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 pl_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 (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
243 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
244 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
245 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
246 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
247 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
248 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
249 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
250 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
251 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
252 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
253 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
254 #define Rop(f) return (pl_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) { \
262 pl_yylval.ival = f; \
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) { \
276 pl_yylval.ival = 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(pl_yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
290 /* how to interpret the pl_yylval associated with the token */
294 TOKENTYPE_OPNUM, /* pl_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 pl_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)pl_yylval.ival);
408 case TOKENTYPE_OPNUM:
409 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
410 PL_op_name[pl_yylval.ival]);
413 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", pl_yylval.pval);
415 case TOKENTYPE_OPVAL:
416 if (pl_yylval.opval) {
417 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
418 PL_op_name[pl_yylval.opval->op_type]);
419 if (pl_yylval.opval->op_type == OP_CONST) {
420 Perl_sv_catpvf(aTHX_ report, " %s",
421 SvPEEK(cSVOPx_sv(pl_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 pl_yylval.ival = OP_ANDASSIGN;
462 else if (toketype == OROR)
463 pl_yylval.ival = OP_ORASSIGN;
464 else if (toketype == DORDOR)
465 pl_yylval.ival = OP_DORASSIGN;
473 * When Perl expects an operator and finds something else, no_op
474 * prints the warning. It always prints "<something> found where
475 * operator expected. It prints "Missing semicolon on previous line?"
476 * if the surprise occurs at the start of the line. "do you need to
477 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
478 * where the compiler doesn't know if foo is a method call or a function.
479 * It prints "Missing operator before end of line" if there's nothing
480 * after the missing operator, or "... before <...>" if there is something
481 * after the missing operator.
485 S_no_op(pTHX_ const char *what, char *s)
488 char * const oldbp = PL_bufptr;
489 const bool is_first = (PL_oldbufptr == PL_linestart);
495 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
496 if (ckWARN_d(WARN_SYNTAX)) {
498 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
499 "\t(Missing semicolon on previous line?)\n");
500 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
502 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
504 if (t < PL_bufptr && isSPACE(*t))
505 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
506 "\t(Do you need to predeclare %.*s?)\n",
507 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
511 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
512 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
520 * Complain about missing quote/regexp/heredoc terminator.
521 * If it's called with NULL then it cauterizes the line buffer.
522 * If we're in a delimited string and the delimiter is a control
523 * character, it's reformatted into a two-char sequence like ^C.
528 S_missingterm(pTHX_ char *s)
534 char * const nl = strrchr(s,'\n');
540 iscntrl(PL_multi_close)
542 PL_multi_close < 32 || PL_multi_close == 127
546 tmpbuf[1] = (char)toCTRL(PL_multi_close);
551 *tmpbuf = (char)PL_multi_close;
555 q = strchr(s,'"') ? '\'' : '"';
556 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
559 #define FEATURE_IS_ENABLED(name) \
560 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
561 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
562 /* The longest string we pass in. */
563 #define MAX_FEATURE_LEN (sizeof("switch")-1)
566 * S_feature_is_enabled
567 * Check whether the named feature is enabled.
570 S_feature_is_enabled(pTHX_ const char *name, STRLEN namelen)
573 HV * const hinthv = GvHV(PL_hintgv);
574 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
575 assert(namelen <= MAX_FEATURE_LEN);
576 memcpy(&he_name[8], name, namelen);
578 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
586 Perl_deprecate(pTHX_ const char *s)
588 if (ckWARN(WARN_DEPRECATED))
589 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
593 Perl_deprecate_old(pTHX_ const char *s)
595 /* This function should NOT be called for any new deprecated warnings */
596 /* Use Perl_deprecate instead */
598 /* It is here to maintain backward compatibility with the pre-5.8 */
599 /* warnings category hierarchy. The "deprecated" category used to */
600 /* live under the "syntax" category. It is now a top-level category */
601 /* in its own right. */
603 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
604 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
605 "Use of %s is deprecated", s);
609 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
610 * utf16-to-utf8-reversed.
613 #ifdef PERL_CR_FILTER
617 register const char *s = SvPVX_const(sv);
618 register const char * const e = s + SvCUR(sv);
619 /* outer loop optimized to do nothing if there are no CR-LFs */
621 if (*s++ == '\r' && *s == '\n') {
622 /* hit a CR-LF, need to copy the rest */
623 register char *d = s - 1;
626 if (*s == '\r' && s[1] == '\n')
637 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
639 const I32 count = FILTER_READ(idx+1, sv, maxlen);
640 if (count > 0 && !maxlen)
651 * Create a parser object and initialise its parser and lexer fields
653 * rsfp is the opened file handle to read from (if any),
655 * line holds any initial content already read from the file (or in
656 * the case of no file, such as an eval, the whole contents);
658 * new_filter indicates that this is a new file and it shouldn't inherit
659 * the filters from the current parser (ie require).
663 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
666 const char *s = NULL;
668 yy_parser *parser, *oparser;
670 /* create and initialise a parser */
672 Newxz(parser, 1, yy_parser);
673 parser->old_parser = oparser = PL_parser;
676 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
677 parser->ps = parser->stack;
678 parser->stack_size = YYINITDEPTH;
680 parser->stack->state = 0;
681 parser->yyerrstatus = 0;
682 parser->yychar = YYEMPTY; /* Cause a token to be read. */
684 /* on scope exit, free this parser and restore any outer one */
686 parser->saved_curcop = PL_curcop;
688 /* initialise lexer state */
691 parser->curforce = -1;
693 parser->nexttoke = 0;
695 parser->copline = NOLINE;
696 parser->lex_state = LEX_NORMAL;
697 parser->expect = XSTATE;
699 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
700 : (AV*)SvREFCNT_inc(oparser->rsfp_filters);
702 Newx(parser->lex_brackstack, 120, char);
703 Newx(parser->lex_casestack, 12, char);
704 *parser->lex_casestack = '\0';
707 s = SvPV_const(line, len);
713 parser->linestr = newSVpvs("\n;");
714 } else if (SvREADONLY(line) || s[len-1] != ';') {
715 parser->linestr = newSVsv(line);
717 sv_catpvs(parser->linestr, "\n;");
720 SvREFCNT_inc_simple_void_NN(line);
721 parser->linestr = line;
723 parser->oldoldbufptr =
726 parser->linestart = SvPVX(parser->linestr);
727 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
728 parser->last_lop = parser->last_uni = NULL;
732 /* delete a parser object */
735 Perl_parser_free(pTHX_ const yy_parser *parser)
737 PL_curcop = parser->saved_curcop;
738 SvREFCNT_dec(parser->linestr);
740 if (parser->rsfp == PerlIO_stdin())
741 PerlIO_clearerr(parser->rsfp);
742 else if (parser->rsfp && parser->old_parser
743 && parser->rsfp != parser->old_parser->rsfp)
744 PerlIO_close(parser->rsfp);
745 SvREFCNT_dec(parser->rsfp_filters);
747 Safefree(parser->stack);
748 Safefree(parser->lex_brackstack);
749 Safefree(parser->lex_casestack);
750 PL_parser = parser->old_parser;
757 * Finalizer for lexing operations. Must be called when the parser is
758 * done with the lexer.
765 PL_doextract = FALSE;
770 * This subroutine has nothing to do with tilting, whether at windmills
771 * or pinball tables. Its name is short for "increment line". It
772 * increments the current line number in CopLINE(PL_curcop) and checks
773 * to see whether the line starts with a comment of the form
774 * # line 500 "foo.pm"
775 * If so, it sets the current line number and file to the values in the comment.
779 S_incline(pTHX_ const char *s)
786 CopLINE_inc(PL_curcop);
789 while (SPACE_OR_TAB(*s))
791 if (strnEQ(s, "line", 4))
795 if (SPACE_OR_TAB(*s))
799 while (SPACE_OR_TAB(*s))
807 while (SPACE_OR_TAB(*s))
809 if (*s == '"' && (t = strchr(s+1, '"'))) {
819 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
821 if (*e != '\n' && *e != '\0')
822 return; /* false alarm */
825 const STRLEN len = t - s;
827 SV *const temp_sv = CopFILESV(PL_curcop);
833 tmplen = SvCUR(temp_sv);
839 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
840 /* must copy *{"::_<(eval N)[oldfilename:L]"}
841 * to *{"::_<newfilename"} */
842 /* However, the long form of evals is only turned on by the
843 debugger - usually they're "(eval %lu)" */
847 STRLEN tmplen2 = len;
848 if (tmplen + 2 <= sizeof smallbuf)
851 Newx(tmpbuf, tmplen + 2, char);
854 memcpy(tmpbuf + 2, cf, tmplen);
856 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
861 if (tmplen2 + 2 <= sizeof smallbuf)
864 Newx(tmpbuf2, tmplen2 + 2, char);
866 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
867 /* Either they malloc'd it, or we malloc'd it,
868 so no prefix is present in ours. */
873 memcpy(tmpbuf2 + 2, s, tmplen2);
876 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
878 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
879 /* adjust ${"::_<newfilename"} to store the new file name */
880 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
881 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
882 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
885 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
887 if (tmpbuf != smallbuf) Safefree(tmpbuf);
890 CopFILE_free(PL_curcop);
891 CopFILE_setn(PL_curcop, s, len);
893 CopLINE_set(PL_curcop, atoi(n)-1);
897 /* skip space before PL_thistoken */
900 S_skipspace0(pTHX_ register char *s)
907 PL_thiswhite = newSVpvs("");
908 sv_catsv(PL_thiswhite, PL_skipwhite);
909 sv_free(PL_skipwhite);
912 PL_realtokenstart = s - SvPVX(PL_linestr);
916 /* skip space after PL_thistoken */
919 S_skipspace1(pTHX_ register char *s)
921 const char *start = s;
922 I32 startoff = start - SvPVX(PL_linestr);
927 start = SvPVX(PL_linestr) + startoff;
928 if (!PL_thistoken && PL_realtokenstart >= 0) {
929 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
930 PL_thistoken = newSVpvn(tstart, start - tstart);
932 PL_realtokenstart = -1;
935 PL_nextwhite = newSVpvs("");
936 sv_catsv(PL_nextwhite, PL_skipwhite);
937 sv_free(PL_skipwhite);
944 S_skipspace2(pTHX_ register char *s, SV **svp)
947 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
948 const I32 startoff = s - SvPVX(PL_linestr);
951 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
952 if (!PL_madskills || !svp)
954 start = SvPVX(PL_linestr) + startoff;
955 if (!PL_thistoken && PL_realtokenstart >= 0) {
956 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
957 PL_thistoken = newSVpvn(tstart, start - tstart);
958 PL_realtokenstart = -1;
963 sv_setsv(*svp, PL_skipwhite);
964 sv_free(PL_skipwhite);
973 S_update_debugger_info(pTHX_ SV *orig_sv, const char *buf, STRLEN len)
975 AV *av = CopFILEAVx(PL_curcop);
977 SV * const sv = newSV_type(SVt_PVMG);
979 sv_setsv(sv, orig_sv);
981 sv_setpvn(sv, buf, len);
984 av_store(av, (I32)CopLINE(PL_curcop), sv);
990 * Called to gobble the appropriate amount and type of whitespace.
991 * Skips comments as well.
995 S_skipspace(pTHX_ register char *s)
1000 int startoff = s - SvPVX(PL_linestr);
1003 sv_free(PL_skipwhite);
1008 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1009 while (s < PL_bufend && SPACE_OR_TAB(*s))
1019 SSize_t oldprevlen, oldoldprevlen;
1020 SSize_t oldloplen = 0, oldunilen = 0;
1021 while (s < PL_bufend && isSPACE(*s)) {
1022 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
1027 if (s < PL_bufend && *s == '#') {
1028 while (s < PL_bufend && *s != '\n')
1030 if (s < PL_bufend) {
1032 if (PL_in_eval && !PL_rsfp) {
1039 /* only continue to recharge the buffer if we're at the end
1040 * of the buffer, we're not reading from a source filter, and
1041 * we're in normal lexing mode
1043 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1044 PL_lex_state == LEX_FORMLINE)
1051 /* try to recharge the buffer */
1053 curoff = s - SvPVX(PL_linestr);
1056 if ((s = filter_gets(PL_linestr, PL_rsfp,
1057 (prevlen = SvCUR(PL_linestr)))) == NULL)
1060 if (PL_madskills && curoff != startoff) {
1062 PL_skipwhite = newSVpvs("");
1063 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1067 /* mustn't throw out old stuff yet if madpropping */
1068 SvCUR(PL_linestr) = curoff;
1069 s = SvPVX(PL_linestr) + curoff;
1071 if (curoff && s[-1] == '\n')
1075 /* end of file. Add on the -p or -n magic */
1076 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1079 sv_catpvs(PL_linestr,
1080 ";}continue{print or die qq(-p destination: $!\\n);}");
1082 sv_setpvs(PL_linestr,
1083 ";}continue{print or die qq(-p destination: $!\\n);}");
1085 PL_minus_n = PL_minus_p = 0;
1087 else if (PL_minus_n) {
1089 sv_catpvn(PL_linestr, ";}", 2);
1091 sv_setpvn(PL_linestr, ";}", 2);
1097 sv_catpvn(PL_linestr,";", 1);
1099 sv_setpvn(PL_linestr,";", 1);
1102 /* reset variables for next time we lex */
1103 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1109 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1110 PL_last_lop = PL_last_uni = NULL;
1112 /* Close the filehandle. Could be from -P preprocessor,
1113 * STDIN, or a regular file. If we were reading code from
1114 * STDIN (because the commandline held no -e or filename)
1115 * then we don't close it, we reset it so the code can
1116 * read from STDIN too.
1119 if (PL_preprocess && !PL_in_eval)
1120 (void)PerlProc_pclose(PL_rsfp);
1121 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1122 PerlIO_clearerr(PL_rsfp);
1124 (void)PerlIO_close(PL_rsfp);
1129 /* not at end of file, so we only read another line */
1130 /* make corresponding updates to old pointers, for yyerror() */
1131 oldprevlen = PL_oldbufptr - PL_bufend;
1132 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1134 oldunilen = PL_last_uni - PL_bufend;
1136 oldloplen = PL_last_lop - PL_bufend;
1137 PL_linestart = PL_bufptr = s + prevlen;
1138 PL_bufend = s + SvCUR(PL_linestr);
1140 PL_oldbufptr = s + oldprevlen;
1141 PL_oldoldbufptr = s + oldoldprevlen;
1143 PL_last_uni = s + oldunilen;
1145 PL_last_lop = s + oldloplen;
1148 /* debugger active and we're not compiling the debugger code,
1149 * so store the line into the debugger's array of lines
1151 if (PERLDB_LINE && PL_curstash != PL_debstash)
1152 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1159 PL_skipwhite = newSVpvs("");
1160 curoff = s - SvPVX(PL_linestr);
1161 if (curoff - startoff)
1162 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1171 * Check the unary operators to ensure there's no ambiguity in how they're
1172 * used. An ambiguous piece of code would be:
1174 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1175 * the +5 is its argument.
1185 if (PL_oldoldbufptr != PL_last_uni)
1187 while (isSPACE(*PL_last_uni))
1190 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1192 if ((t = strchr(s, '(')) && t < PL_bufptr)
1195 if (ckWARN_d(WARN_AMBIGUOUS)){
1196 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1197 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1198 (int)(s - PL_last_uni), PL_last_uni);
1203 * LOP : macro to build a list operator. Its behaviour has been replaced
1204 * with a subroutine, S_lop() for which LOP is just another name.
1207 #define LOP(f,x) return lop(f,x,s)
1211 * Build a list operator (or something that might be one). The rules:
1212 * - if we have a next token, then it's a list operator [why?]
1213 * - if the next thing is an opening paren, then it's a function
1214 * - else it's a list operator
1218 S_lop(pTHX_ I32 f, int x, char *s)
1225 PL_last_lop = PL_oldbufptr;
1226 PL_last_lop_op = (OPCODE)f;
1229 return REPORT(LSTOP);
1232 return REPORT(LSTOP);
1235 return REPORT(FUNC);
1238 return REPORT(FUNC);
1240 return REPORT(LSTOP);
1246 * Sets up for an eventual force_next(). start_force(0) basically does
1247 * an unshift, while start_force(-1) does a push. yylex removes items
1252 S_start_force(pTHX_ int where)
1256 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1257 where = PL_lasttoke;
1258 assert(PL_curforce < 0 || PL_curforce == where);
1259 if (PL_curforce != where) {
1260 for (i = PL_lasttoke; i > where; --i) {
1261 PL_nexttoke[i] = PL_nexttoke[i-1];
1265 if (PL_curforce < 0) /* in case of duplicate start_force() */
1266 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1267 PL_curforce = where;
1270 curmad('^', newSVpvs(""));
1271 CURMAD('_', PL_nextwhite);
1276 S_curmad(pTHX_ char slot, SV *sv)
1282 if (PL_curforce < 0)
1283 where = &PL_thismad;
1285 where = &PL_nexttoke[PL_curforce].next_mad;
1288 sv_setpvn(sv, "", 0);
1291 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1293 else if (PL_encoding) {
1294 sv_recode_to_utf8(sv, PL_encoding);
1299 /* keep a slot open for the head of the list? */
1300 if (slot != '_' && *where && (*where)->mad_key == '^') {
1301 (*where)->mad_key = slot;
1302 sv_free((SV*)((*where)->mad_val));
1303 (*where)->mad_val = (void*)sv;
1306 addmad(newMADsv(slot, sv), where, 0);
1309 # define start_force(where) NOOP
1310 # define curmad(slot, sv) NOOP
1315 * When the lexer realizes it knows the next token (for instance,
1316 * it is reordering tokens for the parser) then it can call S_force_next
1317 * to know what token to return the next time the lexer is called. Caller
1318 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1319 * and possibly PL_expect to ensure the lexer handles the token correctly.
1323 S_force_next(pTHX_ I32 type)
1327 if (PL_curforce < 0)
1328 start_force(PL_lasttoke);
1329 PL_nexttoke[PL_curforce].next_type = type;
1330 if (PL_lex_state != LEX_KNOWNEXT)
1331 PL_lex_defer = PL_lex_state;
1332 PL_lex_state = LEX_KNOWNEXT;
1333 PL_lex_expect = PL_expect;
1336 PL_nexttype[PL_nexttoke] = type;
1338 if (PL_lex_state != LEX_KNOWNEXT) {
1339 PL_lex_defer = PL_lex_state;
1340 PL_lex_expect = PL_expect;
1341 PL_lex_state = LEX_KNOWNEXT;
1347 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1350 SV * const sv = newSVpvn(start,len);
1351 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1358 * When the lexer knows the next thing is a word (for instance, it has
1359 * just seen -> and it knows that the next char is a word char, then
1360 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1364 * char *start : buffer position (must be within PL_linestr)
1365 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1366 * int check_keyword : if true, Perl checks to make sure the word isn't
1367 * a keyword (do this if the word is a label, e.g. goto FOO)
1368 * int allow_pack : if true, : characters will also be allowed (require,
1369 * use, etc. do this)
1370 * int allow_initial_tick : used by the "sub" lexer only.
1374 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1380 start = SKIPSPACE1(start);
1382 if (isIDFIRST_lazy_if(s,UTF) ||
1383 (allow_pack && *s == ':') ||
1384 (allow_initial_tick && *s == '\'') )
1386 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1387 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1389 start_force(PL_curforce);
1391 curmad('X', newSVpvn(start,s-start));
1392 if (token == METHOD) {
1397 PL_expect = XOPERATOR;
1401 curmad('g', newSVpvs( "forced" ));
1402 NEXTVAL_NEXTTOKE.opval
1403 = (OP*)newSVOP(OP_CONST,0,
1404 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1405 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1413 * Called when the lexer wants $foo *foo &foo etc, but the program
1414 * text only contains the "foo" portion. The first argument is a pointer
1415 * to the "foo", and the second argument is the type symbol to prefix.
1416 * Forces the next token to be a "WORD".
1417 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1421 S_force_ident(pTHX_ register const char *s, int kind)
1425 const STRLEN len = strlen(s);
1426 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1427 start_force(PL_curforce);
1428 NEXTVAL_NEXTTOKE.opval = o;
1431 o->op_private = OPpCONST_ENTERED;
1432 /* XXX see note in pp_entereval() for why we forgo typo
1433 warnings if the symbol must be introduced in an eval.
1435 gv_fetchpvn_flags(s, len,
1436 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1438 kind == '$' ? SVt_PV :
1439 kind == '@' ? SVt_PVAV :
1440 kind == '%' ? SVt_PVHV :
1448 Perl_str_to_version(pTHX_ SV *sv)
1453 const char *start = SvPV_const(sv,len);
1454 const char * const end = start + len;
1455 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1456 while (start < end) {
1460 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1465 retval += ((NV)n)/nshift;
1474 * Forces the next token to be a version number.
1475 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1476 * and if "guessing" is TRUE, then no new token is created (and the caller
1477 * must use an alternative parsing method).
1481 S_force_version(pTHX_ char *s, int guessing)
1487 I32 startoff = s - SvPVX(PL_linestr);
1496 while (isDIGIT(*d) || *d == '_' || *d == '.')
1500 start_force(PL_curforce);
1501 curmad('X', newSVpvn(s,d-s));
1504 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1506 s = scan_num(s, &pl_yylval);
1507 version = pl_yylval.opval;
1508 ver = cSVOPx(version)->op_sv;
1509 if (SvPOK(ver) && !SvNIOK(ver)) {
1510 SvUPGRADE(ver, SVt_PVNV);
1511 SvNV_set(ver, str_to_version(ver));
1512 SvNOK_on(ver); /* hint that it is a version */
1515 else if (guessing) {
1518 sv_free(PL_nextwhite); /* let next token collect whitespace */
1520 s = SvPVX(PL_linestr) + startoff;
1528 if (PL_madskills && !version) {
1529 sv_free(PL_nextwhite); /* let next token collect whitespace */
1531 s = SvPVX(PL_linestr) + startoff;
1534 /* NOTE: The parser sees the package name and the VERSION swapped */
1535 start_force(PL_curforce);
1536 NEXTVAL_NEXTTOKE.opval = version;
1544 * Tokenize a quoted string passed in as an SV. It finds the next
1545 * chunk, up to end of string or a backslash. It may make a new
1546 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1551 S_tokeq(pTHX_ SV *sv)
1555 register char *send;
1563 s = SvPV_force(sv, len);
1564 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1567 while (s < send && *s != '\\')
1572 if ( PL_hints & HINT_NEW_STRING ) {
1573 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1579 if (s + 1 < send && (s[1] == '\\'))
1580 s++; /* all that, just for this */
1585 SvCUR_set(sv, d - SvPVX_const(sv));
1587 if ( PL_hints & HINT_NEW_STRING )
1588 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
1593 * Now come three functions related to double-quote context,
1594 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1595 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1596 * interact with PL_lex_state, and create fake ( ... ) argument lists
1597 * to handle functions and concatenation.
1598 * They assume that whoever calls them will be setting up a fake
1599 * join call, because each subthing puts a ',' after it. This lets
1602 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1604 * (I'm not sure whether the spurious commas at the end of lcfirst's
1605 * arguments and join's arguments are created or not).
1610 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1612 * Pattern matching will set PL_lex_op to the pattern-matching op to
1613 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
1615 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1617 * Everything else becomes a FUNC.
1619 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1620 * had an OP_CONST or OP_READLINE). This just sets us up for a
1621 * call to S_sublex_push().
1625 S_sublex_start(pTHX)
1628 register const I32 op_type = pl_yylval.ival;
1630 if (op_type == OP_NULL) {
1631 pl_yylval.opval = PL_lex_op;
1635 if (op_type == OP_CONST || op_type == OP_READLINE) {
1636 SV *sv = tokeq(PL_lex_stuff);
1638 if (SvTYPE(sv) == SVt_PVIV) {
1639 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1641 const char * const p = SvPV_const(sv, len);
1642 SV * const nsv = newSVpvn(p, len);
1648 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1649 PL_lex_stuff = NULL;
1650 /* Allow <FH> // "foo" */
1651 if (op_type == OP_READLINE)
1652 PL_expect = XTERMORDORDOR;
1655 else if (op_type == OP_BACKTICK && PL_lex_op) {
1656 /* readpipe() vas overriden */
1657 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1658 pl_yylval.opval = PL_lex_op;
1660 PL_lex_stuff = NULL;
1664 PL_sublex_info.super_state = PL_lex_state;
1665 PL_sublex_info.sub_inwhat = (U16)op_type;
1666 PL_sublex_info.sub_op = PL_lex_op;
1667 PL_lex_state = LEX_INTERPPUSH;
1671 pl_yylval.opval = PL_lex_op;
1681 * Create a new scope to save the lexing state. The scope will be
1682 * ended in S_sublex_done. Returns a '(', starting the function arguments
1683 * to the uc, lc, etc. found before.
1684 * Sets PL_lex_state to LEX_INTERPCONCAT.
1693 PL_lex_state = PL_sublex_info.super_state;
1694 SAVEBOOL(PL_lex_dojoin);
1695 SAVEI32(PL_lex_brackets);
1696 SAVEI32(PL_lex_casemods);
1697 SAVEI32(PL_lex_starts);
1698 SAVEI8(PL_lex_state);
1699 SAVEVPTR(PL_lex_inpat);
1700 SAVEI16(PL_lex_inwhat);
1701 SAVECOPLINE(PL_curcop);
1702 SAVEPPTR(PL_bufptr);
1703 SAVEPPTR(PL_bufend);
1704 SAVEPPTR(PL_oldbufptr);
1705 SAVEPPTR(PL_oldoldbufptr);
1706 SAVEPPTR(PL_last_lop);
1707 SAVEPPTR(PL_last_uni);
1708 SAVEPPTR(PL_linestart);
1709 SAVESPTR(PL_linestr);
1710 SAVEGENERICPV(PL_lex_brackstack);
1711 SAVEGENERICPV(PL_lex_casestack);
1713 PL_linestr = PL_lex_stuff;
1714 PL_lex_stuff = NULL;
1716 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1717 = SvPVX(PL_linestr);
1718 PL_bufend += SvCUR(PL_linestr);
1719 PL_last_lop = PL_last_uni = NULL;
1720 SAVEFREESV(PL_linestr);
1722 PL_lex_dojoin = FALSE;
1723 PL_lex_brackets = 0;
1724 Newx(PL_lex_brackstack, 120, char);
1725 Newx(PL_lex_casestack, 12, char);
1726 PL_lex_casemods = 0;
1727 *PL_lex_casestack = '\0';
1729 PL_lex_state = LEX_INTERPCONCAT;
1730 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1732 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1733 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1734 PL_lex_inpat = PL_sublex_info.sub_op;
1736 PL_lex_inpat = NULL;
1743 * Restores lexer state after a S_sublex_push.
1750 if (!PL_lex_starts++) {
1751 SV * const sv = newSVpvs("");
1752 if (SvUTF8(PL_linestr))
1754 PL_expect = XOPERATOR;
1755 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1759 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1760 PL_lex_state = LEX_INTERPCASEMOD;
1764 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1765 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1766 PL_linestr = PL_lex_repl;
1768 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1769 PL_bufend += SvCUR(PL_linestr);
1770 PL_last_lop = PL_last_uni = NULL;
1771 SAVEFREESV(PL_linestr);
1772 PL_lex_dojoin = FALSE;
1773 PL_lex_brackets = 0;
1774 PL_lex_casemods = 0;
1775 *PL_lex_casestack = '\0';
1777 if (SvEVALED(PL_lex_repl)) {
1778 PL_lex_state = LEX_INTERPNORMAL;
1780 /* we don't clear PL_lex_repl here, so that we can check later
1781 whether this is an evalled subst; that means we rely on the
1782 logic to ensure sublex_done() is called again only via the
1783 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1786 PL_lex_state = LEX_INTERPCONCAT;
1796 PL_endwhite = newSVpvs("");
1797 sv_catsv(PL_endwhite, PL_thiswhite);
1801 sv_setpvn(PL_thistoken,"",0);
1803 PL_realtokenstart = -1;
1807 PL_bufend = SvPVX(PL_linestr);
1808 PL_bufend += SvCUR(PL_linestr);
1809 PL_expect = XOPERATOR;
1810 PL_sublex_info.sub_inwhat = 0;
1818 Extracts a pattern, double-quoted string, or transliteration. This
1821 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1822 processing a pattern (PL_lex_inpat is true), a transliteration
1823 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1825 Returns a pointer to the character scanned up to. If this is
1826 advanced from the start pointer supplied (i.e. if anything was
1827 successfully parsed), will leave an OP for the substring scanned
1828 in pl_yylval. Caller must intuit reason for not parsing further
1829 by looking at the next characters herself.
1833 double-quoted style: \r and \n
1834 regexp special ones: \D \s
1837 case and quoting: \U \Q \E
1838 stops on @ and $, but not for $ as tail anchor
1840 In transliterations:
1841 characters are VERY literal, except for - not at the start or end
1842 of the string, which indicates a range. If the range is in bytes,
1843 scan_const expands the range to the full set of intermediate
1844 characters. If the range is in utf8, the hyphen is replaced with
1845 a certain range mark which will be handled by pmtrans() in op.c.
1847 In double-quoted strings:
1849 double-quoted style: \r and \n
1851 deprecated backrefs: \1 (in substitution replacements)
1852 case and quoting: \U \Q \E
1855 scan_const does *not* construct ops to handle interpolated strings.
1856 It stops processing as soon as it finds an embedded $ or @ variable
1857 and leaves it to the caller to work out what's going on.
1859 embedded arrays (whether in pattern or not) could be:
1860 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1862 $ in double-quoted strings must be the symbol of an embedded scalar.
1864 $ in pattern could be $foo or could be tail anchor. Assumption:
1865 it's a tail anchor if $ is the last thing in the string, or if it's
1866 followed by one of "()| \r\n\t"
1868 \1 (backreferences) are turned into $1
1870 The structure of the code is
1871 while (there's a character to process) {
1872 handle transliteration ranges
1873 skip regexp comments /(?#comment)/ and codes /(?{code})/
1874 skip #-initiated comments in //x patterns
1875 check for embedded arrays
1876 check for embedded scalars
1878 leave intact backslashes from leaveit (below)
1879 deprecate \1 in substitution replacements
1880 handle string-changing backslashes \l \U \Q \E, etc.
1881 switch (what was escaped) {
1882 handle \- in a transliteration (becomes a literal -)
1883 handle \132 (octal characters)
1884 handle \x15 and \x{1234} (hex characters)
1885 handle \N{name} (named characters)
1886 handle \cV (control characters)
1887 handle printf-style backslashes (\f, \r, \n, etc)
1889 } (end if backslash)
1890 } (end while character to read)
1895 S_scan_const(pTHX_ char *start)
1898 register char *send = PL_bufend; /* end of the constant */
1899 SV *sv = newSV(send - start); /* sv for the constant */
1900 register char *s = start; /* start of the constant */
1901 register char *d = SvPVX(sv); /* destination for copies */
1902 bool dorange = FALSE; /* are we in a translit range? */
1903 bool didrange = FALSE; /* did we just finish a range? */
1904 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1905 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1908 UV literal_endpoint = 0;
1909 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1912 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1913 /* If we are doing a trans and we know we want UTF8 set expectation */
1914 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1915 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1919 while (s < send || dorange) {
1920 /* get transliterations out of the way (they're most literal) */
1921 if (PL_lex_inwhat == OP_TRANS) {
1922 /* expand a range A-Z to the full set of characters. AIE! */
1924 I32 i; /* current expanded character */
1925 I32 min; /* first character in range */
1926 I32 max; /* last character in range */
1937 char * const c = (char*)utf8_hop((U8*)d, -1);
1941 *c = (char)UTF_TO_NATIVE(0xff);
1942 /* mark the range as done, and continue */
1948 i = d - SvPVX_const(sv); /* remember current offset */
1951 SvLEN(sv) + (has_utf8 ?
1952 (512 - UTF_CONTINUATION_MARK +
1955 /* How many two-byte within 0..255: 128 in UTF-8,
1956 * 96 in UTF-8-mod. */
1958 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1960 d = SvPVX(sv) + i; /* refresh d after realloc */
1964 for (j = 0; j <= 1; j++) {
1965 char * const c = (char*)utf8_hop((U8*)d, -1);
1966 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1972 max = (U8)0xff; /* only to \xff */
1973 uvmax = uv; /* \x{100} to uvmax */
1975 d = c; /* eat endpoint chars */
1980 d -= 2; /* eat the first char and the - */
1981 min = (U8)*d; /* first char in range */
1982 max = (U8)d[1]; /* last char in range */
1989 "Invalid range \"%c-%c\" in transliteration operator",
1990 (char)min, (char)max);
1994 if (literal_endpoint == 2 &&
1995 ((isLOWER(min) && isLOWER(max)) ||
1996 (isUPPER(min) && isUPPER(max)))) {
1998 for (i = min; i <= max; i++)
2000 *d++ = NATIVE_TO_NEED(has_utf8,i);
2002 for (i = min; i <= max; i++)
2004 *d++ = NATIVE_TO_NEED(has_utf8,i);
2009 for (i = min; i <= max; i++)
2012 const U8 ch = (U8)NATIVE_TO_UTF(i);
2013 if (UNI_IS_INVARIANT(ch))
2016 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2017 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2026 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2028 *d++ = (char)UTF_TO_NATIVE(0xff);
2030 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2034 /* mark the range as done, and continue */
2038 literal_endpoint = 0;
2043 /* range begins (ignore - as first or last char) */
2044 else if (*s == '-' && s+1 < send && s != start) {
2046 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2053 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2063 literal_endpoint = 0;
2064 native_range = TRUE;
2069 /* if we get here, we're not doing a transliteration */
2071 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2072 except for the last char, which will be done separately. */
2073 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2075 while (s+1 < send && *s != ')')
2076 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2078 else if (s[2] == '{' /* This should match regcomp.c */
2079 || (s[2] == '?' && s[3] == '{'))
2082 char *regparse = s + (s[2] == '{' ? 3 : 4);
2085 while (count && (c = *regparse)) {
2086 if (c == '\\' && regparse[1])
2094 if (*regparse != ')')
2095 regparse--; /* Leave one char for continuation. */
2096 while (s < regparse)
2097 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2101 /* likewise skip #-initiated comments in //x patterns */
2102 else if (*s == '#' && PL_lex_inpat &&
2103 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2104 while (s+1 < send && *s != '\n')
2105 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2108 /* check for embedded arrays
2109 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2111 else if (*s == '@' && s[1]) {
2112 if (isALNUM_lazy_if(s+1,UTF))
2114 if (strchr(":'{$", s[1]))
2116 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2117 break; /* in regexp, neither @+ nor @- are interpolated */
2120 /* check for embedded scalars. only stop if we're sure it's a
2123 else if (*s == '$') {
2124 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2126 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2127 break; /* in regexp, $ might be tail anchor */
2130 /* End of else if chain - OP_TRANS rejoin rest */
2133 if (*s == '\\' && s+1 < send) {
2136 /* deprecate \1 in strings and substitution replacements */
2137 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2138 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2140 if (ckWARN(WARN_SYNTAX))
2141 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2146 /* string-change backslash escapes */
2147 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2151 /* skip any other backslash escapes in a pattern */
2152 else if (PL_lex_inpat) {
2153 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2154 goto default_action;
2157 /* if we get here, it's either a quoted -, or a digit */
2160 /* quoted - in transliterations */
2162 if (PL_lex_inwhat == OP_TRANS) {
2169 if ((isALPHA(*s) || isDIGIT(*s)) &&
2171 Perl_warner(aTHX_ packWARN(WARN_MISC),
2172 "Unrecognized escape \\%c passed through",
2174 /* default action is to copy the quoted character */
2175 goto default_action;
2178 /* \132 indicates an octal constant */
2179 case '0': case '1': case '2': case '3':
2180 case '4': case '5': case '6': case '7':
2184 uv = grok_oct(s, &len, &flags, NULL);
2187 goto NUM_ESCAPE_INSERT;
2189 /* \x24 indicates a hex constant */
2193 char* const e = strchr(s, '}');
2194 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2195 PERL_SCAN_DISALLOW_PREFIX;
2200 yyerror("Missing right brace on \\x{}");
2204 uv = grok_hex(s, &len, &flags, NULL);
2210 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2211 uv = grok_hex(s, &len, &flags, NULL);
2217 /* Insert oct or hex escaped character.
2218 * There will always enough room in sv since such
2219 * escapes will be longer than any UTF-8 sequence
2220 * they can end up as. */
2222 /* We need to map to chars to ASCII before doing the tests
2225 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2226 if (!has_utf8 && uv > 255) {
2227 /* Might need to recode whatever we have
2228 * accumulated so far if it contains any
2231 * (Can't we keep track of that and avoid
2232 * this rescan? --jhi)
2236 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2237 if (!NATIVE_IS_INVARIANT(*c)) {
2242 const STRLEN offset = d - SvPVX_const(sv);
2244 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2248 while (src >= (const U8 *)SvPVX_const(sv)) {
2249 if (!NATIVE_IS_INVARIANT(*src)) {
2250 const U8 ch = NATIVE_TO_ASCII(*src);
2251 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2252 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2262 if (has_utf8 || uv > 255) {
2263 d = (char*)uvchr_to_utf8((U8*)d, uv);
2265 if (PL_lex_inwhat == OP_TRANS &&
2266 PL_sublex_info.sub_op) {
2267 PL_sublex_info.sub_op->op_private |=
2268 (PL_lex_repl ? OPpTRANS_FROM_UTF
2272 if (uv > 255 && !dorange)
2273 native_range = FALSE;
2285 /* \N{LATIN SMALL LETTER A} is a named character */
2289 char* e = strchr(s, '}');
2295 yyerror("Missing right brace on \\N{}");
2299 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2301 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2302 PERL_SCAN_DISALLOW_PREFIX;
2305 uv = grok_hex(s, &len, &flags, NULL);
2306 if ( e > s && len != (STRLEN)(e - s) ) {
2310 goto NUM_ESCAPE_INSERT;
2312 res = newSVpvn(s + 1, e - s - 1);
2313 res = new_constant( NULL, 0, "charnames",
2314 res, NULL, s - 2, e - s + 3 );
2316 sv_utf8_upgrade(res);
2317 str = SvPV_const(res,len);
2318 #ifdef EBCDIC_NEVER_MIND
2319 /* charnames uses pack U and that has been
2320 * recently changed to do the below uni->native
2321 * mapping, so this would be redundant (and wrong,
2322 * the code point would be doubly converted).
2323 * But leave this in just in case the pack U change
2324 * gets revoked, but the semantics is still
2325 * desireable for charnames. --jhi */
2327 UV uv = utf8_to_uvchr((const U8*)str, 0);
2330 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2332 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2333 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2334 str = SvPV_const(res, len);
2338 if (!has_utf8 && SvUTF8(res)) {
2339 const char * const ostart = SvPVX_const(sv);
2340 SvCUR_set(sv, d - ostart);
2343 sv_utf8_upgrade(sv);
2344 /* this just broke our allocation above... */
2345 SvGROW(sv, (STRLEN)(send - start));
2346 d = SvPVX(sv) + SvCUR(sv);
2349 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2350 const char * const odest = SvPVX_const(sv);
2352 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2353 d = SvPVX(sv) + (d - odest);
2357 native_range = FALSE; /* \N{} is guessed to be Unicode */
2359 Copy(str, d, len, char);
2366 yyerror("Missing braces on \\N{}");
2369 /* \c is a control character */
2378 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2381 yyerror("Missing control char name in \\c");
2385 /* printf-style backslashes, formfeeds, newlines, etc */
2387 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2390 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2393 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2396 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2399 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2402 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2405 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2411 } /* end if (backslash) */
2418 /* If we started with encoded form, or already know we want it
2419 and then encode the next character */
2420 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2422 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2423 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2426 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2427 const STRLEN off = d - SvPVX_const(sv);
2428 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2430 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2433 if (uv > 255 && !dorange)
2434 native_range = FALSE;
2438 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2440 } /* while loop to process each character */
2442 /* terminate the string and set up the sv */
2444 SvCUR_set(sv, d - SvPVX_const(sv));
2445 if (SvCUR(sv) >= SvLEN(sv))
2446 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2449 if (PL_encoding && !has_utf8) {
2450 sv_recode_to_utf8(sv, PL_encoding);
2456 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2457 PL_sublex_info.sub_op->op_private |=
2458 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2462 /* shrink the sv if we allocated more than we used */
2463 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2464 SvPV_shrink_to_cur(sv);
2467 /* return the substring (via pl_yylval) only if we parsed anything */
2468 if (s > PL_bufptr) {
2469 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
2470 const char *const key = PL_lex_inpat ? "qr" : "q";
2471 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
2475 if (PL_lex_inwhat == OP_TRANS) {
2478 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
2486 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
2489 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2496 * Returns TRUE if there's more to the expression (e.g., a subscript),
2499 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2501 * ->[ and ->{ return TRUE
2502 * { and [ outside a pattern are always subscripts, so return TRUE
2503 * if we're outside a pattern and it's not { or [, then return FALSE
2504 * if we're in a pattern and the first char is a {
2505 * {4,5} (any digits around the comma) returns FALSE
2506 * if we're in a pattern and the first char is a [
2508 * [SOMETHING] has a funky algorithm to decide whether it's a
2509 * character class or not. It has to deal with things like
2510 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2511 * anything else returns TRUE
2514 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2517 S_intuit_more(pTHX_ register char *s)
2520 if (PL_lex_brackets)
2522 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2524 if (*s != '{' && *s != '[')
2529 /* In a pattern, so maybe we have {n,m}. */
2546 /* On the other hand, maybe we have a character class */
2549 if (*s == ']' || *s == '^')
2552 /* this is terrifying, and it works */
2553 int weight = 2; /* let's weigh the evidence */
2555 unsigned char un_char = 255, last_un_char;
2556 const char * const send = strchr(s,']');
2557 char tmpbuf[sizeof PL_tokenbuf * 4];
2559 if (!send) /* has to be an expression */
2562 Zero(seen,256,char);
2565 else if (isDIGIT(*s)) {
2567 if (isDIGIT(s[1]) && s[2] == ']')
2573 for (; s < send; s++) {
2574 last_un_char = un_char;
2575 un_char = (unsigned char)*s;
2580 weight -= seen[un_char] * 10;
2581 if (isALNUM_lazy_if(s+1,UTF)) {
2583 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2584 len = (int)strlen(tmpbuf);
2585 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2590 else if (*s == '$' && s[1] &&
2591 strchr("[#!%*<>()-=",s[1])) {
2592 if (/*{*/ strchr("])} =",s[2]))
2601 if (strchr("wds]",s[1]))
2603 else if (seen[(U8)'\''] || seen[(U8)'"'])
2605 else if (strchr("rnftbxcav",s[1]))
2607 else if (isDIGIT(s[1])) {
2609 while (s[1] && isDIGIT(s[1]))
2619 if (strchr("aA01! ",last_un_char))
2621 if (strchr("zZ79~",s[1]))
2623 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2624 weight -= 5; /* cope with negative subscript */
2627 if (!isALNUM(last_un_char)
2628 && !(last_un_char == '$' || last_un_char == '@'
2629 || last_un_char == '&')
2630 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2635 if (keyword(tmpbuf, d - tmpbuf, 0))
2638 if (un_char == last_un_char + 1)
2640 weight -= seen[un_char];
2645 if (weight >= 0) /* probably a character class */
2655 * Does all the checking to disambiguate
2657 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2658 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2660 * First argument is the stuff after the first token, e.g. "bar".
2662 * Not a method if bar is a filehandle.
2663 * Not a method if foo is a subroutine prototyped to take a filehandle.
2664 * Not a method if it's really "Foo $bar"
2665 * Method if it's "foo $bar"
2666 * Not a method if it's really "print foo $bar"
2667 * Method if it's really "foo package::" (interpreted as package->foo)
2668 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2669 * Not a method if bar is a filehandle or package, but is quoted with
2674 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2677 char *s = start + (*start == '$');
2678 char tmpbuf[sizeof PL_tokenbuf];
2686 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2690 const char *proto = SvPVX_const(cv);
2701 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2702 /* start is the beginning of the possible filehandle/object,
2703 * and s is the end of it
2704 * tmpbuf is a copy of it
2707 if (*start == '$') {
2708 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
2709 isUPPER(*PL_tokenbuf))
2712 len = start - SvPVX(PL_linestr);
2716 start = SvPVX(PL_linestr) + len;
2720 return *s == '(' ? FUNCMETH : METHOD;
2722 if (!keyword(tmpbuf, len, 0)) {
2723 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2727 soff = s - SvPVX(PL_linestr);
2731 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2732 if (indirgv && GvCVu(indirgv))
2734 /* filehandle or package name makes it a method */
2735 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2737 soff = s - SvPVX(PL_linestr);
2740 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2741 return 0; /* no assumptions -- "=>" quotes bearword */
2743 start_force(PL_curforce);
2744 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2745 newSVpvn(tmpbuf,len));
2746 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2748 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2753 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2755 return *s == '(' ? FUNCMETH : METHOD;
2761 /* Encoded script support. filter_add() effectively inserts a
2762 * 'pre-processing' function into the current source input stream.
2763 * Note that the filter function only applies to the current source file
2764 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2766 * The datasv parameter (which may be NULL) can be used to pass
2767 * private data to this instance of the filter. The filter function
2768 * can recover the SV using the FILTER_DATA macro and use it to
2769 * store private buffers and state information.
2771 * The supplied datasv parameter is upgraded to a PVIO type
2772 * and the IoDIRP/IoANY field is used to store the function pointer,
2773 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2774 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2775 * private use must be set using malloc'd pointers.
2779 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2788 if (!PL_rsfp_filters)
2789 PL_rsfp_filters = newAV();
2792 SvUPGRADE(datasv, SVt_PVIO);
2793 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2794 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2795 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2796 FPTR2DPTR(void *, IoANY(datasv)),
2797 SvPV_nolen(datasv)));
2798 av_unshift(PL_rsfp_filters, 1);
2799 av_store(PL_rsfp_filters, 0, datasv) ;
2804 /* Delete most recently added instance of this filter function. */
2806 Perl_filter_del(pTHX_ filter_t funcp)
2812 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2813 FPTR2DPTR(void*, funcp)));
2815 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2817 /* if filter is on top of stack (usual case) just pop it off */
2818 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2819 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2820 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2821 IoANY(datasv) = (void *)NULL;
2822 sv_free(av_pop(PL_rsfp_filters));
2826 /* we need to search for the correct entry and clear it */
2827 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2831 /* Invoke the idxth filter function for the current rsfp. */
2832 /* maxlen 0 = read one text line */
2834 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2839 /* This API is bad. It should have been using unsigned int for maxlen.
2840 Not sure if we want to change the API, but if not we should sanity
2841 check the value here. */
2842 const unsigned int correct_length
2851 if (!PL_parser || !PL_rsfp_filters)
2853 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2854 /* Provide a default input filter to make life easy. */
2855 /* Note that we append to the line. This is handy. */
2856 DEBUG_P(PerlIO_printf(Perl_debug_log,
2857 "filter_read %d: from rsfp\n", idx));
2858 if (correct_length) {
2861 const int old_len = SvCUR(buf_sv);
2863 /* ensure buf_sv is large enough */
2864 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2865 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2866 correct_length)) <= 0) {
2867 if (PerlIO_error(PL_rsfp))
2868 return -1; /* error */
2870 return 0 ; /* end of file */
2872 SvCUR_set(buf_sv, old_len + len) ;
2875 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2876 if (PerlIO_error(PL_rsfp))
2877 return -1; /* error */
2879 return 0 ; /* end of file */
2882 return SvCUR(buf_sv);
2884 /* Skip this filter slot if filter has been deleted */
2885 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2886 DEBUG_P(PerlIO_printf(Perl_debug_log,
2887 "filter_read %d: skipped (filter deleted)\n",
2889 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2891 /* Get function pointer hidden within datasv */
2892 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2893 DEBUG_P(PerlIO_printf(Perl_debug_log,
2894 "filter_read %d: via function %p (%s)\n",
2895 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2896 /* Call function. The function is expected to */
2897 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2898 /* Return: <0:error, =0:eof, >0:not eof */
2899 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2903 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2906 #ifdef PERL_CR_FILTER
2907 if (!PL_rsfp_filters) {
2908 filter_add(S_cr_textfilter,NULL);
2911 if (PL_rsfp_filters) {
2913 SvCUR_set(sv, 0); /* start with empty line */
2914 if (FILTER_READ(0, sv, 0) > 0)
2915 return ( SvPVX(sv) ) ;
2920 return (sv_gets(sv, fp, append));
2924 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
2929 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2933 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2934 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2936 return GvHV(gv); /* Foo:: */
2939 /* use constant CLASS => 'MyClass' */
2940 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2941 if (gv && GvCV(gv)) {
2942 SV * const sv = cv_const_sv(GvCV(gv));
2944 pkgname = SvPV_const(sv, len);
2947 return gv_stashpvn(pkgname, len, 0);
2951 * S_readpipe_override
2952 * Check whether readpipe() is overriden, and generates the appropriate
2953 * optree, provided sublex_start() is called afterwards.
2956 S_readpipe_override(pTHX)
2959 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2960 pl_yylval.ival = OP_BACKTICK;
2962 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2964 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2965 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
2966 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2968 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2969 append_elem(OP_LIST,
2970 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2971 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2978 * The intent of this yylex wrapper is to minimize the changes to the
2979 * tokener when we aren't interested in collecting madprops. It remains
2980 * to be seen how successful this strategy will be...
2987 char *s = PL_bufptr;
2989 /* make sure PL_thiswhite is initialized */
2993 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2994 if (PL_pending_ident)
2995 return S_pending_ident(aTHX);
2997 /* previous token ate up our whitespace? */
2998 if (!PL_lasttoke && PL_nextwhite) {
2999 PL_thiswhite = PL_nextwhite;
3003 /* isolate the token, and figure out where it is without whitespace */
3004 PL_realtokenstart = -1;
3008 assert(PL_curforce < 0);
3010 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3011 if (!PL_thistoken) {
3012 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3013 PL_thistoken = newSVpvs("");
3015 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3016 PL_thistoken = newSVpvn(tstart, s - tstart);
3019 if (PL_thismad) /* install head */
3020 CURMAD('X', PL_thistoken);
3023 /* last whitespace of a sublex? */
3024 if (optype == ')' && PL_endwhite) {
3025 CURMAD('X', PL_endwhite);
3030 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3031 if (!PL_thiswhite && !PL_endwhite && !optype) {
3032 sv_free(PL_thistoken);
3037 /* put off final whitespace till peg */
3038 if (optype == ';' && !PL_rsfp) {
3039 PL_nextwhite = PL_thiswhite;
3042 else if (PL_thisopen) {
3043 CURMAD('q', PL_thisopen);
3045 sv_free(PL_thistoken);
3049 /* Store actual token text as madprop X */
3050 CURMAD('X', PL_thistoken);
3054 /* add preceding whitespace as madprop _ */
3055 CURMAD('_', PL_thiswhite);
3059 /* add quoted material as madprop = */
3060 CURMAD('=', PL_thisstuff);
3064 /* add terminating quote as madprop Q */
3065 CURMAD('Q', PL_thisclose);
3069 /* special processing based on optype */
3073 /* opval doesn't need a TOKEN since it can already store mp */
3083 if (pl_yylval.opval)
3084 append_madprops(PL_thismad, pl_yylval.opval, 0);
3092 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3101 /* remember any fake bracket that lexer is about to discard */
3102 if (PL_lex_brackets == 1 &&
3103 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3106 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3109 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3110 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3113 break; /* don't bother looking for trailing comment */
3122 /* attach a trailing comment to its statement instead of next token */
3126 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3128 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3130 if (*s == '\n' || *s == '#') {
3131 while (s < PL_bufend && *s != '\n')
3135 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3136 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3153 /* Create new token struct. Note: opvals return early above. */
3154 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
3161 S_tokenize_use(pTHX_ int is_use, char *s) {
3163 if (PL_expect != XSTATE)
3164 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3165 is_use ? "use" : "no"));
3167 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3168 s = force_version(s, TRUE);
3169 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3170 start_force(PL_curforce);
3171 NEXTVAL_NEXTTOKE.opval = NULL;
3174 else if (*s == 'v') {
3175 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3176 s = force_version(s, FALSE);
3180 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3181 s = force_version(s, FALSE);
3183 pl_yylval.ival = is_use;
3187 static const char* const exp_name[] =
3188 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3189 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3196 Works out what to call the token just pulled out of the input
3197 stream. The yacc parser takes care of taking the ops we return and
3198 stitching them into a tree.
3204 if read an identifier
3205 if we're in a my declaration
3206 croak if they tried to say my($foo::bar)
3207 build the ops for a my() declaration
3208 if it's an access to a my() variable
3209 are we in a sort block?
3210 croak if my($a); $a <=> $b
3211 build ops for access to a my() variable
3212 if in a dq string, and they've said @foo and we can't find @foo
3214 build ops for a bareword
3215 if we already built the token before, use it.
3220 #pragma segment Perl_yylex
3226 register char *s = PL_bufptr;
3231 /* orig_keyword, gvp, and gv are initialized here because
3232 * jump to the label just_a_word_zero can bypass their
3233 * initialization later. */
3234 I32 orig_keyword = 0;
3239 SV* tmp = newSVpvs("");
3240 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3241 (IV)CopLINE(PL_curcop),
3242 lex_state_names[PL_lex_state],
3243 exp_name[PL_expect],
3244 pv_display(tmp, s, strlen(s), 0, 60));
3247 /* check if there's an identifier for us to look at */
3248 if (PL_pending_ident)
3249 return REPORT(S_pending_ident(aTHX));
3251 /* no identifier pending identification */
3253 switch (PL_lex_state) {
3255 case LEX_NORMAL: /* Some compilers will produce faster */
3256 case LEX_INTERPNORMAL: /* code if we comment these out. */
3260 /* when we've already built the next token, just pull it out of the queue */
3264 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
3266 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3267 PL_nexttoke[PL_lasttoke].next_mad = 0;
3268 if (PL_thismad && PL_thismad->mad_key == '_') {
3269 PL_thiswhite = (SV*)PL_thismad->mad_val;
3270 PL_thismad->mad_val = 0;
3271 mad_free(PL_thismad);
3276 PL_lex_state = PL_lex_defer;
3277 PL_expect = PL_lex_expect;
3278 PL_lex_defer = LEX_NORMAL;
3279 if (!PL_nexttoke[PL_lasttoke].next_type)
3284 pl_yylval = PL_nextval[PL_nexttoke];
3286 PL_lex_state = PL_lex_defer;
3287 PL_expect = PL_lex_expect;
3288 PL_lex_defer = LEX_NORMAL;
3292 /* FIXME - can these be merged? */
3293 return(PL_nexttoke[PL_lasttoke].next_type);
3295 return REPORT(PL_nexttype[PL_nexttoke]);
3298 /* interpolated case modifiers like \L \U, including \Q and \E.
3299 when we get here, PL_bufptr is at the \
3301 case LEX_INTERPCASEMOD:
3303 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3304 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3306 /* handle \E or end of string */
3307 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3309 if (PL_lex_casemods) {
3310 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3311 PL_lex_casestack[PL_lex_casemods] = '\0';
3313 if (PL_bufptr != PL_bufend
3314 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3316 PL_lex_state = LEX_INTERPCONCAT;
3319 PL_thistoken = newSVpvs("\\E");
3325 while (PL_bufptr != PL_bufend &&
3326 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3328 PL_thiswhite = newSVpvs("");
3329 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3333 if (PL_bufptr != PL_bufend)
3336 PL_lex_state = LEX_INTERPCONCAT;
3340 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3341 "### Saw case modifier\n"); });
3343 if (s[1] == '\\' && s[2] == 'E') {
3346 PL_thiswhite = newSVpvs("");
3347 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3350 PL_lex_state = LEX_INTERPCONCAT;
3355 if (!PL_madskills) /* when just compiling don't need correct */
3356 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3357 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3358 if ((*s == 'L' || *s == 'U') &&
3359 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3360 PL_lex_casestack[--PL_lex_casemods] = '\0';
3363 if (PL_lex_casemods > 10)
3364 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3365 PL_lex_casestack[PL_lex_casemods++] = *s;
3366 PL_lex_casestack[PL_lex_casemods] = '\0';
3367 PL_lex_state = LEX_INTERPCONCAT;
3368 start_force(PL_curforce);
3369 NEXTVAL_NEXTTOKE.ival = 0;
3371 start_force(PL_curforce);
3373 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3375 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3377 NEXTVAL_NEXTTOKE.ival = OP_LC;
3379 NEXTVAL_NEXTTOKE.ival = OP_UC;
3381 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3383 Perl_croak(aTHX_ "panic: yylex");
3385 SV* const tmpsv = newSVpvs("\\ ");
3386 /* replace the space with the character we want to escape
3388 SvPVX(tmpsv)[1] = *s;
3394 if (PL_lex_starts) {
3400 sv_free(PL_thistoken);
3401 PL_thistoken = newSVpvs("");
3404 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3405 if (PL_lex_casemods == 1 && PL_lex_inpat)
3414 case LEX_INTERPPUSH:
3415 return REPORT(sublex_push());
3417 case LEX_INTERPSTART:
3418 if (PL_bufptr == PL_bufend)
3419 return REPORT(sublex_done());
3420 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3421 "### Interpolated variable\n"); });
3423 PL_lex_dojoin = (*PL_bufptr == '@');
3424 PL_lex_state = LEX_INTERPNORMAL;
3425 if (PL_lex_dojoin) {
3426 start_force(PL_curforce);
3427 NEXTVAL_NEXTTOKE.ival = 0;
3429 start_force(PL_curforce);
3430 force_ident("\"", '$');
3431 start_force(PL_curforce);
3432 NEXTVAL_NEXTTOKE.ival = 0;
3434 start_force(PL_curforce);
3435 NEXTVAL_NEXTTOKE.ival = 0;
3437 start_force(PL_curforce);
3438 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3441 if (PL_lex_starts++) {
3446 sv_free(PL_thistoken);
3447 PL_thistoken = newSVpvs("");
3450 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3451 if (!PL_lex_casemods && PL_lex_inpat)
3458 case LEX_INTERPENDMAYBE:
3459 if (intuit_more(PL_bufptr)) {
3460 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3466 if (PL_lex_dojoin) {
3467 PL_lex_dojoin = FALSE;
3468 PL_lex_state = LEX_INTERPCONCAT;
3472 sv_free(PL_thistoken);
3473 PL_thistoken = newSVpvs("");
3478 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3479 && SvEVALED(PL_lex_repl))
3481 if (PL_bufptr != PL_bufend)
3482 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3486 case LEX_INTERPCONCAT:
3488 if (PL_lex_brackets)
3489 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3491 if (PL_bufptr == PL_bufend)
3492 return REPORT(sublex_done());
3494 if (SvIVX(PL_linestr) == '\'') {
3495 SV *sv = newSVsv(PL_linestr);
3498 else if ( PL_hints & HINT_NEW_RE )
3499 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
3500 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3504 s = scan_const(PL_bufptr);
3506 PL_lex_state = LEX_INTERPCASEMOD;
3508 PL_lex_state = LEX_INTERPSTART;
3511 if (s != PL_bufptr) {
3512 start_force(PL_curforce);
3514 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3516 NEXTVAL_NEXTTOKE = pl_yylval;
3519 if (PL_lex_starts++) {
3523 sv_free(PL_thistoken);
3524 PL_thistoken = newSVpvs("");
3527 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3528 if (!PL_lex_casemods && PL_lex_inpat)
3541 PL_lex_state = LEX_NORMAL;
3542 s = scan_formline(PL_bufptr);
3543 if (!PL_lex_formbrack)
3549 PL_oldoldbufptr = PL_oldbufptr;
3555 sv_free(PL_thistoken);
3558 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3562 if (isIDFIRST_lazy_if(s,UTF))
3564 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3565 Perl_croak(aTHX_ "Unrecognized character \\x%02X in column %d", *s & 255, (int) len + 1);
3568 goto fake_eof; /* emulate EOF on ^D or ^Z */
3577 if (PL_lex_brackets) {
3578 yyerror((const char *)
3580 ? "Format not terminated"
3581 : "Missing right curly or square bracket"));
3583 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3584 "### Tokener got EOF\n");
3588 if (s++ < PL_bufend)
3589 goto retry; /* ignore stray nulls */
3592 if (!PL_in_eval && !PL_preambled) {
3593 PL_preambled = TRUE;
3599 /* Generate a string of Perl code to load the debugger.
3600 * If PERL5DB is set, it will return the contents of that,
3601 * otherwise a compile-time require of perl5db.pl. */
3603 const char * const pdb = PerlEnv_getenv("PERL5DB");
3606 sv_setpv(PL_linestr, pdb);
3607 sv_catpvs(PL_linestr,";");
3609 SETERRNO(0,SS_NORMAL);
3610 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
3613 sv_setpvs(PL_linestr,"");
3614 if (PL_preambleav) {
3615 SV **svp = AvARRAY(PL_preambleav);
3616 SV **const end = svp + AvFILLp(PL_preambleav);
3618 sv_catsv(PL_linestr, *svp);
3620 sv_catpvs(PL_linestr, ";");
3622 sv_free((SV*)PL_preambleav);
3623 PL_preambleav = NULL;
3625 if (PL_minus_n || PL_minus_p) {
3626 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3628 sv_catpvs(PL_linestr,"chomp;");
3631 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3632 || *PL_splitstr == '"')
3633 && strchr(PL_splitstr + 1, *PL_splitstr))
3634 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3636 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3637 bytes can be used as quoting characters. :-) */
3638 const char *splits = PL_splitstr;
3639 sv_catpvs(PL_linestr, "our @F=split(q\0");
3642 if (*splits == '\\')
3643 sv_catpvn(PL_linestr, splits, 1);
3644 sv_catpvn(PL_linestr, splits, 1);
3645 } while (*splits++);
3646 /* This loop will embed the trailing NUL of
3647 PL_linestr as the last thing it does before
3649 sv_catpvs(PL_linestr, ");");
3653 sv_catpvs(PL_linestr,"our @F=split(' ');");
3657 sv_catpvs(PL_linestr,"use feature ':5.11';");
3658 sv_catpvs(PL_linestr, "\n");
3659 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3660 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3661 PL_last_lop = PL_last_uni = NULL;
3662 if (PERLDB_LINE && PL_curstash != PL_debstash)
3663 update_debugger_info(PL_linestr, NULL, 0);
3667 bof = PL_rsfp ? TRUE : FALSE;
3668 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3671 PL_realtokenstart = -1;
3674 if (PL_preprocess && !PL_in_eval)
3675 (void)PerlProc_pclose(PL_rsfp);
3676 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3677 PerlIO_clearerr(PL_rsfp);
3679 (void)PerlIO_close(PL_rsfp);
3681 PL_doextract = FALSE;
3683 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3689 sv_setpvs(PL_linestr, ";}continue{print;}");
3691 sv_setpvs(PL_linestr, ";}");
3692 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3693 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3694 PL_last_lop = PL_last_uni = NULL;
3695 PL_minus_n = PL_minus_p = 0;
3698 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3699 PL_last_lop = PL_last_uni = NULL;
3700 sv_setpvn(PL_linestr,"",0);
3701 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3703 /* If it looks like the start of a BOM or raw UTF-16,
3704 * check if it in fact is. */
3710 #ifdef PERLIO_IS_STDIO
3711 # ifdef __GNU_LIBRARY__
3712 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3713 # define FTELL_FOR_PIPE_IS_BROKEN
3717 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3718 # define FTELL_FOR_PIPE_IS_BROKEN
3723 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3724 /* This loses the possibility to detect the bof
3725 * situation on perl -P when the libc5 is being used.
3726 * Workaround? Maybe attach some extra state to PL_rsfp?
3729 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3731 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3734 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3735 s = swallow_bom((U8*)s);
3739 /* Incest with pod. */
3742 sv_catsv(PL_thiswhite, PL_linestr);
3744 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3745 sv_setpvn(PL_linestr, "", 0);
3746 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3747 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3748 PL_last_lop = PL_last_uni = NULL;
3749 PL_doextract = FALSE;
3753 } while (PL_doextract);
3754 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3755 if (PERLDB_LINE && PL_curstash != PL_debstash)
3756 update_debugger_info(PL_linestr, NULL, 0);
3757 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3758 PL_last_lop = PL_last_uni = NULL;
3759 if (CopLINE(PL_curcop) == 1) {
3760 while (s < PL_bufend && isSPACE(*s))
3762 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3766 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3770 if (*s == '#' && *(s+1) == '!')
3772 #ifdef ALTERNATE_SHEBANG
3774 static char const as[] = ALTERNATE_SHEBANG;
3775 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3776 d = s + (sizeof(as) - 1);
3778 #endif /* ALTERNATE_SHEBANG */
3787 while (*d && !isSPACE(*d))
3791 #ifdef ARG_ZERO_IS_SCRIPT
3792 if (ipathend > ipath) {
3794 * HP-UX (at least) sets argv[0] to the script name,
3795 * which makes $^X incorrect. And Digital UNIX and Linux,
3796 * at least, set argv[0] to the basename of the Perl
3797 * interpreter. So, having found "#!", we'll set it right.
3799 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3801 assert(SvPOK(x) || SvGMAGICAL(x));
3802 if (sv_eq(x, CopFILESV(PL_curcop))) {
3803 sv_setpvn(x, ipath, ipathend - ipath);
3809 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3810 const char * const lstart = SvPV_const(x,llen);
3812 bstart += blen - llen;
3813 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3814 sv_setpvn(x, ipath, ipathend - ipath);
3819 TAINT_NOT; /* $^X is always tainted, but that's OK */
3821 #endif /* ARG_ZERO_IS_SCRIPT */
3826 d = instr(s,"perl -");
3828 d = instr(s,"perl");
3830 /* avoid getting into infinite loops when shebang
3831 * line contains "Perl" rather than "perl" */
3833 for (d = ipathend-4; d >= ipath; --d) {
3834 if ((*d == 'p' || *d == 'P')
3835 && !ibcmp(d, "perl", 4))
3845 #ifdef ALTERNATE_SHEBANG
3847 * If the ALTERNATE_SHEBANG on this system starts with a
3848 * character that can be part of a Perl expression, then if
3849 * we see it but not "perl", we're probably looking at the
3850 * start of Perl code, not a request to hand off to some
3851 * other interpreter. Similarly, if "perl" is there, but
3852 * not in the first 'word' of the line, we assume the line
3853 * contains the start of the Perl program.
3855 if (d && *s != '#') {
3856 const char *c = ipath;
3857 while (*c && !strchr("; \t\r\n\f\v#", *c))
3860 d = NULL; /* "perl" not in first word; ignore */
3862 *s = '#'; /* Don't try to parse shebang line */
3864 #endif /* ALTERNATE_SHEBANG */
3865 #ifndef MACOS_TRADITIONAL
3870 !instr(s,"indir") &&
3871 instr(PL_origargv[0],"perl"))
3878 while (s < PL_bufend && isSPACE(*s))
3880 if (s < PL_bufend) {
3881 Newxz(newargv,PL_origargc+3,char*);
3883 while (s < PL_bufend && !isSPACE(*s))
3886 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3889 newargv = PL_origargv;
3892 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3894 Perl_croak(aTHX_ "Can't exec %s", ipath);
3898 while (*d && !isSPACE(*d))
3900 while (SPACE_OR_TAB(*d))
3904 const bool switches_done = PL_doswitches;
3905 const U32 oldpdb = PL_perldb;
3906 const bool oldn = PL_minus_n;
3907 const bool oldp = PL_minus_p;
3911 if (*d1 == 'M' || *d1 == 'm' || *d1 == 'C') {
3912 const char * const m = d1;
3913 while (*d1 && !isSPACE(*d1))
3915 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3918 d1 = moreswitches(d1);
3920 if (PL_doswitches && !switches_done) {
3921 int argc = PL_origargc;
3922 char **argv = PL_origargv;
3925 } while (argc && argv[0][0] == '-' && argv[0][1]);
3926 init_argv_symbols(argc,argv);
3928 if ((PERLDB_LINE && !oldpdb) ||
3929 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3930 /* if we have already added "LINE: while (<>) {",
3931 we must not do it again */
3933 sv_setpvn(PL_linestr, "", 0);
3934 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3935 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3936 PL_last_lop = PL_last_uni = NULL;
3937 PL_preambled = FALSE;
3939 (void)gv_fetchfile(PL_origfilename);
3946 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3948 PL_lex_state = LEX_FORMLINE;
3953 #ifdef PERL_STRICT_CR
3954 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3956 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3958 case ' ': case '\t': case '\f': case 013:
3959 #ifdef MACOS_TRADITIONAL
3963 PL_realtokenstart = -1;
3965 PL_thiswhite = newSVpvs("");
3966 sv_catpvn(PL_thiswhite, s, 1);
3973 PL_realtokenstart = -1;
3977 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3978 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3979 /* handle eval qq[#line 1 "foo"\n ...] */
3980 CopLINE_dec(PL_curcop);
3983 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3985 if (!PL_in_eval || PL_rsfp)
3990 while (d < PL_bufend && *d != '\n')
3994 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3995 Perl_croak(aTHX_ "panic: input overflow");
3998 PL_thiswhite = newSVpvn(s, d - s);
4003 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4005 PL_lex_state = LEX_FORMLINE;
4011 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4012 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4015 TOKEN(PEG); /* make sure any #! line is accessible */
4020 /* if (PL_madskills && PL_lex_formbrack) { */
4022 while (d < PL_bufend && *d != '\n')
4026 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4027 Perl_croak(aTHX_ "panic: input overflow");
4028 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4030 PL_thiswhite = newSVpvs("");
4031 if (CopLINE(PL_curcop) == 1) {
4032 sv_setpvn(PL_thiswhite, "", 0);
4035 sv_catpvn(PL_thiswhite, s, d - s);
4049 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4057 while (s < PL_bufend && SPACE_OR_TAB(*s))
4060 if (strnEQ(s,"=>",2)) {
4061 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4062 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4063 OPERATOR('-'); /* unary minus */
4065 PL_last_uni = PL_oldbufptr;
4067 case 'r': ftst = OP_FTEREAD; break;
4068 case 'w': ftst = OP_FTEWRITE; break;
4069 case 'x': ftst = OP_FTEEXEC; break;
4070 case 'o': ftst = OP_FTEOWNED; break;
4071 case 'R': ftst = OP_FTRREAD; break;
4072 case 'W': ftst = OP_FTRWRITE; break;
4073 case 'X': ftst = OP_FTREXEC; break;
4074 case 'O': ftst = OP_FTROWNED; break;
4075 case 'e': ftst = OP_FTIS; break;
4076 case 'z': ftst = OP_FTZERO; break;
4077 case 's': ftst = OP_FTSIZE; break;
4078 case 'f': ftst = OP_FTFILE; break;
4079 case 'd': ftst = OP_FTDIR; break;
4080 case 'l': ftst = OP_FTLINK; break;
4081 case 'p': ftst = OP_FTPIPE; break;
4082 case 'S': ftst = OP_FTSOCK; break;
4083 case 'u': ftst = OP_FTSUID; break;
4084 case 'g': ftst = OP_FTSGID; break;
4085 case 'k': ftst = OP_FTSVTX; break;
4086 case 'b': ftst = OP_FTBLK; break;
4087 case 'c': ftst = OP_FTCHR; break;
4088 case 't': ftst = OP_FTTTY; break;
4089 case 'T': ftst = OP_FTTEXT; break;
4090 case 'B': ftst = OP_FTBINARY; break;
4091 case 'M': case 'A': case 'C':
4092 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4094 case 'M': ftst = OP_FTMTIME; break;
4095 case 'A': ftst = OP_FTATIME; break;
4096 case 'C': ftst = OP_FTCTIME; break;
4104 PL_last_lop_op = (OPCODE)ftst;
4105 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4106 "### Saw file test %c\n", (int)tmp);
4111 /* Assume it was a minus followed by a one-letter named
4112 * subroutine call (or a -bareword), then. */
4113 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4114 "### '-%c' looked like a file test but was not\n",
4121 const char tmp = *s++;
4124 if (PL_expect == XOPERATOR)
4129 else if (*s == '>') {
4132 if (isIDFIRST_lazy_if(s,UTF)) {
4133 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4141 if (PL_expect == XOPERATOR)
4144 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4146 OPERATOR('-'); /* unary minus */
4152 const char tmp = *s++;
4155 if (PL_expect == XOPERATOR)
4160 if (PL_expect == XOPERATOR)
4163 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4170 if (PL_expect != XOPERATOR) {
4171 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4172 PL_expect = XOPERATOR;
4173 force_ident(PL_tokenbuf, '*');
4186 if (PL_expect == XOPERATOR) {
4190 PL_tokenbuf[0] = '%';
4191 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4192 sizeof PL_tokenbuf - 1, FALSE);
4193 if (!PL_tokenbuf[1]) {
4196 PL_pending_ident = '%';
4207 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4214 const char tmp = *s++;
4220 goto just_a_word_zero_gv;
4223 switch (PL_expect) {
4229 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4231 PL_bufptr = s; /* update in case we back off */
4237 PL_expect = XTERMBLOCK;
4240 stuffstart = s - SvPVX(PL_linestr) - 1;
4244 while (isIDFIRST_lazy_if(s,UTF)) {
4247 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4248 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4249 if (tmp < 0) tmp = -tmp;
4263 sv = newSVpvn(s, len);
4265 d = scan_str(d,TRUE,TRUE);
4267 /* MUST advance bufptr here to avoid bogus
4268 "at end of line" context messages from yyerror().
4270 PL_bufptr = s + len;
4271 yyerror("Unterminated attribute parameter in attribute list");
4275 return REPORT(0); /* EOF indicator */
4279 sv_catsv(sv, PL_lex_stuff);
4280 attrs = append_elem(OP_LIST, attrs,
4281 newSVOP(OP_CONST, 0, sv));
4282 SvREFCNT_dec(PL_lex_stuff);
4283 PL_lex_stuff = NULL;
4286 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4288 if (PL_in_my == KEY_our) {
4290 GvUNIQUE_on(cGVOPx_gv(pl_yylval.opval));
4292 /* skip to avoid loading attributes.pm */
4294 deprecate(":unique");
4297 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4300 /* NOTE: any CV attrs applied here need to be part of
4301 the CVf_BUILTIN_ATTRS define in cv.h! */
4302 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4304 CvLVALUE_on(PL_compcv);
4306 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4308 CvLOCKED_on(PL_compcv);
4310 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4312 CvMETHOD_on(PL_compcv);
4314 /* After we've set the flags, it could be argued that
4315 we don't need to do the attributes.pm-based setting
4316 process, and shouldn't bother appending recognized
4317 flags. To experiment with that, uncomment the
4318 following "else". (Note that's already been
4319 uncommented. That keeps the above-applied built-in
4320 attributes from being intercepted (and possibly
4321 rejected) by a package's attribute routines, but is
4322 justified by the performance win for the common case
4323 of applying only built-in attributes.) */
4325 attrs = append_elem(OP_LIST, attrs,
4326 newSVOP(OP_CONST, 0,
4330 if (*s == ':' && s[1] != ':')
4333 break; /* require real whitespace or :'s */
4334 /* XXX losing whitespace on sequential attributes here */
4338 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4339 if (*s != ';' && *s != '}' && *s != tmp
4340 && (tmp != '=' || *s != ')')) {
4341 const char q = ((*s == '\'') ? '"' : '\'');
4342 /* If here for an expression, and parsed no attrs, back
4344 if (tmp == '=' && !attrs) {
4348 /* MUST advance bufptr here to avoid bogus "at end of line"
4349 context messages from yyerror().
4352 yyerror( (const char *)
4354 ? Perl_form(aTHX_ "Invalid separator character "
4355 "%c%c%c in attribute list", q, *s, q)
4356 : "Unterminated attribute list" ) );
4364 start_force(PL_curforce);
4365 NEXTVAL_NEXTTOKE.opval = attrs;
4366 CURMAD('_', PL_nextwhite);
4371 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4372 (s - SvPVX(PL_linestr)) - stuffstart);
4380 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4381 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4389 const char tmp = *s++;
4394 const char tmp = *s++;
4402 if (PL_lex_brackets <= 0)
4403 yyerror("Unmatched right square bracket");
4406 if (PL_lex_state == LEX_INTERPNORMAL) {
4407 if (PL_lex_brackets == 0) {
4408 if (*s == '-' && s[1] == '>')
4409 PL_lex_state = LEX_INTERPENDMAYBE;
4410 else if (*s != '[' && *s != '{')
4411 PL_lex_state = LEX_INTERPEND;
4418 if (PL_lex_brackets > 100) {
4419 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4421 switch (PL_expect) {
4423 if (PL_lex_formbrack) {
4427 if (PL_oldoldbufptr == PL_last_lop)
4428 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4430 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4431 OPERATOR(HASHBRACK);
4433 while (s < PL_bufend && SPACE_OR_TAB(*s))
4436 PL_tokenbuf[0] = '\0';
4437 if (d < PL_bufend && *d == '-') {
4438 PL_tokenbuf[0] = '-';
4440 while (d < PL_bufend && SPACE_OR_TAB(*d))
4443 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4444 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4446 while (d < PL_bufend && SPACE_OR_TAB(*d))
4449 const char minus = (PL_tokenbuf[0] == '-');
4450 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4458 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4463 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4468 if (PL_oldoldbufptr == PL_last_lop)
4469 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4471 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4474 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4476 /* This hack is to get the ${} in the message. */
4478 yyerror("syntax error");
4481 OPERATOR(HASHBRACK);
4483 /* This hack serves to disambiguate a pair of curlies
4484 * as being a block or an anon hash. Normally, expectation
4485 * determines that, but in cases where we're not in a
4486 * position to expect anything in particular (like inside
4487 * eval"") we have to resolve the ambiguity. This code
4488 * covers the case where the first term in the curlies is a
4489 * quoted string. Most other cases need to be explicitly
4490 * disambiguated by prepending a "+" before the opening
4491 * curly in order to force resolution as an anon hash.
4493 * XXX should probably propagate the outer expectation
4494 * into eval"" to rely less on this hack, but that could
4495 * potentially break current behavior of eval"".
4499 if (*s == '\'' || *s == '"' || *s == '`') {
4500 /* common case: get past first string, handling escapes */
4501 for (t++; t < PL_bufend && *t != *s;)
4502 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4506 else if (*s == 'q') {
4509 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4512 /* skip q//-like construct */
4514 char open, close, term;
4517 while (t < PL_bufend && isSPACE(*t))
4519 /* check for q => */
4520 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4521 OPERATOR(HASHBRACK);
4525 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4529 for (t++; t < PL_bufend; t++) {
4530 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4532 else if (*t == open)
4536 for (t++; t < PL_bufend; t++) {
4537 if (*t == '\\' && t+1 < PL_bufend)
4539 else if (*t == close && --brackets <= 0)
4541 else if (*t == open)
4548 /* skip plain q word */
4549 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4552 else if (isALNUM_lazy_if(t,UTF)) {
4554 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4557 while (t < PL_bufend && isSPACE(*t))
4559 /* if comma follows first term, call it an anon hash */
4560 /* XXX it could be a comma expression with loop modifiers */
4561 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4562 || (*t == '=' && t[1] == '>')))
4563 OPERATOR(HASHBRACK);
4564 if (PL_expect == XREF)
4567 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4573 pl_yylval.ival = CopLINE(PL_curcop);
4574 if (isSPACE(*s) || *s == '#')
4575 PL_copline = NOLINE; /* invalidate current command line number */
4580 if (PL_lex_brackets <= 0)
4581 yyerror("Unmatched right curly bracket");
4583 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4584 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4585 PL_lex_formbrack = 0;
4586 if (PL_lex_state == LEX_INTERPNORMAL) {
4587 if (PL_lex_brackets == 0) {
4588 if (PL_expect & XFAKEBRACK) {
4589 PL_expect &= XENUMMASK;
4590 PL_lex_state = LEX_INTERPEND;
4595 PL_thiswhite = newSVpvs("");
4596 sv_catpvn(PL_thiswhite,"}",1);
4599 return yylex(); /* ignore fake brackets */
4601 if (*s == '-' && s[1] == '>')
4602 PL_lex_state = LEX_INTERPENDMAYBE;
4603 else if (*s != '[' && *s != '{')
4604 PL_lex_state = LEX_INTERPEND;
4607 if (PL_expect & XFAKEBRACK) {
4608 PL_expect &= XENUMMASK;
4610 return yylex(); /* ignore fake brackets */
4612 start_force(PL_curforce);
4614 curmad('X', newSVpvn(s-1,1));
4615 CURMAD('_', PL_thiswhite);
4620 PL_thistoken = newSVpvs("");
4628 if (PL_expect == XOPERATOR) {
4629 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4630 && isIDFIRST_lazy_if(s,UTF))
4632 CopLINE_dec(PL_curcop);
4633 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4634 CopLINE_inc(PL_curcop);
4639 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4641 PL_expect = XOPERATOR;
4642 force_ident(PL_tokenbuf, '&');
4646 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
4658 const char tmp = *s++;
4665 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4666 && strchr("+-*/%.^&|<",tmp))
4667 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4668 "Reversed %c= operator",(int)tmp);
4670 if (PL_expect == XSTATE && isALPHA(tmp) &&
4671 (s == PL_linestart+1 || s[-2] == '\n') )
4673 if (PL_in_eval && !PL_rsfp) {
4678 if (strnEQ(s,"=cut",4)) {
4694 PL_thiswhite = newSVpvs("");
4695 sv_catpvn(PL_thiswhite, PL_linestart,
4696 PL_bufend - PL_linestart);
4700 PL_doextract = TRUE;
4704 if (PL_lex_brackets < PL_lex_formbrack) {
4706 #ifdef PERL_STRICT_CR
4707 while (SPACE_OR_TAB(*t))
4709 while (SPACE_OR_TAB(*t) || *t == '\r')
4712 if (*t == '\n' || *t == '#') {
4723 const char tmp = *s++;
4725 /* was this !=~ where !~ was meant?
4726 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4728 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4729 const char *t = s+1;
4731 while (t < PL_bufend && isSPACE(*t))
4734 if (*t == '/' || *t == '?' ||
4735 ((*t == 'm' || *t == 's' || *t == 'y')
4736 && !isALNUM(t[1])) ||
4737 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4738 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4739 "!=~ should be !~");
4749 if (PL_expect != XOPERATOR) {
4750 if (s[1] != '<' && !strchr(s,'>'))
4753 s = scan_heredoc(s);
4755 s = scan_inputsymbol(s);
4756 TERM(sublex_start());
4762 SHop(OP_LEFT_SHIFT);
4776 const char tmp = *s++;
4778 SHop(OP_RIGHT_SHIFT);
4779 else if (tmp == '=')
4788 if (PL_expect == XOPERATOR) {
4789 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4791 deprecate_old(commaless_variable_list);
4792 return REPORT(','); /* grandfather non-comma-format format */
4796 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4797 PL_tokenbuf[0] = '@';
4798 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4799 sizeof PL_tokenbuf - 1, FALSE);
4800 if (PL_expect == XOPERATOR)
4801 no_op("Array length", s);
4802 if (!PL_tokenbuf[1])
4804 PL_expect = XOPERATOR;
4805 PL_pending_ident = '#';
4809 PL_tokenbuf[0] = '$';
4810 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4811 sizeof PL_tokenbuf - 1, FALSE);
4812 if (PL_expect == XOPERATOR)
4814 if (!PL_tokenbuf[1]) {
4816 yyerror("Final $ should be \\$ or $name");
4820 /* This kludge not intended to be bulletproof. */
4821 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4822 pl_yylval.opval = newSVOP(OP_CONST, 0,
4823 newSViv(CopARYBASE_get(&PL_compiling)));
4824 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
4830 const char tmp = *s;
4831 if (PL_lex_state == LEX_NORMAL)
4834 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4835 && intuit_more(s)) {
4837 PL_tokenbuf[0] = '@';
4838 if (ckWARN(WARN_SYNTAX)) {
4841 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4844 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4845 while (t < PL_bufend && *t != ']')
4847 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4848 "Multidimensional syntax %.*s not supported",
4849 (int)((t - PL_bufptr) + 1), PL_bufptr);
4853 else if (*s == '{') {
4855 PL_tokenbuf[0] = '%';
4856 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4857 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4859 char tmpbuf[sizeof PL_tokenbuf];
4862 } while (isSPACE(*t));
4863 if (isIDFIRST_lazy_if(t,UTF)) {
4865 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4869 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4870 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4871 "You need to quote \"%s\"",
4878 PL_expect = XOPERATOR;
4879 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4880 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4881 if (!islop || PL_last_lop_op == OP_GREPSTART)
4882 PL_expect = XOPERATOR;
4883 else if (strchr("$@\"'`q", *s))
4884 PL_expect = XTERM; /* e.g. print $fh "foo" */
4885 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4886 PL_expect = XTERM; /* e.g. print $fh &sub */
4887 else if (isIDFIRST_lazy_if(s,UTF)) {
4888 char tmpbuf[sizeof PL_tokenbuf];
4890 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4891 if ((t2 = keyword(tmpbuf, len, 0))) {
4892 /* binary operators exclude handle interpretations */
4904 PL_expect = XTERM; /* e.g. print $fh length() */
4909 PL_expect = XTERM; /* e.g. print $fh subr() */
4912 else if (isDIGIT(*s))
4913 PL_expect = XTERM; /* e.g. print $fh 3 */
4914 else if (*s == '.' && isDIGIT(s[1]))
4915 PL_expect = XTERM; /* e.g. print $fh .3 */
4916 else if ((*s == '?' || *s == '-' || *s == '+')
4917 && !isSPACE(s[1]) && s[1] != '=')
4918 PL_expect = XTERM; /* e.g. print $fh -1 */
4919 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4921 PL_expect = XTERM; /* e.g. print $fh /.../
4922 XXX except DORDOR operator
4924 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4926 PL_expect = XTERM; /* print $fh <<"EOF" */
4929 PL_pending_ident = '$';
4933 if (PL_expect == XOPERATOR)
4935 PL_tokenbuf[0] = '@';
4936 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4937 if (!PL_tokenbuf[1]) {
4940 if (PL_lex_state == LEX_NORMAL)
4942 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4944 PL_tokenbuf[0] = '%';
4946 /* Warn about @ where they meant $. */
4947 if (*s == '[' || *s == '{') {
4948 if (ckWARN(WARN_SYNTAX)) {
4949 const char *t = s + 1;
4950 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4952 if (*t == '}' || *t == ']') {
4954 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4955 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4956 "Scalar value %.*s better written as $%.*s",
4957 (int)(t-PL_bufptr), PL_bufptr,
4958 (int)(t-PL_bufptr-1), PL_bufptr+1);
4963 PL_pending_ident = '@';
4966 case '/': /* may be division, defined-or, or pattern */
4967 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4971 case '?': /* may either be conditional or pattern */
4972 if(PL_expect == XOPERATOR) {
4980 /* A // operator. */
4990 /* Disable warning on "study /blah/" */
4991 if (PL_oldoldbufptr == PL_last_uni
4992 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4993 || memNE(PL_last_uni, "study", 5)
4994 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4997 s = scan_pat(s,OP_MATCH);
4998 TERM(sublex_start());
5002 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5003 #ifdef PERL_STRICT_CR
5006 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5008 && (s == PL_linestart || s[-1] == '\n') )
5010 PL_lex_formbrack = 0;
5014 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5020 pl_yylval.ival = OPf_SPECIAL;
5026 if (PL_expect != XOPERATOR)
5031 case '0': case '1': case '2': case '3': case '4':
5032 case '5': case '6': case '7': case '8': case '9':
5033 s = scan_num(s, &pl_yylval);
5034 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5035 if (PL_expect == XOPERATOR)
5040 s = scan_str(s,!!PL_madskills,FALSE);
5041 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5042 if (PL_expect == XOPERATOR) {
5043 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5045 deprecate_old(commaless_variable_list);
5046 return REPORT(','); /* grandfather non-comma-format format */
5053 pl_yylval.ival = OP_CONST;
5054 TERM(sublex_start());
5057 s = scan_str(s,!!PL_madskills,FALSE);
5058 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5059 if (PL_expect == XOPERATOR) {
5060 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5062 deprecate_old(commaless_variable_list);
5063 return REPORT(','); /* grandfather non-comma-format format */
5070 pl_yylval.ival = OP_CONST;
5071 /* FIXME. I think that this can be const if char *d is replaced by
5072 more localised variables. */
5073 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5074 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5075 pl_yylval.ival = OP_STRINGIFY;
5079 TERM(sublex_start());
5082 s = scan_str(s,!!PL_madskills,FALSE);
5083 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5084 if (PL_expect == XOPERATOR)
5085 no_op("Backticks",s);
5088 readpipe_override();
5089 TERM(sublex_start());
5093 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5094 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5096 if (PL_expect == XOPERATOR)
5097 no_op("Backslash",s);
5101 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5102 char *start = s + 2;
5103 while (isDIGIT(*start) || *start == '_')
5105 if (*start == '.' && isDIGIT(start[1])) {
5106 s = scan_num(s, &pl_yylval);
5109 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5110 else if (!isALPHA(*start) && (PL_expect == XTERM
5111 || PL_expect == XREF || PL_expect == XSTATE
5112 || PL_expect == XTERMORDORDOR)) {
5113 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5115 s = scan_num(s, &pl_yylval);
5122 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5164 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5166 /* Some keywords can be followed by any delimiter, including ':' */
5167 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5168 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5169 (PL_tokenbuf[0] == 'q' &&
5170 strchr("qwxr", PL_tokenbuf[1])))));
5172 /* x::* is just a word, unless x is "CORE" */
5173 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5177 while (d < PL_bufend && isSPACE(*d))
5178 d++; /* no comments skipped here, or s### is misparsed */
5180 /* Is this a label? */
5181 if (!tmp && PL_expect == XSTATE
5182 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5184 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5189 /* Check for keywords */
5190 tmp = keyword(PL_tokenbuf, len, 0);
5192 /* Is this a word before a => operator? */
5193 if (*d == '=' && d[1] == '>') {
5196 = (OP*)newSVOP(OP_CONST, 0,
5197 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5198 pl_yylval.opval->op_private = OPpCONST_BARE;
5202 if (tmp < 0) { /* second-class keyword? */
5203 GV *ogv = NULL; /* override (winner) */
5204 GV *hgv = NULL; /* hidden (loser) */
5205 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5207 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5210 if (GvIMPORTED_CV(gv))
5212 else if (! CvMETHOD(cv))
5216 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5217 (gv = *gvp) && isGV_with_GP(gv) &&
5218 GvCVu(gv) && GvIMPORTED_CV(gv))
5225 tmp = 0; /* overridden by import or by GLOBAL */
5228 && -tmp==KEY_lock /* XXX generalizable kludge */
5231 tmp = 0; /* any sub overrides "weak" keyword */
5233 else { /* no override */
5235 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5236 Perl_warner(aTHX_ packWARN(WARN_MISC),
5237 "dump() better written as CORE::dump()");
5241 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5242 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5243 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5244 "Ambiguous call resolved as CORE::%s(), %s",
5245 GvENAME(hgv), "qualify as such or use &");
5252 default: /* not a keyword */
5253 /* Trade off - by using this evil construction we can pull the
5254 variable gv into the block labelled keylookup. If not, then
5255 we have to give it function scope so that the goto from the
5256 earlier ':' case doesn't bypass the initialisation. */
5258 just_a_word_zero_gv:
5266 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5269 SV *nextPL_nextwhite = 0;
5273 /* Get the rest if it looks like a package qualifier */
5275 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5277 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5280 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5281 *s == '\'' ? "'" : "::");
5286 if (PL_expect == XOPERATOR) {
5287 if (PL_bufptr == PL_linestart) {
5288 CopLINE_dec(PL_curcop);
5289 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5290 CopLINE_inc(PL_curcop);
5293 no_op("Bareword",s);
5296 /* Look for a subroutine with this name in current package,
5297 unless name is "Foo::", in which case Foo is a bearword
5298 (and a package name). */
5300 if (len > 2 && !PL_madskills &&
5301 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5303 if (ckWARN(WARN_BAREWORD)
5304 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5305 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5306 "Bareword \"%s\" refers to nonexistent package",
5309 PL_tokenbuf[len] = '\0';
5315 /* Mustn't actually add anything to a symbol table.
5316 But also don't want to "initialise" any placeholder
5317 constants that might already be there into full
5318 blown PVGVs with attached PVCV. */
5319 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5320 GV_NOADD_NOINIT, SVt_PVCV);
5325 /* if we saw a global override before, get the right name */
5328 sv = newSVpvs("CORE::GLOBAL::");
5329 sv_catpv(sv,PL_tokenbuf);
5332 /* If len is 0, newSVpv does strlen(), which is correct.
5333 If len is non-zero, then it will be the true length,
5334 and so the scalar will be created correctly. */
5335 sv = newSVpv(PL_tokenbuf,len);
5338 if (PL_madskills && !PL_thistoken) {
5339 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5340 PL_thistoken = newSVpvn(start,s - start);
5341 PL_realtokenstart = s - SvPVX(PL_linestr);
5345 /* Presume this is going to be a bareword of some sort. */
5348 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5349 pl_yylval.opval->op_private = OPpCONST_BARE;
5350 /* UTF-8 package name? */
5351 if (UTF && !IN_BYTES &&
5352 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5355 /* And if "Foo::", then that's what it certainly is. */
5360 /* Do the explicit type check so that we don't need to force
5361 the initialisation of the symbol table to have a real GV.
5362 Beware - gv may not really be a PVGV, cv may not really be
5363 a PVCV, (because of the space optimisations that gv_init
5364 understands) But they're true if for this symbol there is
5365 respectively a typeglob and a subroutine.
5367 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5368 /* Real typeglob, so get the real subroutine: */
5370 /* A proxy for a subroutine in this package? */
5371 : SvOK(gv) ? (CV *) gv : NULL)
5374 /* See if it's the indirect object for a list operator. */
5376 if (PL_oldoldbufptr &&
5377 PL_oldoldbufptr < PL_bufptr &&
5378 (PL_oldoldbufptr == PL_last_lop
5379 || PL_oldoldbufptr == PL_last_uni) &&
5380 /* NO SKIPSPACE BEFORE HERE! */
5381 (PL_expect == XREF ||
5382 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5384 bool immediate_paren = *s == '(';
5386 /* (Now we can afford to cross potential line boundary.) */
5387 s = SKIPSPACE2(s,nextPL_nextwhite);
5389 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5392 /* Two barewords in a row may indicate method call. */
5394 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5395 (tmp = intuit_method(s, gv, cv)))
5398 /* If not a declared subroutine, it's an indirect object. */
5399 /* (But it's an indir obj regardless for sort.) */
5400 /* Also, if "_" follows a filetest operator, it's a bareword */
5403 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5405 (PL_last_lop_op != OP_MAPSTART &&
5406 PL_last_lop_op != OP_GREPSTART))))
5407 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5408 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5411 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5416 PL_expect = XOPERATOR;
5419 s = SKIPSPACE2(s,nextPL_nextwhite);
5420 PL_nextwhite = nextPL_nextwhite;
5425 /* Is this a word before a => operator? */
5426 if (*s == '=' && s[1] == '>' && !pkgname) {
5428 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
5429 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5430 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
5434 /* If followed by a paren, it's certainly a subroutine. */
5439 while (SPACE_OR_TAB(*d))
5441 if (*d == ')' && (sv = gv_const_sv(gv))) {
5448 PL_nextwhite = PL_thiswhite;
5451 start_force(PL_curforce);
5453 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5454 PL_expect = XOPERATOR;
5457 PL_nextwhite = nextPL_nextwhite;
5458 curmad('X', PL_thistoken);
5459 PL_thistoken = newSVpvs("");
5467 /* If followed by var or block, call it a method (unless sub) */
5469 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5470 PL_last_lop = PL_oldbufptr;
5471 PL_last_lop_op = OP_METHOD;
5475 /* If followed by a bareword, see if it looks like indir obj. */
5478 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5479 && (tmp = intuit_method(s, gv, cv)))
5482 /* Not a method, so call it a subroutine (if defined) */
5485 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5486 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5487 "Ambiguous use of -%s resolved as -&%s()",
5488 PL_tokenbuf, PL_tokenbuf);
5489 /* Check for a constant sub */
5490 if ((sv = gv_const_sv(gv))) {
5492 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
5493 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5494 pl_yylval.opval->op_private = 0;
5498 /* Resolve to GV now. */
5499 if (SvTYPE(gv) != SVt_PVGV) {
5500 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5501 assert (SvTYPE(gv) == SVt_PVGV);
5502 /* cv must have been some sort of placeholder, so
5503 now needs replacing with a real code reference. */
5507 op_free(pl_yylval.opval);
5508 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5509 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5510 PL_last_lop = PL_oldbufptr;
5511 PL_last_lop_op = OP_ENTERSUB;
5512 /* Is there a prototype? */
5520 const char *proto = SvPV_const((SV*)cv, protolen);
5523 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5525 while (*proto == ';')
5527 if (*proto == '&' && *s == '{') {
5529 sv_setpvs(PL_subname, "__ANON__");
5531 sv_setpvs(PL_subname, "__ANON__::__ANON__");
5538 PL_nextwhite = PL_thiswhite;
5541 start_force(PL_curforce);
5542 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5545 PL_nextwhite = nextPL_nextwhite;
5546 curmad('X', PL_thistoken);
5547 PL_thistoken = newSVpvs("");
5554 /* Guess harder when madskills require "best effort". */
5555 if (PL_madskills && (!gv || !GvCVu(gv))) {
5556 int probable_sub = 0;
5557 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5559 else if (isALPHA(*s)) {
5563 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5564 if (!keyword(tmpbuf, tmplen, 0))
5567 while (d < PL_bufend && isSPACE(*d))
5569 if (*d == '=' && d[1] == '>')
5574 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5575 op_free(pl_yylval.opval);
5576 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5577 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5578 PL_last_lop = PL_oldbufptr;
5579 PL_last_lop_op = OP_ENTERSUB;
5580 PL_nextwhite = PL_thiswhite;
5582 start_force(PL_curforce);
5583 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5585 PL_nextwhite = nextPL_nextwhite;
5586 curmad('X', PL_thistoken);
5587 PL_thistoken = newSVpvs("");
5592 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5599 /* Call it a bare word */
5601 if (PL_hints & HINT_STRICT_SUBS)
5602 pl_yylval.opval->op_private |= OPpCONST_STRICT;
5605 if (lastchar != '-') {
5606 if (ckWARN(WARN_RESERVED)) {
5610 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5611 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5618 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5619 && ckWARN_d(WARN_AMBIGUOUS)) {
5620 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5621 "Operator or semicolon missing before %c%s",
5622 lastchar, PL_tokenbuf);
5623 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5624 "Ambiguous use of %c resolved as operator %c",
5625 lastchar, lastchar);
5631 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5632 newSVpv(CopFILE(PL_curcop),0));
5636 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5637 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5640 case KEY___PACKAGE__:
5641 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5643 ? newSVhek(HvNAME_HEK(PL_curstash))
5650 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5651 const char *pname = "main";
5652 if (PL_tokenbuf[2] == 'D')
5653 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5654 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5658 GvIOp(gv) = newIO();
5659 IoIFP(GvIOp(gv)) = PL_rsfp;
5660 #if defined(HAS_FCNTL) && defined(F_SETFD)
5662 const int fd = PerlIO_fileno(PL_rsfp);
5663 fcntl(fd,F_SETFD,fd >= 3);
5666 /* Mark this internal pseudo-handle as clean */
5667 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5669 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5670 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5671 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5673 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5674 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5675 /* if the script was opened in binmode, we need to revert
5676 * it to text mode for compatibility; but only iff it has CRs
5677 * XXX this is a questionable hack at best. */
5678 if (PL_bufend-PL_bufptr > 2
5679 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5682 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5683 loc = PerlIO_tell(PL_rsfp);
5684 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5687 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5689 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5690 #endif /* NETWARE */
5691 #ifdef PERLIO_IS_STDIO /* really? */
5692 # if defined(__BORLANDC__)
5693 /* XXX see note in do_binmode() */
5694 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5698 PerlIO_seek(PL_rsfp, loc, 0);
5702 #ifdef PERLIO_LAYERS
5705 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5706 else if (PL_encoding) {
5713 XPUSHs(PL_encoding);
5715 call_method("name", G_SCALAR);
5719 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5720 Perl_form(aTHX_ ":encoding(%"SVf")",
5729 if (PL_realtokenstart >= 0) {
5730 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5732 PL_endwhite = newSVpvs("");
5733 sv_catsv(PL_endwhite, PL_thiswhite);
5735 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5736 PL_realtokenstart = -1;
5738 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5739 SvCUR(PL_endwhite))) != NULL) ;
5754 if (PL_expect == XSTATE) {
5761 if (*s == ':' && s[1] == ':') {
5764 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5765 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5766 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5769 else if (tmp == KEY_require || tmp == KEY_do)
5770 /* that's a way to remember we saw "CORE::" */
5783 LOP(OP_ACCEPT,XTERM);
5789 LOP(OP_ATAN2,XTERM);
5795 LOP(OP_BINMODE,XTERM);
5798 LOP(OP_BLESS,XTERM);
5807 /* When 'use switch' is in effect, continue has a dual
5808 life as a control operator. */
5810 if (!FEATURE_IS_ENABLED("switch"))
5813 /* We have to disambiguate the two senses of
5814 "continue". If the next token is a '{' then
5815 treat it as the start of a continue block;
5816 otherwise treat it as a control operator.
5828 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5845 if (!PL_cryptseen) {
5846 PL_cryptseen = TRUE;
5850 LOP(OP_CRYPT,XTERM);
5853 LOP(OP_CHMOD,XTERM);
5856 LOP(OP_CHOWN,XTERM);
5859 LOP(OP_CONNECT,XTERM);
5878 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5879 if (orig_keyword == KEY_do) {
5888 PL_hints |= HINT_BLOCK_SCOPE;
5898 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5899 LOP(OP_DBMOPEN,XTERM);
5905 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5912 pl_yylval.ival = CopLINE(PL_curcop);
5928 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5929 UNIBRACK(OP_ENTEREVAL);
5943 case KEY_endhostent:
5949 case KEY_endservent:
5952 case KEY_endprotoent:
5963 pl_yylval.ival = CopLINE(PL_curcop);
5965 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5968 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5971 if ((PL_bufend - p) >= 3 &&
5972 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5974 else if ((PL_bufend - p) >= 4 &&
5975 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5978 if (isIDFIRST_lazy_if(p,UTF)) {
5979 p = scan_ident(p, PL_bufend,
5980 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5984 Perl_croak(aTHX_ "Missing $ on loop variable");
5986 s = SvPVX(PL_linestr) + soff;
5992 LOP(OP_FORMLINE,XTERM);
5998 LOP(OP_FCNTL,XTERM);
6004 LOP(OP_FLOCK,XTERM);
6013 LOP(OP_GREPSTART, XREF);
6016 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6031 case KEY_getpriority:
6032 LOP(OP_GETPRIORITY,XTERM);
6034 case KEY_getprotobyname:
6037 case KEY_getprotobynumber:
6038 LOP(OP_GPBYNUMBER,XTERM);
6040 case KEY_getprotoent:
6052 case KEY_getpeername:
6053 UNI(OP_GETPEERNAME);
6055 case KEY_gethostbyname:
6058 case KEY_gethostbyaddr:
6059 LOP(OP_GHBYADDR,XTERM);
6061 case KEY_gethostent:
6064 case KEY_getnetbyname:
6067 case KEY_getnetbyaddr:
6068 LOP(OP_GNBYADDR,XTERM);
6073 case KEY_getservbyname:
6074 LOP(OP_GSBYNAME,XTERM);
6076 case KEY_getservbyport:
6077 LOP(OP_GSBYPORT,XTERM);
6079 case KEY_getservent:
6082 case KEY_getsockname:
6083 UNI(OP_GETSOCKNAME);
6085 case KEY_getsockopt:
6086 LOP(OP_GSOCKOPT,XTERM);
6101 pl_yylval.ival = CopLINE(PL_curcop);
6111 pl_yylval.ival = CopLINE(PL_curcop);
6115 LOP(OP_INDEX,XTERM);
6121 LOP(OP_IOCTL,XTERM);
6133 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6165 LOP(OP_LISTEN,XTERM);
6174 s = scan_pat(s,OP_MATCH);
6175 TERM(sublex_start());
6178 LOP(OP_MAPSTART, XREF);
6181 LOP(OP_MKDIR,XTERM);
6184 LOP(OP_MSGCTL,XTERM);
6187 LOP(OP_MSGGET,XTERM);
6190 LOP(OP_MSGRCV,XTERM);
6193 LOP(OP_MSGSND,XTERM);
6198 PL_in_my = (U16)tmp;
6200 if (isIDFIRST_lazy_if(s,UTF)) {
6204 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6205 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6207 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6208 if (!PL_in_my_stash) {
6211 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6215 if (PL_madskills) { /* just add type to declarator token */
6216 sv_catsv(PL_thistoken, PL_nextwhite);
6218 sv_catpvn(PL_thistoken, start, s - start);
6226 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6233 s = tokenize_use(0, s);
6237 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6244 if (isIDFIRST_lazy_if(s,UTF)) {
6246 for (d = s; isALNUM_lazy_if(d,UTF);)
6248 for (t=d; isSPACE(*t);)
6250 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6252 && !(t[0] == '=' && t[1] == '>')
6254 int parms_len = (int)(d-s);
6255 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6256 "Precedence problem: open %.*s should be open(%.*s)",
6257 parms_len, s, parms_len, s);
6263 pl_yylval.ival = OP_OR;
6273 LOP(OP_OPEN_DIR,XTERM);
6276 checkcomma(s,PL_tokenbuf,"filehandle");
6280 checkcomma(s,PL_tokenbuf,"filehandle");
6299 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6303 LOP(OP_PIPE_OP,XTERM);
6306 s = scan_str(s,!!PL_madskills,FALSE);
6309 pl_yylval.ival = OP_CONST;
6310 TERM(sublex_start());
6316 s = scan_str(s,!!PL_madskills,FALSE);
6319 PL_expect = XOPERATOR;
6321 if (SvCUR(PL_lex_stuff)) {
6324 d = SvPV_force(PL_lex_stuff, len);
6326 for (; isSPACE(*d) && len; --len, ++d)
6331 if (!warned && ckWARN(WARN_QW)) {
6332 for (; !isSPACE(*d) && len; --len, ++d) {
6334 Perl_warner(aTHX_ packWARN(WARN_QW),
6335 "Possible attempt to separate words with commas");
6338 else if (*d == '#') {
6339 Perl_warner(aTHX_ packWARN(WARN_QW),
6340 "Possible attempt to put comments in qw() list");
6346 for (; !isSPACE(*d) && len; --len, ++d)
6349 sv = newSVpvn(b, d-b);
6350 if (DO_UTF8(PL_lex_stuff))
6352 words = append_elem(OP_LIST, words,
6353 newSVOP(OP_CONST, 0, tokeq(sv)));
6357 start_force(PL_curforce);
6358 NEXTVAL_NEXTTOKE.opval = words;
6363 SvREFCNT_dec(PL_lex_stuff);
6364 PL_lex_stuff = NULL;
6370 s = scan_str(s,!!PL_madskills,FALSE);
6373 pl_yylval.ival = OP_STRINGIFY;
6374 if (SvIVX(PL_lex_stuff) == '\'')
6375 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6376 TERM(sublex_start());
6379 s = scan_pat(s,OP_QR);
6380 TERM(sublex_start());
6383 s = scan_str(s,!!PL_madskills,FALSE);
6386 readpipe_override();
6387 TERM(sublex_start());
6395 s = force_version(s, FALSE);
6397 else if (*s != 'v' || !isDIGIT(s[1])
6398 || (s = force_version(s, TRUE), *s == 'v'))
6400 *PL_tokenbuf = '\0';
6401 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6402 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6403 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6405 yyerror("<> should be quotes");
6407 if (orig_keyword == KEY_require) {
6415 PL_last_uni = PL_oldbufptr;
6416 PL_last_lop_op = OP_REQUIRE;
6418 return REPORT( (int)REQUIRE );
6424 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6428 LOP(OP_RENAME,XTERM);
6437 LOP(OP_RINDEX,XTERM);
6446 UNIDOR(OP_READLINE);
6449 UNIDOR(OP_BACKTICK);
6458 LOP(OP_REVERSE,XTERM);
6461 UNIDOR(OP_READLINK);
6468 if (pl_yylval.opval)
6469 TERM(sublex_start());
6471 TOKEN(1); /* force error */
6474 checkcomma(s,PL_tokenbuf,"filehandle");
6484 LOP(OP_SELECT,XTERM);
6490 LOP(OP_SEMCTL,XTERM);
6493 LOP(OP_SEMGET,XTERM);
6496 LOP(OP_SEMOP,XTERM);
6502 LOP(OP_SETPGRP,XTERM);
6504 case KEY_setpriority:
6505 LOP(OP_SETPRIORITY,XTERM);
6507 case KEY_sethostent:
6513 case KEY_setservent:
6516 case KEY_setprotoent:
6526 LOP(OP_SEEKDIR,XTERM);
6528 case KEY_setsockopt:
6529 LOP(OP_SSOCKOPT,XTERM);
6535 LOP(OP_SHMCTL,XTERM);
6538 LOP(OP_SHMGET,XTERM);
6541 LOP(OP_SHMREAD,XTERM);
6544 LOP(OP_SHMWRITE,XTERM);
6547 LOP(OP_SHUTDOWN,XTERM);
6556 LOP(OP_SOCKET,XTERM);
6558 case KEY_socketpair:
6559 LOP(OP_SOCKPAIR,XTERM);
6562 checkcomma(s,PL_tokenbuf,"subroutine name");
6564 if (*s == ';' || *s == ')') /* probably a close */
6565 Perl_croak(aTHX_ "sort is now a reserved word");
6567 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6571 LOP(OP_SPLIT,XTERM);
6574 LOP(OP_SPRINTF,XTERM);
6577 LOP(OP_SPLICE,XTERM);
6592 LOP(OP_SUBSTR,XTERM);
6598 char tmpbuf[sizeof PL_tokenbuf];
6599 SSize_t tboffset = 0;
6600 expectation attrful;
6601 bool have_name, have_proto;
6602 const int key = tmp;
6607 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6608 SV *subtoken = newSVpvn(tstart, s - tstart);
6612 s = SKIPSPACE2(s,tmpwhite);
6617 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6618 (*s == ':' && s[1] == ':'))
6625 attrful = XATTRBLOCK;
6626 /* remember buffer pos'n for later force_word */
6627 tboffset = s - PL_oldbufptr;
6628 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6631 nametoke = newSVpvn(s, d - s);
6633 if (memchr(tmpbuf, ':', len))
6634 sv_setpvn(PL_subname, tmpbuf, len);
6636 sv_setsv(PL_subname,PL_curstname);
6637 sv_catpvs(PL_subname,"::");
6638 sv_catpvn(PL_subname,tmpbuf,len);
6645 CURMAD('X', nametoke);
6646 CURMAD('_', tmpwhite);
6647 (void) force_word(PL_oldbufptr + tboffset, WORD,
6650 s = SKIPSPACE2(d,tmpwhite);
6657 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6658 PL_expect = XTERMBLOCK;
6659 attrful = XATTRTERM;
6660 sv_setpvn(PL_subname,"?",1);
6664 if (key == KEY_format) {
6666 PL_lex_formbrack = PL_lex_brackets + 1;
6668 PL_thistoken = subtoken;
6672 (void) force_word(PL_oldbufptr + tboffset, WORD,
6678 /* Look for a prototype */
6681 bool bad_proto = FALSE;
6682 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6684 s = scan_str(s,!!PL_madskills,FALSE);
6686 Perl_croak(aTHX_ "Prototype not terminated");
6687 /* strip spaces and check for bad characters */
6688 d = SvPVX(PL_lex_stuff);
6690 for (p = d; *p; ++p) {
6693 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6699 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6700 "Illegal character in prototype for %"SVf" : %s",
6701 SVfARG(PL_subname), d);
6702 SvCUR_set(PL_lex_stuff, tmp);
6707 CURMAD('q', PL_thisopen);
6708 CURMAD('_', tmpwhite);
6709 CURMAD('=', PL_thisstuff);
6710 CURMAD('Q', PL_thisclose);
6711 NEXTVAL_NEXTTOKE.opval =
6712 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6713 PL_lex_stuff = NULL;
6716 s = SKIPSPACE2(s,tmpwhite);
6724 if (*s == ':' && s[1] != ':')
6725 PL_expect = attrful;
6726 else if (*s != '{' && key == KEY_sub) {
6728 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6730 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6737 curmad('^', newSVpvs(""));
6738 CURMAD('_', tmpwhite);
6742 PL_thistoken = subtoken;
6745 NEXTVAL_NEXTTOKE.opval =
6746 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6747 PL_lex_stuff = NULL;
6753 sv_setpvs(PL_subname, "__ANON__");
6755 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6759 (void) force_word(PL_oldbufptr + tboffset, WORD,
6768 LOP(OP_SYSTEM,XREF);
6771 LOP(OP_SYMLINK,XTERM);
6774 LOP(OP_SYSCALL,XTERM);
6777 LOP(OP_SYSOPEN,XTERM);
6780 LOP(OP_SYSSEEK,XTERM);
6783 LOP(OP_SYSREAD,XTERM);
6786 LOP(OP_SYSWRITE,XTERM);
6790 TERM(sublex_start());
6811 LOP(OP_TRUNCATE,XTERM);
6823 pl_yylval.ival = CopLINE(PL_curcop);
6827 pl_yylval.ival = CopLINE(PL_curcop);
6831 LOP(OP_UNLINK,XTERM);
6837 LOP(OP_UNPACK,XTERM);
6840 LOP(OP_UTIME,XTERM);
6846 LOP(OP_UNSHIFT,XTERM);
6849 s = tokenize_use(1, s);
6859 pl_yylval.ival = CopLINE(PL_curcop);
6863 pl_yylval.ival = CopLINE(PL_curcop);
6867 PL_hints |= HINT_BLOCK_SCOPE;
6874 LOP(OP_WAITPID,XTERM);
6883 ctl_l[0] = toCTRL('L');
6885 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6888 /* Make sure $^L is defined */
6889 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6894 if (PL_expect == XOPERATOR)
6900 pl_yylval.ival = OP_XOR;
6905 TERM(sublex_start());
6910 #pragma segment Main
6914 S_pending_ident(pTHX)
6919 /* pit holds the identifier we read and pending_ident is reset */
6920 char pit = PL_pending_ident;
6921 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
6922 /* All routes through this function want to know if there is a colon. */
6923 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
6924 PL_pending_ident = 0;
6926 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6927 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6928 "### Pending identifier '%s'\n", PL_tokenbuf); });
6930 /* if we're in a my(), we can't allow dynamics here.
6931 $foo'bar has already been turned into $foo::bar, so
6932 just check for colons.
6934 if it's a legal name, the OP is a PADANY.
6937 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6939 yyerror(Perl_form(aTHX_ "No package name allowed for "
6940 "variable %s in \"our\"",
6942 tmp = allocmy(PL_tokenbuf);
6946 yyerror(Perl_form(aTHX_ PL_no_myglob,
6947 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6949 pl_yylval.opval = newOP(OP_PADANY, 0);
6950 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf);
6956 build the ops for accesses to a my() variable.
6958 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6959 then used in a comparison. This catches most, but not
6960 all cases. For instance, it catches
6961 sort { my($a); $a <=> $b }
6963 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6964 (although why you'd do that is anyone's guess).
6969 tmp = pad_findmy(PL_tokenbuf);
6970 if (tmp != NOT_IN_PAD) {
6971 /* might be an "our" variable" */
6972 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6973 /* build ops for a bareword */
6974 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6975 HEK * const stashname = HvNAME_HEK(stash);
6976 SV * const sym = newSVhek(stashname);
6977 sv_catpvs(sym, "::");
6978 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
6979 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6980 pl_yylval.opval->op_private = OPpCONST_ENTERED;
6983 ? (GV_ADDMULTI | GV_ADDINEVAL)
6986 ((PL_tokenbuf[0] == '$') ? SVt_PV
6987 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6992 /* if it's a sort block and they're naming $a or $b */
6993 if (PL_last_lop_op == OP_SORT &&
6994 PL_tokenbuf[0] == '$' &&
6995 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6998 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6999 d < PL_bufend && *d != '\n';
7002 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7003 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7009 pl_yylval.opval = newOP(OP_PADANY, 0);
7010 pl_yylval.opval->op_targ = tmp;
7016 Whine if they've said @foo in a doublequoted string,
7017 and @foo isn't a variable we can find in the symbol
7020 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7021 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7023 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7024 && ckWARN(WARN_AMBIGUOUS)
7025 /* DO NOT warn for @- and @+ */
7026 && !( PL_tokenbuf[2] == '\0' &&
7027 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7030 /* Downgraded from fatal to warning 20000522 mjd */
7031 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7032 "Possible unintended interpolation of %s in string",
7037 /* build ops for a bareword */
7038 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7040 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7042 PL_tokenbuf + 1, tokenbuf_len - 1,
7043 /* If the identifier refers to a stash, don't autovivify it.
7044 * Change 24660 had the side effect of causing symbol table
7045 * hashes to always be defined, even if they were freshly
7046 * created and the only reference in the entire program was
7047 * the single statement with the defined %foo::bar:: test.
7048 * It appears that all code in the wild doing this actually
7049 * wants to know whether sub-packages have been loaded, so
7050 * by avoiding auto-vivifying symbol tables, we ensure that
7051 * defined %foo::bar:: continues to be false, and the existing
7052 * tests still give the expected answers, even though what
7053 * they're actually testing has now changed subtly.
7055 (*PL_tokenbuf == '%'
7056 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7059 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7060 ((PL_tokenbuf[0] == '$') ? SVt_PV
7061 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7067 * The following code was generated by perl_keyword.pl.
7071 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7076 case 1: /* 5 tokens of length 1 */
7108 case 2: /* 18 tokens of length 2 */
7254 case 3: /* 29 tokens of length 3 */
7258 if (name[1] == 'N' &&
7321 if (name[1] == 'i' &&
7353 if (name[1] == 'o' &&
7362 if (name[1] == 'e' &&
7371 if (name[1] == 'n' &&
7380 if (name[1] == 'o' &&
7389 if (name[1] == 'a' &&
7398 if (name[1] == 'o' &&
7460 if (name[1] == 'e' &&
7474 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7500 if (name[1] == 'i' &&
7509 if (name[1] == 's' &&
7518 if (name[1] == 'e' &&
7527 if (name[1] == 'o' &&
7539 case 4: /* 41 tokens of length 4 */
7543 if (name[1] == 'O' &&
7553 if (name[1] == 'N' &&
7563 if (name[1] == 'i' &&
7573 if (name[1] == 'h' &&
7583 if (name[1] == 'u' &&
7596 if (name[2] == 'c' &&
7605 if (name[2] == 's' &&
7614 if (name[2] == 'a' &&
7650 if (name[1] == 'o' &&
7663 if (name[2] == 't' &&
7672 if (name[2] == 'o' &&
7681 if (name[2] == 't' &&
7690 if (name[2] == 'e' &&
7703 if (name[1] == 'o' &&
7716 if (name[2] == 'y' &&
7725 if (name[2] == 'l' &&
7741 if (name[2] == 's' &&
7750 if (name[2] == 'n' &&
7759 if (name[2] == 'c' &&
7772 if (name[1] == 'e' &&
7782 if (name[1] == 'p' &&
7795 if (name[2] == 'c' &&
7804 if (name[2] == 'p' &&
7813 if (name[2] == 's' &&
7829 if (name[2] == 'n' &&
7899 if (name[2] == 'r' &&
7908 if (name[2] == 'r' &&
7917 if (name[2] == 'a' &&
7933 if (name[2] == 'l' &&
7995 if (name[2] == 'e' &&
7998 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8011 case 5: /* 39 tokens of length 5 */
8015 if (name[1] == 'E' &&
8026 if (name[1] == 'H' &&
8040 if (name[2] == 'a' &&
8050 if (name[2] == 'a' &&
8067 if (name[2] == 'e' &&
8077 if (name[2] == 'e' &&
8081 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8097 if (name[3] == 'i' &&
8106 if (name[3] == 'o' &&
8142 if (name[2] == 'o' &&
8152 if (name[2] == 'y' &&
8166 if (name[1] == 'l' &&
8180 if (name[2] == 'n' &&
8190 if (name[2] == 'o' &&
8204 if (name[1] == 'i' &&
8209 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8218 if (name[2] == 'd' &&
8228 if (name[2] == 'c' &&
8245 if (name[2] == 'c' &&
8255 if (name[2] == 't' &&
8269 if (name[1] == 'k' &&
8280 if (name[1] == 'r' &&
8294 if (name[2] == 's' &&
8304 if (name[2] == 'd' &&
8321 if (name[2] == 'm' &&
8331 if (name[2] == 'i' &&
8341 if (name[2] == 'e' &&
8351 if (name[2] == 'l' &&
8361 if (name[2] == 'a' &&
8374 if (name[3] == 't' &&
8377 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8383 if (name[3] == 'd' &&
8400 if (name[1] == 'i' &&
8414 if (name[2] == 'a' &&
8427 if (name[3] == 'e' &&
8462 if (name[2] == 'i' &&
8479 if (name[2] == 'i' &&
8489 if (name[2] == 'i' &&
8506 case 6: /* 33 tokens of length 6 */
8510 if (name[1] == 'c' &&
8525 if (name[2] == 'l' &&
8536 if (name[2] == 'r' &&
8551 if (name[1] == 'e' &&
8566 if (name[2] == 's' &&
8571 if(ckWARN_d(WARN_SYNTAX))
8572 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8578 if (name[2] == 'i' &&
8596 if (name[2] == 'l' &&
8607 if (name[2] == 'r' &&
8622 if (name[1] == 'm' &&
8637 if (name[2] == 'n' &&
8648 if (name[2] == 's' &&
8663 if (name[1] == 's' &&
8669 if (name[4] == 't' &&
8678 if (name[4] == 'e' &&
8687 if (name[4] == 'c' &&
8696 if (name[4] == 'n' &&
8712 if (name[1] == 'r' &&
8730 if (name[3] == 'a' &&
8740 if (name[3] == 'u' &&
8754 if (name[2] == 'n' &&
8772 if (name[2] == 'a' &&
8786 if (name[3] == 'e' &&
8799 if (name[4] == 't' &&
8808 if (name[4] == 'e' &&
8830 if (name[4] == 't' &&
8839 if (name[4] == 'e' &&
8855 if (name[2] == 'c' &&
8866 if (name[2] == 'l' &&
8877 if (name[2] == 'b' &&
8888 if (name[2] == 's' &&
8911 if (name[4] == 's' &&
8920 if (name[4] == 'n' &&
8933 if (name[3] == 'a' &&
8950 if (name[1] == 'a' &&
8965 case 7: /* 29 tokens of length 7 */
8969 if (name[1] == 'E' &&
8982 if (name[1] == '_' &&
8995 if (name[1] == 'i' &&
9002 return -KEY_binmode;
9008 if (name[1] == 'o' &&
9015 return -KEY_connect;
9024 if (name[2] == 'm' &&
9030 return -KEY_dbmopen;
9041 if (name[4] == 'u' &&
9045 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9051 if (name[4] == 'n' &&
9072 if (name[1] == 'o' &&
9085 if (name[1] == 'e' &&
9092 if (name[5] == 'r' &&
9095 return -KEY_getpgrp;
9101 if (name[5] == 'i' &&
9104 return -KEY_getppid;
9117 if (name[1] == 'c' &&
9124 return -KEY_lcfirst;
9130 if (name[1] == 'p' &&
9137 return -KEY_opendir;
9143 if (name[1] == 'a' &&
9161 if (name[3] == 'd' &&
9166 return -KEY_readdir;
9172 if (name[3] == 'u' &&
9183 if (name[3] == 'e' &&
9188 return -KEY_reverse;
9207 if (name[3] == 'k' &&
9212 return -KEY_seekdir;
9218 if (name[3] == 'p' &&
9223 return -KEY_setpgrp;
9233 if (name[2] == 'm' &&
9239 return -KEY_shmread;
9245 if (name[2] == 'r' &&
9251 return -KEY_sprintf;
9260 if (name[3] == 'l' &&
9265 return -KEY_symlink;
9274 if (name[4] == 'a' &&
9278 return -KEY_syscall;
9284 if (name[4] == 'p' &&
9288 return -KEY_sysopen;
9294 if (name[4] == 'e' &&
9298 return -KEY_sysread;
9304 if (name[4] == 'e' &&
9308 return -KEY_sysseek;
9326 if (name[1] == 'e' &&
9333 return -KEY_telldir;
9342 if (name[2] == 'f' &&
9348 return -KEY_ucfirst;
9354 if (name[2] == 's' &&
9360 return -KEY_unshift;
9370 if (name[1] == 'a' &&
9377 return -KEY_waitpid;
9386 case 8: /* 26 tokens of length 8 */
9390 if (name[1] == 'U' &&
9398 return KEY_AUTOLOAD;
9409 if (name[3] == 'A' &&
9415 return KEY___DATA__;
9421 if (name[3] == 'I' &&
9427 return -KEY___FILE__;
9433 if (name[3] == 'I' &&
9439 return -KEY___LINE__;
9455 if (name[2] == 'o' &&
9462 return -KEY_closedir;
9468 if (name[2] == 'n' &&
9475 return -KEY_continue;
9485 if (name[1] == 'b' &&
9493 return -KEY_dbmclose;
9499 if (name[1] == 'n' &&
9505 if (name[4] == 'r' &&
9510 return -KEY_endgrent;
9516 if (name[4] == 'w' &&
9521 return -KEY_endpwent;
9534 if (name[1] == 'o' &&
9542 return -KEY_formline;
9548 if (name[1] == 'e' &&
9559 if (name[6] == 'n' &&
9562 return -KEY_getgrent;
9568 if (name[6] == 'i' &&
9571 return -KEY_getgrgid;
9577 if (name[6] == 'a' &&
9580 return -KEY_getgrnam;
9593 if (name[4] == 'o' &&
9598 return -KEY_getlogin;
9609 if (name[6] == 'n' &&
9612 return -KEY_getpwent;
9618 if (name[6] == 'a' &&
9621 return -KEY_getpwnam;
9627 if (name[6] == 'i' &&
9630 return -KEY_getpwuid;
9650 if (name[1] == 'e' &&
9657 if (name[5] == 'i' &&
9664 return -KEY_readline;
9669 return -KEY_readlink;
9680 if (name[5] == 'i' &&
9684 return -KEY_readpipe;
9705 if (name[4] == 'r' &&
9710 return -KEY_setgrent;
9716 if (name[4] == 'w' &&
9721 return -KEY_setpwent;
9737 if (name[3] == 'w' &&
9743 return -KEY_shmwrite;
9749 if (name[3] == 't' &&
9755 return -KEY_shutdown;
9765 if (name[2] == 's' &&
9772 return -KEY_syswrite;
9782 if (name[1] == 'r' &&
9790 return -KEY_truncate;
9799 case 9: /* 9 tokens of length 9 */
9803 if (name[1] == 'N' &&
9812 return KEY_UNITCHECK;
9818 if (name[1] == 'n' &&
9827 return -KEY_endnetent;
9833 if (name[1] == 'e' &&
9842 return -KEY_getnetent;
9848 if (name[1] == 'o' &&
9857 return -KEY_localtime;
9863 if (name[1] == 'r' &&
9872 return KEY_prototype;
9878 if (name[1] == 'u' &&
9887 return -KEY_quotemeta;
9893 if (name[1] == 'e' &&
9902 return -KEY_rewinddir;
9908 if (name[1] == 'e' &&
9917 return -KEY_setnetent;
9923 if (name[1] == 'a' &&
9932 return -KEY_wantarray;
9941 case 10: /* 9 tokens of length 10 */
9945 if (name[1] == 'n' &&
9951 if (name[4] == 'o' &&
9958 return -KEY_endhostent;
9964 if (name[4] == 'e' &&
9971 return -KEY_endservent;
9984 if (name[1] == 'e' &&
9990 if (name[4] == 'o' &&
9997 return -KEY_gethostent;
10006 if (name[5] == 'r' &&
10012 return -KEY_getservent;
10018 if (name[5] == 'c' &&
10024 return -KEY_getsockopt;
10044 if (name[2] == 't')
10049 if (name[4] == 'o' &&
10056 return -KEY_sethostent;
10065 if (name[5] == 'r' &&
10071 return -KEY_setservent;
10077 if (name[5] == 'c' &&
10083 return -KEY_setsockopt;
10100 if (name[2] == 'c' &&
10109 return -KEY_socketpair;
10122 case 11: /* 8 tokens of length 11 */
10126 if (name[1] == '_' &&
10136 { /* __PACKAGE__ */
10137 return -KEY___PACKAGE__;
10143 if (name[1] == 'n' &&
10153 { /* endprotoent */
10154 return -KEY_endprotoent;
10160 if (name[1] == 'e' &&
10169 if (name[5] == 'e' &&
10175 { /* getpeername */
10176 return -KEY_getpeername;
10185 if (name[6] == 'o' &&
10190 { /* getpriority */
10191 return -KEY_getpriority;
10197 if (name[6] == 't' &&
10202 { /* getprotoent */
10203 return -KEY_getprotoent;
10217 if (name[4] == 'o' &&
10224 { /* getsockname */
10225 return -KEY_getsockname;
10238 if (name[1] == 'e' &&
10246 if (name[6] == 'o' &&
10251 { /* setpriority */
10252 return -KEY_setpriority;
10258 if (name[6] == 't' &&
10263 { /* setprotoent */
10264 return -KEY_setprotoent;
10280 case 12: /* 2 tokens of length 12 */
10281 if (name[0] == 'g' &&
10293 if (name[9] == 'd' &&
10296 { /* getnetbyaddr */
10297 return -KEY_getnetbyaddr;
10303 if (name[9] == 'a' &&
10306 { /* getnetbyname */
10307 return -KEY_getnetbyname;
10319 case 13: /* 4 tokens of length 13 */
10320 if (name[0] == 'g' &&
10327 if (name[4] == 'o' &&
10336 if (name[10] == 'd' &&
10339 { /* gethostbyaddr */
10340 return -KEY_gethostbyaddr;
10346 if (name[10] == 'a' &&
10349 { /* gethostbyname */
10350 return -KEY_gethostbyname;
10363 if (name[4] == 'e' &&
10372 if (name[10] == 'a' &&
10375 { /* getservbyname */
10376 return -KEY_getservbyname;
10382 if (name[10] == 'o' &&
10385 { /* getservbyport */
10386 return -KEY_getservbyport;
10405 case 14: /* 1 tokens of length 14 */
10406 if (name[0] == 'g' &&
10420 { /* getprotobyname */
10421 return -KEY_getprotobyname;
10426 case 16: /* 1 tokens of length 16 */
10427 if (name[0] == 'g' &&
10443 { /* getprotobynumber */
10444 return -KEY_getprotobynumber;
10458 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10462 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10463 if (ckWARN(WARN_SYNTAX)) {
10466 for (w = s+2; *w && level; w++) {
10469 else if (*w == ')')
10472 while (isSPACE(*w))
10474 /* the list of chars below is for end of statements or
10475 * block / parens, boolean operators (&&, ||, //) and branch
10476 * constructs (or, and, if, until, unless, while, err, for).
10477 * Not a very solid hack... */
10478 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10479 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10480 "%s (...) interpreted as function",name);
10483 while (s < PL_bufend && isSPACE(*s))
10487 while (s < PL_bufend && isSPACE(*s))
10489 if (isIDFIRST_lazy_if(s,UTF)) {
10490 const char * const w = s++;
10491 while (isALNUM_lazy_if(s,UTF))
10493 while (s < PL_bufend && isSPACE(*s))
10497 if (keyword(w, s - w, 0))
10500 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10501 if (gv && GvCVu(gv))
10503 Perl_croak(aTHX_ "No comma allowed after %s", what);
10508 /* Either returns sv, or mortalizes sv and returns a new SV*.
10509 Best used as sv=new_constant(..., sv, ...).
10510 If s, pv are NULL, calls subroutine with one argument,
10511 and type is used with error messages only. */
10514 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
10515 SV *sv, SV *pv, const char *type, STRLEN typelen)
10518 HV * const table = GvHV(PL_hintgv); /* ^H */
10522 const char *why1 = "", *why2 = "", *why3 = "";
10524 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10527 why2 = (const char *)
10528 (strEQ(key,"charnames")
10529 ? "(possibly a missing \"use charnames ...\")"
10531 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10532 (type ? type: "undef"), why2);
10534 /* This is convoluted and evil ("goto considered harmful")
10535 * but I do not understand the intricacies of all the different
10536 * failure modes of %^H in here. The goal here is to make
10537 * the most probable error message user-friendly. --jhi */
10542 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10543 (type ? type: "undef"), why1, why2, why3);
10545 yyerror(SvPVX_const(msg));
10549 cvp = hv_fetch(table, key, keylen, FALSE);
10550 if (!cvp || !SvOK(*cvp)) {
10553 why3 = "} is not defined";
10556 sv_2mortal(sv); /* Parent created it permanently */
10559 pv = sv_2mortal(newSVpvn(s, len));
10561 typesv = sv_2mortal(newSVpvn(type, typelen));
10563 typesv = &PL_sv_undef;
10565 PUSHSTACKi(PERLSI_OVERLOAD);
10577 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10581 /* Check the eval first */
10582 if (!PL_in_eval && SvTRUE(ERRSV)) {
10583 sv_catpvs(ERRSV, "Propagated");
10584 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10586 res = SvREFCNT_inc_simple(sv);
10590 SvREFCNT_inc_simple_void(res);
10599 why1 = "Call to &{$^H{";
10601 why3 = "}} did not return a defined value";
10609 /* Returns a NUL terminated string, with the length of the string written to
10613 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10616 register char *d = dest;
10617 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10620 Perl_croak(aTHX_ ident_too_long);
10621 if (isALNUM(*s)) /* UTF handled below */
10623 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10628 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10632 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10633 char *t = s + UTF8SKIP(s);
10635 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10639 Perl_croak(aTHX_ ident_too_long);
10640 Copy(s, d, len, char);
10653 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10656 char *bracket = NULL;
10658 register char *d = dest;
10659 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10664 while (isDIGIT(*s)) {
10666 Perl_croak(aTHX_ ident_too_long);
10673 Perl_croak(aTHX_ ident_too_long);
10674 if (isALNUM(*s)) /* UTF handled below */
10676 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10681 else if (*s == ':' && s[1] == ':') {
10685 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10686 char *t = s + UTF8SKIP(s);
10687 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10689 if (d + (t - s) > e)
10690 Perl_croak(aTHX_ ident_too_long);
10691 Copy(s, d, t - s, char);
10702 if (PL_lex_state != LEX_NORMAL)
10703 PL_lex_state = LEX_INTERPENDMAYBE;
10706 if (*s == '$' && s[1] &&
10707 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10720 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10725 if (isSPACE(s[-1])) {
10727 const char ch = *s++;
10728 if (!SPACE_OR_TAB(ch)) {
10734 if (isIDFIRST_lazy_if(d,UTF)) {
10738 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10739 end += UTF8SKIP(end);
10740 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10741 end += UTF8SKIP(end);
10743 Copy(s, d, end - s, char);
10748 while ((isALNUM(*s) || *s == ':') && d < e)
10751 Perl_croak(aTHX_ ident_too_long);
10754 while (s < send && SPACE_OR_TAB(*s))
10756 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10757 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10758 const char * const brack =
10760 ((*s == '[') ? "[...]" : "{...}");
10761 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10762 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10763 funny, dest, brack, funny, dest, brack);
10766 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10770 /* Handle extended ${^Foo} variables
10771 * 1999-02-27 mjd-perl-patch@plover.com */
10772 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10776 while (isALNUM(*s) && d < e) {
10780 Perl_croak(aTHX_ ident_too_long);
10785 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10786 PL_lex_state = LEX_INTERPEND;
10789 if (PL_lex_state == LEX_NORMAL) {
10790 if (ckWARN(WARN_AMBIGUOUS) &&
10791 (keyword(dest, d - dest, 0)
10792 || get_cvn_flags(dest, d - dest, 0)))
10796 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10797 "Ambiguous use of %c{%s} resolved to %c%s",
10798 funny, dest, funny, dest);
10803 s = bracket; /* let the parser handle it */
10807 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10808 PL_lex_state = LEX_INTERPEND;
10813 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10815 PERL_UNUSED_CONTEXT;
10819 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10820 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10821 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10822 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10823 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10829 S_scan_pat(pTHX_ char *start, I32 type)
10833 char *s = scan_str(start,!!PL_madskills,FALSE);
10834 const char * const valid_flags =
10835 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10842 const char * const delimiter = skipspace(start);
10846 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10847 : "Search pattern not terminated" ));
10850 pm = (PMOP*)newPMOP(type, 0);
10851 if (PL_multi_open == '?') {
10852 /* This is the only point in the code that sets PMf_ONCE: */
10853 pm->op_pmflags |= PMf_ONCE;
10855 /* Hence it's safe to do this bit of PMOP book-keeping here, which
10856 allows us to restrict the list needed by reset to just the ??
10858 assert(type != OP_TRANS);
10860 MAGIC *mg = mg_find((SV*)PL_curstash, PERL_MAGIC_symtab);
10863 mg = sv_magicext((SV*)PL_curstash, 0, PERL_MAGIC_symtab, 0, 0,
10866 elements = mg->mg_len / sizeof(PMOP**);
10867 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
10868 ((PMOP**)mg->mg_ptr) [elements++] = pm;
10869 mg->mg_len = elements * sizeof(PMOP**);
10870 PmopSTASH_set(pm,PL_curstash);
10876 while (*s && strchr(valid_flags, *s))
10877 pmflag(&pm->op_pmflags,*s++);
10879 if (PL_madskills && modstart != s) {
10880 SV* tmptoken = newSVpvn(modstart, s - modstart);
10881 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10884 /* issue a warning if /c is specified,but /g is not */
10885 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10886 && ckWARN(WARN_REGEXP))
10888 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
10889 "Use of /c modifier is meaningless without /g" );
10892 PL_lex_op = (OP*)pm;
10893 pl_yylval.ival = OP_MATCH;
10898 S_scan_subst(pTHX_ char *start)
10909 pl_yylval.ival = OP_NULL;
10911 s = scan_str(start,!!PL_madskills,FALSE);
10914 Perl_croak(aTHX_ "Substitution pattern not terminated");
10916 if (s[-1] == PL_multi_open)
10919 if (PL_madskills) {
10920 CURMAD('q', PL_thisopen);
10921 CURMAD('_', PL_thiswhite);
10922 CURMAD('E', PL_thisstuff);
10923 CURMAD('Q', PL_thisclose);
10924 PL_realtokenstart = s - SvPVX(PL_linestr);
10928 first_start = PL_multi_start;
10929 s = scan_str(s,!!PL_madskills,FALSE);
10931 if (PL_lex_stuff) {
10932 SvREFCNT_dec(PL_lex_stuff);
10933 PL_lex_stuff = NULL;
10935 Perl_croak(aTHX_ "Substitution replacement not terminated");
10937 PL_multi_start = first_start; /* so whole substitution is taken together */
10939 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10942 if (PL_madskills) {
10943 CURMAD('z', PL_thisopen);
10944 CURMAD('R', PL_thisstuff);
10945 CURMAD('Z', PL_thisclose);
10951 if (*s == EXEC_PAT_MOD) {
10955 else if (strchr(S_PAT_MODS, *s))
10956 pmflag(&pm->op_pmflags,*s++);
10962 if (PL_madskills) {
10964 curmad('m', newSVpvn(modstart, s - modstart));
10965 append_madprops(PL_thismad, (OP*)pm, 0);
10969 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10970 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10974 SV * const repl = newSVpvs("");
10976 PL_sublex_info.super_bufptr = s;
10977 PL_sublex_info.super_bufend = PL_bufend;
10979 pm->op_pmflags |= PMf_EVAL;
10982 sv_catpvs(repl, "eval ");
10984 sv_catpvs(repl, "do ");
10986 sv_catpvs(repl, "{");
10987 sv_catsv(repl, PL_lex_repl);
10988 if (strchr(SvPVX(PL_lex_repl), '#'))
10989 sv_catpvs(repl, "\n");
10990 sv_catpvs(repl, "}");
10992 SvREFCNT_dec(PL_lex_repl);
10993 PL_lex_repl = repl;
10996 PL_lex_op = (OP*)pm;
10997 pl_yylval.ival = OP_SUBST;
11002 S_scan_trans(pTHX_ char *start)
11015 pl_yylval.ival = OP_NULL;
11017 s = scan_str(start,!!PL_madskills,FALSE);
11019 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11021 if (s[-1] == PL_multi_open)
11024 if (PL_madskills) {
11025 CURMAD('q', PL_thisopen);
11026 CURMAD('_', PL_thiswhite);
11027 CURMAD('E', PL_thisstuff);
11028 CURMAD('Q', PL_thisclose);
11029 PL_realtokenstart = s - SvPVX(PL_linestr);
11033 s = scan_str(s,!!PL_madskills,FALSE);
11035 if (PL_lex_stuff) {
11036 SvREFCNT_dec(PL_lex_stuff);
11037 PL_lex_stuff = NULL;
11039 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11041 if (PL_madskills) {
11042 CURMAD('z', PL_thisopen);
11043 CURMAD('R', PL_thisstuff);
11044 CURMAD('Z', PL_thisclose);
11047 complement = del = squash = 0;
11054 complement = OPpTRANS_COMPLEMENT;
11057 del = OPpTRANS_DELETE;
11060 squash = OPpTRANS_SQUASH;
11069 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11070 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11071 o->op_private &= ~OPpTRANS_ALL;
11072 o->op_private |= del|squash|complement|
11073 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11074 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11077 pl_yylval.ival = OP_TRANS;
11080 if (PL_madskills) {
11082 curmad('m', newSVpvn(modstart, s - modstart));
11083 append_madprops(PL_thismad, o, 0);
11092 S_scan_heredoc(pTHX_ register char *s)
11096 I32 op_type = OP_SCALAR;
11100 const char *found_newline;
11104 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11106 I32 stuffstart = s - SvPVX(PL_linestr);
11109 PL_realtokenstart = -1;
11114 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11118 while (SPACE_OR_TAB(*peek))
11120 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11123 s = delimcpy(d, e, s, PL_bufend, term, &len);
11133 if (!isALNUM_lazy_if(s,UTF))
11134 deprecate_old("bare << to mean <<\"\"");
11135 for (; isALNUM_lazy_if(s,UTF); s++) {
11140 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11141 Perl_croak(aTHX_ "Delimiter for here document is too long");
11144 len = d - PL_tokenbuf;
11147 if (PL_madskills) {
11148 tstart = PL_tokenbuf + !outer;
11149 PL_thisclose = newSVpvn(tstart, len - !outer);
11150 tstart = SvPVX(PL_linestr) + stuffstart;
11151 PL_thisopen = newSVpvn(tstart, s - tstart);
11152 stuffstart = s - SvPVX(PL_linestr);
11155 #ifndef PERL_STRICT_CR
11156 d = strchr(s, '\r');
11158 char * const olds = s;
11160 while (s < PL_bufend) {
11166 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11175 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11182 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11183 herewas = newSVpvn(s,PL_bufend-s);
11187 herewas = newSVpvn(s-1,found_newline-s+1);
11190 herewas = newSVpvn(s,found_newline-s);
11194 if (PL_madskills) {
11195 tstart = SvPVX(PL_linestr) + stuffstart;
11197 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11199 PL_thisstuff = newSVpvn(tstart, s - tstart);
11202 s += SvCUR(herewas);
11205 stuffstart = s - SvPVX(PL_linestr);
11211 tmpstr = newSV_type(SVt_PVIV);
11212 SvGROW(tmpstr, 80);
11213 if (term == '\'') {
11214 op_type = OP_CONST;
11215 SvIV_set(tmpstr, -1);
11217 else if (term == '`') {
11218 op_type = OP_BACKTICK;
11219 SvIV_set(tmpstr, '\\');
11223 PL_multi_start = CopLINE(PL_curcop);
11224 PL_multi_open = PL_multi_close = '<';
11225 term = *PL_tokenbuf;
11226 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11227 char * const bufptr = PL_sublex_info.super_bufptr;
11228 char * const bufend = PL_sublex_info.super_bufend;
11229 char * const olds = s - SvCUR(herewas);
11230 s = strchr(bufptr, '\n');
11234 while (s < bufend &&
11235 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11237 CopLINE_inc(PL_curcop);
11240 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11241 missingterm(PL_tokenbuf);
11243 sv_setpvn(herewas,bufptr,d-bufptr+1);
11244 sv_setpvn(tmpstr,d+1,s-d);
11246 sv_catpvn(herewas,s,bufend-s);
11247 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11254 while (s < PL_bufend &&
11255 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11257 CopLINE_inc(PL_curcop);
11259 if (s >= PL_bufend) {
11260 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11261 missingterm(PL_tokenbuf);
11263 sv_setpvn(tmpstr,d+1,s-d);
11265 if (PL_madskills) {
11267 sv_catpvn(PL_thisstuff, d + 1, s - d);
11269 PL_thisstuff = newSVpvn(d + 1, s - d);
11270 stuffstart = s - SvPVX(PL_linestr);
11274 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11276 sv_catpvn(herewas,s,PL_bufend-s);
11277 sv_setsv(PL_linestr,herewas);
11278 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11279 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11280 PL_last_lop = PL_last_uni = NULL;
11283 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11284 while (s >= PL_bufend) { /* multiple line string? */
11286 if (PL_madskills) {
11287 tstart = SvPVX(PL_linestr) + stuffstart;
11289 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11291 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11295 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11296 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11297 missingterm(PL_tokenbuf);
11300 stuffstart = s - SvPVX(PL_linestr);
11302 CopLINE_inc(PL_curcop);
11303 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11304 PL_last_lop = PL_last_uni = NULL;
11305 #ifndef PERL_STRICT_CR
11306 if (PL_bufend - PL_linestart >= 2) {
11307 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11308 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11310 PL_bufend[-2] = '\n';
11312 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11314 else if (PL_bufend[-1] == '\r')
11315 PL_bufend[-1] = '\n';
11317 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11318 PL_bufend[-1] = '\n';
11320 if (PERLDB_LINE && PL_curstash != PL_debstash)
11321 update_debugger_info(PL_linestr, NULL, 0);
11322 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11323 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11324 *(SvPVX(PL_linestr) + off ) = ' ';
11325 sv_catsv(PL_linestr,herewas);
11326 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11327 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11331 sv_catsv(tmpstr,PL_linestr);
11336 PL_multi_end = CopLINE(PL_curcop);
11337 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11338 SvPV_shrink_to_cur(tmpstr);
11340 SvREFCNT_dec(herewas);
11342 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11344 else if (PL_encoding)
11345 sv_recode_to_utf8(tmpstr, PL_encoding);
11347 PL_lex_stuff = tmpstr;
11348 pl_yylval.ival = op_type;
11352 /* scan_inputsymbol
11353 takes: current position in input buffer
11354 returns: new position in input buffer
11355 side-effects: pl_yylval and lex_op are set.
11360 <FH> read from filehandle
11361 <pkg::FH> read from package qualified filehandle
11362 <pkg'FH> read from package qualified filehandle
11363 <$fh> read from filehandle in $fh
11364 <*.h> filename glob
11369 S_scan_inputsymbol(pTHX_ char *start)
11372 register char *s = start; /* current position in buffer */
11376 char *d = PL_tokenbuf; /* start of temp holding space */
11377 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11379 end = strchr(s, '\n');
11382 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11384 /* die if we didn't have space for the contents of the <>,
11385 or if it didn't end, or if we see a newline
11388 if (len >= (I32)sizeof PL_tokenbuf)
11389 Perl_croak(aTHX_ "Excessively long <> operator");
11391 Perl_croak(aTHX_ "Unterminated <> operator");
11396 Remember, only scalar variables are interpreted as filehandles by
11397 this code. Anything more complex (e.g., <$fh{$num}>) will be
11398 treated as a glob() call.
11399 This code makes use of the fact that except for the $ at the front,
11400 a scalar variable and a filehandle look the same.
11402 if (*d == '$' && d[1]) d++;
11404 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11405 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11408 /* If we've tried to read what we allow filehandles to look like, and
11409 there's still text left, then it must be a glob() and not a getline.
11410 Use scan_str to pull out the stuff between the <> and treat it
11411 as nothing more than a string.
11414 if (d - PL_tokenbuf != len) {
11415 pl_yylval.ival = OP_GLOB;
11416 s = scan_str(start,!!PL_madskills,FALSE);
11418 Perl_croak(aTHX_ "Glob not terminated");
11422 bool readline_overriden = FALSE;
11425 /* we're in a filehandle read situation */
11428 /* turn <> into <ARGV> */
11430 Copy("ARGV",d,5,char);
11432 /* Check whether readline() is overriden */
11433 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11435 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11437 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11438 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11439 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11440 readline_overriden = TRUE;
11442 /* if <$fh>, create the ops to turn the variable into a
11446 /* try to find it in the pad for this block, otherwise find
11447 add symbol table ops
11449 const PADOFFSET tmp = pad_findmy(d);
11450 if (tmp != NOT_IN_PAD) {
11451 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11452 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11453 HEK * const stashname = HvNAME_HEK(stash);
11454 SV * const sym = sv_2mortal(newSVhek(stashname));
11455 sv_catpvs(sym, "::");
11456 sv_catpv(sym, d+1);
11461 OP * const o = newOP(OP_PADSV, 0);
11463 PL_lex_op = readline_overriden
11464 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11465 append_elem(OP_LIST, o,
11466 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11467 : (OP*)newUNOP(OP_READLINE, 0, o);
11476 ? (GV_ADDMULTI | GV_ADDINEVAL)
11479 PL_lex_op = readline_overriden
11480 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11481 append_elem(OP_LIST,
11482 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11483 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11484 : (OP*)newUNOP(OP_READLINE, 0,
11485 newUNOP(OP_RV2SV, 0,
11486 newGVOP(OP_GV, 0, gv)));
11488 if (!readline_overriden)
11489 PL_lex_op->op_flags |= OPf_SPECIAL;
11490 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
11491 pl_yylval.ival = OP_NULL;
11494 /* If it's none of the above, it must be a literal filehandle
11495 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11497 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11498 PL_lex_op = readline_overriden
11499 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11500 append_elem(OP_LIST,
11501 newGVOP(OP_GV, 0, gv),
11502 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11503 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11504 pl_yylval.ival = OP_NULL;
11513 takes: start position in buffer
11514 keep_quoted preserve \ on the embedded delimiter(s)
11515 keep_delims preserve the delimiters around the string
11516 returns: position to continue reading from buffer
11517 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11518 updates the read buffer.
11520 This subroutine pulls a string out of the input. It is called for:
11521 q single quotes q(literal text)
11522 ' single quotes 'literal text'
11523 qq double quotes qq(interpolate $here please)
11524 " double quotes "interpolate $here please"
11525 qx backticks qx(/bin/ls -l)
11526 ` backticks `/bin/ls -l`
11527 qw quote words @EXPORT_OK = qw( func() $spam )
11528 m// regexp match m/this/
11529 s/// regexp substitute s/this/that/
11530 tr/// string transliterate tr/this/that/
11531 y/// string transliterate y/this/that/
11532 ($*@) sub prototypes sub foo ($)
11533 (stuff) sub attr parameters sub foo : attr(stuff)
11534 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11536 In most of these cases (all but <>, patterns and transliterate)
11537 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11538 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11539 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11542 It skips whitespace before the string starts, and treats the first
11543 character as the delimiter. If the delimiter is one of ([{< then
11544 the corresponding "close" character )]}> is used as the closing
11545 delimiter. It allows quoting of delimiters, and if the string has
11546 balanced delimiters ([{<>}]) it allows nesting.
11548 On success, the SV with the resulting string is put into lex_stuff or,
11549 if that is already non-NULL, into lex_repl. The second case occurs only
11550 when parsing the RHS of the special constructs s/// and tr/// (y///).
11551 For convenience, the terminating delimiter character is stuffed into
11556 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11559 SV *sv; /* scalar value: string */
11560 const char *tmps; /* temp string, used for delimiter matching */
11561 register char *s = start; /* current position in the buffer */
11562 register char term; /* terminating character */
11563 register char *to; /* current position in the sv's data */
11564 I32 brackets = 1; /* bracket nesting level */
11565 bool has_utf8 = FALSE; /* is there any utf8 content? */
11566 I32 termcode; /* terminating char. code */
11567 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11568 STRLEN termlen; /* length of terminating string */
11569 int last_off = 0; /* last position for nesting bracket */
11575 /* skip space before the delimiter */
11581 if (PL_realtokenstart >= 0) {
11582 stuffstart = PL_realtokenstart;
11583 PL_realtokenstart = -1;
11586 stuffstart = start - SvPVX(PL_linestr);
11588 /* mark where we are, in case we need to report errors */
11591 /* after skipping whitespace, the next character is the terminator */
11594 termcode = termstr[0] = term;
11598 termcode = utf8_to_uvchr((U8*)s, &termlen);
11599 Copy(s, termstr, termlen, U8);
11600 if (!UTF8_IS_INVARIANT(term))
11604 /* mark where we are */
11605 PL_multi_start = CopLINE(PL_curcop);
11606 PL_multi_open = term;
11608 /* find corresponding closing delimiter */
11609 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11610 termcode = termstr[0] = term = tmps[5];
11612 PL_multi_close = term;
11614 /* create a new SV to hold the contents. 79 is the SV's initial length.
11615 What a random number. */
11616 sv = newSV_type(SVt_PVIV);
11618 SvIV_set(sv, termcode);
11619 (void)SvPOK_only(sv); /* validate pointer */
11621 /* move past delimiter and try to read a complete string */
11623 sv_catpvn(sv, s, termlen);
11626 tstart = SvPVX(PL_linestr) + stuffstart;
11627 if (!PL_thisopen && !keep_delims) {
11628 PL_thisopen = newSVpvn(tstart, s - tstart);
11629 stuffstart = s - SvPVX(PL_linestr);
11633 if (PL_encoding && !UTF) {
11637 int offset = s - SvPVX_const(PL_linestr);
11638 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11639 &offset, (char*)termstr, termlen);
11640 const char * const ns = SvPVX_const(PL_linestr) + offset;
11641 char * const svlast = SvEND(sv) - 1;
11643 for (; s < ns; s++) {
11644 if (*s == '\n' && !PL_rsfp)
11645 CopLINE_inc(PL_curcop);
11648 goto read_more_line;
11650 /* handle quoted delimiters */
11651 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11653 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11655 if ((svlast-1 - t) % 2) {
11656 if (!keep_quoted) {
11657 *(svlast-1) = term;
11659 SvCUR_set(sv, SvCUR(sv) - 1);
11664 if (PL_multi_open == PL_multi_close) {
11670 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11671 /* At here, all closes are "was quoted" one,
11672 so we don't check PL_multi_close. */
11674 if (!keep_quoted && *(t+1) == PL_multi_open)
11679 else if (*t == PL_multi_open)
11687 SvCUR_set(sv, w - SvPVX_const(sv));
11689 last_off = w - SvPVX(sv);
11690 if (--brackets <= 0)
11695 if (!keep_delims) {
11696 SvCUR_set(sv, SvCUR(sv) - 1);
11702 /* extend sv if need be */
11703 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11704 /* set 'to' to the next character in the sv's string */
11705 to = SvPVX(sv)+SvCUR(sv);
11707 /* if open delimiter is the close delimiter read unbridle */
11708 if (PL_multi_open == PL_multi_close) {
11709 for (; s < PL_bufend; s++,to++) {
11710 /* embedded newlines increment the current line number */
11711 if (*s == '\n' && !PL_rsfp)
11712 CopLINE_inc(PL_curcop);
11713 /* handle quoted delimiters */
11714 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11715 if (!keep_quoted && s[1] == term)
11717 /* any other quotes are simply copied straight through */
11721 /* terminate when run out of buffer (the for() condition), or
11722 have found the terminator */
11723 else if (*s == term) {
11726 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11729 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11735 /* if the terminator isn't the same as the start character (e.g.,
11736 matched brackets), we have to allow more in the quoting, and
11737 be prepared for nested brackets.
11740 /* read until we run out of string, or we find the terminator */
11741 for (; s < PL_bufend; s++,to++) {
11742 /* embedded newlines increment the line count */
11743 if (*s == '\n' && !PL_rsfp)
11744 CopLINE_inc(PL_curcop);
11745 /* backslashes can escape the open or closing characters */
11746 if (*s == '\\' && s+1 < PL_bufend) {
11747 if (!keep_quoted &&
11748 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11753 /* allow nested opens and closes */
11754 else if (*s == PL_multi_close && --brackets <= 0)
11756 else if (*s == PL_multi_open)
11758 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11763 /* terminate the copied string and update the sv's end-of-string */
11765 SvCUR_set(sv, to - SvPVX_const(sv));
11768 * this next chunk reads more into the buffer if we're not done yet
11772 break; /* handle case where we are done yet :-) */
11774 #ifndef PERL_STRICT_CR
11775 if (to - SvPVX_const(sv) >= 2) {
11776 if ((to[-2] == '\r' && to[-1] == '\n') ||
11777 (to[-2] == '\n' && to[-1] == '\r'))
11781 SvCUR_set(sv, to - SvPVX_const(sv));
11783 else if (to[-1] == '\r')
11786 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11791 /* if we're out of file, or a read fails, bail and reset the current
11792 line marker so we can report where the unterminated string began
11795 if (PL_madskills) {
11796 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11798 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11800 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11804 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11806 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11812 /* we read a line, so increment our line counter */
11813 CopLINE_inc(PL_curcop);
11815 /* update debugger info */
11816 if (PERLDB_LINE && PL_curstash != PL_debstash)
11817 update_debugger_info(PL_linestr, NULL, 0);
11819 /* having changed the buffer, we must update PL_bufend */
11820 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11821 PL_last_lop = PL_last_uni = NULL;
11824 /* at this point, we have successfully read the delimited string */
11826 if (!PL_encoding || UTF) {
11828 if (PL_madskills) {
11829 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11830 const int len = s - tstart;
11832 sv_catpvn(PL_thisstuff, tstart, len);
11834 PL_thisstuff = newSVpvn(tstart, len);
11835 if (!PL_thisclose && !keep_delims)
11836 PL_thisclose = newSVpvn(s,termlen);
11841 sv_catpvn(sv, s, termlen);
11846 if (PL_madskills) {
11847 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11848 const int len = s - tstart - termlen;
11850 sv_catpvn(PL_thisstuff, tstart, len);
11852 PL_thisstuff = newSVpvn(tstart, len);
11853 if (!PL_thisclose && !keep_delims)
11854 PL_thisclose = newSVpvn(s - termlen,termlen);
11858 if (has_utf8 || PL_encoding)
11861 PL_multi_end = CopLINE(PL_curcop);
11863 /* if we allocated too much space, give some back */
11864 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11865 SvLEN_set(sv, SvCUR(sv) + 1);
11866 SvPV_renew(sv, SvLEN(sv));
11869 /* decide whether this is the first or second quoted string we've read
11882 takes: pointer to position in buffer
11883 returns: pointer to new position in buffer
11884 side-effects: builds ops for the constant in pl_yylval.op
11886 Read a number in any of the formats that Perl accepts:
11888 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11889 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11892 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11894 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11897 If it reads a number without a decimal point or an exponent, it will
11898 try converting the number to an integer and see if it can do so
11899 without loss of precision.
11903 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11906 register const char *s = start; /* current position in buffer */
11907 register char *d; /* destination in temp buffer */
11908 register char *e; /* end of temp buffer */
11909 NV nv; /* number read, as a double */
11910 SV *sv = NULL; /* place to put the converted number */
11911 bool floatit; /* boolean: int or float? */
11912 const char *lastub = NULL; /* position of last underbar */
11913 static char const number_too_long[] = "Number too long";
11915 /* We use the first character to decide what type of number this is */
11919 Perl_croak(aTHX_ "panic: scan_num");
11921 /* if it starts with a 0, it could be an octal number, a decimal in
11922 0.13 disguise, or a hexadecimal number, or a binary number. */
11926 u holds the "number so far"
11927 shift the power of 2 of the base
11928 (hex == 4, octal == 3, binary == 1)
11929 overflowed was the number more than we can hold?
11931 Shift is used when we add a digit. It also serves as an "are
11932 we in octal/hex/binary?" indicator to disallow hex characters
11933 when in octal mode.
11938 bool overflowed = FALSE;
11939 bool just_zero = TRUE; /* just plain 0 or binary number? */
11940 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11941 static const char* const bases[5] =
11942 { "", "binary", "", "octal", "hexadecimal" };
11943 static const char* const Bases[5] =
11944 { "", "Binary", "", "Octal", "Hexadecimal" };
11945 static const char* const maxima[5] =
11947 "0b11111111111111111111111111111111",
11951 const char *base, *Base, *max;
11953 /* check for hex */
11958 } else if (s[1] == 'b') {
11963 /* check for a decimal in disguise */
11964 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11966 /* so it must be octal */
11973 if (ckWARN(WARN_SYNTAX))
11974 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11975 "Misplaced _ in number");
11979 base = bases[shift];
11980 Base = Bases[shift];
11981 max = maxima[shift];
11983 /* read the rest of the number */
11985 /* x is used in the overflow test,
11986 b is the digit we're adding on. */
11991 /* if we don't mention it, we're done */
11995 /* _ are ignored -- but warned about if consecutive */
11997 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11998 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11999 "Misplaced _ in number");
12003 /* 8 and 9 are not octal */
12004 case '8': case '9':
12006 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12010 case '2': case '3': case '4':
12011 case '5': case '6': case '7':
12013 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12016 case '0': case '1':
12017 b = *s++ & 15; /* ASCII digit -> value of digit */
12021 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12022 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12023 /* make sure they said 0x */
12026 b = (*s++ & 7) + 9;
12028 /* Prepare to put the digit we have onto the end
12029 of the number so far. We check for overflows.
12035 x = u << shift; /* make room for the digit */
12037 if ((x >> shift) != u
12038 && !(PL_hints & HINT_NEW_BINARY)) {
12041 if (ckWARN_d(WARN_OVERFLOW))
12042 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12043 "Integer overflow in %s number",
12046 u = x | b; /* add the digit to the end */
12049 n *= nvshift[shift];
12050 /* If an NV has not enough bits in its
12051 * mantissa to represent an UV this summing of
12052 * small low-order numbers is a waste of time
12053 * (because the NV cannot preserve the
12054 * low-order bits anyway): we could just
12055 * remember when did we overflow and in the
12056 * end just multiply n by the right
12064 /* if we get here, we had success: make a scalar value from
12069 /* final misplaced underbar check */
12070 if (s[-1] == '_') {
12071 if (ckWARN(WARN_SYNTAX))
12072 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12077 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12078 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12079 "%s number > %s non-portable",
12085 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12086 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12087 "%s number > %s non-portable",
12092 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12093 sv = new_constant(start, s - start, "integer",
12094 sv, NULL, NULL, 0);
12095 else if (PL_hints & HINT_NEW_BINARY)
12096 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12101 handle decimal numbers.
12102 we're also sent here when we read a 0 as the first digit
12104 case '1': case '2': case '3': case '4': case '5':
12105 case '6': case '7': case '8': case '9': case '.':
12108 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12111 /* read next group of digits and _ and copy into d */
12112 while (isDIGIT(*s) || *s == '_') {
12113 /* skip underscores, checking for misplaced ones
12117 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12118 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12119 "Misplaced _ in number");
12123 /* check for end of fixed-length buffer */
12125 Perl_croak(aTHX_ number_too_long);
12126 /* if we're ok, copy the character */
12131 /* final misplaced underbar check */
12132 if (lastub && s == lastub + 1) {
12133 if (ckWARN(WARN_SYNTAX))
12134 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12137 /* read a decimal portion if there is one. avoid
12138 3..5 being interpreted as the number 3. followed
12141 if (*s == '.' && s[1] != '.') {
12146 if (ckWARN(WARN_SYNTAX))
12147 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12148 "Misplaced _ in number");
12152 /* copy, ignoring underbars, until we run out of digits.
12154 for (; isDIGIT(*s) || *s == '_'; s++) {
12155 /* fixed length buffer check */
12157 Perl_croak(aTHX_ number_too_long);
12159 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12160 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12161 "Misplaced _ in number");
12167 /* fractional part ending in underbar? */
12168 if (s[-1] == '_') {
12169 if (ckWARN(WARN_SYNTAX))
12170 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12171 "Misplaced _ in number");
12173 if (*s == '.' && isDIGIT(s[1])) {
12174 /* oops, it's really a v-string, but without the "v" */
12180 /* read exponent part, if present */
12181 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12185 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12186 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12188 /* stray preinitial _ */
12190 if (ckWARN(WARN_SYNTAX))
12191 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12192 "Misplaced _ in number");
12196 /* allow positive or negative exponent */
12197 if (*s == '+' || *s == '-')
12200 /* stray initial _ */
12202 if (ckWARN(WARN_SYNTAX))
12203 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12204 "Misplaced _ in number");
12208 /* read digits of exponent */
12209 while (isDIGIT(*s) || *s == '_') {
12212 Perl_croak(aTHX_ number_too_long);
12216 if (((lastub && s == lastub + 1) ||
12217 (!isDIGIT(s[1]) && s[1] != '_'))
12218 && ckWARN(WARN_SYNTAX))
12219 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12220 "Misplaced _ in number");
12227 /* make an sv from the string */
12231 We try to do an integer conversion first if no characters
12232 indicating "float" have been found.
12237 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12239 if (flags == IS_NUMBER_IN_UV) {
12241 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12244 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12245 if (uv <= (UV) IV_MIN)
12246 sv_setiv(sv, -(IV)uv);
12253 /* terminate the string */
12255 nv = Atof(PL_tokenbuf);
12260 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12261 const char *const key = floatit ? "float" : "integer";
12262 const STRLEN keylen = floatit ? 5 : 7;
12263 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12264 key, keylen, sv, NULL, NULL, 0);
12268 /* if it starts with a v, it could be a v-string */
12271 sv = newSV(5); /* preallocate storage space */
12272 s = scan_vstring(s, PL_bufend, sv);
12276 /* make the op for the constant and return */
12279 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12281 lvalp->opval = NULL;
12287 S_scan_formline(pTHX_ register char *s)
12290 register char *eol;
12292 SV * const stuff = newSVpvs("");
12293 bool needargs = FALSE;
12294 bool eofmt = FALSE;
12296 char *tokenstart = s;
12299 if (PL_madskills) {
12300 savewhite = PL_thiswhite;
12305 while (!needargs) {
12308 #ifdef PERL_STRICT_CR
12309 while (SPACE_OR_TAB(*t))
12312 while (SPACE_OR_TAB(*t) || *t == '\r')
12315 if (*t == '\n' || t == PL_bufend) {
12320 if (PL_in_eval && !PL_rsfp) {
12321 eol = (char *) memchr(s,'\n',PL_bufend-s);
12326 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12328 for (t = s; t < eol; t++) {
12329 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12331 goto enough; /* ~~ must be first line in formline */
12333 if (*t == '@' || *t == '^')
12337 sv_catpvn(stuff, s, eol-s);
12338 #ifndef PERL_STRICT_CR
12339 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12340 char *end = SvPVX(stuff) + SvCUR(stuff);
12343 SvCUR_set(stuff, SvCUR(stuff) - 1);
12353 if (PL_madskills) {
12355 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12357 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12360 s = filter_gets(PL_linestr, PL_rsfp, 0);
12362 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12364 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12366 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12367 PL_last_lop = PL_last_uni = NULL;
12376 if (SvCUR(stuff)) {
12379 PL_lex_state = LEX_NORMAL;
12380 start_force(PL_curforce);
12381 NEXTVAL_NEXTTOKE.ival = 0;
12385 PL_lex_state = LEX_FORMLINE;
12387 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12389 else if (PL_encoding)
12390 sv_recode_to_utf8(stuff, PL_encoding);
12392 start_force(PL_curforce);
12393 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12395 start_force(PL_curforce);
12396 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12400 SvREFCNT_dec(stuff);
12402 PL_lex_formbrack = 0;
12406 if (PL_madskills) {
12408 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12410 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12411 PL_thiswhite = savewhite;
12418 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12421 const I32 oldsavestack_ix = PL_savestack_ix;
12422 CV* const outsidecv = PL_compcv;
12425 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12427 SAVEI32(PL_subline);
12428 save_item(PL_subname);
12429 SAVESPTR(PL_compcv);
12431 PL_compcv = (CV*)newSV_type(is_format ? SVt_PVFM : SVt_PVCV);
12432 CvFLAGS(PL_compcv) |= flags;
12434 PL_subline = CopLINE(PL_curcop);
12435 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12436 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12437 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12439 return oldsavestack_ix;
12443 #pragma segment Perl_yylex
12446 Perl_yywarn(pTHX_ const char *s)
12449 PL_in_eval |= EVAL_WARNONLY;
12451 PL_in_eval &= ~EVAL_WARNONLY;
12456 Perl_yyerror(pTHX_ const char *s)
12459 const char *where = NULL;
12460 const char *context = NULL;
12463 int yychar = PL_parser->yychar;
12465 if (!yychar || (yychar == ';' && !PL_rsfp))
12467 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12468 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12469 PL_oldbufptr != PL_bufptr) {
12472 The code below is removed for NetWare because it abends/crashes on NetWare
12473 when the script has error such as not having the closing quotes like:
12474 if ($var eq "value)
12475 Checking of white spaces is anyway done in NetWare code.
12478 while (isSPACE(*PL_oldoldbufptr))
12481 context = PL_oldoldbufptr;
12482 contlen = PL_bufptr - PL_oldoldbufptr;
12484 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12485 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12488 The code below is removed for NetWare because it abends/crashes on NetWare
12489 when the script has error such as not having the closing quotes like:
12490 if ($var eq "value)
12491 Checking of white spaces is anyway done in NetWare code.
12494 while (isSPACE(*PL_oldbufptr))
12497 context = PL_oldbufptr;
12498 contlen = PL_bufptr - PL_oldbufptr;
12500 else if (yychar > 255)
12501 where = "next token ???";
12502 else if (yychar == -2) { /* YYEMPTY */
12503 if (PL_lex_state == LEX_NORMAL ||
12504 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12505 where = "at end of line";
12506 else if (PL_lex_inpat)
12507 where = "within pattern";
12509 where = "within string";
12512 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12514 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12515 else if (isPRINT_LC(yychar)) {
12516 const char string = yychar;
12517 sv_catpvn(where_sv, &string, 1);
12520 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12521 where = SvPVX_const(where_sv);
12523 msg = sv_2mortal(newSVpv(s, 0));
12524 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12525 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12527 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12529 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12530 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12531 Perl_sv_catpvf(aTHX_ msg,
12532 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12533 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12536 if (PL_in_eval & EVAL_WARNONLY) {
12537 if (ckWARN_d(WARN_SYNTAX))
12538 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12542 if (PL_error_count >= 10) {
12543 if (PL_in_eval && SvCUR(ERRSV))
12544 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12545 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12547 Perl_croak(aTHX_ "%s has too many errors.\n",
12548 OutCopFILE(PL_curcop));
12551 PL_in_my_stash = NULL;
12555 #pragma segment Main
12559 S_swallow_bom(pTHX_ U8 *s)
12562 const STRLEN slen = SvCUR(PL_linestr);
12565 if (s[1] == 0xFE) {
12566 /* UTF-16 little-endian? (or UTF32-LE?) */
12567 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12568 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12569 #ifndef PERL_NO_UTF16_FILTER
12570 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12573 if (PL_bufend > (char*)s) {
12577 filter_add(utf16rev_textfilter, NULL);
12578 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12579 utf16_to_utf8_reversed(s, news,
12580 PL_bufend - (char*)s - 1,
12582 sv_setpvn(PL_linestr, (const char*)news, newlen);
12584 s = (U8*)SvPVX(PL_linestr);
12585 Copy(news, s, newlen, U8);
12589 SvUTF8_on(PL_linestr);
12590 s = (U8*)SvPVX(PL_linestr);
12592 /* FIXME - is this a general bug fix? */
12595 PL_bufend = SvPVX(PL_linestr) + newlen;
12598 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12603 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12604 #ifndef PERL_NO_UTF16_FILTER
12605 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12608 if (PL_bufend > (char *)s) {
12612 filter_add(utf16_textfilter, NULL);
12613 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12614 utf16_to_utf8(s, news,
12615 PL_bufend - (char*)s,
12617 sv_setpvn(PL_linestr, (const char*)news, newlen);
12619 SvUTF8_on(PL_linestr);
12620 s = (U8*)SvPVX(PL_linestr);
12621 PL_bufend = SvPVX(PL_linestr) + newlen;
12624 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12629 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12630 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12631 s += 3; /* UTF-8 */
12637 if (s[2] == 0xFE && s[3] == 0xFF) {
12638 /* UTF-32 big-endian */
12639 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12642 else if (s[2] == 0 && s[3] != 0) {
12645 * are a good indicator of UTF-16BE. */
12646 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12652 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12653 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12654 s += 4; /* UTF-8 */
12660 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12663 * are a good indicator of UTF-16LE. */
12664 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12672 #ifndef PERL_NO_UTF16_FILTER
12674 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12677 const STRLEN old = SvCUR(sv);
12678 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12679 DEBUG_P(PerlIO_printf(Perl_debug_log,
12680 "utf16_textfilter(%p): %d %d (%d)\n",
12681 FPTR2DPTR(void *, utf16_textfilter),
12682 idx, maxlen, (int) count));
12686 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12687 Copy(SvPVX_const(sv), tmps, old, char);
12688 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12689 SvCUR(sv) - old, &newlen);
12690 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12692 DEBUG_P({sv_dump(sv);});
12697 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12700 const STRLEN old = SvCUR(sv);
12701 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12702 DEBUG_P(PerlIO_printf(Perl_debug_log,
12703 "utf16rev_textfilter(%p): %d %d (%d)\n",
12704 FPTR2DPTR(void *, utf16rev_textfilter),
12705 idx, maxlen, (int) count));
12709 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12710 Copy(SvPVX_const(sv), tmps, old, char);
12711 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12712 SvCUR(sv) - old, &newlen);
12713 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12715 DEBUG_P({ sv_dump(sv); });
12721 Returns a pointer to the next character after the parsed
12722 vstring, as well as updating the passed in sv.
12724 Function must be called like
12727 s = scan_vstring(s,e,sv);
12729 where s and e are the start and end of the string.
12730 The sv should already be large enough to store the vstring
12731 passed in, for performance reasons.
12736 Perl_scan_vstring(pTHX_ const char *s, const char *e, SV *sv)
12739 const char *pos = s;
12740 const char *start = s;
12741 if (*pos == 'v') pos++; /* get past 'v' */
12742 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12744 if ( *pos != '.') {
12745 /* this may not be a v-string if followed by => */
12746 const char *next = pos;
12747 while (next < e && isSPACE(*next))
12749 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12750 /* return string not v-string */
12751 sv_setpvn(sv,(char *)s,pos-s);
12752 return (char *)pos;
12756 if (!isALPHA(*pos)) {
12757 U8 tmpbuf[UTF8_MAXBYTES+1];
12760 s++; /* get past 'v' */
12762 sv_setpvn(sv, "", 0);
12765 /* this is atoi() that tolerates underscores */
12768 const char *end = pos;
12770 while (--end >= s) {
12772 const UV orev = rev;
12773 rev += (*end - '0') * mult;
12775 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12776 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12777 "Integer overflow in decimal number");
12781 if (rev > 0x7FFFFFFF)
12782 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12784 /* Append native character for the rev point */
12785 tmpend = uvchr_to_utf8(tmpbuf, rev);
12786 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12787 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12789 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12795 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12799 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12807 * c-indentation-style: bsd
12808 * c-basic-offset: 4
12809 * indent-tabs-mode: t
12812 * ex: set ts=8 sts=4 sw=4 noet: