL_linestr = PL_lex_stuff;
1714 PL_lex_stuff = NULL;
1716 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1717 = SvPVX(PL_linestr);
1718 PL_bufend += SvCUR(PL_linestr);
1719 PL_last_lop = PL_last_uni = NULL;
1720 SAVEFREESV(PL_linestr);
1722 PL_lex_dojoin = FALSE;
1723 PL_lex_brackets = 0;
1724 Newx(PL_lex_brackstack, 120, char);
1725 Newx(PL_lex_casestack, 12, char);
1726 PL_lex_casemods = 0;
1727 *PL_lex_casestack = '\0';
1729 PL_lex_state = LEX_INTERPCONCAT;
1730 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1732 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1733 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1734 PL_lex_inpat = PL_sublex_info.sub_op;
1736 PL_lex_inpat = NULL;
1743 * Restores lexer state after a S_sublex_push.
1750 if (!PL_lex_starts++) {
1751 SV * const sv = newSVpvs("");
1752 if (SvUTF8(PL_linestr))
1754 PL_expect = XOPERATOR;
1755 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1759 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1760 PL_lex_state = LEX_INTERPCASEMOD;
1764 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1765 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1766 PL_linestr = PL_lex_repl;
1768 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1769 PL_bufend += SvCUR(PL_linestr);
1770 PL_last_lop = PL_last_uni = NULL;
1771 SAVEFREESV(PL_linestr);
1772 PL_lex_dojoin = FALSE;
1773 PL_lex_brackets = 0;
1774 PL_lex_casemods = 0;
1775 *PL_lex_casestack = '\0';
1777 if (SvEVALED(PL_lex_repl)) {
1778 PL_lex_state = LEX_INTERPNORMAL;
1780 /* we don't clear PL_lex_repl here, so that we can check later
1781 whether this is an evalled subst; that means we rely on the
1782 logic to ensure sublex_done() is called again only via the
1783 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1786 PL_lex_state = LEX_INTERPCONCAT;
1796 PL_endwhite = newSVpvs("");
1797 sv_catsv(PL_endwhite, PL_thiswhite);
1801 sv_setpvn(PL_thistoken,"",0);
1803 PL_realtokenstart = -1;
1807 PL_bufend = SvPVX(PL_linestr);
1808 PL_bufend += SvCUR(PL_linestr);
1809 PL_expect = XOPERATOR;
1810 PL_sublex_info.sub_inwhat = 0;
1818 Extracts a pattern, double-quoted string, or transliteration. This
1821 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1822 processing a pattern (PL_lex_inpat is true), a transliteration
1823 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1825 Returns a pointer to the character scanned up to. If this is
1826 advanced from the start pointer supplied (i.e. if anything was
1827 successfully parsed), will leave an OP for the substring scanned
1828 in yylval. Caller must intuit reason for not parsing further
1829 by looking at the next characters herself.
1833 double-quoted style: \r and \n
1834 regexp special ones: \D \s
1837 case and quoting: \U \Q \E
1838 stops on @ and $, but not for $ as tail anchor
1840 In transliterations:
1841 characters are VERY literal, except for - not at the start or end
1842 of the string, which indicates a range. If the range is in bytes,
1843 scan_const expands the range to the full set of intermediate
1844 characters. If the range is in utf8, the hyphen is replaced with
1845 a certain range mark which will be handled by pmtrans() in op.c.
1847 In double-quoted strings:
1849 double-quoted style: \r and \n
1851 deprecated backrefs: \1 (in substitution replacements)
1852 case and quoting: \U \Q \E
1855 scan_const does *not* construct ops to handle interpolated strings.
1856 It stops processing as soon as it finds an embedded $ or @ variable
1857 and leaves it to the caller to work out what's going on.
1859 embedded arrays (whether in pattern or not) could be:
1860 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1862 $ in double-quoted strings must be the symbol of an embedded scalar.
1864 $ in pattern could be $foo or could be tail anchor. Assumption:
1865 it's a tail anchor if $ is the last thing in the string, or if it's
1866 followed by one of "()| \r\n\t"
1868 \1 (backreferences) are turned into $1
1870 The structure of the code is
1871 while (there's a character to process) {
1872 handle transliteration ranges
1873 skip regexp comments /(?#comment)/ and codes /(?{code})/
1874 skip #-initiated comments in //x patterns
1875 check for embedded arrays
1876 check for embedded scalars
1878 leave intact backslashes from leaveit (below)
1879 deprecate \1 in substitution replacements
1880 handle string-changing backslashes \l \U \Q \E, etc.
1881 switch (what was escaped) {
1882 handle \- in a transliteration (becomes a literal -)
1883 handle \132 (octal characters)
1884 handle \x15 and \x{1234} (hex characters)
1885 handle \N{name} (named characters)
1886 handle \cV (control characters)
1887 handle printf-style backslashes (\f, \r, \n, etc)
1889 } (end if backslash)
1890 } (end while character to read)
1895 S_scan_const(pTHX_ char *start)
1898 register char *send = PL_bufend; /* end of the constant */
1899 SV *sv = newSV(send - start); /* sv for the constant */
1900 register char *s = start; /* start of the constant */
1901 register char *d = SvPVX(sv); /* destination for copies */
1902 bool dorange = FALSE; /* are we in a translit range? */
1903 bool didrange = FALSE; /* did we just finish a range? */
1904 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1905 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1908 UV literal_endpoint = 0;
1909 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1912 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1913 /* If we are doing a trans and we know we want UTF8 set expectation */
1914 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1915 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1919 while (s < send || dorange) {
1920 /* get transliterations out of the way (they're most literal) */
1921 if (PL_lex_inwhat == OP_TRANS) {
1922 /* expand a range A-Z to the full set of characters. AIE! */
1924 I32 i; /* current expanded character */
1925 I32 min; /* first character in range */
1926 I32 max; /* last character in range */
1937 char * const c = (char*)utf8_hop((U8*)d, -1);
1941 *c = (char)UTF_TO_NATIVE(0xff);
1942 /* mark the range as done, and continue */
1948 i = d - SvPVX_const(sv); /* remember current offset */
1951 SvLEN(sv) + (has_utf8 ?
1952 (512 - UTF_CONTINUATION_MARK +
1955 /* How many two-byte within 0..255: 128 in UTF-8,
1956 * 96 in UTF-8-mod. */
1958 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1960 d = SvPVX(sv) + i; /* refresh d after realloc */
1964 for (j = 0; j <= 1; j++) {
1965 char * const c = (char*)utf8_hop((U8*)d, -1);
1966 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1972 max = (U8)0xff; /* only to \xff */
1973 uvmax = uv; /* \x{100} to uvmax */
1975 d = c; /* eat endpoint chars */
1980 d -= 2; /* eat the first char and the - */
1981 min = (U8)*d; /* first char in range */
1982 max = (U8)d[1]; /* last char in range */
1989 "Invalid range \"%c-%c\" in transliteration operator",
1990 (char)min, (char)max);
1994 if (literal_endpoint == 2 &&
1995 ((isLOWER(min) && isLOWER(max)) ||
1996 (isUPPER(min) && isUPPER(max)))) {
1998 for (i = min; i <= max; i++)
2000 *d++ = NATIVE_TO_NEED(has_utf8,i);
2002 for (i = min; i <= max; i++)
2004 *d++ = NATIVE_TO_NEED(has_utf8,i);
2009 for (i = min; i <= max; i++)
2012 const U8 ch = (U8)NATIVE_TO_UTF(i);
2013 if (UNI_IS_INVARIANT(ch))
2016 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2017 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2026 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2028 *d++ = (char)UTF_TO_NATIVE(0xff);
2030 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2034 /* mark the range as done, and continue */
2038 literal_endpoint = 0;
2043 /* range begins (ignore - as first or last char) */
2044 else if (*s == '-' && s+1 < send && s != start) {
2046 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2053 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2063 literal_endpoint = 0;
2064 native_range = TRUE;
2069 /* if we get here, we're not doing a transliteration */
2071 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2072 except for the last char, which will be done separately. */
2073 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2075 while (s+1 < send && *s != ')')
2076 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2078 else if (s[2] == '{' /* This should match regcomp.c */
2079 || (s[2] == '?' && s[3] == '{'))
2082 char *regparse = s + (s[2] == '{' ? 3 : 4);
2085 while (count && (c = *regparse)) {
2086 if (c == '\\' && regparse[1])
2094 if (*regparse != ')')
2095 regparse--; /* Leave one char for continuation. */
2096 while (s < regparse)
2097 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2101 /* likewise skip #-initiated comments in //x patterns */
2102 else if (*s == '#' && PL_lex_inpat &&
2103 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2104 while (s+1 < send && *s != '\n')
2105 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2108 /* check for embedded arrays
2109 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2111 else if (*s == '@' && s[1]) {
2112 if (isALNUM_lazy_if(s+1,UTF))
2114 if (strchr(":'{$", s[1]))
2116 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2117 break; /* in regexp, neither @+ nor @- are interpolated */
2120 /* check for embedded scalars. only stop if we're sure it's a
2123 else if (*s == '$') {
2124 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2126 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2127 break; /* in regexp, $ might be tail anchor */
2130 /* End of else if chain - OP_TRANS rejoin rest */
2133 if (*s == '\\' && s+1 < send) {
2136 /* deprecate \1 in strings and substitution replacements */
2137 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2138 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2140 if (ckWARN(WARN_SYNTAX))
2141 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2146 /* string-change backslash escapes */
2147 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2151 /* skip any other backslash escapes in a pattern */
2152 else if (PL_lex_inpat) {
2153 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2154 goto default_action;
2157 /* if we get here, it's either a quoted -, or a digit */
2160 /* quoted - in transliterations */
2162 if (PL_lex_inwhat == OP_TRANS) {
2169 if ((isALPHA(*s) || isDIGIT(*s)) &&
2171 Perl_warner(aTHX_ packWARN(WARN_MISC),
2172 "Unrecognized escape \\%c passed through",
2174 /* default action is to copy the quoted character */
2175 goto default_action;
2178 /* \132 indicates an octal constant */
2179 case '0': case '1': case '2': case '3':
2180 case '4': case '5': case '6': case '7':
2184 uv = grok_oct(s, &len, &flags, NULL);
2187 goto NUM_ESCAPE_INSERT;
2189 /* \x24 indicates a hex constant */
2193 char* const e = strchr(s, '}');
2194 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2195 PERL_SCAN_DISALLOW_PREFIX;
2200 yyerror("Missing right brace on \\x{}");
2204 uv = grok_hex(s, &len, &flags, NULL);
2210 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2211 uv = grok_hex(s, &len, &flags, NULL);
2217 /* Insert oct or hex escaped character.
2218 * There will always enough room in sv since such
2219 * escapes will be longer than any UTF-8 sequence
2220 * they can end up as. */
2222 /* We need to map to chars to ASCII before doing the tests
2225 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2226 if (!has_utf8 && uv > 255) {
2227 /* Might need to recode whatever we have
2228 * accumulated so far if it contains any
2231 * (Can't we keep track of that and avoid
2232 * this rescan? --jhi)
2236 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2237 if (!NATIVE_IS_INVARIANT(*c)) {
2242 const STRLEN offset = d - SvPVX_const(sv);
2244 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2248 while (src >= (const U8 *)SvPVX_const(sv)) {
2249 if (!NATIVE_IS_INVARIANT(*src)) {
2250 const U8 ch = NATIVE_TO_ASCII(*src);
2251 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2252 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2262 if (has_utf8 || uv > 255) {
2263 d = (char*)uvchr_to_utf8((U8*)d, uv);
2265 if (PL_lex_inwhat == OP_TRANS &&
2266 PL_sublex_info.sub_op) {
2267 PL_sublex_info.sub_op->op_private |=
2268 (PL_lex_repl ? OPpTRANS_FROM_UTF
2272 if (uv > 255 && !dorange)
2273 native_range = FALSE;
2285 /* \N{LATIN SMALL LETTER A} is a named character */
2289 char* e = strchr(s, '}');
2295 yyerror("Missing right brace on \\N{}");
2299 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2301 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2302 PERL_SCAN_DISALLOW_PREFIX;
2305 uv = grok_hex(s, &len, &flags, NULL);
2306 if ( e > s && len != (STRLEN)(e - s) ) {
2310 goto NUM_ESCAPE_INSERT;
2312 res = newSVpvn(s + 1, e - s - 1);
2313 res = new_constant( NULL, 0, "charnames",
2314 res, NULL, s - 2, e - s + 3 );
2316 sv_utf8_upgrade(res);
2317 str = SvPV_const(res,len);
2318 #ifdef EBCDIC_NEVER_MIND
2319 /* charnames uses pack U and that has been
2320 * recently changed to do the below uni->native
2321 * mapping, so this would be redundant (and wrong,
2322 * the code point would be doubly converted).
2323 * But leave this in just in case the pack U change
2324 * gets revoked, but the semantics is still
2325 * desireable for charnames. --jhi */
2327 UV uv = utf8_to_uvchr((const U8*)str, 0);
2330 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2332 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2333 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2334 str = SvPV_const(res, len);
2338 if (!has_utf8 && SvUTF8(res)) {
2339 const char * const ostart = SvPVX_const(sv);
2340 SvCUR_set(sv, d - ostart);
2343 sv_utf8_upgrade(sv);
2344 /* this just broke our allocation above... */
2345 SvGROW(sv, (STRLEN)(send - start));
2346 d = SvPVX(sv) + SvCUR(sv);
2349 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2350 const char * const odest = SvPVX_const(sv);
2352 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2353 d = SvPVX(sv) + (d - odest);
2357 native_range = FALSE; /* \N{} is guessed to be Unicode */
2359 Copy(str, d, len, char);
2366 yyerror("Missing braces on \\N{}");
2369 /* \c is a control character */
2378 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2381 yyerror("Missing control char name in \\c");
2385 /* printf-style backslashes, formfeeds, newlines, etc */
2387 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2390 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2393 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2396 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2399 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2402 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2405 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2411 } /* end if (backslash) */
2418 /* If we started with encoded form, or already know we want it
2419 and then encode the next character */
2420 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2422 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2423 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2426 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2427 const STRLEN off = d - SvPVX_const(sv);
2428 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2430 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2433 if (uv > 255 && !dorange)
2434 native_range = FALSE;
2438 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2440 } /* while loop to process each character */
2442 /* terminate the string and set up the sv */
2444 SvCUR_set(sv, d - SvPVX_const(sv));
2445 if (SvCUR(sv) >= SvLEN(sv))
2446 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2449 if (PL_encoding && !has_utf8) {
2450 sv_recode_to_utf8(sv, PL_encoding);
2456 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2457 PL_sublex_info.sub_op->op_private |=
2458 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2462 /* shrink the sv if we allocated more than we used */
2463 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2464 SvPV_shrink_to_cur(sv);
2467 /* return the substring (via yylval) only if we parsed anything */
2468 if (s > PL_bufptr) {
2469 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
2470 const char *const key = PL_lex_inpat ? "qr" : "q";
2471 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
2475 if (PL_lex_inwhat == OP_TRANS) {
2478 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
2486 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
2489 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2496 * Returns TRUE if there's more to the expression (e.g., a subscript),
2499 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2501 * ->[ and ->{ return TRUE
2502 * { and [ outside a pattern are always subscripts, so return TRUE
2503 * if we're outside a pattern and it's not { or [, then return FALSE
2504 * if we're in a pattern and the first char is a {
2505 * {4,5} (any digits around the comma) returns FALSE
2506 * if we're in a pattern and the first char is a [
2508 * [SOMETHING] has a funky algorithm to decide whether it's a
2509 * character class or not. It has to deal with things like
2510 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2511 * anything else returns TRUE
2514 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2517 S_intuit_more(pTHX_ register char *s)
2520 if (PL_lex_brackets)
2522 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2524 if (*s != '{' && *s != '[')
2529 /* In a pattern, so maybe we have {n,m}. */
2546 /* On the other hand, maybe we have a character class */
2549 if (*s == ']' || *s == '^')
2552 /* this is terrifying, and it works */
2553 int weight = 2; /* let's weigh the evidence */
2555 unsigned char un_char = 255, last_un_char;
2556 const char * const send = strchr(s,']');
2557 char tmpbuf[sizeof PL_tokenbuf * 4];
2559 if (!send) /* has to be an expression */
2562 Zero(seen,256,char);
2565 else if (isDIGIT(*s)) {
2567 if (isDIGIT(s[1]) && s[2] == ']')
2573 for (; s < send; s++) {
2574 last_un_char = un_char;
2575 un_char = (unsigned char)*s;
2580 weight -= seen[un_char] * 10;
2581 if (isALNUM_lazy_if(s+1,UTF)) {
2583 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2584 len = (int)strlen(tmpbuf);
2585 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2590 else if (*s == '$' && s[1] &&
2591 strchr("[#!%*<>()-=",s[1])) {
2592 if (/*{*/ strchr("])} =",s[2]))
2601 if (strchr("wds]",s[1]))
2603 else if (seen[(U8)'\''] || seen[(U8)'"'])
2605 else if (strchr("rnftbxcav",s[1]))
2607 else if (isDIGIT(s[1])) {
2609 while (s[1] && isDIGIT(s[1]))
2619 if (strchr("aA01! ",last_un_char))
2621 if (strchr("zZ79~",s[1]))
2623 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2624 weight -= 5; /* cope with negative subscript */
2627 if (!isALNUM(last_un_char)
2628 && !(last_un_char == '$' || last_un_char == '@'
2629 || last_un_char == '&')
2630 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2635 if (keyword(tmpbuf, d - tmpbuf, 0))
2638 if (un_char == last_un_char + 1)
2640 weight -= seen[un_char];
2645 if (weight >= 0) /* probably a character class */
2655 * Does all the checking to disambiguate
2657 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2658 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2660 * First argument is the stuff after the first token, e.g. "bar".
2662 * Not a method if bar is a filehandle.
2663 * Not a method if foo is a subroutine prototyped to take a filehandle.
2664 * Not a method if it's really "Foo $bar"
2665 * Method if it's "foo $bar"
2666 * Not a method if it's really "print foo $bar"
2667 * Method if it's really "foo package::" (interpreted as package->foo)
2668 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2669 * Not a method if bar is a filehandle or package, but is quoted with
2674 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2677 char *s = start + (*start == '$');
2678 char tmpbuf[sizeof PL_tokenbuf];
2686 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2690 const char *proto = SvPVX_const(cv);
2701 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2702 /* start is the beginning of the possible filehandle/object,
2703 * and s is the end of it
2704 * tmpbuf is a copy of it
2707 if (*start == '$') {
2708 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
2709 isUPPER(*PL_tokenbuf))
2712 len = start - SvPVX(PL_linestr);
2716 start = SvPVX(PL_linestr) + len;
2720 return *s == '(' ? FUNCMETH : METHOD;
2722 if (!keyword(tmpbuf, len, 0)) {
2723 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2727 soff = s - SvPVX(PL_linestr);
2731 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2732 if (indirgv && GvCVu(indirgv))
2734 /* filehandle or package name makes it a method */
2735 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2737 soff = s - SvPVX(PL_linestr);
2740 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2741 return 0; /* no assumptions -- "=>" quotes bearword */
2743 start_force(PL_curforce);
2744 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2745 newSVpvn(tmpbuf,len));
2746 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2748 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2753 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2755 return *s == '(' ? FUNCMETH : METHOD;
2761 /* Encoded script support. filter_add() effectively inserts a
2762 * 'pre-processing' function into the current source input stream.
2763 * Note that the filter function only applies to the current source file
2764 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2766 * The datasv parameter (which may be NULL) can be used to pass
2767 * private data to this instance of the filter. The filter function
2768 * can recover the SV using the FILTER_DATA macro and use it to
2769 * store private buffers and state information.
2771 * The supplied datasv parameter is upgraded to a PVIO type
2772 * and the IoDIRP/IoANY field is used to store the function pointer,
2773 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2774 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2775 * private use must be set using malloc'd pointers.
2779 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2788 if (!PL_rsfp_filters)
2789 PL_rsfp_filters = newAV();
2792 SvUPGRADE(datasv, SVt_PVIO);
2793 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2794 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2795 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2796 FPTR2DPTR(void *, IoANY(datasv)),
2797 SvPV_nolen(datasv)));
2798 av_unshift(PL_rsfp_filters, 1);
2799 av_store(PL_rsfp_filters, 0, datasv) ;
2804 /* Delete most recently added instance of this filter function. */
2806 Perl_filter_del(pTHX_ filter_t funcp)
2812 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2813 FPTR2DPTR(void*, funcp)));
2815 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2817 /* if filter is on top of stack (usual case) just pop it off */
2818 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2819 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2820 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2821 IoANY(datasv) = (void *)NULL;
2822 sv_free(av_pop(PL_rsfp_filters));
2826 /* we need to search for the correct entry and clear it */
2827 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2831 /* Invoke the idxth filter function for the current rsfp. */
2832 /* maxlen 0 = read one text line */
2834 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2839 /* This API is bad. It should have been using unsigned int for maxlen.
2840 Not sure if we want to change the API, but if not we should sanity
2841 check the value here. */
2842 const unsigned int correct_length
2851 if (!PL_parser || !PL_rsfp_filters)
2853 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2854 /* Provide a default input filter to make life easy. */
2855 /* Note that we append to the line. This is handy. */
2856 DEBUG_P(PerlIO_printf(Perl_debug_log,
2857 "filter_read %d: from rsfp\n", idx));
2858 if (correct_length) {
2861 const int old_len = SvCUR(buf_sv);
2863 /* ensure buf_sv is large enough */
2864 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2865 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2866 correct_length)) <= 0) {
2867 if (PerlIO_error(PL_rsfp))
2868 return -1; /* error */
2870 return 0 ; /* end of file */
2872 SvCUR_set(buf_sv, old_len + len) ;
2875 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2876 if (PerlIO_error(PL_rsfp))
2877 return -1; /* error */
2879 return 0 ; /* end of file */
2882 return SvCUR(buf_sv);
2884 /* Skip this filter slot if filter has been deleted */
2885 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2886 DEBUG_P(PerlIO_printf(Perl_debug_log,
2887 "filter_read %d: skipped (filter deleted)\n",
2889 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2891 /* Get function pointer hidden within datasv */
2892 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2893 DEBUG_P(PerlIO_printf(Perl_debug_log,
2894 "filter_read %d: via function %p (%s)\n",
2895 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2896 /* Call function. The function is expected to */
2897 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2898 /* Return: <0:error, =0:eof, >0:not eof */
2899 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2903 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2906 #ifdef PERL_CR_FILTER
2907 if (!PL_rsfp_filters) {
2908 filter_add(S_cr_textfilter,NULL);
2911 if (PL_rsfp_filters) {
2913 SvCUR_set(sv, 0); /* start with empty line */
2914 if (FILTER_READ(0, sv, 0) > 0)
2915 return ( SvPVX(sv) ) ;
2920 return (sv_gets(sv, fp, append));
2924 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
2929 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2933 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2934 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2936 return GvHV(gv); /* Foo:: */
2939 /* use constant CLASS => 'MyClass' */
2940 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2941 if (gv && GvCV(gv)) {
2942 SV * const sv = cv_const_sv(GvCV(gv));
2944 pkgname = SvPV_const(sv, len);
2947 return gv_stashpvn(pkgname, len, 0);
2951 * S_readpipe_override
2952 * Check whether readpipe() is overriden, and generates the appropriate
2953 * optree, provided sublex_start() is called afterwards.
2956 S_readpipe_override(pTHX)
2959 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2960 yylval.ival = OP_BACKTICK;
2962 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2964 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2965 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
2966 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2968 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2969 append_elem(OP_LIST,
2970 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2971 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2978 * The intent of this yylex wrapper is to minimize the changes to the
2979 * tokener when we aren't interested in collecting madprops. It remains
2980 * to be seen how successful this strategy will be...
2987 char *s = PL_bufptr;
2989 /* make sure PL_thiswhite is initialized */
2993 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2994 if (PL_pending_ident)
2995 return S_pending_ident(aTHX);
2997 /* previous token ate up our whitespace? */
2998 if (!PL_lasttoke && PL_nextwhite) {
2999 PL_thiswhite = PL_nextwhite;
3003 /* isolate the token, and figure out where it is without whitespace */
3004 PL_realtokenstart = -1;
3008 assert(PL_curforce < 0);
3010 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3011 if (!PL_thistoken) {
3012 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3013 PL_thistoken = newSVpvs("");
3015 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3016 PL_thistoken = newSVpvn(tstart, s - tstart);
3019 if (PL_thismad) /* install head */
3020 CURMAD('X', PL_thistoken);
3023 /* last whitespace of a sublex? */
3024 if (optype == ')' && PL_endwhite) {
3025 CURMAD('X', PL_endwhite);
3030 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3031 if (!PL_thiswhite && !PL_endwhite && !optype) {
3032 sv_free(PL_thistoken);
3037 /* put off final whitespace till peg */
3038 if (optype == ';' && !PL_rsfp) {
3039 PL_nextwhite = PL_thiswhite;
3042 else if (PL_thisopen) {
3043 CURMAD('q', PL_thisopen);
3045 sv_free(PL_thistoken);
3049 /* Store actual token text as madprop X */
3050 CURMAD('X', PL_thistoken);
3054 /* add preceding whitespace as madprop _ */
3055 CURMAD('_', PL_thiswhite);
3059 /* add quoted material as madprop = */
3060 CURMAD('=', PL_thisstuff);
3064 /* add terminating quote as madprop Q */
3065 CURMAD('Q', PL_thisclose);
3069 /* special processing based on optype */
3073 /* opval doesn't need a TOKEN since it can already store mp */
3084 append_madprops(PL_thismad, yylval.opval, 0);
3092 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3101 /* remember any fake bracket that lexer is about to discard */
3102 if (PL_lex_brackets == 1 &&
3103 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3106 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3109 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3110 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3113 break; /* don't bother looking for trailing comment */
3122 /* attach a trailing comment to its statement instead of next token */
3126 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3128 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3130 if (*s == '\n' || *s == '#') {
3131 while (s < PL_bufend && *s != '\n')
3135 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3136 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3153 /* Create new token struct. Note: opvals return early above. */
3154 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3161 S_tokenize_use(pTHX_ int is_use, char *s) {
3163 if (PL_expect != XSTATE)
3164 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3165 is_use ? "use" : "no"));
3167 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3168 s = force_version(s, TRUE);
3169 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3170 start_force(PL_curforce);
3171 NEXTVAL_NEXTTOKE.opval = NULL;
3174 else if (*s == 'v') {
3175 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3176 s = force_version(s, FALSE);
3180 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3181 s = force_version(s, FALSE);
3183 yylval.ival = is_use;
3187 static const char* const exp_name[] =
3188 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3189 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3196 Works out what to call the token just pulled out of the input
3197 stream. The yacc parser takes care of taking the ops we return and
3198 stitching them into a tree.
3204 if read an identifier
3205 if we're in a my declaration
3206 croak if they tried to say my($foo::bar)
3207 build the ops for a my() declaration
3208 if it's an access to a my() variable
3209 are we in a sort block?
3210 croak if my($a); $a <=> $b
3211 build ops for access to a my() variable
3212 if in a dq string, and they've said @foo and we can't find @foo
3214 build ops for a bareword
3215 if we already built the token before, use it.
3220 #pragma segment Perl_yylex
3226 register char *s = PL_bufptr;
3231 /* orig_keyword, gvp, and gv are initialized here because
3232 * jump to the label just_a_word_zero can bypass their
3233 * initialization later. */
3234 I32 orig_keyword = 0;
3239 SV* tmp = newSVpvs("");
3240 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3241 (IV)CopLINE(PL_curcop),
3242 lex_state_names[PL_lex_state],
3243 exp_name[PL_expect],
3244 pv_display(tmp, s, strlen(s), 0, 60));
3247 /* check if there's an identifier for us to look at */
3248 if (PL_pending_ident)
3249 return REPORT(S_pending_ident(aTHX));
3251 /* no identifier pending identification */
3253 switch (PL_lex_state) {
3255 case LEX_NORMAL: /* Some compilers will produce faster */
3256 case LEX_INTERPNORMAL: /* code if we comment these out. */
3260 /* when we've already built the next token, just pull it out of the queue */
3264 yylval = PL_nexttoke[PL_lasttoke].next_val;
3266 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3267 PL_nexttoke[PL_lasttoke].next_mad = 0;
3268 if (PL_thismad && PL_thismad->mad_key == '_') {
3269 PL_thiswhite = (SV*)PL_thismad->mad_val;
3270 PL_thismad->mad_val = 0;
3271 mad_free(PL_thismad);
3276 PL_lex_state = PL_lex_defer;
3277 PL_expect = PL_lex_expect;
3278 PL_lex_defer = LEX_NORMAL;
3279 if (!PL_nexttoke[PL_lasttoke].next_type)
3284 yylval = PL_nextval[PL_nexttoke];
3286 PL_lex_state = PL_lex_defer;
3287 PL_expect = PL_lex_expect;
3288 PL_lex_defer = LEX_NORMAL;
3292 /* FIXME - can these be merged? */
3293 return(PL_nexttoke[PL_lasttoke].next_type);
3295 return REPORT(PL_nexttype[PL_nexttoke]);
3298 /* interpolated case modifiers like \L \U, including \Q and \E.
3299 when we get here, PL_bufptr is at the \
3301 case LEX_INTERPCASEMOD:
3303 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3304 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3306 /* handle \E or end of string */
3307 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3309 if (PL_lex_casemods) {
3310 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3311 PL_lex_casestack[PL_lex_casemods] = '\0';
3313 if (PL_bufptr != PL_bufend
3314 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3316 PL_lex_state = LEX_INTERPCONCAT;
3319 PL_thistoken = newSVpvs("\\E");
3325 while (PL_bufptr != PL_bufend &&
3326 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3328 PL_thiswhite = newSVpvs("");
3329 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3333 if (PL_bufptr != PL_bufend)
3336 PL_lex_state = LEX_INTERPCONCAT;
3340 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3341 "### Saw case modifier\n"); });
3343 if (s[1] == '\\' && s[2] == 'E') {
3346 PL_thiswhite = newSVpvs("");
3347 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3350 PL_lex_state = LEX_INTERPCONCAT;
3355 if (!PL_madskills) /* when just compiling don't need correct */
3356 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3357 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3358 if ((*s == 'L' || *s == 'U') &&
3359 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3360 PL_lex_casestack[--PL_lex_casemods] = '\0';
3363 if (PL_lex_casemods > 10)
3364 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3365 PL_lex_casestack[PL_lex_casemods++] = *s;
3366 PL_lex_casestack[PL_lex_casemods] = '\0';
3367 PL_lex_state = LEX_INTERPCONCAT;
3368 start_force(PL_curforce);
3369 NEXTVAL_NEXTTOKE.ival = 0;
3371 start_force(PL_curforce);
3373 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3375 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3377 NEXTVAL_NEXTTOKE.ival = OP_LC;
3379 NEXTVAL_NEXTTOKE.ival = OP_UC;
3381 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3383 Perl_croak(aTHX_ "panic: yylex");
3385 SV* const tmpsv = newSVpvs("\\ ");
3386 /* replace the space with the character we want to escape
3388 SvPVX(tmpsv)[1] = *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.$c)) */
3451 if (!PL_lex_casemods && PL_lex_inpat)
3458 case LEX_INTERPENDMAYBE:
3459 if (intuit_more(PL_bufptr)) {
3460 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3466 if (PL_lex_dojoin) {
3467 PL_lex_dojoin = FALSE;
3468 PL_lex_state = LEX_INTERPCONCAT;
3472 sv_free(PL_thistoken);
3473 PL_thistoken = newSVpvs("");
3478 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3479 && SvEVALED(PL_lex_repl))
3481 if (PL_bufptr != PL_bufend)
3482 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3486 case LEX_INTERPCONCAT:
3488 if (PL_lex_brackets)
3489 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3491 if (PL_bufptr == PL_bufend)
3492 return REPORT(sublex_done());
3494 if (SvIVX(PL_linestr) == '\'') {
3495 SV *sv = newSVsv(PL_linestr);
3498 else if ( PL_hints & HINT_NEW_RE )
3499 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
3500 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3504 s = scan_const(PL_bufptr);
3506 PL_lex_state = LEX_INTERPCASEMOD;
3508 PL_lex_state = LEX_INTERPSTART;
3511 if (s != PL_bufptr) {
3512 start_force(PL_curforce);
3514 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3516 NEXTVAL_NEXTTOKE = yylval;
3519 if (PL_lex_starts++) {
3523 sv_free(PL_thistoken);
3524 PL_thistoken = newSVpvs("");
3527 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3528 if (!PL_lex_casemods && PL_lex_inpat)
3541 PL_lex_state = LEX_NORMAL;
3542 s = scan_formline(PL_bufptr);
3543 if (!PL_lex_formbrack)
3549 PL_oldoldbufptr = PL_oldbufptr;
3555 sv_free(PL_thistoken);
3558 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3562 if (isIDFIRST_lazy_if(s,UTF))
3564 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3565 Perl_croak(aTHX_ "Unrecognized character \\x%02X in column %d", *s & 255, (int) len + 1);
3568 goto fake_eof; /* emulate EOF on ^D or ^Z */
3577 if (PL_lex_brackets) {
3578 yyerror((const char *)
3580 ? "Format not terminated"
3581 : "Missing right curly or square bracket"));
3583 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3584 "### Tokener got EOF\n");
3588 if (s++ < PL_bufend)
3589 goto retry; /* ignore stray nulls */
3592 if (!PL_in_eval && !PL_preambled) {
3593 PL_preambled = TRUE;
3599 /* Generate a string of Perl code to load the debugger.
3600 * If PERL5DB is set, it will return the contents of that,
3601 * otherwise a compile-time require of perl5db.pl. */
3603 const char * const pdb = PerlEnv_getenv("PERL5DB");
3606 sv_setpv(PL_linestr, pdb);
3607 sv_catpvs(PL_linestr,";");
3609 SETERRNO(0,SS_NORMAL);
3610 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
3613 sv_setpvs(PL_linestr,"");
3615 while(AvFILLp(PL_preambleav) >= 0) {
3616 SV *tmpsv = av_shift(PL_preambleav);
3617 sv_catsv(PL_linestr, tmpsv);
3618 sv_catpvs(PL_linestr, ";");
3621 sv_free((SV*)PL_preambleav);
3622 PL_preambleav = NULL;
3624 if (PL_minus_n || PL_minus_p) {
3625 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3627 sv_catpvs(PL_linestr,"chomp;");
3630 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3631 || *PL_splitstr == '"')
3632 && strchr(PL_splitstr + 1, *PL_splitstr))
3633 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3635 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3636 bytes can be used as quoting characters. :-) */
3637 const char *splits = PL_splitstr;
3638 sv_catpvs(PL_linestr, "our @F=split(q\0");
3641 if (*splits == '\\')
3642 sv_catpvn(PL_linestr, splits, 1);
3643 sv_catpvn(PL_linestr, splits, 1);
3644 } while (*splits++);
3645 /* This loop will embed the trailing NUL of
3646 PL_linestr as the last thing it does before
3648 sv_catpvs(PL_linestr, ");");
3652 sv_catpvs(PL_linestr,"our @F=split(' ');");
3656 sv_catpvs(PL_linestr,"use feature ':5.10';");
3657 sv_catpvs(PL_linestr, "\n");
3658 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3659 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3660 PL_last_lop = PL_last_uni = NULL;
3661 if (PERLDB_LINE && PL_curstash != PL_debstash)
3662 update_debugger_info(PL_linestr, NULL, 0);
3666 bof = PL_rsfp ? TRUE : FALSE;
3667 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3670 PL_realtokenstart = -1;
3673 if (PL_preprocess && !PL_in_eval)
3674 (void)PerlProc_pclose(PL_rsfp);
3675 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3676 PerlIO_clearerr(PL_rsfp);
3678 (void)PerlIO_close(PL_rsfp);
3680 PL_doextract = FALSE;
3682 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3688 sv_setpvs(PL_linestr, ";}continue{print;}");
3690 sv_setpvs(PL_linestr, ";}");
3691 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3692 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3693 PL_last_lop = PL_last_uni = NULL;
3694 PL_minus_n = PL_minus_p = 0;
3697 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3698 PL_last_lop = PL_last_uni = NULL;
3699 sv_setpvn(PL_linestr,"",0);
3700 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3702 /* If it looks like the start of a BOM or raw UTF-16,
3703 * check if it in fact is. */
3709 #ifdef PERLIO_IS_STDIO
3710 # ifdef __GNU_LIBRARY__
3711 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3712 # define FTELL_FOR_PIPE_IS_BROKEN
3716 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3717 # define FTELL_FOR_PIPE_IS_BROKEN
3722 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3723 /* This loses the possibility to detect the bof
3724 * situation on perl -P when the libc5 is being used.
3725 * Workaround? Maybe attach some extra state to PL_rsfp?
3728 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3730 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3733 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3734 s = swallow_bom((U8*)s);
3738 /* Incest with pod. */
3741 sv_catsv(PL_thiswhite, PL_linestr);
3743 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3744 sv_setpvn(PL_linestr, "", 0);
3745 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3746 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3747 PL_last_lop = PL_last_uni = NULL;
3748 PL_doextract = FALSE;
3752 } while (PL_doextract);
3753 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3754 if (PERLDB_LINE && PL_curstash != PL_debstash)
3755 update_debugger_info(PL_linestr, NULL, 0);
3756 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3757 PL_last_lop = PL_last_uni = NULL;
3758 if (CopLINE(PL_curcop) == 1) {
3759 while (s < PL_bufend && isSPACE(*s))
3761 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3765 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3769 if (*s == '#' && *(s+1) == '!')
3771 #ifdef ALTERNATE_SHEBANG
3773 static char const as[] = ALTERNATE_SHEBANG;
3774 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3775 d = s + (sizeof(as) - 1);
3777 #endif /* ALTERNATE_SHEBANG */
3786 while (*d && !isSPACE(*d))
3790 #ifdef ARG_ZERO_IS_SCRIPT
3791 if (ipathend > ipath) {
3793 * HP-UX (at least) sets argv[0] to the script name,
3794 * which makes $^X incorrect. And Digital UNIX and Linux,
3795 * at least, set argv[0] to the basename of the Perl
3796 * interpreter. So, having found "#!", we'll set it right.
3798 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3800 assert(SvPOK(x) || SvGMAGICAL(x));
3801 if (sv_eq(x, CopFILESV(PL_curcop))) {
3802 sv_setpvn(x, ipath, ipathend - ipath);
3808 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3809 const char * const lstart = SvPV_const(x,llen);
3811 bstart += blen - llen;
3812 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3813 sv_setpvn(x, ipath, ipathend - ipath);
3818 TAINT_NOT; /* $^X is always tainted, but that's OK */
3820 #endif /* ARG_ZERO_IS_SCRIPT */
3825 d = instr(s,"perl -");
3827 d = instr(s,"perl");
3829 /* avoid getting into infinite loops when shebang
3830 * line contains "Perl" rather than "perl" */
3832 for (d = ipathend-4; d >= ipath; --d) {
3833 if ((*d == 'p' || *d == 'P')
3834 && !ibcmp(d, "perl", 4))
3844 #ifdef ALTERNATE_SHEBANG
3846 * If the ALTERNATE_SHEBANG on this system starts with a
3847 * character that can be part of a Perl expression, then if
3848 * we see it but not "perl", we're probably looking at the
3849 * start of Perl code, not a request to hand off to some
3850 * other interpreter. Similarly, if "perl" is there, but
3851 * not in the first 'word' of the line, we assume the line
3852 * contains the start of the Perl program.
3854 if (d && *s != '#') {
3855 const char *c = ipath;
3856 while (*c && !strchr("; \t\r\n\f\v#", *c))
3859 d = NULL; /* "perl" not in first word; ignore */
3861 *s = '#'; /* Don't try to parse shebang line */
3863 #endif /* ALTERNATE_SHEBANG */
3864 #ifndef MACOS_TRADITIONAL
3869 !instr(s,"indir") &&
3870 instr(PL_origargv[0],"perl"))
3877 while (s < PL_bufend && isSPACE(*s))
3879 if (s < PL_bufend) {
3880 Newxz(newargv,PL_origargc+3,char*);
3882 while (s < PL_bufend && !isSPACE(*s))
3885 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3888 newargv = PL_origargv;
3891 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3893 Perl_croak(aTHX_ "Can't exec %s", ipath);
3897 while (*d && !isSPACE(*d))
3899 while (SPACE_OR_TAB(*d))
3903 const bool switches_done = PL_doswitches;
3904 const U32 oldpdb = PL_perldb;
3905 const bool oldn = PL_minus_n;
3906 const bool oldp = PL_minus_p;
3910 if (*d1 == 'M' || *d1 == 'm' || *d1 == 'C') {
3911 const char * const m = d1;
3912 while (*d1 && !isSPACE(*d1))
3914 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3917 d1 = moreswitches(d1);
3919 if (PL_doswitches && !switches_done) {
3920 int argc = PL_origargc;
3921 char **argv = PL_origargv;
3924 } while (argc && argv[0][0] == '-' && argv[0][1]);
3925 init_argv_symbols(argc,argv);
3927 if ((PERLDB_LINE && !oldpdb) ||
3928 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3929 /* if we have already added "LINE: while (<>) {",
3930 we must not do it again */
3932 sv_setpvn(PL_linestr, "", 0);
3933 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3934 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3935 PL_last_lop = PL_last_uni = NULL;
3936 PL_preambled = FALSE;
3938 (void)gv_fetchfile(PL_origfilename);
3945 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3947 PL_lex_state = LEX_FORMLINE;
3952 #ifdef PERL_STRICT_CR
3953 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3955 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3957 case ' ': case '\t': case '\f': case 013:
3958 #ifdef MACOS_TRADITIONAL
3962 PL_realtokenstart = -1;
3964 PL_thiswhite = newSVpvs("");
3965 sv_catpvn(PL_thiswhite, s, 1);
3972 PL_realtokenstart = -1;
3976 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3977 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3978 /* handle eval qq[#line 1 "foo"\n ...] */
3979 CopLINE_dec(PL_curcop);
3982 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3984 if (!PL_in_eval || PL_rsfp)
3989 while (d < PL_bufend && *d != '\n')
3993 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3994 Perl_croak(aTHX_ "panic: input overflow");
3997 PL_thiswhite = newSVpvn(s, d - s);
4002 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4004 PL_lex_state = LEX_FORMLINE;
4010 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4011 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4014 TOKEN(PEG); /* make sure any #! line is accessible */
4019 /* if (PL_madskills && PL_lex_formbrack) { */
4021 while (d < PL_bufend && *d != '\n')
4025 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4026 Perl_croak(aTHX_ "panic: input overflow");
4027 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4029 PL_thiswhite = newSVpvs("");
4030 if (CopLINE(PL_curcop) == 1) {
4031 sv_setpvn(PL_thiswhite, "", 0);
4034 sv_catpvn(PL_thiswhite, s, d - s);
4048 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4056 while (s < PL_bufend && SPACE_OR_TAB(*s))
4059 if (strnEQ(s,"=>",2)) {
4060 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4061 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4062 OPERATOR('-'); /* unary minus */
4064 PL_last_uni = PL_oldbufptr;
4066 case 'r': ftst = OP_FTEREAD; break;
4067 case 'w': ftst = OP_FTEWRITE; break;
4068 case 'x': ftst = OP_FTEEXEC; break;
4069 case 'o': ftst = OP_FTEOWNED; break;
4070 case 'R': ftst = OP_FTRREAD; break;
4071 case 'W': ftst = OP_FTRWRITE; break;
4072 case 'X': ftst = OP_FTREXEC; break;
4073 case 'O': ftst = OP_FTROWNED; break;
4074 case 'e': ftst = OP_FTIS; break;
4075 case 'z': ftst = OP_FTZERO; break;
4076 case 's': ftst = OP_FTSIZE; break;
4077 case 'f': ftst = OP_FTFILE; break;
4078 case 'd': ftst = OP_FTDIR; break;
4079 case 'l': ftst = OP_FTLINK; break;
4080 case 'p': ftst = OP_FTPIPE; break;
4081 case 'S': ftst = OP_FTSOCK; break;
4082 case 'u': ftst = OP_FTSUID; break;
4083 case 'g': ftst = OP_FTSGID; break;
4084 case 'k': ftst = OP_FTSVTX; break;
4085 case 'b': ftst = OP_FTBLK; break;
4086 case 'c': ftst = OP_FTCHR; break;
4087 case 't': ftst = OP_FTTTY; break;
4088 case 'T': ftst = OP_FTTEXT; break;
4089 case 'B': ftst = OP_FTBINARY; break;
4090 case 'M': case 'A': case 'C':
4091 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4093 case 'M': ftst = OP_FTMTIME; break;
4094 case 'A': ftst = OP_FTATIME; break;
4095 case 'C': ftst = OP_FTCTIME; break;
4103 PL_last_lop_op = (OPCODE)ftst;
4104 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4105 "### Saw file test %c\n", (int)tmp);
4110 /* Assume it was a minus followed by a one-letter named
4111 * subroutine call (or a -bareword), then. */
4112 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4113 "### '-%c' looked like a file test but was not\n",
4120 const char tmp = *s++;
4123 if (PL_expect == XOPERATOR)
4128 else if (*s == '>') {
4131 if (isIDFIRST_lazy_if(s,UTF)) {
4132 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4140 if (PL_expect == XOPERATOR)
4143 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4145 OPERATOR('-'); /* unary minus */
4151 const char tmp = *s++;
4154 if (PL_expect == XOPERATOR)
4159 if (PL_expect == XOPERATOR)
4162 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4169 if (PL_expect != XOPERATOR) {
4170 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4171 PL_expect = XOPERATOR;
4172 force_ident(PL_tokenbuf, '*');
4185 if (PL_expect == XOPERATOR) {
4189 PL_tokenbuf[0] = '%';
4190 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4191 sizeof PL_tokenbuf - 1, FALSE);
4192 if (!PL_tokenbuf[1]) {
4195 PL_pending_ident = '%';
4206 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4213 const char tmp = *s++;
4219 goto just_a_word_zero_gv;
4222 switch (PL_expect) {
4228 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4230 PL_bufptr = s; /* update in case we back off */
4236 PL_expect = XTERMBLOCK;
4239 stuffstart = s - SvPVX(PL_linestr) - 1;
4243 while (isIDFIRST_lazy_if(s,UTF)) {
4246 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4247 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4248 if (tmp < 0) tmp = -tmp;
4262 sv = newSVpvn(s, len);
4264 d = scan_str(d,TRUE,TRUE);
4266 /* MUST advance bufptr here to avoid bogus
4267 "at end of line" context messages from yyerror().
4269 PL_bufptr = s + len;
4270 yyerror("Unterminated attribute parameter in attribute list");
4274 return REPORT(0); /* EOF indicator */
4278 sv_catsv(sv, PL_lex_stuff);
4279 attrs = append_elem(OP_LIST, attrs,
4280 newSVOP(OP_CONST, 0, sv));
4281 SvREFCNT_dec(PL_lex_stuff);
4282 PL_lex_stuff = NULL;
4285 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4287 if (PL_in_my == KEY_our) {
4289 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4291 /* skip to avoid loading attributes.pm */
4293 deprecate(":unique");
4296 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4299 /* NOTE: any CV attrs applied here need to be part of
4300 the CVf_BUILTIN_ATTRS define in cv.h! */
4301 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4303 CvLVALUE_on(PL_compcv);
4305 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4307 CvLOCKED_on(PL_compcv);
4309 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4311 CvMETHOD_on(PL_compcv);
4313 /* After we've set the flags, it could be argued that
4314 we don't need to do the attributes.pm-based setting
4315 process, and shouldn't bother appending recognized
4316 flags. To experiment with that, uncomment the
4317 following "else". (Note that's already been
4318 uncommented. That keeps the above-applied built-in
4319 attributes from being intercepted (and possibly
4320 rejected) by a package's attribute routines, but is
4321 justified by the performance win for the common case
4322 of applying only built-in attributes.) */
4324 attrs = append_elem(OP_LIST, attrs,
4325 newSVOP(OP_CONST, 0,
4329 if (*s == ':' && s[1] != ':')
4332 break; /* require real whitespace or :'s */
4333 /* XXX losing whitespace on sequential attributes here */
4337 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4338 if (*s != ';' && *s != '}' && *s != tmp
4339 && (tmp != '=' || *s != ')')) {
4340 const char q = ((*s == '\'') ? '"' : '\'');
4341 /* If here for an expression, and parsed no attrs, back
4343 if (tmp == '=' && !attrs) {
4347 /* MUST advance bufptr here to avoid bogus "at end of line"
4348 context messages from yyerror().
4351 yyerror( (const char *)
4353 ? Perl_form(aTHX_ "Invalid separator character "
4354 "%c%c%c in attribute list", q, *s, q)
4355 : "Unterminated attribute list" ) );
4363 start_force(PL_curforce);
4364 NEXTVAL_NEXTTOKE.opval = attrs;
4365 CURMAD('_', PL_nextwhite);
4370 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4371 (s - SvPVX(PL_linestr)) - stuffstart);
4379 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4380 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4388 const char tmp = *s++;
4393 const char tmp = *s++;
4401 if (PL_lex_brackets <= 0)
4402 yyerror("Unmatched right square bracket");
4405 if (PL_lex_state == LEX_INTERPNORMAL) {
4406 if (PL_lex_brackets == 0) {
4407 if (*s == '-' && s[1] == '>')
4408 PL_lex_state = LEX_INTERPENDMAYBE;
4409 else if (*s != '[' && *s != '{')
4410 PL_lex_state = LEX_INTERPEND;
4417 if (PL_lex_brackets > 100) {
4418 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4420 switch (PL_expect) {
4422 if (PL_lex_formbrack) {
4426 if (PL_oldoldbufptr == PL_last_lop)
4427 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4429 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4430 OPERATOR(HASHBRACK);
4432 while (s < PL_bufend && SPACE_OR_TAB(*s))
4435 PL_tokenbuf[0] = '\0';
4436 if (d < PL_bufend && *d == '-') {
4437 PL_tokenbuf[0] = '-';
4439 while (d < PL_bufend && SPACE_OR_TAB(*d))
4442 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4443 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4445 while (d < PL_bufend && SPACE_OR_TAB(*d))
4448 const char minus = (PL_tokenbuf[0] == '-');
4449 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4457 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4462 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4467 if (PL_oldoldbufptr == PL_last_lop)
4468 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4470 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4473 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4475 /* This hack is to get the ${} in the message. */
4477 yyerror("syntax error");
4480 OPERATOR(HASHBRACK);
4482 /* This hack serves to disambiguate a pair of curlies
4483 * as being a block or an anon hash. Normally, expectation
4484 * determines that, but in cases where we're not in a
4485 * position to expect anything in particular (like inside
4486 * eval"") we have to resolve the ambiguity. This code
4487 * covers the case where the first term in the curlies is a
4488 * quoted string. Most other cases need to be explicitly
4489 * disambiguated by prepending a "+" before the opening
4490 * curly in order to force resolution as an anon hash.
4492 * XXX should probably propagate the outer expectation
4493 * into eval"" to rely less on this hack, but that could
4494 * potentially break current behavior of eval"".
4498 if (*s == '\'' || *s == '"' || *s == '`') {
4499 /* common case: get past first string, handling escapes */
4500 for (t++; t < PL_bufend && *t != *s;)
4501 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4505 else if (*s == 'q') {
4508 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4511 /* skip q//-like construct */
4513 char open, close, term;
4516 while (t < PL_bufend && isSPACE(*t))
4518 /* check for q => */
4519 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4520 OPERATOR(HASHBRACK);
4524 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4528 for (t++; t < PL_bufend; t++) {
4529 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4531 else if (*t == open)
4535 for (t++; t < PL_bufend; t++) {
4536 if (*t == '\\' && t+1 < PL_bufend)
4538 else if (*t == close && --brackets <= 0)
4540 else if (*t == open)
4547 /* skip plain q word */
4548 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4551 else if (isALNUM_lazy_if(t,UTF)) {
4553 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4556 while (t < PL_bufend && isSPACE(*t))
4558 /* if comma follows first term, call it an anon hash */
4559 /* XXX it could be a comma expression with loop modifiers */
4560 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4561 || (*t == '=' && t[1] == '>')))
4562 OPERATOR(HASHBRACK);
4563 if (PL_expect == XREF)
4566 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4572 yylval.ival = CopLINE(PL_curcop);
4573 if (isSPACE(*s) || *s == '#')
4574 PL_copline = NOLINE; /* invalidate current command line number */
4579 if (PL_lex_brackets <= 0)
4580 yyerror("Unmatched right curly bracket");
4582 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4583 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4584 PL_lex_formbrack = 0;
4585 if (PL_lex_state == LEX_INTERPNORMAL) {
4586 if (PL_lex_brackets == 0) {
4587 if (PL_expect & XFAKEBRACK) {
4588 PL_expect &= XENUMMASK;
4589 PL_lex_state = LEX_INTERPEND;
4594 PL_thiswhite = newSVpvs("");
4595 sv_catpvn(PL_thiswhite,"}",1);
4598 return yylex(); /* ignore fake brackets */
4600 if (*s == '-' && s[1] == '>')
4601 PL_lex_state = LEX_INTERPENDMAYBE;
4602 else if (*s != '[' && *s != '{')
4603 PL_lex_state = LEX_INTERPEND;
4606 if (PL_expect & XFAKEBRACK) {
4607 PL_expect &= XENUMMASK;
4609 return yylex(); /* ignore fake brackets */
4611 start_force(PL_curforce);
4613 curmad('X', newSVpvn(s-1,1));
4614 CURMAD('_', PL_thiswhite);
4619 PL_thistoken = newSVpvs("");
4627 if (PL_expect == XOPERATOR) {
4628 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4629 && isIDFIRST_lazy_if(s,UTF))
4631 CopLINE_dec(PL_curcop);
4632 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4633 CopLINE_inc(PL_curcop);
4638 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4640 PL_expect = XOPERATOR;
4641 force_ident(PL_tokenbuf, '&');
4645 yylval.ival = (OPpENTERSUB_AMPER<<8);
4657 const char tmp = *s++;
4664 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4665 && strchr("+-*/%.^&|<",tmp))
4666 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4667 "Reversed %c= operator",(int)tmp);
4669 if (PL_expect == XSTATE && isALPHA(tmp) &&
4670 (s == PL_linestart+1 || s[-2] == '\n') )
4672 if (PL_in_eval && !PL_rsfp) {
4677 if (strnEQ(s,"=cut",4)) {
4693 PL_thiswhite = newSVpvs("");
4694 sv_catpvn(PL_thiswhite, PL_linestart,
4695 PL_bufend - PL_linestart);
4699 PL_doextract = TRUE;
4703 if (PL_lex_brackets < PL_lex_formbrack) {
4705 #ifdef PERL_STRICT_CR
4706 while (SPACE_OR_TAB(*t))
4708 while (SPACE_OR_TAB(*t) || *t == '\r')
4711 if (*t == '\n' || *t == '#') {
4722 const char tmp = *s++;
4724 /* was this !=~ where !~ was meant?
4725 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4727 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4728 const char *t = s+1;
4730 while (t < PL_bufend && isSPACE(*t))
4733 if (*t == '/' || *t == '?' ||
4734 ((*t == 'm' || *t == 's' || *t == 'y')
4735 && !isALNUM(t[1])) ||
4736 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4737 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4738 "!=~ should be !~");
4748 if (PL_expect != XOPERATOR) {
4749 if (s[1] != '<' && !strchr(s,'>'))
4752 s = scan_heredoc(s);
4754 s = scan_inputsymbol(s);
4755 TERM(sublex_start());
4761 SHop(OP_LEFT_SHIFT);
4775 const char tmp = *s++;
4777 SHop(OP_RIGHT_SHIFT);
4778 else if (tmp == '=')
4787 if (PL_expect == XOPERATOR) {
4788 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4790 deprecate_old(commaless_variable_list);
4791 return REPORT(','); /* grandfather non-comma-format format */
4795 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4796 PL_tokenbuf[0] = '@';
4797 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4798 sizeof PL_tokenbuf - 1, FALSE);
4799 if (PL_expect == XOPERATOR)
4800 no_op("Array length", s);
4801 if (!PL_tokenbuf[1])
4803 PL_expect = XOPERATOR;
4804 PL_pending_ident = '#';
4808 PL_tokenbuf[0] = '$';
4809 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4810 sizeof PL_tokenbuf - 1, FALSE);
4811 if (PL_expect == XOPERATOR)
4813 if (!PL_tokenbuf[1]) {
4815 yyerror("Final $ should be \\$ or $name");
4819 /* This kludge not intended to be bulletproof. */
4820 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4821 yylval.opval = newSVOP(OP_CONST, 0,
4822 newSViv(CopARYBASE_get(&PL_compiling)));
4823 yylval.opval->op_private = OPpCONST_ARYBASE;
4829 const char tmp = *s;
4830 if (PL_lex_state == LEX_NORMAL)
4833 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4834 && intuit_more(s)) {
4836 PL_tokenbuf[0] = '@';
4837 if (ckWARN(WARN_SYNTAX)) {
4840 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4843 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4844 while (t < PL_bufend && *t != ']')
4846 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4847 "Multidimensional syntax %.*s not supported",
4848 (int)((t - PL_bufptr) + 1), PL_bufptr);
4852 else if (*s == '{') {
4854 PL_tokenbuf[0] = '%';
4855 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4856 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4858 char tmpbuf[sizeof PL_tokenbuf];
4861 } while (isSPACE(*t));
4862 if (isIDFIRST_lazy_if(t,UTF)) {
4864 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4868 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4869 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4870 "You need to quote \"%s\"",
4877 PL_expect = XOPERATOR;
4878 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4879 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4880 if (!islop || PL_last_lop_op == OP_GREPSTART)
4881 PL_expect = XOPERATOR;
4882 else if (strchr("$@\"'`q", *s))
4883 PL_expect = XTERM; /* e.g. print $fh "foo" */
4884 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4885 PL_expect = XTERM; /* e.g. print $fh &sub */
4886 else if (isIDFIRST_lazy_if(s,UTF)) {
4887 char tmpbuf[sizeof PL_tokenbuf];
4889 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4890 if ((t2 = keyword(tmpbuf, len, 0))) {
4891 /* binary operators exclude handle interpretations */
4903 PL_expect = XTERM; /* e.g. print $fh length() */
4908 PL_expect = XTERM; /* e.g. print $fh subr() */
4911 else if (isDIGIT(*s))
4912 PL_expect = XTERM; /* e.g. print $fh 3 */
4913 else if (*s == '.' && isDIGIT(s[1]))
4914 PL_expect = XTERM; /* e.g. print $fh .3 */
4915 else if ((*s == '?' || *s == '-' || *s == '+')
4916 && !isSPACE(s[1]) && s[1] != '=')
4917 PL_expect = XTERM; /* e.g. print $fh -1 */
4918 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4920 PL_expect = XTERM; /* e.g. print $fh /.../
4921 XXX except DORDOR operator
4923 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4925 PL_expect = XTERM; /* print $fh <<"EOF" */
4928 PL_pending_ident = '$';
4932 if (PL_expect == XOPERATOR)
4934 PL_tokenbuf[0] = '@';
4935 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4936 if (!PL_tokenbuf[1]) {
4939 if (PL_lex_state == LEX_NORMAL)
4941 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4943 PL_tokenbuf[0] = '%';
4945 /* Warn about @ where they meant $. */
4946 if (*s == '[' || *s == '{') {
4947 if (ckWARN(WARN_SYNTAX)) {
4948 const char *t = s + 1;
4949 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4951 if (*t == '}' || *t == ']') {
4953 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4954 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4955 "Scalar value %.*s better written as $%.*s",
4956 (int)(t-PL_bufptr), PL_bufptr,
4957 (int)(t-PL_bufptr-1), PL_bufptr+1);
4962 PL_pending_ident = '@';
4965 case '/': /* may be division, defined-or, or pattern */
4966 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4970 case '?': /* may either be conditional or pattern */
4971 if(PL_expect == XOPERATOR) {
4979 /* A // operator. */
4989 /* Disable warning on "study /blah/" */
4990 if (PL_oldoldbufptr == PL_last_uni
4991 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4992 || memNE(PL_last_uni, "study", 5)
4993 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4996 s = scan_pat(s,OP_MATCH);
4997 TERM(sublex_start());
5001 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5002 #ifdef PERL_STRICT_CR
5005 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5007 && (s == PL_linestart || s[-1] == '\n') )
5009 PL_lex_formbrack = 0;
5013 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5019 yylval.ival = OPf_SPECIAL;
5025 if (PL_expect != XOPERATOR)
5030 case '0': case '1': case '2': case '3': case '4':
5031 case '5': case '6': case '7': case '8': case '9':
5032 s = scan_num(s, &yylval);
5033 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5034 if (PL_expect == XOPERATOR)
5039 s = scan_str(s,!!PL_madskills,FALSE);
5040 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5041 if (PL_expect == XOPERATOR) {
5042 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5044 deprecate_old(commaless_variable_list);
5045 return REPORT(','); /* grandfather non-comma-format format */
5052 yylval.ival = OP_CONST;
5053 TERM(sublex_start());
5056 s = scan_str(s,!!PL_madskills,FALSE);
5057 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5058 if (PL_expect == XOPERATOR) {
5059 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5061 deprecate_old(commaless_variable_list);
5062 return REPORT(','); /* grandfather non-comma-format format */
5069 yylval.ival = OP_CONST;
5070 /* FIXME. I think that this can be const if char *d is replaced by
5071 more localised variables. */
5072 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5073 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5074 yylval.ival = OP_STRINGIFY;
5078 TERM(sublex_start());
5081 s = scan_str(s,!!PL_madskills,FALSE);
5082 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5083 if (PL_expect == XOPERATOR)
5084 no_op("Backticks",s);
5087 readpipe_override();
5088 TERM(sublex_start());
5092 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5093 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5095 if (PL_expect == XOPERATOR)
5096 no_op("Backslash",s);
5100 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5101 char *start = s + 2;
5102 while (isDIGIT(*start) || *start == '_')
5104 if (*start == '.' && isDIGIT(start[1])) {
5105 s = scan_num(s, &yylval);
5108 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5109 else if (!isALPHA(*start) && (PL_expect == XTERM
5110 || PL_expect == XREF || PL_expect == XSTATE
5111 || PL_expect == XTERMORDORDOR)) {
5112 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5114 s = scan_num(s, &yylval);
5121 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5163 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5165 /* Some keywords can be followed by any delimiter, including ':' */
5166 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5167 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5168 (PL_tokenbuf[0] == 'q' &&
5169 strchr("qwxr", PL_tokenbuf[1])))));
5171 /* x::* is just a word, unless x is "CORE" */
5172 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5176 while (d < PL_bufend && isSPACE(*d))
5177 d++; /* no comments skipped here, or s### is misparsed */
5179 /* Is this a label? */
5180 if (!tmp && PL_expect == XSTATE
5181 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5183 yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5188 /* Check for keywords */
5189 tmp = keyword(PL_tokenbuf, len, 0);
5191 /* Is this a word before a => operator? */
5192 if (*d == '=' && d[1] == '>') {
5195 = (OP*)newSVOP(OP_CONST, 0,
5196 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5197 yylval.opval->op_private = OPpCONST_BARE;
5201 if (tmp < 0) { /* second-class keyword? */
5202 GV *ogv = NULL; /* override (winner) */
5203 GV *hgv = NULL; /* hidden (loser) */
5204 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5206 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5209 if (GvIMPORTED_CV(gv))
5211 else if (! CvMETHOD(cv))
5215 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5216 (gv = *gvp) && isGV_with_GP(gv) &&
5217 GvCVu(gv) && GvIMPORTED_CV(gv))
5224 tmp = 0; /* overridden by import or by GLOBAL */
5227 && -tmp==KEY_lock /* XXX generalizable kludge */
5230 tmp = 0; /* any sub overrides "weak" keyword */
5232 else { /* no override */
5234 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5235 Perl_warner(aTHX_ packWARN(WARN_MISC),
5236 "dump() better written as CORE::dump()");
5240 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5241 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5242 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5243 "Ambiguous call resolved as CORE::%s(), %s",
5244 GvENAME(hgv), "qualify as such or use &");
5251 default: /* not a keyword */
5252 /* Trade off - by using this evil construction we can pull the
5253 variable gv into the block labelled keylookup. If not, then
5254 we have to give it function scope so that the goto from the
5255 earlier ':' case doesn't bypass the initialisation. */
5257 just_a_word_zero_gv:
5265 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5268 SV *nextPL_nextwhite = 0;
5272 /* Get the rest if it looks like a package qualifier */
5274 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5276 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5279 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5280 *s == '\'' ? "'" : "::");
5285 if (PL_expect == XOPERATOR) {
5286 if (PL_bufptr == PL_linestart) {
5287 CopLINE_dec(PL_curcop);
5288 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5289 CopLINE_inc(PL_curcop);
5292 no_op("Bareword",s);
5295 /* Look for a subroutine with this name in current package,
5296 unless name is "Foo::", in which case Foo is a bearword
5297 (and a package name). */
5299 if (len > 2 && !PL_madskills &&
5300 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5302 if (ckWARN(WARN_BAREWORD)
5303 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5304 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5305 "Bareword \"%s\" refers to nonexistent package",
5308 PL_tokenbuf[len] = '\0';
5314 /* Mustn't actually add anything to a symbol table.
5315 But also don't want to "initialise" any placeholder
5316 constants that might already be there into full
5317 blown PVGVs with attached PVCV. */
5318 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5319 GV_NOADD_NOINIT, SVt_PVCV);
5324 /* if we saw a global override before, get the right name */
5327 sv = newSVpvs("CORE::GLOBAL::");
5328 sv_catpv(sv,PL_tokenbuf);
5331 /* If len is 0, newSVpv does strlen(), which is correct.
5332 If len is non-zero, then it will be the true length,
5333 and so the scalar will be created correctly. */
5334 sv = newSVpv(PL_tokenbuf,len);
5337 if (PL_madskills && !PL_thistoken) {
5338 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5339 PL_thistoken = newSVpvn(start,s - start);
5340 PL_realtokenstart = s - SvPVX(PL_linestr);
5344 /* Presume this is going to be a bareword of some sort. */
5347 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5348 yylval.opval->op_private = OPpCONST_BARE;
5349 /* UTF-8 package name? */
5350 if (UTF && !IN_BYTES &&
5351 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5354 /* And if "Foo::", then that's what it certainly is. */
5359 /* Do the explicit type check so that we don't need to force
5360 the initialisation of the symbol table to have a real GV.
5361 Beware - gv may not really be a PVGV, cv may not really be
5362 a PVCV, (because of the space optimisations that gv_init
5363 understands) But they're true if for this symbol there is
5364 respectively a typeglob and a subroutine.
5366 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5367 /* Real typeglob, so get the real subroutine: */
5369 /* A proxy for a subroutine in this package? */
5370 : SvOK(gv) ? (CV *) gv : NULL)
5373 /* See if it's the indirect object for a list operator. */
5375 if (PL_oldoldbufptr &&
5376 PL_oldoldbufptr < PL_bufptr &&
5377 (PL_oldoldbufptr == PL_last_lop
5378 || PL_oldoldbufptr == PL_last_uni) &&
5379 /* NO SKIPSPACE BEFORE HERE! */
5380 (PL_expect == XREF ||
5381 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5383 bool immediate_paren = *s == '(';
5385 /* (Now we can afford to cross potential line boundary.) */
5386 s = SKIPSPACE2(s,nextPL_nextwhite);
5388 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5391 /* Two barewords in a row may indicate method call. */
5393 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5394 (tmp = intuit_method(s, gv, cv)))
5397 /* If not a declared subroutine, it's an indirect object. */
5398 /* (But it's an indir obj regardless for sort.) */
5399 /* Also, if "_" follows a filetest operator, it's a bareword */
5402 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5404 (PL_last_lop_op != OP_MAPSTART &&
5405 PL_last_lop_op != OP_GREPSTART))))
5406 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5407 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5410 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5415 PL_expect = XOPERATOR;
5418 s = SKIPSPACE2(s,nextPL_nextwhite);
5419 PL_nextwhite = nextPL_nextwhite;
5424 /* Is this a word before a => operator? */
5425 if (*s == '=' && s[1] == '>' && !pkgname) {
5427 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5428 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5429 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5433 /* If followed by a paren, it's certainly a subroutine. */
5438 while (SPACE_OR_TAB(*d))
5440 if (*d == ')' && (sv = gv_const_sv(gv))) {
5447 PL_nextwhite = PL_thiswhite;
5450 start_force(PL_curforce);
5452 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5453 PL_expect = XOPERATOR;
5456 PL_nextwhite = nextPL_nextwhite;
5457 curmad('X', PL_thistoken);
5458 PL_thistoken = newSVpvs("");
5466 /* If followed by var or block, call it a method (unless sub) */
5468 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5469 PL_last_lop = PL_oldbufptr;
5470 PL_last_lop_op = OP_METHOD;
5474 /* If followed by a bareword, see if it looks like indir obj. */
5477 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5478 && (tmp = intuit_method(s, gv, cv)))
5481 /* Not a method, so call it a subroutine (if defined) */
5484 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5485 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5486 "Ambiguous use of -%s resolved as -&%s()",
5487 PL_tokenbuf, PL_tokenbuf);
5488 /* Check for a constant sub */
5489 if ((sv = gv_const_sv(gv))) {
5491 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5492 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5493 yylval.opval->op_private = 0;
5497 /* Resolve to GV now. */
5498 if (SvTYPE(gv) != SVt_PVGV) {
5499 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5500 assert (SvTYPE(gv) == SVt_PVGV);
5501 /* cv must have been some sort of placeholder, so
5502 now needs replacing with a real code reference. */
5506 op_free(yylval.opval);
5507 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5508 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5509 PL_last_lop = PL_oldbufptr;
5510 PL_last_lop_op = OP_ENTERSUB;
5511 /* Is there a prototype? */
5519 const char *proto = SvPV_const((SV*)cv, protolen);
5522 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5524 while (*proto == ';')
5526 if (*proto == '&' && *s == '{') {
5528 sv_setpvs(PL_subname, "__ANON__");
5530 sv_setpvs(PL_subname, "__ANON__::__ANON__");
5537 PL_nextwhite = PL_thiswhite;
5540 start_force(PL_curforce);
5541 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5544 PL_nextwhite = nextPL_nextwhite;
5545 curmad('X', PL_thistoken);
5546 PL_thistoken = newSVpvs("");
5553 /* Guess harder when madskills require "best effort". */
5554 if (PL_madskills && (!gv || !GvCVu(gv))) {
5555 int probable_sub = 0;
5556 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5558 else if (isALPHA(*s)) {
5562 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5563 if (!keyword(tmpbuf, tmplen, 0))
5566 while (d < PL_bufend && isSPACE(*d))
5568 if (*d == '=' && d[1] == '>')
5573 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5574 op_free(yylval.opval);
5575 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5576 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5577 PL_last_lop = PL_oldbufptr;
5578 PL_last_lop_op = OP_ENTERSUB;
5579 PL_nextwhite = PL_thiswhite;
5581 start_force(PL_curforce);
5582 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5584 PL_nextwhite = nextPL_nextwhite;
5585 curmad('X', PL_thistoken);
5586 PL_thistoken = newSVpvs("");
5591 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5598 /* Call it a bare word */
5600 if (PL_hints & HINT_STRICT_SUBS)
5601 yylval.opval->op_private |= OPpCONST_STRICT;
5604 if (lastchar != '-') {
5605 if (ckWARN(WARN_RESERVED)) {
5609 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5610 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5617 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5618 && ckWARN_d(WARN_AMBIGUOUS)) {
5619 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5620 "Operator or semicolon missing before %c%s",
5621 lastchar, PL_tokenbuf);
5622 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5623 "Ambiguous use of %c resolved as operator %c",
5624 lastchar, lastchar);
5630 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5631 newSVpv(CopFILE(PL_curcop),0));
5635 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5636 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5639 case KEY___PACKAGE__:
5640 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5642 ? newSVhek(HvNAME_HEK(PL_curstash))
5649 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5650 const char *pname = "main";
5651 if (PL_tokenbuf[2] == 'D')
5652 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5653 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5657 GvIOp(gv) = newIO();
5658 IoIFP(GvIOp(gv)) = PL_rsfp;
5659 #if defined(HAS_FCNTL) && defined(F_SETFD)
5661 const int fd = PerlIO_fileno(PL_rsfp);
5662 fcntl(fd,F_SETFD,fd >= 3);
5665 /* Mark this internal pseudo-handle as clean */
5666 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5668 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5669 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5670 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5672 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5673 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5674 /* if the script was opened in binmode, we need to revert
5675 * it to text mode for compatibility; but only iff it has CRs
5676 * XXX this is a questionable hack at best. */
5677 if (PL_bufend-PL_bufptr > 2
5678 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5681 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5682 loc = PerlIO_tell(PL_rsfp);
5683 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5686 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5688 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5689 #endif /* NETWARE */
5690 #ifdef PERLIO_IS_STDIO /* really? */
5691 # if defined(__BORLANDC__)
5692 /* XXX see note in do_binmode() */
5693 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5697 PerlIO_seek(PL_rsfp, loc, 0);
5701 #ifdef PERLIO_LAYERS
5704 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5705 else if (PL_encoding) {
5712 XPUSHs(PL_encoding);
5714 call_method("name", G_SCALAR);
5718 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5719 Perl_form(aTHX_ ":encoding(%"SVf")",
5728 if (PL_realtokenstart >= 0) {
5729 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5731 PL_endwhite = newSVpvs("");
5732 sv_catsv(PL_endwhite, PL_thiswhite);
5734 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5735 PL_realtokenstart = -1;
5737 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5738 SvCUR(PL_endwhite))) != NULL) ;
5753 if (PL_expect == XSTATE) {
5760 if (*s == ':' && s[1] == ':') {
5763 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5764 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5765 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5768 else if (tmp == KEY_require || tmp == KEY_do)
5769 /* that's a way to remember we saw "CORE::" */
5782 LOP(OP_ACCEPT,XTERM);
5788 LOP(OP_ATAN2,XTERM);
5794 LOP(OP_BINMODE,XTERM);
5797 LOP(OP_BLESS,XTERM);
5806 /* When 'use switch' is in effect, continue has a dual
5807 life as a control operator. */
5809 if (!FEATURE_IS_ENABLED("switch"))
5812 /* We have to disambiguate the two senses of
5813 "continue". If the next token is a '{' then
5814 treat it as the start of a continue block;
5815 otherwise treat it as a control operator.
5827 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5844 if (!PL_cryptseen) {
5845 PL_cryptseen = TRUE;
5849 LOP(OP_CRYPT,XTERM);
5852 LOP(OP_CHMOD,XTERM);
5855 LOP(OP_CHOWN,XTERM);
5858 LOP(OP_CONNECT,XTERM);
5877 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5878 if (orig_keyword == KEY_do) {
5887 PL_hints |= HINT_BLOCK_SCOPE;
5897 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5898 LOP(OP_DBMOPEN,XTERM);
5904 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5911 yylval.ival = CopLINE(PL_curcop);
5927 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5928 UNIBRACK(OP_ENTEREVAL);
5942 case KEY_endhostent:
5948 case KEY_endservent:
5951 case KEY_endprotoent:
5962 yylval.ival = CopLINE(PL_curcop);
5964 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5967 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5970 if ((PL_bufend - p) >= 3 &&
5971 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5973 else if ((PL_bufend - p) >= 4 &&
5974 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5977 if (isIDFIRST_lazy_if(p,UTF)) {
5978 p = scan_ident(p, PL_bufend,
5979 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5983 Perl_croak(aTHX_ "Missing $ on loop variable");
5985 s = SvPVX(PL_linestr) + soff;
5991 LOP(OP_FORMLINE,XTERM);
5997 LOP(OP_FCNTL,XTERM);
6003 LOP(OP_FLOCK,XTERM);
6012 LOP(OP_GREPSTART, XREF);
6015 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6030 case KEY_getpriority:
6031 LOP(OP_GETPRIORITY,XTERM);
6033 case KEY_getprotobyname:
6036 case KEY_getprotobynumber:
6037 LOP(OP_GPBYNUMBER,XTERM);
6039 case KEY_getprotoent:
6051 case KEY_getpeername:
6052 UNI(OP_GETPEERNAME);
6054 case KEY_gethostbyname:
6057 case KEY_gethostbyaddr:
6058 LOP(OP_GHBYADDR,XTERM);
6060 case KEY_gethostent:
6063 case KEY_getnetbyname:
6066 case KEY_getnetbyaddr:
6067 LOP(OP_GNBYADDR,XTERM);
6072 case KEY_getservbyname:
6073 LOP(OP_GSBYNAME,XTERM);
6075 case KEY_getservbyport:
6076 LOP(OP_GSBYPORT,XTERM);
6078 case KEY_getservent:
6081 case KEY_getsockname:
6082 UNI(OP_GETSOCKNAME);
6084 case KEY_getsockopt:
6085 LOP(OP_GSOCKOPT,XTERM);
6100 yylval.ival = CopLINE(PL_curcop);
6110 yylval.ival = CopLINE(PL_curcop);
6114 LOP(OP_INDEX,XTERM);
6120 LOP(OP_IOCTL,XTERM);
6132 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6164 LOP(OP_LISTEN,XTERM);
6173 s = scan_pat(s,OP_MATCH);
6174 TERM(sublex_start());
6177 LOP(OP_MAPSTART, XREF);
6180 LOP(OP_MKDIR,XTERM);
6183 LOP(OP_MSGCTL,XTERM);
6186 LOP(OP_MSGGET,XTERM);
6189 LOP(OP_MSGRCV,XTERM);
6192 LOP(OP_MSGSND,XTERM);
6197 PL_in_my = (U16)tmp;
6199 if (isIDFIRST_lazy_if(s,UTF)) {
6203 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6204 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6206 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6207 if (!PL_in_my_stash) {
6210 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6214 if (PL_madskills) { /* just add type to declarator token */
6215 sv_catsv(PL_thistoken, PL_nextwhite);
6217 sv_catpvn(PL_thistoken, start, s - start);
6225 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6232 s = tokenize_use(0, s);
6236 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6243 if (isIDFIRST_lazy_if(s,UTF)) {
6245 for (d = s; isALNUM_lazy_if(d,UTF);)
6247 for (t=d; isSPACE(*t);)
6249 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6251 && !(t[0] == '=' && t[1] == '>')
6253 int parms_len = (int)(d-s);
6254 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6255 "Precedence problem: open %.*s should be open(%.*s)",
6256 parms_len, s, parms_len, s);
6262 yylval.ival = OP_OR;
6272 LOP(OP_OPEN_DIR,XTERM);
6275 checkcomma(s,PL_tokenbuf,"filehandle");
6279 checkcomma(s,PL_tokenbuf,"filehandle");
6298 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6302 LOP(OP_PIPE_OP,XTERM);
6305 s = scan_str(s,!!PL_madskills,FALSE);
6308 yylval.ival = OP_CONST;
6309 TERM(sublex_start());
6315 s = scan_str(s,!!PL_madskills,FALSE);
6318 PL_expect = XOPERATOR;
6320 if (SvCUR(PL_lex_stuff)) {
6323 d = SvPV_force(PL_lex_stuff, len);
6325 for (; isSPACE(*d) && len; --len, ++d)
6330 if (!warned && ckWARN(WARN_QW)) {
6331 for (; !isSPACE(*d) && len; --len, ++d) {
6333 Perl_warner(aTHX_ packWARN(WARN_QW),
6334 "Possible attempt to separate words with commas");
6337 else if (*d == '#') {
6338 Perl_warner(aTHX_ packWARN(WARN_QW),
6339 "Possible attempt to put comments in qw() list");
6345 for (; !isSPACE(*d) && len; --len, ++d)
6348 sv = newSVpvn(b, d-b);
6349 if (DO_UTF8(PL_lex_stuff))
6351 words = append_elem(OP_LIST, words,
6352 newSVOP(OP_CONST, 0, tokeq(sv)));
6356 start_force(PL_curforce);
6357 NEXTVAL_NEXTTOKE.opval = words;
6362 SvREFCNT_dec(PL_lex_stuff);
6363 PL_lex_stuff = NULL;
6369 s = scan_str(s,!!PL_madskills,FALSE);
6372 yylval.ival = OP_STRINGIFY;
6373 if (SvIVX(PL_lex_stuff) == '\'')
6374 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6375 TERM(sublex_start());
6378 s = scan_pat(s,OP_QR);
6379 TERM(sublex_start());
6382 s = scan_str(s,!!PL_madskills,FALSE);
6385 readpipe_override();
6386 TERM(sublex_start());
6394 s = force_version(s, FALSE);
6396 else if (*s != 'v' || !isDIGIT(s[1])
6397 || (s = force_version(s, TRUE), *s == 'v'))
6399 *PL_tokenbuf = '\0';
6400 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6401 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6402 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6404 yyerror("<> should be quotes");
6406 if (orig_keyword == KEY_require) {
6414 PL_last_uni = PL_oldbufptr;
6415 PL_last_lop_op = OP_REQUIRE;
6417 return REPORT( (int)REQUIRE );
6423 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6427 LOP(OP_RENAME,XTERM);
6436 LOP(OP_RINDEX,XTERM);
6445 UNIDOR(OP_READLINE);
6448 UNIDOR(OP_BACKTICK);
6457 LOP(OP_REVERSE,XTERM);
6460 UNIDOR(OP_READLINK);
6468 TERM(sublex_start());
6470 TOKEN(1); /* force error */
6473 checkcomma(s,PL_tokenbuf,"filehandle");
6483 LOP(OP_SELECT,XTERM);
6489 LOP(OP_SEMCTL,XTERM);
6492 LOP(OP_SEMGET,XTERM);
6495 LOP(OP_SEMOP,XTERM);
6501 LOP(OP_SETPGRP,XTERM);
6503 case KEY_setpriority:
6504 LOP(OP_SETPRIORITY,XTERM);
6506 case KEY_sethostent:
6512 case KEY_setservent:
6515 case KEY_setprotoent:
6525 LOP(OP_SEEKDIR,XTERM);
6527 case KEY_setsockopt:
6528 LOP(OP_SSOCKOPT,XTERM);
6534 LOP(OP_SHMCTL,XTERM);
6537 LOP(OP_SHMGET,XTERM);
6540 LOP(OP_SHMREAD,XTERM);
6543 LOP(OP_SHMWRITE,XTERM);
6546 LOP(OP_SHUTDOWN,XTERM);
6555 LOP(OP_SOCKET,XTERM);
6557 case KEY_socketpair:
6558 LOP(OP_SOCKPAIR,XTERM);
6561 checkcomma(s,PL_tokenbuf,"subroutine name");
6563 if (*s == ';' || *s == ')') /* probably a close */
6564 Perl_croak(aTHX_ "sort is now a reserved word");
6566 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6570 LOP(OP_SPLIT,XTERM);
6573 LOP(OP_SPRINTF,XTERM);
6576 LOP(OP_SPLICE,XTERM);
6591 LOP(OP_SUBSTR,XTERM);
6597 char tmpbuf[sizeof PL_tokenbuf];
6598 SSize_t tboffset = 0;
6599 expectation attrful;
6600 bool have_name, have_proto;
6601 const int key = tmp;
6606 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6607 SV *subtoken = newSVpvn(tstart, s - tstart);
6611 s = SKIPSPACE2(s,tmpwhite);
6616 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6617 (*s == ':' && s[1] == ':'))
6624 attrful = XATTRBLOCK;
6625 /* remember buffer pos'n for later force_word */
6626 tboffset = s - PL_oldbufptr;
6627 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6630 nametoke = newSVpvn(s, d - s);
6632 if (memchr(tmpbuf, ':', len))
6633 sv_setpvn(PL_subname, tmpbuf, len);
6635 sv_setsv(PL_subname,PL_curstname);
6636 sv_catpvs(PL_subname,"::");
6637 sv_catpvn(PL_subname,tmpbuf,len);
6644 CURMAD('X', nametoke);
6645 CURMAD('_', tmpwhite);
6646 (void) force_word(PL_oldbufptr + tboffset, WORD,
6649 s = SKIPSPACE2(d,tmpwhite);
6656 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6657 PL_expect = XTERMBLOCK;
6658 attrful = XATTRTERM;
6659 sv_setpvn(PL_subname,"?",1);
6663 if (key == KEY_format) {
6665 PL_lex_formbrack = PL_lex_brackets + 1;
6667 PL_thistoken = subtoken;
6671 (void) force_word(PL_oldbufptr + tboffset, WORD,
6677 /* Look for a prototype */
6680 bool bad_proto = FALSE;
6681 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6683 s = scan_str(s,!!PL_madskills,FALSE);
6685 Perl_croak(aTHX_ "Prototype not terminated");
6686 /* strip spaces and check for bad characters */
6687 d = SvPVX(PL_lex_stuff);
6689 for (p = d; *p; ++p) {
6692 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6698 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6699 "Illegal character in prototype for %"SVf" : %s",
6700 SVfARG(PL_subname), d);
6701 SvCUR_set(PL_lex_stuff, tmp);
6706 CURMAD('q', PL_thisopen);
6707 CURMAD('_', tmpwhite);
6708 CURMAD('=', PL_thisstuff);
6709 CURMAD('Q', PL_thisclose);
6710 NEXTVAL_NEXTTOKE.opval =
6711 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6712 PL_lex_stuff = NULL;
6715 s = SKIPSPACE2(s,tmpwhite);
6723 if (*s == ':' && s[1] != ':')
6724 PL_expect = attrful;
6725 else if (*s != '{' && key == KEY_sub) {
6727 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6729 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6736 curmad('^', newSVpvs(""));
6737 CURMAD('_', tmpwhite);
6741 PL_thistoken = subtoken;
6744 NEXTVAL_NEXTTOKE.opval =
6745 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6746 PL_lex_stuff = NULL;
6752 sv_setpvs(PL_subname, "__ANON__");
6754 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6758 (void) force_word(PL_oldbufptr + tboffset, WORD,
6767 LOP(OP_SYSTEM,XREF);
6770 LOP(OP_SYMLINK,XTERM);
6773 LOP(OP_SYSCALL,XTERM);
6776 LOP(OP_SYSOPEN,XTERM);
6779 LOP(OP_SYSSEEK,XTERM);
6782 LOP(OP_SYSREAD,XTERM);
6785 LOP(OP_SYSWRITE,XTERM);
6789 TERM(sublex_start());
6810 LOP(OP_TRUNCATE,XTERM);
6822 yylval.ival = CopLINE(PL_curcop);
6826 yylval.ival = CopLINE(PL_curcop);
6830 LOP(OP_UNLINK,XTERM);
6836 LOP(OP_UNPACK,XTERM);
6839 LOP(OP_UTIME,XTERM);
6845 LOP(OP_UNSHIFT,XTERM);
6848 s = tokenize_use(1, s);
6858 yylval.ival = CopLINE(PL_curcop);
6862 yylval.ival = CopLINE(PL_curcop);
6866 PL_hints |= HINT_BLOCK_SCOPE;
6873 LOP(OP_WAITPID,XTERM);
6882 ctl_l[0] = toCTRL('L');
6884 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6887 /* Make sure $^L is defined */
6888 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6893 if (PL_expect == XOPERATOR)
6899 yylval.ival = OP_XOR;
6904 TERM(sublex_start());
6909 #pragma segment Main
6913 S_pending_ident(pTHX)
6918 /* pit holds the identifier we read and pending_ident is reset */
6919 char pit = PL_pending_ident;
6920 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
6921 /* All routes through this function want to know if there is a colon. */
6922 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
6923 PL_pending_ident = 0;
6925 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6926 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6927 "### Pending identifier '%s'\n", PL_tokenbuf); });
6929 /* if we're in a my(), we can't allow dynamics here.
6930 $foo'bar has already been turned into $foo::bar, so
6931 just check for colons.
6933 if it's a legal name, the OP is a PADANY.
6936 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6938 yyerror(Perl_form(aTHX_ "No package name allowed for "
6939 "variable %s in \"our\"",
6941 tmp = allocmy(PL_tokenbuf);
6945 yyerror(Perl_form(aTHX_ PL_no_myglob,
6946 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6948 yylval.opval = newOP(OP_PADANY, 0);
6949 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6955 build the ops for accesses to a my() variable.
6957 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6958 then used in a comparison. This catches most, but not
6959 all cases. For instance, it catches
6960 sort { my($a); $a <=> $b }
6962 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6963 (although why you'd do that is anyone's guess).
6968 tmp = pad_findmy(PL_tokenbuf);
6969 if (tmp != NOT_IN_PAD) {
6970 /* might be an "our" variable" */
6971 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6972 /* build ops for a bareword */
6973 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6974 HEK * const stashname = HvNAME_HEK(stash);
6975 SV * const sym = newSVhek(stashname);
6976 sv_catpvs(sym, "::");
6977 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
6978 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6979 yylval.opval->op_private = OPpCONST_ENTERED;
6982 ? (GV_ADDMULTI | GV_ADDINEVAL)
6985 ((PL_tokenbuf[0] == '$') ? SVt_PV
6986 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6991 /* if it's a sort block and they're naming $a or $b */
6992 if (PL_last_lop_op == OP_SORT &&
6993 PL_tokenbuf[0] == '$' &&
6994 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6997 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6998 d < PL_bufend && *d != '\n';
7001 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7002 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7008 yylval.opval = newOP(OP_PADANY, 0);
7009 yylval.opval->op_targ = tmp;
7015 Whine if they've said @foo in a doublequoted string,
7016 and @foo isn't a variable we can find in the symbol
7019 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7020 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7022 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7023 && ckWARN(WARN_AMBIGUOUS)
7024 /* DO NOT warn for @- and @+ */
7025 && !( PL_tokenbuf[2] == '\0' &&
7026 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7029 /* Downgraded from fatal to warning 20000522 mjd */
7030 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7031 "Possible unintended interpolation of %s in string",
7036 /* build ops for a bareword */
7037 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7039 yylval.opval->op_private = OPpCONST_ENTERED;
7041 PL_tokenbuf + 1, tokenbuf_len - 1,
7042 /* If the identifier refers to a stash, don't autovivify it.
7043 * Change 24660 had the side effect of causing symbol table
7044 * hashes to always be defined, even if they were freshly
7045 * created and the only reference in the entire program was
7046 * the single statement with the defined %foo::bar:: test.
7047 * It appears that all code in the wild doing this actually
7048 * wants to know whether sub-packages have been loaded, so
7049 * by avoiding auto-vivifying symbol tables, we ensure that
7050 * defined %foo::bar:: continues to be false, and the existing
7051 * tests still give the expected answers, even though what
7052 * they're actually testing has now changed subtly.
7054 (*PL_tokenbuf == '%'
7055 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7058 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7059 ((PL_tokenbuf[0] == '$') ? SVt_PV
7060 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7066 * The following code was generated by perl_keyword.pl.
7070 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7075 case 1: /* 5 tokens of length 1 */
7107 case 2: /* 18 tokens of length 2 */
7253 case 3: /* 29 tokens of length 3 */
7257 if (name[1] == 'N' &&
7320 if (name[1] == 'i' &&
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, STRLEN keylen,
10514 SV *sv, SV *pv, const char *type, STRLEN typelen)
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, keylen, 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(newSVpvn(type, typelen));
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;
10981 sv_catpvs(repl, "eval ");
10983 sv_catpvs(repl, "do ");
10985 sv_catpvs(repl, "{");
10986 sv_catsv(repl, PL_lex_repl);
10987 if (strchr(SvPVX(PL_lex_repl), '#'))
10988 sv_catpvs(repl, "\n");
10989 sv_catpvs(repl, "}");
10991 SvREFCNT_dec(PL_lex_repl);
10992 PL_lex_repl = repl;
10995 PL_lex_op = (OP*)pm;
10996 yylval.ival = OP_SUBST;
11001 S_scan_trans(pTHX_ char *start)
11014 yylval.ival = OP_NULL;
11016 s = scan_str(start,!!PL_madskills,FALSE);
11018 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11020 if (s[-1] == PL_multi_open)
11023 if (PL_madskills) {
11024 CURMAD('q', PL_thisopen);
11025 CURMAD('_', PL_thiswhite);
11026 CURMAD('E', PL_thisstuff);
11027 CURMAD('Q', PL_thisclose);
11028 PL_realtokenstart = s - SvPVX(PL_linestr);
11032 s = scan_str(s,!!PL_madskills,FALSE);
11034 if (PL_lex_stuff) {
11035 SvREFCNT_dec(PL_lex_stuff);
11036 PL_lex_stuff = NULL;
11038 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11040 if (PL_madskills) {
11041 CURMAD('z', PL_thisopen);
11042 CURMAD('R', PL_thisstuff);
11043 CURMAD('Z', PL_thisclose);
11046 complement = del = squash = 0;
11053 complement = OPpTRANS_COMPLEMENT;
11056 del = OPpTRANS_DELETE;
11059 squash = OPpTRANS_SQUASH;
11068 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11069 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11070 o->op_private &= ~OPpTRANS_ALL;
11071 o->op_private |= del|squash|complement|
11072 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11073 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11076 yylval.ival = OP_TRANS;
11079 if (PL_madskills) {
11081 curmad('m', newSVpvn(modstart, s - modstart));
11082 append_madprops(PL_thismad, o, 0);
11091 S_scan_heredoc(pTHX_ register char *s)
11095 I32 op_type = OP_SCALAR;
11099 const char *found_newline;
11103 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11105 I32 stuffstart = s - SvPVX(PL_linestr);
11108 PL_realtokenstart = -1;
11113 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11117 while (SPACE_OR_TAB(*peek))
11119 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11122 s = delimcpy(d, e, s, PL_bufend, term, &len);
11132 if (!isALNUM_lazy_if(s,UTF))
11133 deprecate_old("bare << to mean <<\"\"");
11134 for (; isALNUM_lazy_if(s,UTF); s++) {
11139 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11140 Perl_croak(aTHX_ "Delimiter for here document is too long");
11143 len = d - PL_tokenbuf;
11146 if (PL_madskills) {
11147 tstart = PL_tokenbuf + !outer;
11148 PL_thisclose = newSVpvn(tstart, len - !outer);
11149 tstart = SvPVX(PL_linestr) + stuffstart;
11150 PL_thisopen = newSVpvn(tstart, s - tstart);
11151 stuffstart = s - SvPVX(PL_linestr);
11154 #ifndef PERL_STRICT_CR
11155 d = strchr(s, '\r');
11157 char * const olds = s;
11159 while (s < PL_bufend) {
11165 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11174 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11181 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11182 herewas = newSVpvn(s,PL_bufend-s);
11186 herewas = newSVpvn(s-1,found_newline-s+1);
11189 herewas = newSVpvn(s,found_newline-s);
11193 if (PL_madskills) {
11194 tstart = SvPVX(PL_linestr) + stuffstart;
11196 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11198 PL_thisstuff = newSVpvn(tstart, s - tstart);
11201 s += SvCUR(herewas);
11204 stuffstart = s - SvPVX(PL_linestr);
11210 tmpstr = newSV_type(SVt_PVIV);
11211 SvGROW(tmpstr, 80);
11212 if (term == '\'') {
11213 op_type = OP_CONST;
11214 SvIV_set(tmpstr, -1);
11216 else if (term == '`') {
11217 op_type = OP_BACKTICK;
11218 SvIV_set(tmpstr, '\\');
11222 PL_multi_start = CopLINE(PL_curcop);
11223 PL_multi_open = PL_multi_close = '<';
11224 term = *PL_tokenbuf;
11225 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11226 char * const bufptr = PL_sublex_info.super_bufptr;
11227 char * const bufend = PL_sublex_info.super_bufend;
11228 char * const olds = s - SvCUR(herewas);
11229 s = strchr(bufptr, '\n');
11233 while (s < bufend &&
11234 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11236 CopLINE_inc(PL_curcop);
11239 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11240 missingterm(PL_tokenbuf);
11242 sv_setpvn(herewas,bufptr,d-bufptr+1);
11243 sv_setpvn(tmpstr,d+1,s-d);
11245 sv_catpvn(herewas,s,bufend-s);
11246 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11253 while (s < PL_bufend &&
11254 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11256 CopLINE_inc(PL_curcop);
11258 if (s >= PL_bufend) {
11259 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11260 missingterm(PL_tokenbuf);
11262 sv_setpvn(tmpstr,d+1,s-d);
11264 if (PL_madskills) {
11266 sv_catpvn(PL_thisstuff, d + 1, s - d);
11268 PL_thisstuff = newSVpvn(d + 1, s - d);
11269 stuffstart = s - SvPVX(PL_linestr);
11273 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11275 sv_catpvn(herewas,s,PL_bufend-s);
11276 sv_setsv(PL_linestr,herewas);
11277 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11278 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11279 PL_last_lop = PL_last_uni = NULL;
11282 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11283 while (s >= PL_bufend) { /* multiple line string? */
11285 if (PL_madskills) {
11286 tstart = SvPVX(PL_linestr) + stuffstart;
11288 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11290 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11294 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11295 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11296 missingterm(PL_tokenbuf);
11299 stuffstart = s - SvPVX(PL_linestr);
11301 CopLINE_inc(PL_curcop);
11302 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11303 PL_last_lop = PL_last_uni = NULL;
11304 #ifndef PERL_STRICT_CR
11305 if (PL_bufend - PL_linestart >= 2) {
11306 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11307 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11309 PL_bufend[-2] = '\n';
11311 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11313 else if (PL_bufend[-1] == '\r')
11314 PL_bufend[-1] = '\n';
11316 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11317 PL_bufend[-1] = '\n';
11319 if (PERLDB_LINE && PL_curstash != PL_debstash)
11320 update_debugger_info(PL_linestr, NULL, 0);
11321 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11322 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11323 *(SvPVX(PL_linestr) + off ) = ' ';
11324 sv_catsv(PL_linestr,herewas);
11325 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11326 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11330 sv_catsv(tmpstr,PL_linestr);
11335 PL_multi_end = CopLINE(PL_curcop);
11336 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11337 SvPV_shrink_to_cur(tmpstr);
11339 SvREFCNT_dec(herewas);
11341 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11343 else if (PL_encoding)
11344 sv_recode_to_utf8(tmpstr, PL_encoding);
11346 PL_lex_stuff = tmpstr;
11347 yylval.ival = op_type;
11351 /* scan_inputsymbol
11352 takes: current position in input buffer
11353 returns: new position in input buffer
11354 side-effects: yylval and lex_op are set.
11359 <FH> read from filehandle
11360 <pkg::FH> read from package qualified filehandle
11361 <pkg'FH> read from package qualified filehandle
11362 <$fh> read from filehandle in $fh
11363 <*.h> filename glob
11368 S_scan_inputsymbol(pTHX_ char *start)
11371 register char *s = start; /* current position in buffer */
11375 char *d = PL_tokenbuf; /* start of temp holding space */
11376 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11378 end = strchr(s, '\n');
11381 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11383 /* die if we didn't have space for the contents of the <>,
11384 or if it didn't end, or if we see a newline
11387 if (len >= (I32)sizeof PL_tokenbuf)
11388 Perl_croak(aTHX_ "Excessively long <> operator");
11390 Perl_croak(aTHX_ "Unterminated <> operator");
11395 Remember, only scalar variables are interpreted as filehandles by
11396 this code. Anything more complex (e.g., <$fh{$num}>) will be
11397 treated as a glob() call.
11398 This code makes use of the fact that except for the $ at the front,
11399 a scalar variable and a filehandle look the same.
11401 if (*d == '$' && d[1]) d++;
11403 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11404 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11407 /* If we've tried to read what we allow filehandles to look like, and
11408 there's still text left, then it must be a glob() and not a getline.
11409 Use scan_str to pull out the stuff between the <> and treat it
11410 as nothing more than a string.
11413 if (d - PL_tokenbuf != len) {
11414 yylval.ival = OP_GLOB;
11415 s = scan_str(start,!!PL_madskills,FALSE);
11417 Perl_croak(aTHX_ "Glob not terminated");
11421 bool readline_overriden = FALSE;
11424 /* we're in a filehandle read situation */
11427 /* turn <> into <ARGV> */
11429 Copy("ARGV",d,5,char);
11431 /* Check whether readline() is overriden */
11432 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11434 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11436 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11437 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11438 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11439 readline_overriden = TRUE;
11441 /* if <$fh>, create the ops to turn the variable into a
11445 /* try to find it in the pad for this block, otherwise find
11446 add symbol table ops
11448 const PADOFFSET tmp = pad_findmy(d);
11449 if (tmp != NOT_IN_PAD) {
11450 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11451 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11452 HEK * const stashname = HvNAME_HEK(stash);
11453 SV * const sym = sv_2mortal(newSVhek(stashname));
11454 sv_catpvs(sym, "::");
11455 sv_catpv(sym, d+1);
11460 OP * const o = newOP(OP_PADSV, 0);
11462 PL_lex_op = readline_overriden
11463 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11464 append_elem(OP_LIST, o,
11465 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11466 : (OP*)newUNOP(OP_READLINE, 0, o);
11475 ? (GV_ADDMULTI | GV_ADDINEVAL)
11478 PL_lex_op = readline_overriden
11479 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11480 append_elem(OP_LIST,
11481 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11482 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11483 : (OP*)newUNOP(OP_READLINE, 0,
11484 newUNOP(OP_RV2SV, 0,
11485 newGVOP(OP_GV, 0, gv)));
11487 if (!readline_overriden)
11488 PL_lex_op->op_flags |= OPf_SPECIAL;
11489 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11490 yylval.ival = OP_NULL;
11493 /* If it's none of the above, it must be a literal filehandle
11494 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11496 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11497 PL_lex_op = readline_overriden
11498 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11499 append_elem(OP_LIST,
11500 newGVOP(OP_GV, 0, gv),
11501 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11502 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11503 yylval.ival = OP_NULL;
11512 takes: start position in buffer
11513 keep_quoted preserve \ on the embedded delimiter(s)
11514 keep_delims preserve the delimiters around the string
11515 returns: position to continue reading from buffer
11516 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11517 updates the read buffer.
11519 This subroutine pulls a string out of the input. It is called for:
11520 q single quotes q(literal text)
11521 ' single quotes 'literal text'
11522 qq double quotes qq(interpolate $here please)
11523 " double quotes "interpolate $here please"
11524 qx backticks qx(/bin/ls -l)
11525 ` backticks `/bin/ls -l`
11526 qw quote words @EXPORT_OK = qw( func() $spam )
11527 m// regexp match m/this/
11528 s/// regexp substitute s/this/that/
11529 tr/// string transliterate tr/this/that/
11530 y/// string transliterate y/this/that/
11531 ($*@) sub prototypes sub foo ($)
11532 (stuff) sub attr parameters sub foo : attr(stuff)
11533 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11535 In most of these cases (all but <>, patterns and transliterate)
11536 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11537 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11538 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11541 It skips whitespace before the string starts, and treats the first
11542 character as the delimiter. If the delimiter is one of ([{< then
11543 the corresponding "close" character )]}> is used as the closing
11544 delimiter. It allows quoting of delimiters, and if the string has
11545 balanced delimiters ([{<>}]) it allows nesting.
11547 On success, the SV with the resulting string is put into lex_stuff or,
11548 if that is already non-NULL, into lex_repl. The second case occurs only
11549 when parsing the RHS of the special constructs s/// and tr/// (y///).
11550 For convenience, the terminating delimiter character is stuffed into
11555 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11558 SV *sv; /* scalar value: string */
11559 const char *tmps; /* temp string, used for delimiter matching */
11560 register char *s = start; /* current position in the buffer */
11561 register char term; /* terminating character */
11562 register char *to; /* current position in the sv's data */
11563 I32 brackets = 1; /* bracket nesting level */
11564 bool has_utf8 = FALSE; /* is there any utf8 content? */
11565 I32 termcode; /* terminating char. code */
11566 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11567 STRLEN termlen; /* length of terminating string */
11568 int last_off = 0; /* last position for nesting bracket */
11574 /* skip space before the delimiter */
11580 if (PL_realtokenstart >= 0) {
11581 stuffstart = PL_realtokenstart;
11582 PL_realtokenstart = -1;
11585 stuffstart = start - SvPVX(PL_linestr);
11587 /* mark where we are, in case we need to report errors */
11590 /* after skipping whitespace, the next character is the terminator */
11593 termcode = termstr[0] = term;
11597 termcode = utf8_to_uvchr((U8*)s, &termlen);
11598 Copy(s, termstr, termlen, U8);
11599 if (!UTF8_IS_INVARIANT(term))
11603 /* mark where we are */
11604 PL_multi_start = CopLINE(PL_curcop);
11605 PL_multi_open = term;
11607 /* find corresponding closing delimiter */
11608 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11609 termcode = termstr[0] = term = tmps[5];
11611 PL_multi_close = term;
11613 /* create a new SV to hold the contents. 79 is the SV's initial length.
11614 What a random number. */
11615 sv = newSV_type(SVt_PVIV);
11617 SvIV_set(sv, termcode);
11618 (void)SvPOK_only(sv); /* validate pointer */
11620 /* move past delimiter and try to read a complete string */
11622 sv_catpvn(sv, s, termlen);
11625 tstart = SvPVX(PL_linestr) + stuffstart;
11626 if (!PL_thisopen && !keep_delims) {
11627 PL_thisopen = newSVpvn(tstart, s - tstart);
11628 stuffstart = s - SvPVX(PL_linestr);
11632 if (PL_encoding && !UTF) {
11636 int offset = s - SvPVX_const(PL_linestr);
11637 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11638 &offset, (char*)termstr, termlen);
11639 const char * const ns = SvPVX_const(PL_linestr) + offset;
11640 char * const svlast = SvEND(sv) - 1;
11642 for (; s < ns; s++) {
11643 if (*s == '\n' && !PL_rsfp)
11644 CopLINE_inc(PL_curcop);
11647 goto read_more_line;
11649 /* handle quoted delimiters */
11650 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11652 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11654 if ((svlast-1 - t) % 2) {
11655 if (!keep_quoted) {
11656 *(svlast-1) = term;
11658 SvCUR_set(sv, SvCUR(sv) - 1);
11663 if (PL_multi_open == PL_multi_close) {
11669 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11670 /* At here, all closes are "was quoted" one,
11671 so we don't check PL_multi_close. */
11673 if (!keep_quoted && *(t+1) == PL_multi_open)
11678 else if (*t == PL_multi_open)
11686 SvCUR_set(sv, w - SvPVX_const(sv));
11688 last_off = w - SvPVX(sv);
11689 if (--brackets <= 0)
11694 if (!keep_delims) {
11695 SvCUR_set(sv, SvCUR(sv) - 1);
11701 /* extend sv if need be */
11702 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11703 /* set 'to' to the next character in the sv's string */
11704 to = SvPVX(sv)+SvCUR(sv);
11706 /* if open delimiter is the close delimiter read unbridle */
11707 if (PL_multi_open == PL_multi_close) {
11708 for (; s < PL_bufend; s++,to++) {
11709 /* embedded newlines increment the current line number */
11710 if (*s == '\n' && !PL_rsfp)
11711 CopLINE_inc(PL_curcop);
11712 /* handle quoted delimiters */
11713 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11714 if (!keep_quoted && s[1] == term)
11716 /* any other quotes are simply copied straight through */
11720 /* terminate when run out of buffer (the for() condition), or
11721 have found the terminator */
11722 else if (*s == term) {
11725 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11728 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11734 /* if the terminator isn't the same as the start character (e.g.,
11735 matched brackets), we have to allow more in the quoting, and
11736 be prepared for nested brackets.
11739 /* read until we run out of string, or we find the terminator */
11740 for (; s < PL_bufend; s++,to++) {
11741 /* embedded newlines increment the line count */
11742 if (*s == '\n' && !PL_rsfp)
11743 CopLINE_inc(PL_curcop);
11744 /* backslashes can escape the open or closing characters */
11745 if (*s == '\\' && s+1 < PL_bufend) {
11746 if (!keep_quoted &&
11747 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11752 /* allow nested opens and closes */
11753 else if (*s == PL_multi_close && --brackets <= 0)
11755 else if (*s == PL_multi_open)
11757 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11762 /* terminate the copied string and update the sv's end-of-string */
11764 SvCUR_set(sv, to - SvPVX_const(sv));
11767 * this next chunk reads more into the buffer if we're not done yet
11771 break; /* handle case where we are done yet :-) */
11773 #ifndef PERL_STRICT_CR
11774 if (to - SvPVX_const(sv) >= 2) {
11775 if ((to[-2] == '\r' && to[-1] == '\n') ||
11776 (to[-2] == '\n' && to[-1] == '\r'))
11780 SvCUR_set(sv, to - SvPVX_const(sv));
11782 else if (to[-1] == '\r')
11785 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11790 /* if we're out of file, or a read fails, bail and reset the current
11791 line marker so we can report where the unterminated string began
11794 if (PL_madskills) {
11795 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11797 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11799 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11803 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11805 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11811 /* we read a line, so increment our line counter */
11812 CopLINE_inc(PL_curcop);
11814 /* update debugger info */
11815 if (PERLDB_LINE && PL_curstash != PL_debstash)
11816 update_debugger_info(PL_linestr, NULL, 0);
11818 /* having changed the buffer, we must update PL_bufend */
11819 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11820 PL_last_lop = PL_last_uni = NULL;
11823 /* at this point, we have successfully read the delimited string */
11825 if (!PL_encoding || UTF) {
11827 if (PL_madskills) {
11828 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11829 const int len = s - tstart;
11831 sv_catpvn(PL_thisstuff, tstart, len);
11833 PL_thisstuff = newSVpvn(tstart, len);
11834 if (!PL_thisclose && !keep_delims)
11835 PL_thisclose = newSVpvn(s,termlen);
11840 sv_catpvn(sv, s, termlen);
11845 if (PL_madskills) {
11846 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11847 const int len = s - tstart - termlen;
11849 sv_catpvn(PL_thisstuff, tstart, len);
11851 PL_thisstuff = newSVpvn(tstart, len);
11852 if (!PL_thisclose && !keep_delims)
11853 PL_thisclose = newSVpvn(s - termlen,termlen);
11857 if (has_utf8 || PL_encoding)
11860 PL_multi_end = CopLINE(PL_curcop);
11862 /* if we allocated too much space, give some back */
11863 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11864 SvLEN_set(sv, SvCUR(sv) + 1);
11865 SvPV_renew(sv, SvLEN(sv));
11868 /* decide whether this is the first or second quoted string we've read
11881 takes: pointer to position in buffer
11882 returns: pointer to new position in buffer
11883 side-effects: builds ops for the constant in yylval.op
11885 Read a number in any of the formats that Perl accepts:
11887 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11888 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11891 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11893 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11896 If it reads a number without a decimal point or an exponent, it will
11897 try converting the number to an integer and see if it can do so
11898 without loss of precision.
11902 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11905 register const char *s = start; /* current position in buffer */
11906 register char *d; /* destination in temp buffer */
11907 register char *e; /* end of temp buffer */
11908 NV nv; /* number read, as a double */
11909 SV *sv = NULL; /* place to put the converted number */
11910 bool floatit; /* boolean: int or float? */
11911 const char *lastub = NULL; /* position of last underbar */
11912 static char const number_too_long[] = "Number too long";
11914 /* We use the first character to decide what type of number this is */
11918 Perl_croak(aTHX_ "panic: scan_num");
11920 /* if it starts with a 0, it could be an octal number, a decimal in
11921 0.13 disguise, or a hexadecimal number, or a binary number. */
11925 u holds the "number so far"
11926 shift the power of 2 of the base
11927 (hex == 4, octal == 3, binary == 1)
11928 overflowed was the number more than we can hold?
11930 Shift is used when we add a digit. It also serves as an "are
11931 we in octal/hex/binary?" indicator to disallow hex characters
11932 when in octal mode.
11937 bool overflowed = FALSE;
11938 bool just_zero = TRUE; /* just plain 0 or binary number? */
11939 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11940 static const char* const bases[5] =
11941 { "", "binary", "", "octal", "hexadecimal" };
11942 static const char* const Bases[5] =
11943 { "", "Binary", "", "Octal", "Hexadecimal" };
11944 static const char* const maxima[5] =
11946 "0b11111111111111111111111111111111",
11950 const char *base, *Base, *max;
11952 /* check for hex */
11957 } else if (s[1] == 'b') {
11962 /* check for a decimal in disguise */
11963 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11965 /* so it must be octal */
11972 if (ckWARN(WARN_SYNTAX))
11973 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11974 "Misplaced _ in number");
11978 base = bases[shift];
11979 Base = Bases[shift];
11980 max = maxima[shift];
11982 /* read the rest of the number */
11984 /* x is used in the overflow test,
11985 b is the digit we're adding on. */
11990 /* if we don't mention it, we're done */
11994 /* _ are ignored -- but warned about if consecutive */
11996 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11997 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11998 "Misplaced _ in number");
12002 /* 8 and 9 are not octal */
12003 case '8': case '9':
12005 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12009 case '2': case '3': case '4':
12010 case '5': case '6': case '7':
12012 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12015 case '0': case '1':
12016 b = *s++ & 15; /* ASCII digit -> value of digit */
12020 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12021 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12022 /* make sure they said 0x */
12025 b = (*s++ & 7) + 9;
12027 /* Prepare to put the digit we have onto the end
12028 of the number so far. We check for overflows.
12034 x = u << shift; /* make room for the digit */
12036 if ((x >> shift) != u
12037 && !(PL_hints & HINT_NEW_BINARY)) {
12040 if (ckWARN_d(WARN_OVERFLOW))
12041 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12042 "Integer overflow in %s number",
12045 u = x | b; /* add the digit to the end */
12048 n *= nvshift[shift];
12049 /* If an NV has not enough bits in its
12050 * mantissa to represent an UV this summing of
12051 * small low-order numbers is a waste of time
12052 * (because the NV cannot preserve the
12053 * low-order bits anyway): we could just
12054 * remember when did we overflow and in the
12055 * end just multiply n by the right
12063 /* if we get here, we had success: make a scalar value from
12068 /* final misplaced underbar check */
12069 if (s[-1] == '_') {
12070 if (ckWARN(WARN_SYNTAX))
12071 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12076 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12077 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12078 "%s number > %s non-portable",
12084 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12085 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12086 "%s number > %s non-portable",
12091 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12092 sv = new_constant(start, s - start, "integer",
12093 sv, NULL, NULL, 0);
12094 else if (PL_hints & HINT_NEW_BINARY)
12095 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12100 handle decimal numbers.
12101 we're also sent here when we read a 0 as the first digit
12103 case '1': case '2': case '3': case '4': case '5':
12104 case '6': case '7': case '8': case '9': case '.':
12107 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12110 /* read next group of digits and _ and copy into d */
12111 while (isDIGIT(*s) || *s == '_') {
12112 /* skip underscores, checking for misplaced ones
12116 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12117 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12118 "Misplaced _ in number");
12122 /* check for end of fixed-length buffer */
12124 Perl_croak(aTHX_ number_too_long);
12125 /* if we're ok, copy the character */
12130 /* final misplaced underbar check */
12131 if (lastub && s == lastub + 1) {
12132 if (ckWARN(WARN_SYNTAX))
12133 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12136 /* read a decimal portion if there is one. avoid
12137 3..5 being interpreted as the number 3. followed
12140 if (*s == '.' && s[1] != '.') {
12145 if (ckWARN(WARN_SYNTAX))
12146 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12147 "Misplaced _ in number");
12151 /* copy, ignoring underbars, until we run out of digits.
12153 for (; isDIGIT(*s) || *s == '_'; s++) {
12154 /* fixed length buffer check */
12156 Perl_croak(aTHX_ number_too_long);
12158 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12159 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12160 "Misplaced _ in number");
12166 /* fractional part ending in underbar? */
12167 if (s[-1] == '_') {
12168 if (ckWARN(WARN_SYNTAX))
12169 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12170 "Misplaced _ in number");
12172 if (*s == '.' && isDIGIT(s[1])) {
12173 /* oops, it's really a v-string, but without the "v" */
12179 /* read exponent part, if present */
12180 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12184 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12185 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12187 /* stray preinitial _ */
12189 if (ckWARN(WARN_SYNTAX))
12190 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12191 "Misplaced _ in number");
12195 /* allow positive or negative exponent */
12196 if (*s == '+' || *s == '-')
12199 /* stray initial _ */
12201 if (ckWARN(WARN_SYNTAX))
12202 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12203 "Misplaced _ in number");
12207 /* read digits of exponent */
12208 while (isDIGIT(*s) || *s == '_') {
12211 Perl_croak(aTHX_ number_too_long);
12215 if (((lastub && s == lastub + 1) ||
12216 (!isDIGIT(s[1]) && s[1] != '_'))
12217 && ckWARN(WARN_SYNTAX))
12218 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12219 "Misplaced _ in number");
12226 /* make an sv from the string */
12230 We try to do an integer conversion first if no characters
12231 indicating "float" have been found.
12236 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12238 if (flags == IS_NUMBER_IN_UV) {
12240 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12243 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12244 if (uv <= (UV) IV_MIN)
12245 sv_setiv(sv, -(IV)uv);
12252 /* terminate the string */
12254 nv = Atof(PL_tokenbuf);
12259 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12260 const char *const key = floatit ? "float" : "integer";
12261 const STRLEN keylen = floatit ? 5 : 7;
12262 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12263 key, keylen, sv, NULL, NULL, 0);
12267 /* if it starts with a v, it could be a v-string */
12270 sv = newSV(5); /* preallocate storage space */
12271 s = scan_vstring(s, PL_bufend, sv);
12275 /* make the op for the constant and return */
12278 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12280 lvalp->opval = NULL;
12286 S_scan_formline(pTHX_ register char *s)
12289 register char *eol;
12291 SV * const stuff = newSVpvs("");
12292 bool needargs = FALSE;
12293 bool eofmt = FALSE;
12295 char *tokenstart = s;
12298 if (PL_madskills) {
12299 savewhite = PL_thiswhite;
12304 while (!needargs) {
12307 #ifdef PERL_STRICT_CR
12308 while (SPACE_OR_TAB(*t))
12311 while (SPACE_OR_TAB(*t) || *t == '\r')
12314 if (*t == '\n' || t == PL_bufend) {
12319 if (PL_in_eval && !PL_rsfp) {
12320 eol = (char *) memchr(s,'\n',PL_bufend-s);
12325 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12327 for (t = s; t < eol; t++) {
12328 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12330 goto enough; /* ~~ must be first line in formline */
12332 if (*t == '@' || *t == '^')
12336 sv_catpvn(stuff, s, eol-s);
12337 #ifndef PERL_STRICT_CR
12338 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12339 char *end = SvPVX(stuff) + SvCUR(stuff);
12342 SvCUR_set(stuff, SvCUR(stuff) - 1);
12352 if (PL_madskills) {
12354 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12356 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12359 s = filter_gets(PL_linestr, PL_rsfp, 0);
12361 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12363 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12365 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12366 PL_last_lop = PL_last_uni = NULL;
12375 if (SvCUR(stuff)) {
12378 PL_lex_state = LEX_NORMAL;
12379 start_force(PL_curforce);
12380 NEXTVAL_NEXTTOKE.ival = 0;
12384 PL_lex_state = LEX_FORMLINE;
12386 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12388 else if (PL_encoding)
12389 sv_recode_to_utf8(stuff, PL_encoding);
12391 start_force(PL_curforce);
12392 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12394 start_force(PL_curforce);
12395 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12399 SvREFCNT_dec(stuff);
12401 PL_lex_formbrack = 0;
12405 if (PL_madskills) {
12407 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12409 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12410 PL_thiswhite = savewhite;
12417 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12420 const I32 oldsavestack_ix = PL_savestack_ix;
12421 CV* const outsidecv = PL_compcv;
12424 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12426 SAVEI32(PL_subline);
12427 save_item(PL_subname);
12428 SAVESPTR(PL_compcv);
12430 PL_compcv = (CV*)newSV_type(is_format ? SVt_PVFM : SVt_PVCV);
12431 CvFLAGS(PL_compcv) |= flags;
12433 PL_subline = CopLINE(PL_curcop);
12434 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12435 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12436 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12438 return oldsavestack_ix;
12442 #pragma segment Perl_yylex
12445 Perl_yywarn(pTHX_ const char *s)
12448 PL_in_eval |= EVAL_WARNONLY;
12450 PL_in_eval &= ~EVAL_WARNONLY;
12455 Perl_yyerror(pTHX_ const char *s)
12458 const char *where = NULL;
12459 const char *context = NULL;
12462 int yychar = PL_parser->yychar;
12464 if (!yychar || (yychar == ';' && !PL_rsfp))
12466 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12467 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12468 PL_oldbufptr != PL_bufptr) {
12471 The code below is removed for NetWare because it abends/crashes on NetWare
12472 when the script has error such as not having the closing quotes like:
12473 if ($var eq "value)
12474 Checking of white spaces is anyway done in NetWare code.
12477 while (isSPACE(*PL_oldoldbufptr))
12480 context = PL_oldoldbufptr;
12481 contlen = PL_bufptr - PL_oldoldbufptr;
12483 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12484 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12487 The code below is removed for NetWare because it abends/crashes on NetWare
12488 when the script has error such as not having the closing quotes like:
12489 if ($var eq "value)
12490 Checking of white spaces is anyway done in NetWare code.
12493 while (isSPACE(*PL_oldbufptr))
12496 context = PL_oldbufptr;
12497 contlen = PL_bufptr - PL_oldbufptr;
12499 else if (yychar > 255)
12500 where = "next token ???";
12501 else if (yychar == -2) { /* YYEMPTY */
12502 if (PL_lex_state == LEX_NORMAL ||
12503 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12504 where = "at end of line";
12505 else if (PL_lex_inpat)
12506 where = "within pattern";
12508 where = "within string";
12511 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12513 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12514 else if (isPRINT_LC(yychar)) {
12515 const char string = yychar;
12516 sv_catpvn(where_sv, &string, 1);
12519 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12520 where = SvPVX_const(where_sv);
12522 msg = sv_2mortal(newSVpv(s, 0));
12523 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12524 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12526 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12528 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12529 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12530 Perl_sv_catpvf(aTHX_ msg,
12531 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12532 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12535 if (PL_in_eval & EVAL_WARNONLY) {
12536 if (ckWARN_d(WARN_SYNTAX))
12537 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12541 if (PL_error_count >= 10) {
12542 if (PL_in_eval && SvCUR(ERRSV))
12543 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12544 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12546 Perl_croak(aTHX_ "%s has too many errors.\n",
12547 OutCopFILE(PL_curcop));
12550 PL_in_my_stash = NULL;
12554 #pragma segment Main
12558 S_swallow_bom(pTHX_ U8 *s)
12561 const STRLEN slen = SvCUR(PL_linestr);
12564 if (s[1] == 0xFE) {
12565 /* UTF-16 little-endian? (or UTF32-LE?) */
12566 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12567 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12568 #ifndef PERL_NO_UTF16_FILTER
12569 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12572 if (PL_bufend > (char*)s) {
12576 filter_add(utf16rev_textfilter, NULL);
12577 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12578 utf16_to_utf8_reversed(s, news,
12579 PL_bufend - (char*)s - 1,
12581 sv_setpvn(PL_linestr, (const char*)news, newlen);
12583 s = (U8*)SvPVX(PL_linestr);
12584 Copy(news, s, newlen, U8);
12588 SvUTF8_on(PL_linestr);
12589 s = (U8*)SvPVX(PL_linestr);
12591 /* FIXME - is this a general bug fix? */
12594 PL_bufend = SvPVX(PL_linestr) + newlen;
12597 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12602 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12603 #ifndef PERL_NO_UTF16_FILTER
12604 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12607 if (PL_bufend > (char *)s) {
12611 filter_add(utf16_textfilter, NULL);
12612 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12613 utf16_to_utf8(s, news,
12614 PL_bufend - (char*)s,
12616 sv_setpvn(PL_linestr, (const char*)news, newlen);
12618 SvUTF8_on(PL_linestr);
12619 s = (U8*)SvPVX(PL_linestr);
12620 PL_bufend = SvPVX(PL_linestr) + newlen;
12623 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12628 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12629 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12630 s += 3; /* UTF-8 */
12636 if (s[2] == 0xFE && s[3] == 0xFF) {
12637 /* UTF-32 big-endian */
12638 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12641 else if (s[2] == 0 && s[3] != 0) {
12644 * are a good indicator of UTF-16BE. */
12645 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12651 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12652 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12653 s += 4; /* UTF-8 */
12659 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12662 * are a good indicator of UTF-16LE. */
12663 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12671 #ifndef PERL_NO_UTF16_FILTER
12673 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12676 const STRLEN old = SvCUR(sv);
12677 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12678 DEBUG_P(PerlIO_printf(Perl_debug_log,
12679 "utf16_textfilter(%p): %d %d (%d)\n",
12680 FPTR2DPTR(void *, utf16_textfilter),
12681 idx, maxlen, (int) count));
12685 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12686 Copy(SvPVX_const(sv), tmps, old, char);
12687 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12688 SvCUR(sv) - old, &newlen);
12689 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12691 DEBUG_P({sv_dump(sv);});
12696 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12699 const STRLEN old = SvCUR(sv);
12700 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12701 DEBUG_P(PerlIO_printf(Perl_debug_log,
12702 "utf16rev_textfilter(%p): %d %d (%d)\n",
12703 FPTR2DPTR(void *, utf16rev_textfilter),
12704 idx, maxlen, (int) count));
12708 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12709 Copy(SvPVX_const(sv), tmps, old, char);
12710 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12711 SvCUR(sv) - old, &newlen);
12712 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12714 DEBUG_P({ sv_dump(sv); });
12720 Returns a pointer to the next character after the parsed
12721 vstring, as well as updating the passed in sv.
12723 Function must be called like
12726 s = scan_vstring(s,e,sv);
12728 where s and e are the start and end of the string.
12729 The sv should already be large enough to store the vstring
12730 passed in, for performance reasons.
12735 Perl_scan_vstring(pTHX_ const char *s, const char *e, SV *sv)
12738 const char *pos = s;
12739 const char *start = s;
12740 if (*pos == 'v') pos++; /* get past 'v' */
12741 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12743 if ( *pos != '.') {
12744 /* this may not be a v-string if followed by => */
12745 const char *next = pos;
12746 while (next < e && isSPACE(*next))
12748 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12749 /* return string not v-string */
12750 sv_setpvn(sv,(char *)s,pos-s);
12751 return (char *)pos;
12755 if (!isALPHA(*pos)) {
12756 U8 tmpbuf[UTF8_MAXBYTES+1];
12759 s++; /* get past 'v' */
12761 sv_setpvn(sv, "", 0);
12764 /* this is atoi() that tolerates underscores */
12767 const char *end = pos;
12769 while (--end >= s) {
12771 const UV orev = rev;
12772 rev += (*end - '0') * mult;
12774 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12775 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12776 "Integer overflow in decimal number");
12780 if (rev > 0x7FFFFFFF)
12781 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12783 /* Append native character for the rev point */
12784 tmpend = uvchr_to_utf8(tmpbuf, rev);
12785 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12786 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12788 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12794 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12798 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12806 * c-indentation-style: bsd
12807 * c-basic-offset: 4
12808 * indent-tabs-mode: t
12811 * ex: set ts=8 sts=4 sw=4 noet: