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 const char * const cf = CopFILE(PL_curcop);
828 STRLEN tmplen = cf ? strlen(cf) : 0;
829 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
830 /* must copy *{"::_<(eval N)[oldfilename:L]"}
831 * to *{"::_<newfilename"} */
832 /* However, the long form of evals is only turned on by the
833 debugger - usually they're "(eval %lu)" */
837 STRLEN tmplen2 = len;
838 if (tmplen + 2 <= sizeof smallbuf)
841 Newx(tmpbuf, tmplen + 2, char);
844 memcpy(tmpbuf + 2, cf, tmplen);
846 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
851 if (tmplen2 + 2 <= sizeof smallbuf)
854 Newx(tmpbuf2, tmplen2 + 2, char);
856 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
857 /* Either they malloc'd it, or we malloc'd it,
858 so no prefix is present in ours. */
863 memcpy(tmpbuf2 + 2, s, tmplen2);
866 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
868 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
869 /* adjust ${"::_<newfilename"} to store the new file name */
870 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
871 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
872 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
875 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
877 if (tmpbuf != smallbuf) Safefree(tmpbuf);
880 CopFILE_free(PL_curcop);
881 CopFILE_setn(PL_curcop, s, len);
883 CopLINE_set(PL_curcop, atoi(n)-1);
887 /* skip space before PL_thistoken */
890 S_skipspace0(pTHX_ register char *s)
897 PL_thiswhite = newSVpvs("");
898 sv_catsv(PL_thiswhite, PL_skipwhite);
899 sv_free(PL_skipwhite);
902 PL_realtokenstart = s - SvPVX(PL_linestr);
906 /* skip space after PL_thistoken */
909 S_skipspace1(pTHX_ register char *s)
911 const char *start = s;
912 I32 startoff = start - SvPVX(PL_linestr);
917 start = SvPVX(PL_linestr) + startoff;
918 if (!PL_thistoken && PL_realtokenstart >= 0) {
919 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
920 PL_thistoken = newSVpvn(tstart, start - tstart);
922 PL_realtokenstart = -1;
925 PL_nextwhite = newSVpvs("");
926 sv_catsv(PL_nextwhite, PL_skipwhite);
927 sv_free(PL_skipwhite);
934 S_skipspace2(pTHX_ register char *s, SV **svp)
937 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
938 const I32 startoff = s - SvPVX(PL_linestr);
941 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
942 if (!PL_madskills || !svp)
944 start = SvPVX(PL_linestr) + startoff;
945 if (!PL_thistoken && PL_realtokenstart >= 0) {
946 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
947 PL_thistoken = newSVpvn(tstart, start - tstart);
948 PL_realtokenstart = -1;
953 sv_setsv(*svp, PL_skipwhite);
954 sv_free(PL_skipwhite);
963 S_update_debugger_info(pTHX_ SV *orig_sv, const char *buf, STRLEN len)
965 AV *av = CopFILEAVx(PL_curcop);
967 SV * const sv = newSV_type(SVt_PVMG);
969 sv_setsv(sv, orig_sv);
971 sv_setpvn(sv, buf, len);
974 av_store(av, (I32)CopLINE(PL_curcop), sv);
980 * Called to gobble the appropriate amount and type of whitespace.
981 * Skips comments as well.
985 S_skipspace(pTHX_ register char *s)
990 int startoff = s - SvPVX(PL_linestr);
993 sv_free(PL_skipwhite);
998 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
999 while (s < PL_bufend && SPACE_OR_TAB(*s))
1009 SSize_t oldprevlen, oldoldprevlen;
1010 SSize_t oldloplen = 0, oldunilen = 0;
1011 while (s < PL_bufend && isSPACE(*s)) {
1012 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
1017 if (s < PL_bufend && *s == '#') {
1018 while (s < PL_bufend && *s != '\n')
1020 if (s < PL_bufend) {
1022 if (PL_in_eval && !PL_rsfp) {
1029 /* only continue to recharge the buffer if we're at the end
1030 * of the buffer, we're not reading from a source filter, and
1031 * we're in normal lexing mode
1033 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1034 PL_lex_state == LEX_FORMLINE)
1041 /* try to recharge the buffer */
1043 curoff = s - SvPVX(PL_linestr);
1046 if ((s = filter_gets(PL_linestr, PL_rsfp,
1047 (prevlen = SvCUR(PL_linestr)))) == NULL)
1050 if (PL_madskills && curoff != startoff) {
1052 PL_skipwhite = newSVpvs("");
1053 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1057 /* mustn't throw out old stuff yet if madpropping */
1058 SvCUR(PL_linestr) = curoff;
1059 s = SvPVX(PL_linestr) + curoff;
1061 if (curoff && s[-1] == '\n')
1065 /* end of file. Add on the -p or -n magic */
1066 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1069 sv_catpvs(PL_linestr,
1070 ";}continue{print or die qq(-p destination: $!\\n);}");
1072 sv_setpvs(PL_linestr,
1073 ";}continue{print or die qq(-p destination: $!\\n);}");
1075 PL_minus_n = PL_minus_p = 0;
1077 else if (PL_minus_n) {
1079 sv_catpvn(PL_linestr, ";}", 2);
1081 sv_setpvn(PL_linestr, ";}", 2);
1087 sv_catpvn(PL_linestr,";", 1);
1089 sv_setpvn(PL_linestr,";", 1);
1092 /* reset variables for next time we lex */
1093 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1099 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1100 PL_last_lop = PL_last_uni = NULL;
1102 /* Close the filehandle. Could be from -P preprocessor,
1103 * STDIN, or a regular file. If we were reading code from
1104 * STDIN (because the commandline held no -e or filename)
1105 * then we don't close it, we reset it so the code can
1106 * read from STDIN too.
1109 if (PL_preprocess && !PL_in_eval)
1110 (void)PerlProc_pclose(PL_rsfp);
1111 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1112 PerlIO_clearerr(PL_rsfp);
1114 (void)PerlIO_close(PL_rsfp);
1119 /* not at end of file, so we only read another line */
1120 /* make corresponding updates to old pointers, for yyerror() */
1121 oldprevlen = PL_oldbufptr - PL_bufend;
1122 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1124 oldunilen = PL_last_uni - PL_bufend;
1126 oldloplen = PL_last_lop - PL_bufend;
1127 PL_linestart = PL_bufptr = s + prevlen;
1128 PL_bufend = s + SvCUR(PL_linestr);
1130 PL_oldbufptr = s + oldprevlen;
1131 PL_oldoldbufptr = s + oldoldprevlen;
1133 PL_last_uni = s + oldunilen;
1135 PL_last_lop = s + oldloplen;
1138 /* debugger active and we're not compiling the debugger code,
1139 * so store the line into the debugger's array of lines
1141 if (PERLDB_LINE && PL_curstash != PL_debstash)
1142 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1149 PL_skipwhite = newSVpvs("");
1150 curoff = s - SvPVX(PL_linestr);
1151 if (curoff - startoff)
1152 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1161 * Check the unary operators to ensure there's no ambiguity in how they're
1162 * used. An ambiguous piece of code would be:
1164 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1165 * the +5 is its argument.
1175 if (PL_oldoldbufptr != PL_last_uni)
1177 while (isSPACE(*PL_last_uni))
1180 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1182 if ((t = strchr(s, '(')) && t < PL_bufptr)
1185 if (ckWARN_d(WARN_AMBIGUOUS)){
1186 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1187 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1188 (int)(s - PL_last_uni), PL_last_uni);
1193 * LOP : macro to build a list operator. Its behaviour has been replaced
1194 * with a subroutine, S_lop() for which LOP is just another name.
1197 #define LOP(f,x) return lop(f,x,s)
1201 * Build a list operator (or something that might be one). The rules:
1202 * - if we have a next token, then it's a list operator [why?]
1203 * - if the next thing is an opening paren, then it's a function
1204 * - else it's a list operator
1208 S_lop(pTHX_ I32 f, int x, char *s)
1215 PL_last_lop = PL_oldbufptr;
1216 PL_last_lop_op = (OPCODE)f;
1219 return REPORT(LSTOP);
1222 return REPORT(LSTOP);
1225 return REPORT(FUNC);
1228 return REPORT(FUNC);
1230 return REPORT(LSTOP);
1236 * Sets up for an eventual force_next(). start_force(0) basically does
1237 * an unshift, while start_force(-1) does a push. yylex removes items
1242 S_start_force(pTHX_ int where)
1246 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1247 where = PL_lasttoke;
1248 assert(PL_curforce < 0 || PL_curforce == where);
1249 if (PL_curforce != where) {
1250 for (i = PL_lasttoke; i > where; --i) {
1251 PL_nexttoke[i] = PL_nexttoke[i-1];
1255 if (PL_curforce < 0) /* in case of duplicate start_force() */
1256 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1257 PL_curforce = where;
1260 curmad('^', newSVpvs(""));
1261 CURMAD('_', PL_nextwhite);
1266 S_curmad(pTHX_ char slot, SV *sv)
1272 if (PL_curforce < 0)
1273 where = &PL_thismad;
1275 where = &PL_nexttoke[PL_curforce].next_mad;
1278 sv_setpvn(sv, "", 0);
1281 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1283 else if (PL_encoding) {
1284 sv_recode_to_utf8(sv, PL_encoding);
1289 /* keep a slot open for the head of the list? */
1290 if (slot != '_' && *where && (*where)->mad_key == '^') {
1291 (*where)->mad_key = slot;
1292 sv_free((SV*)((*where)->mad_val));
1293 (*where)->mad_val = (void*)sv;
1296 addmad(newMADsv(slot, sv), where, 0);
1299 # define start_force(where) NOOP
1300 # define curmad(slot, sv) NOOP
1305 * When the lexer realizes it knows the next token (for instance,
1306 * it is reordering tokens for the parser) then it can call S_force_next
1307 * to know what token to return the next time the lexer is called. Caller
1308 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1309 * and possibly PL_expect to ensure the lexer handles the token correctly.
1313 S_force_next(pTHX_ I32 type)
1317 if (PL_curforce < 0)
1318 start_force(PL_lasttoke);
1319 PL_nexttoke[PL_curforce].next_type = type;
1320 if (PL_lex_state != LEX_KNOWNEXT)
1321 PL_lex_defer = PL_lex_state;
1322 PL_lex_state = LEX_KNOWNEXT;
1323 PL_lex_expect = PL_expect;
1326 PL_nexttype[PL_nexttoke] = type;
1328 if (PL_lex_state != LEX_KNOWNEXT) {
1329 PL_lex_defer = PL_lex_state;
1330 PL_lex_expect = PL_expect;
1331 PL_lex_state = LEX_KNOWNEXT;
1337 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1340 SV * const sv = newSVpvn(start,len);
1341 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1348 * When the lexer knows the next thing is a word (for instance, it has
1349 * just seen -> and it knows that the next char is a word char, then
1350 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1354 * char *start : buffer position (must be within PL_linestr)
1355 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1356 * int check_keyword : if true, Perl checks to make sure the word isn't
1357 * a keyword (do this if the word is a label, e.g. goto FOO)
1358 * int allow_pack : if true, : characters will also be allowed (require,
1359 * use, etc. do this)
1360 * int allow_initial_tick : used by the "sub" lexer only.
1364 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1370 start = SKIPSPACE1(start);
1372 if (isIDFIRST_lazy_if(s,UTF) ||
1373 (allow_pack && *s == ':') ||
1374 (allow_initial_tick && *s == '\'') )
1376 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1377 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1379 start_force(PL_curforce);
1381 curmad('X', newSVpvn(start,s-start));
1382 if (token == METHOD) {
1387 PL_expect = XOPERATOR;
1391 curmad('g', newSVpvs( "forced" ));
1392 NEXTVAL_NEXTTOKE.opval
1393 = (OP*)newSVOP(OP_CONST,0,
1394 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1395 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1403 * Called when the lexer wants $foo *foo &foo etc, but the program
1404 * text only contains the "foo" portion. The first argument is a pointer
1405 * to the "foo", and the second argument is the type symbol to prefix.
1406 * Forces the next token to be a "WORD".
1407 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1411 S_force_ident(pTHX_ register const char *s, int kind)
1415 const STRLEN len = strlen(s);
1416 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1417 start_force(PL_curforce);
1418 NEXTVAL_NEXTTOKE.opval = o;
1421 o->op_private = OPpCONST_ENTERED;
1422 /* XXX see note in pp_entereval() for why we forgo typo
1423 warnings if the symbol must be introduced in an eval.
1425 gv_fetchpvn_flags(s, len,
1426 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1428 kind == '$' ? SVt_PV :
1429 kind == '@' ? SVt_PVAV :
1430 kind == '%' ? SVt_PVHV :
1438 Perl_str_to_version(pTHX_ SV *sv)
1443 const char *start = SvPV_const(sv,len);
1444 const char * const end = start + len;
1445 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1446 while (start < end) {
1450 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1455 retval += ((NV)n)/nshift;
1464 * Forces the next token to be a version number.
1465 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1466 * and if "guessing" is TRUE, then no new token is created (and the caller
1467 * must use an alternative parsing method).
1471 S_force_version(pTHX_ char *s, int guessing)
1477 I32 startoff = s - SvPVX(PL_linestr);
1486 while (isDIGIT(*d) || *d == '_' || *d == '.')
1490 start_force(PL_curforce);
1491 curmad('X', newSVpvn(s,d-s));
1494 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1496 s = scan_num(s, &yylval);
1497 version = yylval.opval;
1498 ver = cSVOPx(version)->op_sv;
1499 if (SvPOK(ver) && !SvNIOK(ver)) {
1500 SvUPGRADE(ver, SVt_PVNV);
1501 SvNV_set(ver, str_to_version(ver));
1502 SvNOK_on(ver); /* hint that it is a version */
1505 else if (guessing) {
1508 sv_free(PL_nextwhite); /* let next token collect whitespace */
1510 s = SvPVX(PL_linestr) + startoff;
1518 if (PL_madskills && !version) {
1519 sv_free(PL_nextwhite); /* let next token collect whitespace */
1521 s = SvPVX(PL_linestr) + startoff;
1524 /* NOTE: The parser sees the package name and the VERSION swapped */
1525 start_force(PL_curforce);
1526 NEXTVAL_NEXTTOKE.opval = version;
1534 * Tokenize a quoted string passed in as an SV. It finds the next
1535 * chunk, up to end of string or a backslash. It may make a new
1536 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1541 S_tokeq(pTHX_ SV *sv)
1545 register char *send;
1553 s = SvPV_force(sv, len);
1554 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1557 while (s < send && *s != '\\')
1562 if ( PL_hints & HINT_NEW_STRING ) {
1563 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1569 if (s + 1 < send && (s[1] == '\\'))
1570 s++; /* all that, just for this */
1575 SvCUR_set(sv, d - SvPVX_const(sv));
1577 if ( PL_hints & HINT_NEW_STRING )
1578 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
1583 * Now come three functions related to double-quote context,
1584 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1585 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1586 * interact with PL_lex_state, and create fake ( ... ) argument lists
1587 * to handle functions and concatenation.
1588 * They assume that whoever calls them will be setting up a fake
1589 * join call, because each subthing puts a ',' after it. This lets
1592 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1594 * (I'm not sure whether the spurious commas at the end of lcfirst's
1595 * arguments and join's arguments are created or not).
1600 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1602 * Pattern matching will set PL_lex_op to the pattern-matching op to
1603 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1605 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1607 * Everything else becomes a FUNC.
1609 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1610 * had an OP_CONST or OP_READLINE). This just sets us up for a
1611 * call to S_sublex_push().
1615 S_sublex_start(pTHX)
1618 register const I32 op_type = yylval.ival;
1620 if (op_type == OP_NULL) {
1621 yylval.opval = PL_lex_op;
1625 if (op_type == OP_CONST || op_type == OP_READLINE) {
1626 SV *sv = tokeq(PL_lex_stuff);
1628 if (SvTYPE(sv) == SVt_PVIV) {
1629 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1631 const char * const p = SvPV_const(sv, len);
1632 SV * const nsv = newSVpvn(p, len);
1638 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1639 PL_lex_stuff = NULL;
1640 /* Allow <FH> // "foo" */
1641 if (op_type == OP_READLINE)
1642 PL_expect = XTERMORDORDOR;
1645 else if (op_type == OP_BACKTICK && PL_lex_op) {
1646 /* readpipe() vas overriden */
1647 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1648 yylval.opval = PL_lex_op;
1650 PL_lex_stuff = NULL;
1654 PL_sublex_info.super_state = PL_lex_state;
1655 PL_sublex_info.sub_inwhat = (U16)op_type;
1656 PL_sublex_info.sub_op = PL_lex_op;
1657 PL_lex_state = LEX_INTERPPUSH;
1661 yylval.opval = PL_lex_op;
1671 * Create a new scope to save the lexing state. The scope will be
1672 * ended in S_sublex_done. Returns a '(', starting the function arguments
1673 * to the uc, lc, etc. found before.
1674 * Sets PL_lex_state to LEX_INTERPCONCAT.
1683 PL_lex_state = PL_sublex_info.super_state;
1684 SAVEBOOL(PL_lex_dojoin);
1685 SAVEI32(PL_lex_brackets);
1686 SAVEI32(PL_lex_casemods);
1687 SAVEI32(PL_lex_starts);
1688 SAVEI8(PL_lex_state);
1689 SAVEVPTR(PL_lex_inpat);
1690 SAVEI16(PL_lex_inwhat);
1691 SAVECOPLINE(PL_curcop);
1692 SAVEPPTR(PL_bufptr);
1693 SAVEPPTR(PL_bufend);
1694 SAVEPPTR(PL_oldbufptr);
1695 SAVEPPTR(PL_oldoldbufptr);
1696 SAVEPPTR(PL_last_lop);
1697 SAVEPPTR(PL_last_uni);
1698 SAVEPPTR(PL_linestart);
1699 SAVESPTR(PL_linestr);
1700 SAVEGENERICPV(PL_lex_brackstack);
1701 SAVEGENERICPV(PL_lex_casestack);
1703 PL_linestr = PL_lex_stuff;
1704 PL_lex_stuff = NULL;
1706 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1707 = SvPVX(PL_linestr);
1708 PL_bufend += SvCUR(PL_linestr);
1709 PL_last_lop = PL_last_uni = NULL;
1710 SAVEFREESV(PL_linestr);
1712 PL_lex_dojoin = FALSE;
1713 PL_lex_brackets = 0;
1714 Newx(PL_lex_brackstack, 120, char);
1715 Newx(PL_lex_casestack, 12, char);
1716 PL_lex_casemods = 0;
1717 *PL_lex_casestack = '\0';
1719 PL_lex_state = LEX_INTERPCONCAT;
1720 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1722 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1723 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1724 PL_lex_inpat = PL_sublex_info.sub_op;
1726 PL_lex_inpat = NULL;
1733 * Restores lexer state after a S_sublex_push.
1740 if (!PL_lex_starts++) {
1741 SV * const sv = newSVpvs("");
1742 if (SvUTF8(PL_linestr))
1744 PL_expect = XOPERATOR;
1745 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1749 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1750 PL_lex_state = LEX_INTERPCASEMOD;
1754 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1755 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1756 PL_linestr = PL_lex_repl;
1758 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1759 PL_bufend += SvCUR(PL_linestr);
1760 PL_last_lop = PL_last_uni = NULL;
1761 SAVEFREESV(PL_linestr);
1762 PL_lex_dojoin = FALSE;
1763 PL_lex_brackets = 0;
1764 PL_lex_casemods = 0;
1765 *PL_lex_casestack = '\0';
1767 if (SvEVALED(PL_lex_repl)) {
1768 PL_lex_state = LEX_INTERPNORMAL;
1770 /* we don't clear PL_lex_repl here, so that we can check later
1771 whether this is an evalled subst; that means we rely on the
1772 logic to ensure sublex_done() is called again only via the
1773 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1776 PL_lex_state = LEX_INTERPCONCAT;
1786 PL_endwhite = newSVpvs("");
1787 sv_catsv(PL_endwhite, PL_thiswhite);
1791 sv_setpvn(PL_thistoken,"",0);
1793 PL_realtokenstart = -1;
1797 PL_bufend = SvPVX(PL_linestr);
1798 PL_bufend += SvCUR(PL_linestr);
1799 PL_expect = XOPERATOR;
1800 PL_sublex_info.sub_inwhat = 0;
1808 Extracts a pattern, double-quoted string, or transliteration. This
1811 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1812 processing a pattern (PL_lex_inpat is true), a transliteration
1813 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1815 Returns a pointer to the character scanned up to. If this is
1816 advanced from the start pointer supplied (i.e. if anything was
1817 successfully parsed), will leave an OP for the substring scanned
1818 in yylval. Caller must intuit reason for not parsing further
1819 by looking at the next characters herself.
1823 double-quoted style: \r and \n
1824 regexp special ones: \D \s
1827 case and quoting: \U \Q \E
1828 stops on @ and $, but not for $ as tail anchor
1830 In transliterations:
1831 characters are VERY literal, except for - not at the start or end
1832 of the string, which indicates a range. If the range is in bytes,
1833 scan_const expands the range to the full set of intermediate
1834 characters. If the range is in utf8, the hyphen is replaced with
1835 a certain range mark which will be handled by pmtrans() in op.c.
1837 In double-quoted strings:
1839 double-quoted style: \r and \n
1841 deprecated backrefs: \1 (in substitution replacements)
1842 case and quoting: \U \Q \E
1845 scan_const does *not* construct ops to handle interpolated strings.
1846 It stops processing as soon as it finds an embedded $ or @ variable
1847 and leaves it to the caller to work out what's going on.
1849 embedded arrays (whether in pattern or not) could be:
1850 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1852 $ in double-quoted strings must be the symbol of an embedded scalar.
1854 $ in pattern could be $foo or could be tail anchor. Assumption:
1855 it's a tail anchor if $ is the last thing in the string, or if it's
1856 followed by one of "()| \r\n\t"
1858 \1 (backreferences) are turned into $1
1860 The structure of the code is
1861 while (there's a character to process) {
1862 handle transliteration ranges
1863 skip regexp comments /(?#comment)/ and codes /(?{code})/
1864 skip #-initiated comments in //x patterns
1865 check for embedded arrays
1866 check for embedded scalars
1868 leave intact backslashes from leaveit (below)
1869 deprecate \1 in substitution replacements
1870 handle string-changing backslashes \l \U \Q \E, etc.
1871 switch (what was escaped) {
1872 handle \- in a transliteration (becomes a literal -)
1873 handle \132 (octal characters)
1874 handle \x15 and \x{1234} (hex characters)
1875 handle \N{name} (named characters)
1876 handle \cV (control characters)
1877 handle printf-style backslashes (\f, \r, \n, etc)
1879 } (end if backslash)
1880 } (end while character to read)
1885 S_scan_const(pTHX_ char *start)
1888 register char *send = PL_bufend; /* end of the constant */
1889 SV *sv = newSV(send - start); /* sv for the constant */
1890 register char *s = start; /* start of the constant */
1891 register char *d = SvPVX(sv); /* destination for copies */
1892 bool dorange = FALSE; /* are we in a translit range? */
1893 bool didrange = FALSE; /* did we just finish a range? */
1894 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1895 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1898 UV literal_endpoint = 0;
1899 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1902 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1903 /* If we are doing a trans and we know we want UTF8 set expectation */
1904 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1905 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1909 while (s < send || dorange) {
1910 /* get transliterations out of the way (they're most literal) */
1911 if (PL_lex_inwhat == OP_TRANS) {
1912 /* expand a range A-Z to the full set of characters. AIE! */
1914 I32 i; /* current expanded character */
1915 I32 min; /* first character in range */
1916 I32 max; /* last character in range */
1927 char * const c = (char*)utf8_hop((U8*)d, -1);
1931 *c = (char)UTF_TO_NATIVE(0xff);
1932 /* mark the range as done, and continue */
1938 i = d - SvPVX_const(sv); /* remember current offset */
1941 SvLEN(sv) + (has_utf8 ?
1942 (512 - UTF_CONTINUATION_MARK +
1945 /* How many two-byte within 0..255: 128 in UTF-8,
1946 * 96 in UTF-8-mod. */
1948 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1950 d = SvPVX(sv) + i; /* refresh d after realloc */
1954 for (j = 0; j <= 1; j++) {
1955 char * const c = (char*)utf8_hop((U8*)d, -1);
1956 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1962 max = (U8)0xff; /* only to \xff */
1963 uvmax = uv; /* \x{100} to uvmax */
1965 d = c; /* eat endpoint chars */
1970 d -= 2; /* eat the first char and the - */
1971 min = (U8)*d; /* first char in range */
1972 max = (U8)d[1]; /* last char in range */
1979 "Invalid range \"%c-%c\" in transliteration operator",
1980 (char)min, (char)max);
1984 if (literal_endpoint == 2 &&
1985 ((isLOWER(min) && isLOWER(max)) ||
1986 (isUPPER(min) && isUPPER(max)))) {
1988 for (i = min; i <= max; i++)
1990 *d++ = NATIVE_TO_NEED(has_utf8,i);
1992 for (i = min; i <= max; i++)
1994 *d++ = NATIVE_TO_NEED(has_utf8,i);
1999 for (i = min; i <= max; i++)
2002 const U8 ch = (U8)NATIVE_TO_UTF(i);
2003 if (UNI_IS_INVARIANT(ch))
2006 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2007 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2016 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2018 *d++ = (char)UTF_TO_NATIVE(0xff);
2020 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2024 /* mark the range as done, and continue */
2028 literal_endpoint = 0;
2033 /* range begins (ignore - as first or last char) */
2034 else if (*s == '-' && s+1 < send && s != start) {
2036 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2043 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2053 literal_endpoint = 0;
2054 native_range = TRUE;
2059 /* if we get here, we're not doing a transliteration */
2061 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2062 except for the last char, which will be done separately. */
2063 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2065 while (s+1 < send && *s != ')')
2066 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2068 else if (s[2] == '{' /* This should match regcomp.c */
2069 || (s[2] == '?' && s[3] == '{'))
2072 char *regparse = s + (s[2] == '{' ? 3 : 4);
2075 while (count && (c = *regparse)) {
2076 if (c == '\\' && regparse[1])
2084 if (*regparse != ')')
2085 regparse--; /* Leave one char for continuation. */
2086 while (s < regparse)
2087 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2091 /* likewise skip #-initiated comments in //x patterns */
2092 else if (*s == '#' && PL_lex_inpat &&
2093 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2094 while (s+1 < send && *s != '\n')
2095 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2098 /* check for embedded arrays
2099 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2101 else if (*s == '@' && s[1]) {
2102 if (isALNUM_lazy_if(s+1,UTF))
2104 if (strchr(":'{$", s[1]))
2106 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2107 break; /* in regexp, neither @+ nor @- are interpolated */
2110 /* check for embedded scalars. only stop if we're sure it's a
2113 else if (*s == '$') {
2114 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2116 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2117 break; /* in regexp, $ might be tail anchor */
2120 /* End of else if chain - OP_TRANS rejoin rest */
2123 if (*s == '\\' && s+1 < send) {
2126 /* deprecate \1 in strings and substitution replacements */
2127 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2128 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2130 if (ckWARN(WARN_SYNTAX))
2131 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2136 /* string-change backslash escapes */
2137 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2141 /* skip any other backslash escapes in a pattern */
2142 else if (PL_lex_inpat) {
2143 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2144 goto default_action;
2147 /* if we get here, it's either a quoted -, or a digit */
2150 /* quoted - in transliterations */
2152 if (PL_lex_inwhat == OP_TRANS) {
2159 if ((isALPHA(*s) || isDIGIT(*s)) &&
2161 Perl_warner(aTHX_ packWARN(WARN_MISC),
2162 "Unrecognized escape \\%c passed through",
2164 /* default action is to copy the quoted character */
2165 goto default_action;
2168 /* \132 indicates an octal constant */
2169 case '0': case '1': case '2': case '3':
2170 case '4': case '5': case '6': case '7':
2174 uv = grok_oct(s, &len, &flags, NULL);
2177 goto NUM_ESCAPE_INSERT;
2179 /* \x24 indicates a hex constant */
2183 char* const e = strchr(s, '}');
2184 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2185 PERL_SCAN_DISALLOW_PREFIX;
2190 yyerror("Missing right brace on \\x{}");
2194 uv = grok_hex(s, &len, &flags, NULL);
2200 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2201 uv = grok_hex(s, &len, &flags, NULL);
2207 /* Insert oct or hex escaped character.
2208 * There will always enough room in sv since such
2209 * escapes will be longer than any UTF-8 sequence
2210 * they can end up as. */
2212 /* We need to map to chars to ASCII before doing the tests
2215 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2216 if (!has_utf8 && uv > 255) {
2217 /* Might need to recode whatever we have
2218 * accumulated so far if it contains any
2221 * (Can't we keep track of that and avoid
2222 * this rescan? --jhi)
2226 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2227 if (!NATIVE_IS_INVARIANT(*c)) {
2232 const STRLEN offset = d - SvPVX_const(sv);
2234 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2238 while (src >= (const U8 *)SvPVX_const(sv)) {
2239 if (!NATIVE_IS_INVARIANT(*src)) {
2240 const U8 ch = NATIVE_TO_ASCII(*src);
2241 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2242 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2252 if (has_utf8 || uv > 255) {
2253 d = (char*)uvchr_to_utf8((U8*)d, uv);
2255 if (PL_lex_inwhat == OP_TRANS &&
2256 PL_sublex_info.sub_op) {
2257 PL_sublex_info.sub_op->op_private |=
2258 (PL_lex_repl ? OPpTRANS_FROM_UTF
2262 if (uv > 255 && !dorange)
2263 native_range = FALSE;
2275 /* \N{LATIN SMALL LETTER A} is a named character */
2279 char* e = strchr(s, '}');
2285 yyerror("Missing right brace on \\N{}");
2289 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2291 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2292 PERL_SCAN_DISALLOW_PREFIX;
2295 uv = grok_hex(s, &len, &flags, NULL);
2296 if ( e > s && len != (STRLEN)(e - s) ) {
2300 goto NUM_ESCAPE_INSERT;
2302 res = newSVpvn(s + 1, e - s - 1);
2303 res = new_constant( NULL, 0, "charnames",
2304 res, NULL, s - 2, e - s + 3 );
2306 sv_utf8_upgrade(res);
2307 str = SvPV_const(res,len);
2308 #ifdef EBCDIC_NEVER_MIND
2309 /* charnames uses pack U and that has been
2310 * recently changed to do the below uni->native
2311 * mapping, so this would be redundant (and wrong,
2312 * the code point would be doubly converted).
2313 * But leave this in just in case the pack U change
2314 * gets revoked, but the semantics is still
2315 * desireable for charnames. --jhi */
2317 UV uv = utf8_to_uvchr((const U8*)str, 0);
2320 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2322 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2323 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2324 str = SvPV_const(res, len);
2328 if (!has_utf8 && SvUTF8(res)) {
2329 const char * const ostart = SvPVX_const(sv);
2330 SvCUR_set(sv, d - ostart);
2333 sv_utf8_upgrade(sv);
2334 /* this just broke our allocation above... */
2335 SvGROW(sv, (STRLEN)(send - start));
2336 d = SvPVX(sv) + SvCUR(sv);
2339 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2340 const char * const odest = SvPVX_const(sv);
2342 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2343 d = SvPVX(sv) + (d - odest);
2347 native_range = FALSE; /* \N{} is guessed to be Unicode */
2349 Copy(str, d, len, char);
2356 yyerror("Missing braces on \\N{}");
2359 /* \c is a control character */
2368 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2371 yyerror("Missing control char name in \\c");
2375 /* printf-style backslashes, formfeeds, newlines, etc */
2377 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2380 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2383 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2386 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2389 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2392 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2395 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2401 } /* end if (backslash) */
2408 /* If we started with encoded form, or already know we want it
2409 and then encode the next character */
2410 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2412 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2413 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2416 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2417 const STRLEN off = d - SvPVX_const(sv);
2418 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2420 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2423 if (uv > 255 && !dorange)
2424 native_range = FALSE;
2428 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2430 } /* while loop to process each character */
2432 /* terminate the string and set up the sv */
2434 SvCUR_set(sv, d - SvPVX_const(sv));
2435 if (SvCUR(sv) >= SvLEN(sv))
2436 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2439 if (PL_encoding && !has_utf8) {
2440 sv_recode_to_utf8(sv, PL_encoding);
2446 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2447 PL_sublex_info.sub_op->op_private |=
2448 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2452 /* shrink the sv if we allocated more than we used */
2453 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2454 SvPV_shrink_to_cur(sv);
2457 /* return the substring (via yylval) only if we parsed anything */
2458 if (s > PL_bufptr) {
2459 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
2460 const char *const key = PL_lex_inpat ? "qr" : "q";
2461 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
2465 if (PL_lex_inwhat == OP_TRANS) {
2468 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
2476 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
2479 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2486 * Returns TRUE if there's more to the expression (e.g., a subscript),
2489 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2491 * ->[ and ->{ return TRUE
2492 * { and [ outside a pattern are always subscripts, so return TRUE
2493 * if we're outside a pattern and it's not { or [, then return FALSE
2494 * if we're in a pattern and the first char is a {
2495 * {4,5} (any digits around the comma) returns FALSE
2496 * if we're in a pattern and the first char is a [
2498 * [SOMETHING] has a funky algorithm to decide whether it's a
2499 * character class or not. It has to deal with things like
2500 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2501 * anything else returns TRUE
2504 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2507 S_intuit_more(pTHX_ register char *s)
2510 if (PL_lex_brackets)
2512 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2514 if (*s != '{' && *s != '[')
2519 /* In a pattern, so maybe we have {n,m}. */
2536 /* On the other hand, maybe we have a character class */
2539 if (*s == ']' || *s == '^')
2542 /* this is terrifying, and it works */
2543 int weight = 2; /* let's weigh the evidence */
2545 unsigned char un_char = 255, last_un_char;
2546 const char * const send = strchr(s,']');
2547 char tmpbuf[sizeof PL_tokenbuf * 4];
2549 if (!send) /* has to be an expression */
2552 Zero(seen,256,char);
2555 else if (isDIGIT(*s)) {
2557 if (isDIGIT(s[1]) && s[2] == ']')
2563 for (; s < send; s++) {
2564 last_un_char = un_char;
2565 un_char = (unsigned char)*s;
2570 weight -= seen[un_char] * 10;
2571 if (isALNUM_lazy_if(s+1,UTF)) {
2573 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2574 len = (int)strlen(tmpbuf);
2575 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2580 else if (*s == '$' && s[1] &&
2581 strchr("[#!%*<>()-=",s[1])) {
2582 if (/*{*/ strchr("])} =",s[2]))
2591 if (strchr("wds]",s[1]))
2593 else if (seen[(U8)'\''] || seen[(U8)'"'])
2595 else if (strchr("rnftbxcav",s[1]))
2597 else if (isDIGIT(s[1])) {
2599 while (s[1] && isDIGIT(s[1]))
2609 if (strchr("aA01! ",last_un_char))
2611 if (strchr("zZ79~",s[1]))
2613 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2614 weight -= 5; /* cope with negative subscript */
2617 if (!isALNUM(last_un_char)
2618 && !(last_un_char == '$' || last_un_char == '@'
2619 || last_un_char == '&')
2620 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2625 if (keyword(tmpbuf, d - tmpbuf, 0))
2628 if (un_char == last_un_char + 1)
2630 weight -= seen[un_char];
2635 if (weight >= 0) /* probably a character class */
2645 * Does all the checking to disambiguate
2647 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2648 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2650 * First argument is the stuff after the first token, e.g. "bar".
2652 * Not a method if bar is a filehandle.
2653 * Not a method if foo is a subroutine prototyped to take a filehandle.
2654 * Not a method if it's really "Foo $bar"
2655 * Method if it's "foo $bar"
2656 * Not a method if it's really "print foo $bar"
2657 * Method if it's really "foo package::" (interpreted as package->foo)
2658 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2659 * Not a method if bar is a filehandle or package, but is quoted with
2664 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2667 char *s = start + (*start == '$');
2668 char tmpbuf[sizeof PL_tokenbuf];
2676 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2680 const char *proto = SvPVX_const(cv);
2691 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2692 /* start is the beginning of the possible filehandle/object,
2693 * and s is the end of it
2694 * tmpbuf is a copy of it
2697 if (*start == '$') {
2698 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
2699 isUPPER(*PL_tokenbuf))
2702 len = start - SvPVX(PL_linestr);
2706 start = SvPVX(PL_linestr) + len;
2710 return *s == '(' ? FUNCMETH : METHOD;
2712 if (!keyword(tmpbuf, len, 0)) {
2713 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2717 soff = s - SvPVX(PL_linestr);
2721 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2722 if (indirgv && GvCVu(indirgv))
2724 /* filehandle or package name makes it a method */
2725 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2727 soff = s - SvPVX(PL_linestr);
2730 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2731 return 0; /* no assumptions -- "=>" quotes bearword */
2733 start_force(PL_curforce);
2734 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2735 newSVpvn(tmpbuf,len));
2736 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2738 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2743 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2745 return *s == '(' ? FUNCMETH : METHOD;
2753 * Return a string of Perl code to load the debugger. If PERL5DB
2754 * is set, it will return the contents of that, otherwise a
2755 * compile-time require of perl5db.pl.
2763 const char * const pdb = PerlEnv_getenv("PERL5DB");
2767 SETERRNO(0,SS_NORMAL);
2768 return "BEGIN { require 'perl5db.pl' }";
2774 /* Encoded script support. filter_add() effectively inserts a
2775 * 'pre-processing' function into the current source input stream.
2776 * Note that the filter function only applies to the current source file
2777 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2779 * The datasv parameter (which may be NULL) can be used to pass
2780 * private data to this instance of the filter. The filter function
2781 * can recover the SV using the FILTER_DATA macro and use it to
2782 * store private buffers and state information.
2784 * The supplied datasv parameter is upgraded to a PVIO type
2785 * and the IoDIRP/IoANY field is used to store the function pointer,
2786 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2787 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2788 * private use must be set using malloc'd pointers.
2792 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2801 if (!PL_rsfp_filters)
2802 PL_rsfp_filters = newAV();
2805 SvUPGRADE(datasv, SVt_PVIO);
2806 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2807 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2808 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2809 FPTR2DPTR(void *, IoANY(datasv)),
2810 SvPV_nolen(datasv)));
2811 av_unshift(PL_rsfp_filters, 1);
2812 av_store(PL_rsfp_filters, 0, datasv) ;
2817 /* Delete most recently added instance of this filter function. */
2819 Perl_filter_del(pTHX_ filter_t funcp)
2825 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2826 FPTR2DPTR(void*, funcp)));
2828 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2830 /* if filter is on top of stack (usual case) just pop it off */
2831 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2832 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2833 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2834 IoANY(datasv) = (void *)NULL;
2835 sv_free(av_pop(PL_rsfp_filters));
2839 /* we need to search for the correct entry and clear it */
2840 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2844 /* Invoke the idxth filter function for the current rsfp. */
2845 /* maxlen 0 = read one text line */
2847 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2852 /* This API is bad. It should have been using unsigned int for maxlen.
2853 Not sure if we want to change the API, but if not we should sanity
2854 check the value here. */
2855 const unsigned int correct_length
2864 if (!PL_parser || !PL_rsfp_filters)
2866 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2867 /* Provide a default input filter to make life easy. */
2868 /* Note that we append to the line. This is handy. */
2869 DEBUG_P(PerlIO_printf(Perl_debug_log,
2870 "filter_read %d: from rsfp\n", idx));
2871 if (correct_length) {
2874 const int old_len = SvCUR(buf_sv);
2876 /* ensure buf_sv is large enough */
2877 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2878 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2879 correct_length)) <= 0) {
2880 if (PerlIO_error(PL_rsfp))
2881 return -1; /* error */
2883 return 0 ; /* end of file */
2885 SvCUR_set(buf_sv, old_len + len) ;
2888 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2889 if (PerlIO_error(PL_rsfp))
2890 return -1; /* error */
2892 return 0 ; /* end of file */
2895 return SvCUR(buf_sv);
2897 /* Skip this filter slot if filter has been deleted */
2898 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2899 DEBUG_P(PerlIO_printf(Perl_debug_log,
2900 "filter_read %d: skipped (filter deleted)\n",
2902 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2904 /* Get function pointer hidden within datasv */
2905 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2906 DEBUG_P(PerlIO_printf(Perl_debug_log,
2907 "filter_read %d: via function %p (%s)\n",
2908 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2909 /* Call function. The function is expected to */
2910 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2911 /* Return: <0:error, =0:eof, >0:not eof */
2912 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2916 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2919 #ifdef PERL_CR_FILTER
2920 if (!PL_rsfp_filters) {
2921 filter_add(S_cr_textfilter,NULL);
2924 if (PL_rsfp_filters) {
2926 SvCUR_set(sv, 0); /* start with empty line */
2927 if (FILTER_READ(0, sv, 0) > 0)
2928 return ( SvPVX(sv) ) ;
2933 return (sv_gets(sv, fp, append));
2937 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
2942 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2946 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2947 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2949 return GvHV(gv); /* Foo:: */
2952 /* use constant CLASS => 'MyClass' */
2953 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2954 if (gv && GvCV(gv)) {
2955 SV * const sv = cv_const_sv(GvCV(gv));
2957 pkgname = SvPV_const(sv, len);
2960 return gv_stashpvn(pkgname, len, 0);
2964 * S_readpipe_override
2965 * Check whether readpipe() is overriden, and generates the appropriate
2966 * optree, provided sublex_start() is called afterwards.
2969 S_readpipe_override(pTHX)
2972 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2973 yylval.ival = OP_BACKTICK;
2975 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2977 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2978 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
2979 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2981 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2982 append_elem(OP_LIST,
2983 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2984 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2991 * The intent of this yylex wrapper is to minimize the changes to the
2992 * tokener when we aren't interested in collecting madprops. It remains
2993 * to be seen how successful this strategy will be...
3000 char *s = PL_bufptr;
3002 /* make sure PL_thiswhite is initialized */
3006 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3007 if (PL_pending_ident)
3008 return S_pending_ident(aTHX);
3010 /* previous token ate up our whitespace? */
3011 if (!PL_lasttoke && PL_nextwhite) {
3012 PL_thiswhite = PL_nextwhite;
3016 /* isolate the token, and figure out where it is without whitespace */
3017 PL_realtokenstart = -1;
3021 assert(PL_curforce < 0);
3023 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3024 if (!PL_thistoken) {
3025 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3026 PL_thistoken = newSVpvs("");
3028 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3029 PL_thistoken = newSVpvn(tstart, s - tstart);
3032 if (PL_thismad) /* install head */
3033 CURMAD('X', PL_thistoken);
3036 /* last whitespace of a sublex? */
3037 if (optype == ')' && PL_endwhite) {
3038 CURMAD('X', PL_endwhite);
3043 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3044 if (!PL_thiswhite && !PL_endwhite && !optype) {
3045 sv_free(PL_thistoken);
3050 /* put off final whitespace till peg */
3051 if (optype == ';' && !PL_rsfp) {
3052 PL_nextwhite = PL_thiswhite;
3055 else if (PL_thisopen) {
3056 CURMAD('q', PL_thisopen);
3058 sv_free(PL_thistoken);
3062 /* Store actual token text as madprop X */
3063 CURMAD('X', PL_thistoken);
3067 /* add preceding whitespace as madprop _ */
3068 CURMAD('_', PL_thiswhite);
3072 /* add quoted material as madprop = */
3073 CURMAD('=', PL_thisstuff);
3077 /* add terminating quote as madprop Q */
3078 CURMAD('Q', PL_thisclose);
3082 /* special processing based on optype */
3086 /* opval doesn't need a TOKEN since it can already store mp */
3097 append_madprops(PL_thismad, yylval.opval, 0);
3105 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3114 /* remember any fake bracket that lexer is about to discard */
3115 if (PL_lex_brackets == 1 &&
3116 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3119 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3122 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3123 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3126 break; /* don't bother looking for trailing comment */
3135 /* attach a trailing comment to its statement instead of next token */
3139 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3141 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3143 if (*s == '\n' || *s == '#') {
3144 while (s < PL_bufend && *s != '\n')
3148 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3149 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3166 /* Create new token struct. Note: opvals return early above. */
3167 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3174 S_tokenize_use(pTHX_ int is_use, char *s) {
3176 if (PL_expect != XSTATE)
3177 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3178 is_use ? "use" : "no"));
3180 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3181 s = force_version(s, TRUE);
3182 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3183 start_force(PL_curforce);
3184 NEXTVAL_NEXTTOKE.opval = NULL;
3187 else if (*s == 'v') {
3188 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3189 s = force_version(s, FALSE);
3193 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3194 s = force_version(s, FALSE);
3196 yylval.ival = is_use;
3200 static const char* const exp_name[] =
3201 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3202 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3209 Works out what to call the token just pulled out of the input
3210 stream. The yacc parser takes care of taking the ops we return and
3211 stitching them into a tree.
3217 if read an identifier
3218 if we're in a my declaration
3219 croak if they tried to say my($foo::bar)
3220 build the ops for a my() declaration
3221 if it's an access to a my() variable
3222 are we in a sort block?
3223 croak if my($a); $a <=> $b
3224 build ops for access to a my() variable
3225 if in a dq string, and they've said @foo and we can't find @foo
3227 build ops for a bareword
3228 if we already built the token before, use it.
3233 #pragma segment Perl_yylex
3239 register char *s = PL_bufptr;
3244 /* orig_keyword, gvp, and gv are initialized here because
3245 * jump to the label just_a_word_zero can bypass their
3246 * initialization later. */
3247 I32 orig_keyword = 0;
3252 SV* tmp = newSVpvs("");
3253 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3254 (IV)CopLINE(PL_curcop),
3255 lex_state_names[PL_lex_state],
3256 exp_name[PL_expect],
3257 pv_display(tmp, s, strlen(s), 0, 60));
3260 /* check if there's an identifier for us to look at */
3261 if (PL_pending_ident)
3262 return REPORT(S_pending_ident(aTHX));
3264 /* no identifier pending identification */
3266 switch (PL_lex_state) {
3268 case LEX_NORMAL: /* Some compilers will produce faster */
3269 case LEX_INTERPNORMAL: /* code if we comment these out. */
3273 /* when we've already built the next token, just pull it out of the queue */
3277 yylval = PL_nexttoke[PL_lasttoke].next_val;
3279 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3280 PL_nexttoke[PL_lasttoke].next_mad = 0;
3281 if (PL_thismad && PL_thismad->mad_key == '_') {
3282 PL_thiswhite = (SV*)PL_thismad->mad_val;
3283 PL_thismad->mad_val = 0;
3284 mad_free(PL_thismad);
3289 PL_lex_state = PL_lex_defer;
3290 PL_expect = PL_lex_expect;
3291 PL_lex_defer = LEX_NORMAL;
3292 if (!PL_nexttoke[PL_lasttoke].next_type)
3297 yylval = PL_nextval[PL_nexttoke];
3299 PL_lex_state = PL_lex_defer;
3300 PL_expect = PL_lex_expect;
3301 PL_lex_defer = LEX_NORMAL;
3305 /* FIXME - can these be merged? */
3306 return(PL_nexttoke[PL_lasttoke].next_type);
3308 return REPORT(PL_nexttype[PL_nexttoke]);
3311 /* interpolated case modifiers like \L \U, including \Q and \E.
3312 when we get here, PL_bufptr is at the \
3314 case LEX_INTERPCASEMOD:
3316 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3317 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3319 /* handle \E or end of string */
3320 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3322 if (PL_lex_casemods) {
3323 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3324 PL_lex_casestack[PL_lex_casemods] = '\0';
3326 if (PL_bufptr != PL_bufend
3327 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3329 PL_lex_state = LEX_INTERPCONCAT;
3332 PL_thistoken = newSVpvs("\\E");
3338 while (PL_bufptr != PL_bufend &&
3339 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3341 PL_thiswhite = newSVpvs("");
3342 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3346 if (PL_bufptr != PL_bufend)
3349 PL_lex_state = LEX_INTERPCONCAT;
3353 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3354 "### Saw case modifier\n"); });
3356 if (s[1] == '\\' && s[2] == 'E') {
3359 PL_thiswhite = newSVpvs("");
3360 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3363 PL_lex_state = LEX_INTERPCONCAT;
3368 if (!PL_madskills) /* when just compiling don't need correct */
3369 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3370 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3371 if ((*s == 'L' || *s == 'U') &&
3372 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3373 PL_lex_casestack[--PL_lex_casemods] = '\0';
3376 if (PL_lex_casemods > 10)
3377 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3378 PL_lex_casestack[PL_lex_casemods++] = *s;
3379 PL_lex_casestack[PL_lex_casemods] = '\0';
3380 PL_lex_state = LEX_INTERPCONCAT;
3381 start_force(PL_curforce);
3382 NEXTVAL_NEXTTOKE.ival = 0;
3384 start_force(PL_curforce);
3386 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3388 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3390 NEXTVAL_NEXTTOKE.ival = OP_LC;
3392 NEXTVAL_NEXTTOKE.ival = OP_UC;
3394 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3396 Perl_croak(aTHX_ "panic: yylex");
3398 SV* const tmpsv = newSVpvs("\\ ");
3399 /* replace the space with the character we want to escape
3401 SvPVX(tmpsv)[1] = *s;
3407 if (PL_lex_starts) {
3413 sv_free(PL_thistoken);
3414 PL_thistoken = newSVpvs("");
3417 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3418 if (PL_lex_casemods == 1 && PL_lex_inpat)
3427 case LEX_INTERPPUSH:
3428 return REPORT(sublex_push());
3430 case LEX_INTERPSTART:
3431 if (PL_bufptr == PL_bufend)
3432 return REPORT(sublex_done());
3433 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3434 "### Interpolated variable\n"); });
3436 PL_lex_dojoin = (*PL_bufptr == '@');
3437 PL_lex_state = LEX_INTERPNORMAL;
3438 if (PL_lex_dojoin) {
3439 start_force(PL_curforce);
3440 NEXTVAL_NEXTTOKE.ival = 0;
3442 start_force(PL_curforce);
3443 force_ident("\"", '$');
3444 start_force(PL_curforce);
3445 NEXTVAL_NEXTTOKE.ival = 0;
3447 start_force(PL_curforce);
3448 NEXTVAL_NEXTTOKE.ival = 0;
3450 start_force(PL_curforce);
3451 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3454 if (PL_lex_starts++) {
3459 sv_free(PL_thistoken);
3460 PL_thistoken = newSVpvs("");
3463 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3464 if (!PL_lex_casemods && PL_lex_inpat)
3471 case LEX_INTERPENDMAYBE:
3472 if (intuit_more(PL_bufptr)) {
3473 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3479 if (PL_lex_dojoin) {
3480 PL_lex_dojoin = FALSE;
3481 PL_lex_state = LEX_INTERPCONCAT;
3485 sv_free(PL_thistoken);
3486 PL_thistoken = newSVpvs("");
3491 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3492 && SvEVALED(PL_lex_repl))
3494 if (PL_bufptr != PL_bufend)
3495 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3499 case LEX_INTERPCONCAT:
3501 if (PL_lex_brackets)
3502 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3504 if (PL_bufptr == PL_bufend)
3505 return REPORT(sublex_done());
3507 if (SvIVX(PL_linestr) == '\'') {
3508 SV *sv = newSVsv(PL_linestr);
3511 else if ( PL_hints & HINT_NEW_RE )
3512 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
3513 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3517 s = scan_const(PL_bufptr);
3519 PL_lex_state = LEX_INTERPCASEMOD;
3521 PL_lex_state = LEX_INTERPSTART;
3524 if (s != PL_bufptr) {
3525 start_force(PL_curforce);
3527 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3529 NEXTVAL_NEXTTOKE = yylval;
3532 if (PL_lex_starts++) {
3536 sv_free(PL_thistoken);
3537 PL_thistoken = newSVpvs("");
3540 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3541 if (!PL_lex_casemods && PL_lex_inpat)
3554 PL_lex_state = LEX_NORMAL;
3555 s = scan_formline(PL_bufptr);
3556 if (!PL_lex_formbrack)
3562 PL_oldoldbufptr = PL_oldbufptr;
3568 sv_free(PL_thistoken);
3571 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3575 if (isIDFIRST_lazy_if(s,UTF))
3577 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3578 Perl_croak(aTHX_ "Unrecognized character \\x%02X in column %d", *s & 255, (int) len + 1);
3581 goto fake_eof; /* emulate EOF on ^D or ^Z */
3590 if (PL_lex_brackets) {
3591 yyerror((const char *)
3593 ? "Format not terminated"
3594 : "Missing right curly or square bracket"));
3596 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3597 "### Tokener got EOF\n");
3601 if (s++ < PL_bufend)
3602 goto retry; /* ignore stray nulls */
3605 if (!PL_in_eval && !PL_preambled) {
3606 PL_preambled = TRUE;
3611 sv_setpv(PL_linestr,incl_perldb());
3612 if (SvCUR(PL_linestr))
3613 sv_catpvs(PL_linestr,";");
3615 while(AvFILLp(PL_preambleav) >= 0) {
3616 SV *tmpsv = av_shift(PL_preambleav);
3617 sv_catsv(PL_linestr, tmpsv);
3618 sv_catpvs(PL_linestr, ";");
3621 sv_free((SV*)PL_preambleav);
3622 PL_preambleav = NULL;
3624 if (PL_minus_n || PL_minus_p) {
3625 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3627 sv_catpvs(PL_linestr,"chomp;");
3630 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3631 || *PL_splitstr == '"')
3632 && strchr(PL_splitstr + 1, *PL_splitstr))
3633 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3635 /* "q\0${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)) {
3687 sv_setpv(PL_linestr,
3690 ? ";}continue{print;}" : ";}"));
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 = newSVpv(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 == '$' || *proto == '_') && proto[1] == '\0')
5524 while (*proto == ';')
5526 if (*proto == '&' && *s == '{') {
5527 sv_setpv(PL_subname,
5530 "__ANON__" : "__ANON__::__ANON__"));
5537 PL_nextwhite = PL_thiswhite;
5540 start_force(PL_curforce);
5541 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5544 PL_nextwhite = nextPL_nextwhite;
5545 curmad('X', PL_thistoken);
5546 PL_thistoken = newSVpvs("");
5553 /* Guess harder when madskills require "best effort". */
5554 if (PL_madskills && (!gv || !GvCVu(gv))) {
5555 int probable_sub = 0;
5556 if (strchr("\"'`$@%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;
6751 sv_setpv(PL_subname,
6753 (PL_curstash ? "__ANON__" : "__ANON__::__ANON__"));
6757 (void) force_word(PL_oldbufptr + tboffset, WORD,
6766 LOP(OP_SYSTEM,XREF);
6769 LOP(OP_SYMLINK,XTERM);
6772 LOP(OP_SYSCALL,XTERM);
6775 LOP(OP_SYSOPEN,XTERM);
6778 LOP(OP_SYSSEEK,XTERM);
6781 LOP(OP_SYSREAD,XTERM);
6784 LOP(OP_SYSWRITE,XTERM);
6788 TERM(sublex_start());
6809 LOP(OP_TRUNCATE,XTERM);
6821 yylval.ival = CopLINE(PL_curcop);
6825 yylval.ival = CopLINE(PL_curcop);
6829 LOP(OP_UNLINK,XTERM);
6835 LOP(OP_UNPACK,XTERM);
6838 LOP(OP_UTIME,XTERM);
6844 LOP(OP_UNSHIFT,XTERM);
6847 s = tokenize_use(1, s);
6857 yylval.ival = CopLINE(PL_curcop);
6861 yylval.ival = CopLINE(PL_curcop);
6865 PL_hints |= HINT_BLOCK_SCOPE;
6872 LOP(OP_WAITPID,XTERM);
6881 ctl_l[0] = toCTRL('L');
6883 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6886 /* Make sure $^L is defined */
6887 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6892 if (PL_expect == XOPERATOR)
6898 yylval.ival = OP_XOR;
6903 TERM(sublex_start());
6908 #pragma segment Main
6912 S_pending_ident(pTHX)
6917 /* pit holds the identifier we read and pending_ident is reset */
6918 char pit = PL_pending_ident;
6919 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
6920 /* All routes through this function want to know if there is a colon. */
6921 const char *const has_colon = memchr (PL_tokenbuf, ':', tokenbuf_len);
6922 PL_pending_ident = 0;
6924 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6925 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6926 "### Pending identifier '%s'\n", PL_tokenbuf); });
6928 /* if we're in a my(), we can't allow dynamics here.
6929 $foo'bar has already been turned into $foo::bar, so
6930 just check for colons.
6932 if it's a legal name, the OP is a PADANY.
6935 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6937 yyerror(Perl_form(aTHX_ "No package name allowed for "
6938 "variable %s in \"our\"",
6940 tmp = allocmy(PL_tokenbuf);
6944 yyerror(Perl_form(aTHX_ PL_no_myglob,
6945 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6947 yylval.opval = newOP(OP_PADANY, 0);
6948 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6954 build the ops for accesses to a my() variable.
6956 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6957 then used in a comparison. This catches most, but not
6958 all cases. For instance, it catches
6959 sort { my($a); $a <=> $b }
6961 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6962 (although why you'd do that is anyone's guess).
6967 tmp = pad_findmy(PL_tokenbuf);
6968 if (tmp != NOT_IN_PAD) {
6969 /* might be an "our" variable" */
6970 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6971 /* build ops for a bareword */
6972 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6973 HEK * const stashname = HvNAME_HEK(stash);
6974 SV * const sym = newSVhek(stashname);
6975 sv_catpvs(sym, "::");
6976 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
6977 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6978 yylval.opval->op_private = OPpCONST_ENTERED;
6981 ? (GV_ADDMULTI | GV_ADDINEVAL)
6984 ((PL_tokenbuf[0] == '$') ? SVt_PV
6985 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6990 /* if it's a sort block and they're naming $a or $b */
6991 if (PL_last_lop_op == OP_SORT &&
6992 PL_tokenbuf[0] == '$' &&
6993 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6996 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6997 d < PL_bufend && *d != '\n';
7000 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7001 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7007 yylval.opval = newOP(OP_PADANY, 0);
7008 yylval.opval->op_targ = tmp;
7014 Whine if they've said @foo in a doublequoted string,
7015 and @foo isn't a variable we can find in the symbol
7018 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7019 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7021 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7022 && ckWARN(WARN_AMBIGUOUS)
7023 /* DO NOT warn for @- and @+ */
7024 && !( PL_tokenbuf[2] == '\0' &&
7025 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7028 /* Downgraded from fatal to warning 20000522 mjd */
7029 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7030 "Possible unintended interpolation of %s in string",
7035 /* build ops for a bareword */
7036 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7038 yylval.opval->op_private = OPpCONST_ENTERED;
7040 PL_tokenbuf + 1, tokenbuf_len - 1,
7041 /* If the identifier refers to a stash, don't autovivify it.
7042 * Change 24660 had the side effect of causing symbol table
7043 * hashes to always be defined, even if they were freshly
7044 * created and the only reference in the entire program was
7045 * the single statement with the defined %foo::bar:: test.
7046 * It appears that all code in the wild doing this actually
7047 * wants to know whether sub-packages have been loaded, so
7048 * by avoiding auto-vivifying symbol tables, we ensure that
7049 * defined %foo::bar:: continues to be false, and the existing
7050 * tests still give the expected answers, even though what
7051 * they're actually testing has now changed subtly.
7053 (*PL_tokenbuf == '%'
7054 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7057 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7058 ((PL_tokenbuf[0] == '$') ? SVt_PV
7059 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7065 * The following code was generated by perl_keyword.pl.
7069 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7074 case 1: /* 5 tokens of length 1 */
7106 case 2: /* 18 tokens of length 2 */
7252 case 3: /* 29 tokens of length 3 */
7256 if (name[1] == 'N' &&
7319 if (name[1] == 'i' &&
7351 if (name[1] == 'o' &&
7360 if (name[1] == 'e' &&
7369 if (name[1] == 'n' &&
7378 if (name[1] == 'o' &&
7387 if (name[1] == 'a' &&
7396 if (name[1] == 'o' &&
7458 if (name[1] == 'e' &&
7472 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7498 if (name[1] == 'i' &&
7507 if (name[1] == 's' &&
7516 if (name[1] == 'e' &&
7525 if (name[1] == 'o' &&
7537 case 4: /* 41 tokens of length 4 */
7541 if (name[1] == 'O' &&
7551 if (name[1] == 'N' &&
7561 if (name[1] == 'i' &&
7571 if (name[1] == 'h' &&
7581 if (name[1] == 'u' &&
7594 if (name[2] == 'c' &&
7603 if (name[2] == 's' &&
7612 if (name[2] == 'a' &&
7648 if (name[1] == 'o' &&
7661 if (name[2] == 't' &&
7670 if (name[2] == 'o' &&
7679 if (name[2] == 't' &&
7688 if (name[2] == 'e' &&
7701 if (name[1] == 'o' &&
7714 if (name[2] == 'y' &&
7723 if (name[2] == 'l' &&
7739 if (name[2] == 's' &&
7748 if (name[2] == 'n' &&
7757 if (name[2] == 'c' &&
7770 if (name[1] == 'e' &&
7780 if (name[1] == 'p' &&
7793 if (name[2] == 'c' &&
7802 if (name[2] == 'p' &&
7811 if (name[2] == 's' &&
7827 if (name[2] == 'n' &&
7897 if (name[2] == 'r' &&
7906 if (name[2] == 'r' &&
7915 if (name[2] == 'a' &&
7931 if (name[2] == 'l' &&
7993 if (name[2] == 'e' &&
7996 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8009 case 5: /* 39 tokens of length 5 */
8013 if (name[1] == 'E' &&
8024 if (name[1] == 'H' &&
8038 if (name[2] == 'a' &&
8048 if (name[2] == 'a' &&
8065 if (name[2] == 'e' &&
8075 if (name[2] == 'e' &&
8079 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8095 if (name[3] == 'i' &&
8104 if (name[3] == 'o' &&
8140 if (name[2] == 'o' &&
8150 if (name[2] == 'y' &&
8164 if (name[1] == 'l' &&
8178 if (name[2] == 'n' &&
8188 if (name[2] == 'o' &&
8202 if (name[1] == 'i' &&
8207 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8216 if (name[2] == 'd' &&
8226 if (name[2] == 'c' &&
8243 if (name[2] == 'c' &&
8253 if (name[2] == 't' &&
8267 if (name[1] == 'k' &&
8278 if (name[1] == 'r' &&
8292 if (name[2] == 's' &&
8302 if (name[2] == 'd' &&
8319 if (name[2] == 'm' &&
8329 if (name[2] == 'i' &&
8339 if (name[2] == 'e' &&
8349 if (name[2] == 'l' &&
8359 if (name[2] == 'a' &&
8372 if (name[3] == 't' &&
8375 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8381 if (name[3] == 'd' &&
8398 if (name[1] == 'i' &&
8412 if (name[2] == 'a' &&
8425 if (name[3] == 'e' &&
8460 if (name[2] == 'i' &&
8477 if (name[2] == 'i' &&
8487 if (name[2] == 'i' &&
8504 case 6: /* 33 tokens of length 6 */
8508 if (name[1] == 'c' &&
8523 if (name[2] == 'l' &&
8534 if (name[2] == 'r' &&
8549 if (name[1] == 'e' &&
8564 if (name[2] == 's' &&
8569 if(ckWARN_d(WARN_SYNTAX))
8570 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8576 if (name[2] == 'i' &&
8594 if (name[2] == 'l' &&
8605 if (name[2] == 'r' &&
8620 if (name[1] == 'm' &&
8635 if (name[2] == 'n' &&
8646 if (name[2] == 's' &&
8661 if (name[1] == 's' &&
8667 if (name[4] == 't' &&
8676 if (name[4] == 'e' &&
8685 if (name[4] == 'c' &&
8694 if (name[4] == 'n' &&
8710 if (name[1] == 'r' &&
8728 if (name[3] == 'a' &&
8738 if (name[3] == 'u' &&
8752 if (name[2] == 'n' &&
8770 if (name[2] == 'a' &&
8784 if (name[3] == 'e' &&
8797 if (name[4] == 't' &&
8806 if (name[4] == 'e' &&
8828 if (name[4] == 't' &&
8837 if (name[4] == 'e' &&
8853 if (name[2] == 'c' &&
8864 if (name[2] == 'l' &&
8875 if (name[2] == 'b' &&
8886 if (name[2] == 's' &&
8909 if (name[4] == 's' &&
8918 if (name[4] == 'n' &&
8931 if (name[3] == 'a' &&
8948 if (name[1] == 'a' &&
8963 case 7: /* 29 tokens of length 7 */
8967 if (name[1] == 'E' &&
8980 if (name[1] == '_' &&
8993 if (name[1] == 'i' &&
9000 return -KEY_binmode;
9006 if (name[1] == 'o' &&
9013 return -KEY_connect;
9022 if (name[2] == 'm' &&
9028 return -KEY_dbmopen;
9039 if (name[4] == 'u' &&
9043 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9049 if (name[4] == 'n' &&
9070 if (name[1] == 'o' &&
9083 if (name[1] == 'e' &&
9090 if (name[5] == 'r' &&
9093 return -KEY_getpgrp;
9099 if (name[5] == 'i' &&
9102 return -KEY_getppid;
9115 if (name[1] == 'c' &&
9122 return -KEY_lcfirst;
9128 if (name[1] == 'p' &&
9135 return -KEY_opendir;
9141 if (name[1] == 'a' &&
9159 if (name[3] == 'd' &&
9164 return -KEY_readdir;
9170 if (name[3] == 'u' &&
9181 if (name[3] == 'e' &&
9186 return -KEY_reverse;
9205 if (name[3] == 'k' &&
9210 return -KEY_seekdir;
9216 if (name[3] == 'p' &&
9221 return -KEY_setpgrp;
9231 if (name[2] == 'm' &&
9237 return -KEY_shmread;
9243 if (name[2] == 'r' &&
9249 return -KEY_sprintf;
9258 if (name[3] == 'l' &&
9263 return -KEY_symlink;
9272 if (name[4] == 'a' &&
9276 return -KEY_syscall;
9282 if (name[4] == 'p' &&
9286 return -KEY_sysopen;
9292 if (name[4] == 'e' &&
9296 return -KEY_sysread;
9302 if (name[4] == 'e' &&
9306 return -KEY_sysseek;
9324 if (name[1] == 'e' &&
9331 return -KEY_telldir;
9340 if (name[2] == 'f' &&
9346 return -KEY_ucfirst;
9352 if (name[2] == 's' &&
9358 return -KEY_unshift;
9368 if (name[1] == 'a' &&
9375 return -KEY_waitpid;
9384 case 8: /* 26 tokens of length 8 */
9388 if (name[1] == 'U' &&
9396 return KEY_AUTOLOAD;
9407 if (name[3] == 'A' &&
9413 return KEY___DATA__;
9419 if (name[3] == 'I' &&
9425 return -KEY___FILE__;
9431 if (name[3] == 'I' &&
9437 return -KEY___LINE__;
9453 if (name[2] == 'o' &&
9460 return -KEY_closedir;
9466 if (name[2] == 'n' &&
9473 return -KEY_continue;
9483 if (name[1] == 'b' &&
9491 return -KEY_dbmclose;
9497 if (name[1] == 'n' &&
9503 if (name[4] == 'r' &&
9508 return -KEY_endgrent;
9514 if (name[4] == 'w' &&
9519 return -KEY_endpwent;
9532 if (name[1] == 'o' &&
9540 return -KEY_formline;
9546 if (name[1] == 'e' &&
9557 if (name[6] == 'n' &&
9560 return -KEY_getgrent;
9566 if (name[6] == 'i' &&
9569 return -KEY_getgrgid;
9575 if (name[6] == 'a' &&
9578 return -KEY_getgrnam;
9591 if (name[4] == 'o' &&
9596 return -KEY_getlogin;
9607 if (name[6] == 'n' &&
9610 return -KEY_getpwent;
9616 if (name[6] == 'a' &&
9619 return -KEY_getpwnam;
9625 if (name[6] == 'i' &&
9628 return -KEY_getpwuid;
9648 if (name[1] == 'e' &&
9655 if (name[5] == 'i' &&
9662 return -KEY_readline;
9667 return -KEY_readlink;
9678 if (name[5] == 'i' &&
9682 return -KEY_readpipe;
9703 if (name[4] == 'r' &&
9708 return -KEY_setgrent;
9714 if (name[4] == 'w' &&
9719 return -KEY_setpwent;
9735 if (name[3] == 'w' &&
9741 return -KEY_shmwrite;
9747 if (name[3] == 't' &&
9753 return -KEY_shutdown;
9763 if (name[2] == 's' &&
9770 return -KEY_syswrite;
9780 if (name[1] == 'r' &&
9788 return -KEY_truncate;
9797 case 9: /* 9 tokens of length 9 */
9801 if (name[1] == 'N' &&
9810 return KEY_UNITCHECK;
9816 if (name[1] == 'n' &&
9825 return -KEY_endnetent;
9831 if (name[1] == 'e' &&
9840 return -KEY_getnetent;
9846 if (name[1] == 'o' &&
9855 return -KEY_localtime;
9861 if (name[1] == 'r' &&
9870 return KEY_prototype;
9876 if (name[1] == 'u' &&
9885 return -KEY_quotemeta;
9891 if (name[1] == 'e' &&
9900 return -KEY_rewinddir;
9906 if (name[1] == 'e' &&
9915 return -KEY_setnetent;
9921 if (name[1] == 'a' &&
9930 return -KEY_wantarray;
9939 case 10: /* 9 tokens of length 10 */
9943 if (name[1] == 'n' &&
9949 if (name[4] == 'o' &&
9956 return -KEY_endhostent;
9962 if (name[4] == 'e' &&
9969 return -KEY_endservent;
9982 if (name[1] == 'e' &&
9988 if (name[4] == 'o' &&
9995 return -KEY_gethostent;
10004 if (name[5] == 'r' &&
10010 return -KEY_getservent;
10016 if (name[5] == 'c' &&
10022 return -KEY_getsockopt;
10042 if (name[2] == 't')
10047 if (name[4] == 'o' &&
10054 return -KEY_sethostent;
10063 if (name[5] == 'r' &&
10069 return -KEY_setservent;
10075 if (name[5] == 'c' &&
10081 return -KEY_setsockopt;
10098 if (name[2] == 'c' &&
10107 return -KEY_socketpair;
10120 case 11: /* 8 tokens of length 11 */
10124 if (name[1] == '_' &&
10134 { /* __PACKAGE__ */
10135 return -KEY___PACKAGE__;
10141 if (name[1] == 'n' &&
10151 { /* endprotoent */
10152 return -KEY_endprotoent;
10158 if (name[1] == 'e' &&
10167 if (name[5] == 'e' &&
10173 { /* getpeername */
10174 return -KEY_getpeername;
10183 if (name[6] == 'o' &&
10188 { /* getpriority */
10189 return -KEY_getpriority;
10195 if (name[6] == 't' &&
10200 { /* getprotoent */
10201 return -KEY_getprotoent;
10215 if (name[4] == 'o' &&
10222 { /* getsockname */
10223 return -KEY_getsockname;
10236 if (name[1] == 'e' &&
10244 if (name[6] == 'o' &&
10249 { /* setpriority */
10250 return -KEY_setpriority;
10256 if (name[6] == 't' &&
10261 { /* setprotoent */
10262 return -KEY_setprotoent;
10278 case 12: /* 2 tokens of length 12 */
10279 if (name[0] == 'g' &&
10291 if (name[9] == 'd' &&
10294 { /* getnetbyaddr */
10295 return -KEY_getnetbyaddr;
10301 if (name[9] == 'a' &&
10304 { /* getnetbyname */
10305 return -KEY_getnetbyname;
10317 case 13: /* 4 tokens of length 13 */
10318 if (name[0] == 'g' &&
10325 if (name[4] == 'o' &&
10334 if (name[10] == 'd' &&
10337 { /* gethostbyaddr */
10338 return -KEY_gethostbyaddr;
10344 if (name[10] == 'a' &&
10347 { /* gethostbyname */
10348 return -KEY_gethostbyname;
10361 if (name[4] == 'e' &&
10370 if (name[10] == 'a' &&
10373 { /* getservbyname */
10374 return -KEY_getservbyname;
10380 if (name[10] == 'o' &&
10383 { /* getservbyport */
10384 return -KEY_getservbyport;
10403 case 14: /* 1 tokens of length 14 */
10404 if (name[0] == 'g' &&
10418 { /* getprotobyname */
10419 return -KEY_getprotobyname;
10424 case 16: /* 1 tokens of length 16 */
10425 if (name[0] == 'g' &&
10441 { /* getprotobynumber */
10442 return -KEY_getprotobynumber;
10456 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10460 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10461 if (ckWARN(WARN_SYNTAX)) {
10464 for (w = s+2; *w && level; w++) {
10467 else if (*w == ')')
10470 while (isSPACE(*w))
10472 /* the list of chars below is for end of statements or
10473 * block / parens, boolean operators (&&, ||, //) and branch
10474 * constructs (or, and, if, until, unless, while, err, for).
10475 * Not a very solid hack... */
10476 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10477 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10478 "%s (...) interpreted as function",name);
10481 while (s < PL_bufend && isSPACE(*s))
10485 while (s < PL_bufend && isSPACE(*s))
10487 if (isIDFIRST_lazy_if(s,UTF)) {
10488 const char * const w = s++;
10489 while (isALNUM_lazy_if(s,UTF))
10491 while (s < PL_bufend && isSPACE(*s))
10495 if (keyword(w, s - w, 0))
10498 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10499 if (gv && GvCVu(gv))
10501 Perl_croak(aTHX_ "No comma allowed after %s", what);
10506 /* Either returns sv, or mortalizes sv and returns a new SV*.
10507 Best used as sv=new_constant(..., sv, ...).
10508 If s, pv are NULL, calls subroutine with one argument,
10509 and type is used with error messages only. */
10512 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
10513 SV *sv, SV *pv, const char *type, STRLEN typelen)
10516 HV * const table = GvHV(PL_hintgv); /* ^H */
10520 const char *why1 = "", *why2 = "", *why3 = "";
10522 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10525 why2 = (const char *)
10526 (strEQ(key,"charnames")
10527 ? "(possibly a missing \"use charnames ...\")"
10529 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10530 (type ? type: "undef"), why2);
10532 /* This is convoluted and evil ("goto considered harmful")
10533 * but I do not understand the intricacies of all the different
10534 * failure modes of %^H in here. The goal here is to make
10535 * the most probable error message user-friendly. --jhi */
10540 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10541 (type ? type: "undef"), why1, why2, why3);
10543 yyerror(SvPVX_const(msg));
10547 cvp = hv_fetch(table, key, keylen, FALSE);
10548 if (!cvp || !SvOK(*cvp)) {
10551 why3 = "} is not defined";
10554 sv_2mortal(sv); /* Parent created it permanently */
10557 pv = sv_2mortal(newSVpvn(s, len));
10559 typesv = sv_2mortal(newSVpvn(type, typelen));
10561 typesv = &PL_sv_undef;
10563 PUSHSTACKi(PERLSI_OVERLOAD);
10575 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10579 /* Check the eval first */
10580 if (!PL_in_eval && SvTRUE(ERRSV)) {
10581 sv_catpvs(ERRSV, "Propagated");
10582 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10584 res = SvREFCNT_inc_simple(sv);
10588 SvREFCNT_inc_simple_void(res);
10597 why1 = "Call to &{$^H{";
10599 why3 = "}} did not return a defined value";
10607 /* Returns a NUL terminated string, with the length of the string written to
10611 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10614 register char *d = dest;
10615 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10618 Perl_croak(aTHX_ ident_too_long);
10619 if (isALNUM(*s)) /* UTF handled below */
10621 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10626 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10630 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10631 char *t = s + UTF8SKIP(s);
10633 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10637 Perl_croak(aTHX_ ident_too_long);
10638 Copy(s, d, len, char);
10651 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10654 char *bracket = NULL;
10656 register char *d = dest;
10657 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10662 while (isDIGIT(*s)) {
10664 Perl_croak(aTHX_ ident_too_long);
10671 Perl_croak(aTHX_ ident_too_long);
10672 if (isALNUM(*s)) /* UTF handled below */
10674 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10679 else if (*s == ':' && s[1] == ':') {
10683 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10684 char *t = s + UTF8SKIP(s);
10685 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10687 if (d + (t - s) > e)
10688 Perl_croak(aTHX_ ident_too_long);
10689 Copy(s, d, t - s, char);
10700 if (PL_lex_state != LEX_NORMAL)
10701 PL_lex_state = LEX_INTERPENDMAYBE;
10704 if (*s == '$' && s[1] &&
10705 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10718 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10723 if (isSPACE(s[-1])) {
10725 const char ch = *s++;
10726 if (!SPACE_OR_TAB(ch)) {
10732 if (isIDFIRST_lazy_if(d,UTF)) {
10736 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10737 end += UTF8SKIP(end);
10738 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10739 end += UTF8SKIP(end);
10741 Copy(s, d, end - s, char);
10746 while ((isALNUM(*s) || *s == ':') && d < e)
10749 Perl_croak(aTHX_ ident_too_long);
10752 while (s < send && SPACE_OR_TAB(*s))
10754 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10755 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10756 const char * const brack =
10758 ((*s == '[') ? "[...]" : "{...}");
10759 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10760 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10761 funny, dest, brack, funny, dest, brack);
10764 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10768 /* Handle extended ${^Foo} variables
10769 * 1999-02-27 mjd-perl-patch@plover.com */
10770 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10774 while (isALNUM(*s) && d < e) {
10778 Perl_croak(aTHX_ ident_too_long);
10783 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10784 PL_lex_state = LEX_INTERPEND;
10787 if (PL_lex_state == LEX_NORMAL) {
10788 if (ckWARN(WARN_AMBIGUOUS) &&
10789 (keyword(dest, d - dest, 0)
10790 || get_cvn_flags(dest, d - dest, 0)))
10794 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10795 "Ambiguous use of %c{%s} resolved to %c%s",
10796 funny, dest, funny, dest);
10801 s = bracket; /* let the parser handle it */
10805 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10806 PL_lex_state = LEX_INTERPEND;
10811 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10813 PERL_UNUSED_CONTEXT;
10817 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10818 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10819 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10820 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10821 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10827 S_scan_pat(pTHX_ char *start, I32 type)
10831 char *s = scan_str(start,!!PL_madskills,FALSE);
10832 const char * const valid_flags =
10833 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10840 const char * const delimiter = skipspace(start);
10844 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10845 : "Search pattern not terminated" ));
10848 pm = (PMOP*)newPMOP(type, 0);
10849 if (PL_multi_open == '?') {
10850 /* This is the only point in the code that sets PMf_ONCE: */
10851 pm->op_pmflags |= PMf_ONCE;
10853 /* Hence it's safe to do this bit of PMOP book-keeping here, which
10854 allows us to restrict the list needed by reset to just the ??
10856 assert(type != OP_TRANS);
10858 MAGIC *mg = mg_find((SV*)PL_curstash, PERL_MAGIC_symtab);
10861 mg = sv_magicext((SV*)PL_curstash, 0, PERL_MAGIC_symtab, 0, 0,
10864 elements = mg->mg_len / sizeof(PMOP**);
10865 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
10866 ((PMOP**)mg->mg_ptr) [elements++] = pm;
10867 mg->mg_len = elements * sizeof(PMOP**);
10868 PmopSTASH_set(pm,PL_curstash);
10874 while (*s && strchr(valid_flags, *s))
10875 pmflag(&pm->op_pmflags,*s++);
10877 if (PL_madskills && modstart != s) {
10878 SV* tmptoken = newSVpvn(modstart, s - modstart);
10879 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10882 /* issue a warning if /c is specified,but /g is not */
10883 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10884 && ckWARN(WARN_REGEXP))
10886 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
10887 "Use of /c modifier is meaningless without /g" );
10890 PL_lex_op = (OP*)pm;
10891 yylval.ival = OP_MATCH;
10896 S_scan_subst(pTHX_ char *start)
10907 yylval.ival = OP_NULL;
10909 s = scan_str(start,!!PL_madskills,FALSE);
10912 Perl_croak(aTHX_ "Substitution pattern not terminated");
10914 if (s[-1] == PL_multi_open)
10917 if (PL_madskills) {
10918 CURMAD('q', PL_thisopen);
10919 CURMAD('_', PL_thiswhite);
10920 CURMAD('E', PL_thisstuff);
10921 CURMAD('Q', PL_thisclose);
10922 PL_realtokenstart = s - SvPVX(PL_linestr);
10926 first_start = PL_multi_start;
10927 s = scan_str(s,!!PL_madskills,FALSE);
10929 if (PL_lex_stuff) {
10930 SvREFCNT_dec(PL_lex_stuff);
10931 PL_lex_stuff = NULL;
10933 Perl_croak(aTHX_ "Substitution replacement not terminated");
10935 PL_multi_start = first_start; /* so whole substitution is taken together */
10937 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10940 if (PL_madskills) {
10941 CURMAD('z', PL_thisopen);
10942 CURMAD('R', PL_thisstuff);
10943 CURMAD('Z', PL_thisclose);
10949 if (*s == EXEC_PAT_MOD) {
10953 else if (strchr(S_PAT_MODS, *s))
10954 pmflag(&pm->op_pmflags,*s++);
10960 if (PL_madskills) {
10962 curmad('m', newSVpvn(modstart, s - modstart));
10963 append_madprops(PL_thismad, (OP*)pm, 0);
10967 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10968 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10972 SV * const repl = newSVpvs("");
10974 PL_sublex_info.super_bufptr = s;
10975 PL_sublex_info.super_bufend = PL_bufend;
10977 pm->op_pmflags |= PMf_EVAL;
10980 sv_catpvs(repl, "eval ");
10982 sv_catpvs(repl, "do ");
10984 sv_catpvs(repl, "{");
10985 sv_catsv(repl, PL_lex_repl);
10986 if (strchr(SvPVX(PL_lex_repl), '#'))
10987 sv_catpvs(repl, "\n");
10988 sv_catpvs(repl, "}");
10990 SvREFCNT_dec(PL_lex_repl);
10991 PL_lex_repl = repl;
10994 PL_lex_op = (OP*)pm;
10995 yylval.ival = OP_SUBST;
11000 S_scan_trans(pTHX_ char *start)
11013 yylval.ival = OP_NULL;
11015 s = scan_str(start,!!PL_madskills,FALSE);
11017 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11019 if (s[-1] == PL_multi_open)
11022 if (PL_madskills) {
11023 CURMAD('q', PL_thisopen);
11024 CURMAD('_', PL_thiswhite);
11025 CURMAD('E', PL_thisstuff);
11026 CURMAD('Q', PL_thisclose);
11027 PL_realtokenstart = s - SvPVX(PL_linestr);
11031 s = scan_str(s,!!PL_madskills,FALSE);
11033 if (PL_lex_stuff) {
11034 SvREFCNT_dec(PL_lex_stuff);
11035 PL_lex_stuff = NULL;
11037 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11039 if (PL_madskills) {
11040 CURMAD('z', PL_thisopen);
11041 CURMAD('R', PL_thisstuff);
11042 CURMAD('Z', PL_thisclose);
11045 complement = del = squash = 0;
11052 complement = OPpTRANS_COMPLEMENT;
11055 del = OPpTRANS_DELETE;
11058 squash = OPpTRANS_SQUASH;
11067 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11068 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11069 o->op_private &= ~OPpTRANS_ALL;
11070 o->op_private |= del|squash|complement|
11071 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11072 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11075 yylval.ival = OP_TRANS;
11078 if (PL_madskills) {
11080 curmad('m', newSVpvn(modstart, s - modstart));
11081 append_madprops(PL_thismad, o, 0);
11090 S_scan_heredoc(pTHX_ register char *s)
11094 I32 op_type = OP_SCALAR;
11098 const char *found_newline;
11102 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11104 I32 stuffstart = s - SvPVX(PL_linestr);
11107 PL_realtokenstart = -1;
11112 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11116 while (SPACE_OR_TAB(*peek))
11118 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11121 s = delimcpy(d, e, s, PL_bufend, term, &len);
11131 if (!isALNUM_lazy_if(s,UTF))
11132 deprecate_old("bare << to mean <<\"\"");
11133 for (; isALNUM_lazy_if(s,UTF); s++) {
11138 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11139 Perl_croak(aTHX_ "Delimiter for here document is too long");
11142 len = d - PL_tokenbuf;
11145 if (PL_madskills) {
11146 tstart = PL_tokenbuf + !outer;
11147 PL_thisclose = newSVpvn(tstart, len - !outer);
11148 tstart = SvPVX(PL_linestr) + stuffstart;
11149 PL_thisopen = newSVpvn(tstart, s - tstart);
11150 stuffstart = s - SvPVX(PL_linestr);
11153 #ifndef PERL_STRICT_CR
11154 d = strchr(s, '\r');
11156 char * const olds = s;
11158 while (s < PL_bufend) {
11164 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11173 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11180 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11181 herewas = newSVpvn(s,PL_bufend-s);
11185 herewas = newSVpvn(s-1,found_newline-s+1);
11188 herewas = newSVpvn(s,found_newline-s);
11192 if (PL_madskills) {
11193 tstart = SvPVX(PL_linestr) + stuffstart;
11195 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11197 PL_thisstuff = newSVpvn(tstart, s - tstart);
11200 s += SvCUR(herewas);
11203 stuffstart = s - SvPVX(PL_linestr);
11209 tmpstr = newSV_type(SVt_PVIV);
11210 SvGROW(tmpstr, 80);
11211 if (term == '\'') {
11212 op_type = OP_CONST;
11213 SvIV_set(tmpstr, -1);
11215 else if (term == '`') {
11216 op_type = OP_BACKTICK;
11217 SvIV_set(tmpstr, '\\');
11221 PL_multi_start = CopLINE(PL_curcop);
11222 PL_multi_open = PL_multi_close = '<';
11223 term = *PL_tokenbuf;
11224 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11225 char * const bufptr = PL_sublex_info.super_bufptr;
11226 char * const bufend = PL_sublex_info.super_bufend;
11227 char * const olds = s - SvCUR(herewas);
11228 s = strchr(bufptr, '\n');
11232 while (s < bufend &&
11233 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11235 CopLINE_inc(PL_curcop);
11238 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11239 missingterm(PL_tokenbuf);
11241 sv_setpvn(herewas,bufptr,d-bufptr+1);
11242 sv_setpvn(tmpstr,d+1,s-d);
11244 sv_catpvn(herewas,s,bufend-s);
11245 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11252 while (s < PL_bufend &&
11253 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11255 CopLINE_inc(PL_curcop);
11257 if (s >= PL_bufend) {
11258 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11259 missingterm(PL_tokenbuf);
11261 sv_setpvn(tmpstr,d+1,s-d);
11263 if (PL_madskills) {
11265 sv_catpvn(PL_thisstuff, d + 1, s - d);
11267 PL_thisstuff = newSVpvn(d + 1, s - d);
11268 stuffstart = s - SvPVX(PL_linestr);
11272 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11274 sv_catpvn(herewas,s,PL_bufend-s);
11275 sv_setsv(PL_linestr,herewas);
11276 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11277 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11278 PL_last_lop = PL_last_uni = NULL;
11281 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11282 while (s >= PL_bufend) { /* multiple line string? */
11284 if (PL_madskills) {
11285 tstart = SvPVX(PL_linestr) + stuffstart;
11287 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11289 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11293 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11294 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11295 missingterm(PL_tokenbuf);
11298 stuffstart = s - SvPVX(PL_linestr);
11300 CopLINE_inc(PL_curcop);
11301 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11302 PL_last_lop = PL_last_uni = NULL;
11303 #ifndef PERL_STRICT_CR
11304 if (PL_bufend - PL_linestart >= 2) {
11305 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11306 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11308 PL_bufend[-2] = '\n';
11310 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11312 else if (PL_bufend[-1] == '\r')
11313 PL_bufend[-1] = '\n';
11315 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11316 PL_bufend[-1] = '\n';
11318 if (PERLDB_LINE && PL_curstash != PL_debstash)
11319 update_debugger_info(PL_linestr, NULL, 0);
11320 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11321 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11322 *(SvPVX(PL_linestr) + off ) = ' ';
11323 sv_catsv(PL_linestr,herewas);
11324 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11325 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11329 sv_catsv(tmpstr,PL_linestr);
11334 PL_multi_end = CopLINE(PL_curcop);
11335 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11336 SvPV_shrink_to_cur(tmpstr);
11338 SvREFCNT_dec(herewas);
11340 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11342 else if (PL_encoding)
11343 sv_recode_to_utf8(tmpstr, PL_encoding);
11345 PL_lex_stuff = tmpstr;
11346 yylval.ival = op_type;
11350 /* scan_inputsymbol
11351 takes: current position in input buffer
11352 returns: new position in input buffer
11353 side-effects: yylval and lex_op are set.
11358 <FH> read from filehandle
11359 <pkg::FH> read from package qualified filehandle
11360 <pkg'FH> read from package qualified filehandle
11361 <$fh> read from filehandle in $fh
11362 <*.h> filename glob
11367 S_scan_inputsymbol(pTHX_ char *start)
11370 register char *s = start; /* current position in buffer */
11374 char *d = PL_tokenbuf; /* start of temp holding space */
11375 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11377 end = strchr(s, '\n');
11380 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11382 /* die if we didn't have space for the contents of the <>,
11383 or if it didn't end, or if we see a newline
11386 if (len >= (I32)sizeof PL_tokenbuf)
11387 Perl_croak(aTHX_ "Excessively long <> operator");
11389 Perl_croak(aTHX_ "Unterminated <> operator");
11394 Remember, only scalar variables are interpreted as filehandles by
11395 this code. Anything more complex (e.g., <$fh{$num}>) will be
11396 treated as a glob() call.
11397 This code makes use of the fact that except for the $ at the front,
11398 a scalar variable and a filehandle look the same.
11400 if (*d == '$' && d[1]) d++;
11402 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11403 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11406 /* If we've tried to read what we allow filehandles to look like, and
11407 there's still text left, then it must be a glob() and not a getline.
11408 Use scan_str to pull out the stuff between the <> and treat it
11409 as nothing more than a string.
11412 if (d - PL_tokenbuf != len) {
11413 yylval.ival = OP_GLOB;
11414 s = scan_str(start,!!PL_madskills,FALSE);
11416 Perl_croak(aTHX_ "Glob not terminated");
11420 bool readline_overriden = FALSE;
11423 /* we're in a filehandle read situation */
11426 /* turn <> into <ARGV> */
11428 Copy("ARGV",d,5,char);
11430 /* Check whether readline() is overriden */
11431 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11433 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11435 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11436 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11437 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11438 readline_overriden = TRUE;
11440 /* if <$fh>, create the ops to turn the variable into a
11444 /* try to find it in the pad for this block, otherwise find
11445 add symbol table ops
11447 const PADOFFSET tmp = pad_findmy(d);
11448 if (tmp != NOT_IN_PAD) {
11449 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11450 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11451 HEK * const stashname = HvNAME_HEK(stash);
11452 SV * const sym = sv_2mortal(newSVhek(stashname));
11453 sv_catpvs(sym, "::");
11454 sv_catpv(sym, d+1);
11459 OP * const o = newOP(OP_PADSV, 0);
11461 PL_lex_op = readline_overriden
11462 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11463 append_elem(OP_LIST, o,
11464 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11465 : (OP*)newUNOP(OP_READLINE, 0, o);
11474 ? (GV_ADDMULTI | GV_ADDINEVAL)
11477 PL_lex_op = readline_overriden
11478 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11479 append_elem(OP_LIST,
11480 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11481 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11482 : (OP*)newUNOP(OP_READLINE, 0,
11483 newUNOP(OP_RV2SV, 0,
11484 newGVOP(OP_GV, 0, gv)));
11486 if (!readline_overriden)
11487 PL_lex_op->op_flags |= OPf_SPECIAL;
11488 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11489 yylval.ival = OP_NULL;
11492 /* If it's none of the above, it must be a literal filehandle
11493 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11495 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11496 PL_lex_op = readline_overriden
11497 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11498 append_elem(OP_LIST,
11499 newGVOP(OP_GV, 0, gv),
11500 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11501 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11502 yylval.ival = OP_NULL;
11511 takes: start position in buffer
11512 keep_quoted preserve \ on the embedded delimiter(s)
11513 keep_delims preserve the delimiters around the string
11514 returns: position to continue reading from buffer
11515 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11516 updates the read buffer.
11518 This subroutine pulls a string out of the input. It is called for:
11519 q single quotes q(literal text)
11520 ' single quotes 'literal text'
11521 qq double quotes qq(interpolate $here please)
11522 " double quotes "interpolate $here please"
11523 qx backticks qx(/bin/ls -l)
11524 ` backticks `/bin/ls -l`
11525 qw quote words @EXPORT_OK = qw( func() $spam )
11526 m// regexp match m/this/
11527 s/// regexp substitute s/this/that/
11528 tr/// string transliterate tr/this/that/
11529 y/// string transliterate y/this/that/
11530 ($*@) sub prototypes sub foo ($)
11531 (stuff) sub attr parameters sub foo : attr(stuff)
11532 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11534 In most of these cases (all but <>, patterns and transliterate)
11535 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11536 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11537 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11540 It skips whitespace before the string starts, and treats the first
11541 character as the delimiter. If the delimiter is one of ([{< then
11542 the corresponding "close" character )]}> is used as the closing
11543 delimiter. It allows quoting of delimiters, and if the string has
11544 balanced delimiters ([{<>}]) it allows nesting.
11546 On success, the SV with the resulting string is put into lex_stuff or,
11547 if that is already non-NULL, into lex_repl. The second case occurs only
11548 when parsing the RHS of the special constructs s/// and tr/// (y///).
11549 For convenience, the terminating delimiter character is stuffed into
11554 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11557 SV *sv; /* scalar value: string */
11558 const char *tmps; /* temp string, used for delimiter matching */
11559 register char *s = start; /* current position in the buffer */
11560 register char term; /* terminating character */
11561 register char *to; /* current position in the sv's data */
11562 I32 brackets = 1; /* bracket nesting level */
11563 bool has_utf8 = FALSE; /* is there any utf8 content? */
11564 I32 termcode; /* terminating char. code */
11565 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11566 STRLEN termlen; /* length of terminating string */
11567 int last_off = 0; /* last position for nesting bracket */
11573 /* skip space before the delimiter */
11579 if (PL_realtokenstart >= 0) {
11580 stuffstart = PL_realtokenstart;
11581 PL_realtokenstart = -1;
11584 stuffstart = start - SvPVX(PL_linestr);
11586 /* mark where we are, in case we need to report errors */
11589 /* after skipping whitespace, the next character is the terminator */
11592 termcode = termstr[0] = term;
11596 termcode = utf8_to_uvchr((U8*)s, &termlen);
11597 Copy(s, termstr, termlen, U8);
11598 if (!UTF8_IS_INVARIANT(term))
11602 /* mark where we are */
11603 PL_multi_start = CopLINE(PL_curcop);
11604 PL_multi_open = term;
11606 /* find corresponding closing delimiter */
11607 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11608 termcode = termstr[0] = term = tmps[5];
11610 PL_multi_close = term;
11612 /* create a new SV to hold the contents. 79 is the SV's initial length.
11613 What a random number. */
11614 sv = newSV_type(SVt_PVIV);
11616 SvIV_set(sv, termcode);
11617 (void)SvPOK_only(sv); /* validate pointer */
11619 /* move past delimiter and try to read a complete string */
11621 sv_catpvn(sv, s, termlen);
11624 tstart = SvPVX(PL_linestr) + stuffstart;
11625 if (!PL_thisopen && !keep_delims) {
11626 PL_thisopen = newSVpvn(tstart, s - tstart);
11627 stuffstart = s - SvPVX(PL_linestr);
11631 if (PL_encoding && !UTF) {
11635 int offset = s - SvPVX_const(PL_linestr);
11636 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11637 &offset, (char*)termstr, termlen);
11638 const char * const ns = SvPVX_const(PL_linestr) + offset;
11639 char * const svlast = SvEND(sv) - 1;
11641 for (; s < ns; s++) {
11642 if (*s == '\n' && !PL_rsfp)
11643 CopLINE_inc(PL_curcop);
11646 goto read_more_line;
11648 /* handle quoted delimiters */
11649 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11651 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11653 if ((svlast-1 - t) % 2) {
11654 if (!keep_quoted) {
11655 *(svlast-1) = term;
11657 SvCUR_set(sv, SvCUR(sv) - 1);
11662 if (PL_multi_open == PL_multi_close) {
11668 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11669 /* At here, all closes are "was quoted" one,
11670 so we don't check PL_multi_close. */
11672 if (!keep_quoted && *(t+1) == PL_multi_open)
11677 else if (*t == PL_multi_open)
11685 SvCUR_set(sv, w - SvPVX_const(sv));
11687 last_off = w - SvPVX(sv);
11688 if (--brackets <= 0)
11693 if (!keep_delims) {
11694 SvCUR_set(sv, SvCUR(sv) - 1);
11700 /* extend sv if need be */
11701 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11702 /* set 'to' to the next character in the sv's string */
11703 to = SvPVX(sv)+SvCUR(sv);
11705 /* if open delimiter is the close delimiter read unbridle */
11706 if (PL_multi_open == PL_multi_close) {
11707 for (; s < PL_bufend; s++,to++) {
11708 /* embedded newlines increment the current line number */
11709 if (*s == '\n' && !PL_rsfp)
11710 CopLINE_inc(PL_curcop);
11711 /* handle quoted delimiters */
11712 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11713 if (!keep_quoted && s[1] == term)
11715 /* any other quotes are simply copied straight through */
11719 /* terminate when run out of buffer (the for() condition), or
11720 have found the terminator */
11721 else if (*s == term) {
11724 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11727 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11733 /* if the terminator isn't the same as the start character (e.g.,
11734 matched brackets), we have to allow more in the quoting, and
11735 be prepared for nested brackets.
11738 /* read until we run out of string, or we find the terminator */
11739 for (; s < PL_bufend; s++,to++) {
11740 /* embedded newlines increment the line count */
11741 if (*s == '\n' && !PL_rsfp)
11742 CopLINE_inc(PL_curcop);
11743 /* backslashes can escape the open or closing characters */
11744 if (*s == '\\' && s+1 < PL_bufend) {
11745 if (!keep_quoted &&
11746 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11751 /* allow nested opens and closes */
11752 else if (*s == PL_multi_close && --brackets <= 0)
11754 else if (*s == PL_multi_open)
11756 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11761 /* terminate the copied string and update the sv's end-of-string */
11763 SvCUR_set(sv, to - SvPVX_const(sv));
11766 * this next chunk reads more into the buffer if we're not done yet
11770 break; /* handle case where we are done yet :-) */
11772 #ifndef PERL_STRICT_CR
11773 if (to - SvPVX_const(sv) >= 2) {
11774 if ((to[-2] == '\r' && to[-1] == '\n') ||
11775 (to[-2] == '\n' && to[-1] == '\r'))
11779 SvCUR_set(sv, to - SvPVX_const(sv));
11781 else if (to[-1] == '\r')
11784 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11789 /* if we're out of file, or a read fails, bail and reset the current
11790 line marker so we can report where the unterminated string began
11793 if (PL_madskills) {
11794 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11796 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11798 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11802 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11804 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11810 /* we read a line, so increment our line counter */
11811 CopLINE_inc(PL_curcop);
11813 /* update debugger info */
11814 if (PERLDB_LINE && PL_curstash != PL_debstash)
11815 update_debugger_info(PL_linestr, NULL, 0);
11817 /* having changed the buffer, we must update PL_bufend */
11818 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11819 PL_last_lop = PL_last_uni = NULL;
11822 /* at this point, we have successfully read the delimited string */
11824 if (!PL_encoding || UTF) {
11826 if (PL_madskills) {
11827 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11828 const int len = s - tstart;
11830 sv_catpvn(PL_thisstuff, tstart, len);
11832 PL_thisstuff = newSVpvn(tstart, len);
11833 if (!PL_thisclose && !keep_delims)
11834 PL_thisclose = newSVpvn(s,termlen);
11839 sv_catpvn(sv, s, termlen);
11844 if (PL_madskills) {
11845 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11846 const int len = s - tstart - termlen;
11848 sv_catpvn(PL_thisstuff, tstart, len);
11850 PL_thisstuff = newSVpvn(tstart, len);
11851 if (!PL_thisclose && !keep_delims)
11852 PL_thisclose = newSVpvn(s - termlen,termlen);
11856 if (has_utf8 || PL_encoding)
11859 PL_multi_end = CopLINE(PL_curcop);
11861 /* if we allocated too much space, give some back */
11862 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11863 SvLEN_set(sv, SvCUR(sv) + 1);
11864 SvPV_renew(sv, SvLEN(sv));
11867 /* decide whether this is the first or second quoted string we've read
11880 takes: pointer to position in buffer
11881 returns: pointer to new position in buffer
11882 side-effects: builds ops for the constant in yylval.op
11884 Read a number in any of the formats that Perl accepts:
11886 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11887 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11890 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11892 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11895 If it reads a number without a decimal point or an exponent, it will
11896 try converting the number to an integer and see if it can do so
11897 without loss of precision.
11901 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11904 register const char *s = start; /* current position in buffer */
11905 register char *d; /* destination in temp buffer */
11906 register char *e; /* end of temp buffer */
11907 NV nv; /* number read, as a double */
11908 SV *sv = NULL; /* place to put the converted number */
11909 bool floatit; /* boolean: int or float? */
11910 const char *lastub = NULL; /* position of last underbar */
11911 static char const number_too_long[] = "Number too long";
11913 /* We use the first character to decide what type of number this is */
11917 Perl_croak(aTHX_ "panic: scan_num");
11919 /* if it starts with a 0, it could be an octal number, a decimal in
11920 0.13 disguise, or a hexadecimal number, or a binary number. */
11924 u holds the "number so far"
11925 shift the power of 2 of the base
11926 (hex == 4, octal == 3, binary == 1)
11927 overflowed was the number more than we can hold?
11929 Shift is used when we add a digit. It also serves as an "are
11930 we in octal/hex/binary?" indicator to disallow hex characters
11931 when in octal mode.
11936 bool overflowed = FALSE;
11937 bool just_zero = TRUE; /* just plain 0 or binary number? */
11938 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11939 static const char* const bases[5] =
11940 { "", "binary", "", "octal", "hexadecimal" };
11941 static const char* const Bases[5] =
11942 { "", "Binary", "", "Octal", "Hexadecimal" };
11943 static const char* const maxima[5] =
11945 "0b11111111111111111111111111111111",
11949 const char *base, *Base, *max;
11951 /* check for hex */
11956 } else if (s[1] == 'b') {
11961 /* check for a decimal in disguise */
11962 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11964 /* so it must be octal */
11971 if (ckWARN(WARN_SYNTAX))
11972 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11973 "Misplaced _ in number");
11977 base = bases[shift];
11978 Base = Bases[shift];
11979 max = maxima[shift];
11981 /* read the rest of the number */
11983 /* x is used in the overflow test,
11984 b is the digit we're adding on. */
11989 /* if we don't mention it, we're done */
11993 /* _ are ignored -- but warned about if consecutive */
11995 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11996 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11997 "Misplaced _ in number");
12001 /* 8 and 9 are not octal */
12002 case '8': case '9':
12004 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12008 case '2': case '3': case '4':
12009 case '5': case '6': case '7':
12011 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12014 case '0': case '1':
12015 b = *s++ & 15; /* ASCII digit -> value of digit */
12019 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12020 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12021 /* make sure they said 0x */
12024 b = (*s++ & 7) + 9;
12026 /* Prepare to put the digit we have onto the end
12027 of the number so far. We check for overflows.
12033 x = u << shift; /* make room for the digit */
12035 if ((x >> shift) != u
12036 && !(PL_hints & HINT_NEW_BINARY)) {
12039 if (ckWARN_d(WARN_OVERFLOW))
12040 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12041 "Integer overflow in %s number",
12044 u = x | b; /* add the digit to the end */
12047 n *= nvshift[shift];
12048 /* If an NV has not enough bits in its
12049 * mantissa to represent an UV this summing of
12050 * small low-order numbers is a waste of time
12051 * (because the NV cannot preserve the
12052 * low-order bits anyway): we could just
12053 * remember when did we overflow and in the
12054 * end just multiply n by the right
12062 /* if we get here, we had success: make a scalar value from
12067 /* final misplaced underbar check */
12068 if (s[-1] == '_') {
12069 if (ckWARN(WARN_SYNTAX))
12070 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12075 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12076 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12077 "%s number > %s non-portable",
12083 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12084 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12085 "%s number > %s non-portable",
12090 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12091 sv = new_constant(start, s - start, "integer",
12092 sv, NULL, NULL, 0);
12093 else if (PL_hints & HINT_NEW_BINARY)
12094 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12099 handle decimal numbers.
12100 we're also sent here when we read a 0 as the first digit
12102 case '1': case '2': case '3': case '4': case '5':
12103 case '6': case '7': case '8': case '9': case '.':
12106 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12109 /* read next group of digits and _ and copy into d */
12110 while (isDIGIT(*s) || *s == '_') {
12111 /* skip underscores, checking for misplaced ones
12115 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12116 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12117 "Misplaced _ in number");
12121 /* check for end of fixed-length buffer */
12123 Perl_croak(aTHX_ number_too_long);
12124 /* if we're ok, copy the character */
12129 /* final misplaced underbar check */
12130 if (lastub && s == lastub + 1) {
12131 if (ckWARN(WARN_SYNTAX))
12132 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12135 /* read a decimal portion if there is one. avoid
12136 3..5 being interpreted as the number 3. followed
12139 if (*s == '.' && s[1] != '.') {
12144 if (ckWARN(WARN_SYNTAX))
12145 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12146 "Misplaced _ in number");
12150 /* copy, ignoring underbars, until we run out of digits.
12152 for (; isDIGIT(*s) || *s == '_'; s++) {
12153 /* fixed length buffer check */
12155 Perl_croak(aTHX_ number_too_long);
12157 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12158 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12159 "Misplaced _ in number");
12165 /* fractional part ending in underbar? */
12166 if (s[-1] == '_') {
12167 if (ckWARN(WARN_SYNTAX))
12168 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12169 "Misplaced _ in number");
12171 if (*s == '.' && isDIGIT(s[1])) {
12172 /* oops, it's really a v-string, but without the "v" */
12178 /* read exponent part, if present */
12179 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12183 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12184 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12186 /* stray preinitial _ */
12188 if (ckWARN(WARN_SYNTAX))
12189 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12190 "Misplaced _ in number");
12194 /* allow positive or negative exponent */
12195 if (*s == '+' || *s == '-')
12198 /* stray initial _ */
12200 if (ckWARN(WARN_SYNTAX))
12201 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12202 "Misplaced _ in number");
12206 /* read digits of exponent */
12207 while (isDIGIT(*s) || *s == '_') {
12210 Perl_croak(aTHX_ number_too_long);
12214 if (((lastub && s == lastub + 1) ||
12215 (!isDIGIT(s[1]) && s[1] != '_'))
12216 && ckWARN(WARN_SYNTAX))
12217 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12218 "Misplaced _ in number");
12225 /* make an sv from the string */
12229 We try to do an integer conversion first if no characters
12230 indicating "float" have been found.
12235 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12237 if (flags == IS_NUMBER_IN_UV) {
12239 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12242 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12243 if (uv <= (UV) IV_MIN)
12244 sv_setiv(sv, -(IV)uv);
12251 /* terminate the string */
12253 nv = Atof(PL_tokenbuf);
12258 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12259 const char *const key = floatit ? "float" : "integer";
12260 const STRLEN keylen = floatit ? 5 : 7;
12261 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12262 key, keylen, sv, NULL, NULL, 0);
12266 /* if it starts with a v, it could be a v-string */
12269 sv = newSV(5); /* preallocate storage space */
12270 s = scan_vstring(s, PL_bufend, sv);
12274 /* make the op for the constant and return */
12277 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12279 lvalp->opval = NULL;
12285 S_scan_formline(pTHX_ register char *s)
12288 register char *eol;
12290 SV * const stuff = newSVpvs("");
12291 bool needargs = FALSE;
12292 bool eofmt = FALSE;
12294 char *tokenstart = s;
12297 if (PL_madskills) {
12298 savewhite = PL_thiswhite;
12303 while (!needargs) {
12306 #ifdef PERL_STRICT_CR
12307 while (SPACE_OR_TAB(*t))
12310 while (SPACE_OR_TAB(*t) || *t == '\r')
12313 if (*t == '\n' || t == PL_bufend) {
12318 if (PL_in_eval && !PL_rsfp) {
12319 eol = (char *) memchr(s,'\n',PL_bufend-s);
12324 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12326 for (t = s; t < eol; t++) {
12327 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12329 goto enough; /* ~~ must be first line in formline */
12331 if (*t == '@' || *t == '^')
12335 sv_catpvn(stuff, s, eol-s);
12336 #ifndef PERL_STRICT_CR
12337 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12338 char *end = SvPVX(stuff) + SvCUR(stuff);
12341 SvCUR_set(stuff, SvCUR(stuff) - 1);
12351 if (PL_madskills) {
12353 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12355 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12358 s = filter_gets(PL_linestr, PL_rsfp, 0);
12360 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12362 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12364 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12365 PL_last_lop = PL_last_uni = NULL;
12374 if (SvCUR(stuff)) {
12377 PL_lex_state = LEX_NORMAL;
12378 start_force(PL_curforce);
12379 NEXTVAL_NEXTTOKE.ival = 0;
12383 PL_lex_state = LEX_FORMLINE;
12385 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12387 else if (PL_encoding)
12388 sv_recode_to_utf8(stuff, PL_encoding);
12390 start_force(PL_curforce);
12391 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12393 start_force(PL_curforce);
12394 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12398 SvREFCNT_dec(stuff);
12400 PL_lex_formbrack = 0;
12404 if (PL_madskills) {
12406 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12408 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12409 PL_thiswhite = savewhite;
12416 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12419 const I32 oldsavestack_ix = PL_savestack_ix;
12420 CV* const outsidecv = PL_compcv;
12423 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12425 SAVEI32(PL_subline);
12426 save_item(PL_subname);
12427 SAVESPTR(PL_compcv);
12429 PL_compcv = (CV*)newSV_type(is_format ? SVt_PVFM : SVt_PVCV);
12430 CvFLAGS(PL_compcv) |= flags;
12432 PL_subline = CopLINE(PL_curcop);
12433 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12434 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12435 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12437 return oldsavestack_ix;
12441 #pragma segment Perl_yylex
12444 Perl_yywarn(pTHX_ const char *s)
12447 PL_in_eval |= EVAL_WARNONLY;
12449 PL_in_eval &= ~EVAL_WARNONLY;
12454 Perl_yyerror(pTHX_ const char *s)
12457 const char *where = NULL;
12458 const char *context = NULL;
12461 int yychar = PL_parser->yychar;
12463 if (!yychar || (yychar == ';' && !PL_rsfp))
12465 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12466 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12467 PL_oldbufptr != PL_bufptr) {
12470 The code below is removed for NetWare because it abends/crashes on NetWare
12471 when the script has error such as not having the closing quotes like:
12472 if ($var eq "value)
12473 Checking of white spaces is anyway done in NetWare code.
12476 while (isSPACE(*PL_oldoldbufptr))
12479 context = PL_oldoldbufptr;
12480 contlen = PL_bufptr - PL_oldoldbufptr;
12482 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12483 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12486 The code below is removed for NetWare because it abends/crashes on NetWare
12487 when the script has error such as not having the closing quotes like:
12488 if ($var eq "value)
12489 Checking of white spaces is anyway done in NetWare code.
12492 while (isSPACE(*PL_oldbufptr))
12495 context = PL_oldbufptr;
12496 contlen = PL_bufptr - PL_oldbufptr;
12498 else if (yychar > 255)
12499 where = "next token ???";
12500 else if (yychar == -2) { /* YYEMPTY */
12501 if (PL_lex_state == LEX_NORMAL ||
12502 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12503 where = "at end of line";
12504 else if (PL_lex_inpat)
12505 where = "within pattern";
12507 where = "within string";
12510 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12512 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12513 else if (isPRINT_LC(yychar)) {
12514 const char string = yychar;
12515 sv_catpvn(where_sv, &string, 1);
12518 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12519 where = SvPVX_const(where_sv);
12521 msg = sv_2mortal(newSVpv(s, 0));
12522 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12523 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12525 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12527 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12528 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12529 Perl_sv_catpvf(aTHX_ msg,
12530 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12531 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12534 if (PL_in_eval & EVAL_WARNONLY) {
12535 if (ckWARN_d(WARN_SYNTAX))
12536 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12540 if (PL_error_count >= 10) {
12541 if (PL_in_eval && SvCUR(ERRSV))
12542 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12543 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12545 Perl_croak(aTHX_ "%s has too many errors.\n",
12546 OutCopFILE(PL_curcop));
12549 PL_in_my_stash = NULL;
12553 #pragma segment Main
12557 S_swallow_bom(pTHX_ U8 *s)
12560 const STRLEN slen = SvCUR(PL_linestr);
12563 if (s[1] == 0xFE) {
12564 /* UTF-16 little-endian? (or UTF32-LE?) */
12565 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12566 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12567 #ifndef PERL_NO_UTF16_FILTER
12568 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12571 if (PL_bufend > (char*)s) {
12575 filter_add(utf16rev_textfilter, NULL);
12576 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12577 utf16_to_utf8_reversed(s, news,
12578 PL_bufend - (char*)s - 1,
12580 sv_setpvn(PL_linestr, (const char*)news, newlen);
12582 s = (U8*)SvPVX(PL_linestr);
12583 Copy(news, s, newlen, U8);
12587 SvUTF8_on(PL_linestr);
12588 s = (U8*)SvPVX(PL_linestr);
12590 /* FIXME - is this a general bug fix? */
12593 PL_bufend = SvPVX(PL_linestr) + newlen;
12596 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12601 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12602 #ifndef PERL_NO_UTF16_FILTER
12603 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12606 if (PL_bufend > (char *)s) {
12610 filter_add(utf16_textfilter, NULL);
12611 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12612 utf16_to_utf8(s, news,
12613 PL_bufend - (char*)s,
12615 sv_setpvn(PL_linestr, (const char*)news, newlen);
12617 SvUTF8_on(PL_linestr);
12618 s = (U8*)SvPVX(PL_linestr);
12619 PL_bufend = SvPVX(PL_linestr) + newlen;
12622 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12627 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12628 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12629 s += 3; /* UTF-8 */
12635 if (s[2] == 0xFE && s[3] == 0xFF) {
12636 /* UTF-32 big-endian */
12637 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12640 else if (s[2] == 0 && s[3] != 0) {
12643 * are a good indicator of UTF-16BE. */
12644 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12650 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12651 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12652 s += 4; /* UTF-8 */
12658 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12661 * are a good indicator of UTF-16LE. */
12662 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12670 #ifndef PERL_NO_UTF16_FILTER
12672 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12675 const STRLEN old = SvCUR(sv);
12676 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12677 DEBUG_P(PerlIO_printf(Perl_debug_log,
12678 "utf16_textfilter(%p): %d %d (%d)\n",
12679 FPTR2DPTR(void *, utf16_textfilter),
12680 idx, maxlen, (int) count));
12684 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12685 Copy(SvPVX_const(sv), tmps, old, char);
12686 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12687 SvCUR(sv) - old, &newlen);
12688 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12690 DEBUG_P({sv_dump(sv);});
12695 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12698 const STRLEN old = SvCUR(sv);
12699 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12700 DEBUG_P(PerlIO_printf(Perl_debug_log,
12701 "utf16rev_textfilter(%p): %d %d (%d)\n",
12702 FPTR2DPTR(void *, utf16rev_textfilter),
12703 idx, maxlen, (int) count));
12707 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12708 Copy(SvPVX_const(sv), tmps, old, char);
12709 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12710 SvCUR(sv) - old, &newlen);
12711 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12713 DEBUG_P({ sv_dump(sv); });
12719 Returns a pointer to the next character after the parsed
12720 vstring, as well as updating the passed in sv.
12722 Function must be called like
12725 s = scan_vstring(s,e,sv);
12727 where s and e are the start and end of the string.
12728 The sv should already be large enough to store the vstring
12729 passed in, for performance reasons.
12734 Perl_scan_vstring(pTHX_ const char *s, const char *e, SV *sv)
12737 const char *pos = s;
12738 const char *start = s;
12739 if (*pos == 'v') pos++; /* get past 'v' */
12740 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12742 if ( *pos != '.') {
12743 /* this may not be a v-string if followed by => */
12744 const char *next = pos;
12745 while (next < e && isSPACE(*next))
12747 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12748 /* return string not v-string */
12749 sv_setpvn(sv,(char *)s,pos-s);
12750 return (char *)pos;
12754 if (!isALPHA(*pos)) {
12755 U8 tmpbuf[UTF8_MAXBYTES+1];
12758 s++; /* get past 'v' */
12760 sv_setpvn(sv, "", 0);
12763 /* this is atoi() that tolerates underscores */
12766 const char *end = pos;
12768 while (--end >= s) {
12770 const UV orev = rev;
12771 rev += (*end - '0') * mult;
12773 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12774 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12775 "Integer overflow in decimal number");
12779 if (rev > 0x7FFFFFFF)
12780 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12782 /* Append native character for the rev point */
12783 tmpend = uvchr_to_utf8(tmpbuf, rev);
12784 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12785 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12787 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12793 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12797 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12805 * c-indentation-style: bsd
12806 * c-basic-offset: 4
12807 * indent-tabs-mode: t
12810 * ex: set ts=8 sts=4 sw=4 noet: