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, &pl_yylval)
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, const YYSTYPE* lvalp)
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)lvalp->ival);
408 case TOKENTYPE_OPNUM:
409 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
410 PL_op_name[lvalp->ival]);
413 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
415 case TOKENTYPE_OPVAL:
417 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
418 PL_op_name[lvalp->opval->op_type]);
419 if (lvalp->opval->op_type == OP_CONST) {
420 Perl_sv_catpvf(aTHX_ report, " %s",
421 SvPEEK(cSVOPx_sv(lvalp->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 *const fmt, const char *const 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 *const 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 *const 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 *const 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 *const 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->error_count = oparser ? oparser->error_count : 0;
696 parser->copline = NOLINE;
697 parser->lex_state = LEX_NORMAL;
698 parser->expect = XSTATE;
700 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
701 : (AV*)SvREFCNT_inc(oparser->rsfp_filters);
703 Newx(parser->lex_brackstack, 120, char);
704 Newx(parser->lex_casestack, 12, char);
705 *parser->lex_casestack = '\0';
708 s = SvPV_const(line, len);
714 parser->linestr = newSVpvs("\n;");
715 } else if (SvREADONLY(line) || s[len-1] != ';') {
716 parser->linestr = newSVsv(line);
718 sv_catpvs(parser->linestr, "\n;");
721 SvREFCNT_inc_simple_void_NN(line);
722 parser->linestr = line;
724 parser->oldoldbufptr =
727 parser->linestart = SvPVX(parser->linestr);
728 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
729 parser->last_lop = parser->last_uni = NULL;
733 /* delete a parser object */
736 Perl_parser_free(pTHX_ const yy_parser *parser)
738 PL_curcop = parser->saved_curcop;
739 SvREFCNT_dec(parser->linestr);
741 if (parser->rsfp == PerlIO_stdin())
742 PerlIO_clearerr(parser->rsfp);
743 else if (parser->rsfp && parser->old_parser
744 && parser->rsfp != parser->old_parser->rsfp)
745 PerlIO_close(parser->rsfp);
746 SvREFCNT_dec(parser->rsfp_filters);
748 Safefree(parser->stack);
749 Safefree(parser->lex_brackstack);
750 Safefree(parser->lex_casestack);
751 PL_parser = parser->old_parser;
758 * Finalizer for lexing operations. Must be called when the parser is
759 * done with the lexer.
766 PL_doextract = FALSE;
771 * This subroutine has nothing to do with tilting, whether at windmills
772 * or pinball tables. Its name is short for "increment line". It
773 * increments the current line number in CopLINE(PL_curcop) and checks
774 * to see whether the line starts with a comment of the form
775 * # line 500 "foo.pm"
776 * If so, it sets the current line number and file to the values in the comment.
780 S_incline(pTHX_ const char *s)
787 CopLINE_inc(PL_curcop);
790 while (SPACE_OR_TAB(*s))
792 if (strnEQ(s, "line", 4))
796 if (SPACE_OR_TAB(*s))
800 while (SPACE_OR_TAB(*s))
808 while (SPACE_OR_TAB(*s))
810 if (*s == '"' && (t = strchr(s+1, '"'))) {
820 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
822 if (*e != '\n' && *e != '\0')
823 return; /* false alarm */
826 const STRLEN len = t - s;
828 SV *const temp_sv = CopFILESV(PL_curcop);
834 tmplen = SvCUR(temp_sv);
840 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
841 /* must copy *{"::_<(eval N)[oldfilename:L]"}
842 * to *{"::_<newfilename"} */
843 /* However, the long form of evals is only turned on by the
844 debugger - usually they're "(eval %lu)" */
848 STRLEN tmplen2 = len;
849 if (tmplen + 2 <= sizeof smallbuf)
852 Newx(tmpbuf, tmplen + 2, char);
855 memcpy(tmpbuf + 2, cf, tmplen);
857 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
862 if (tmplen2 + 2 <= sizeof smallbuf)
865 Newx(tmpbuf2, tmplen2 + 2, char);
867 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
868 /* Either they malloc'd it, or we malloc'd it,
869 so no prefix is present in ours. */
874 memcpy(tmpbuf2 + 2, s, tmplen2);
877 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
879 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
880 /* adjust ${"::_<newfilename"} to store the new file name */
881 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
882 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
883 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
886 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
888 if (tmpbuf != smallbuf) Safefree(tmpbuf);
891 CopFILE_free(PL_curcop);
892 CopFILE_setn(PL_curcop, s, len);
894 CopLINE_set(PL_curcop, atoi(n)-1);
898 /* skip space before PL_thistoken */
901 S_skipspace0(pTHX_ register char *s)
908 PL_thiswhite = newSVpvs("");
909 sv_catsv(PL_thiswhite, PL_skipwhite);
910 sv_free(PL_skipwhite);
913 PL_realtokenstart = s - SvPVX(PL_linestr);
917 /* skip space after PL_thistoken */
920 S_skipspace1(pTHX_ register char *s)
922 const char *start = s;
923 I32 startoff = start - SvPVX(PL_linestr);
928 start = SvPVX(PL_linestr) + startoff;
929 if (!PL_thistoken && PL_realtokenstart >= 0) {
930 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
931 PL_thistoken = newSVpvn(tstart, start - tstart);
933 PL_realtokenstart = -1;
936 PL_nextwhite = newSVpvs("");
937 sv_catsv(PL_nextwhite, PL_skipwhite);
938 sv_free(PL_skipwhite);
945 S_skipspace2(pTHX_ register char *s, SV **svp)
948 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
949 const I32 startoff = s - SvPVX(PL_linestr);
952 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
953 if (!PL_madskills || !svp)
955 start = SvPVX(PL_linestr) + startoff;
956 if (!PL_thistoken && PL_realtokenstart >= 0) {
957 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
958 PL_thistoken = newSVpvn(tstart, start - tstart);
959 PL_realtokenstart = -1;
964 sv_setsv(*svp, PL_skipwhite);
965 sv_free(PL_skipwhite);
974 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
976 AV *av = CopFILEAVx(PL_curcop);
978 SV * const sv = newSV_type(SVt_PVMG);
980 sv_setsv(sv, orig_sv);
982 sv_setpvn(sv, buf, len);
985 av_store(av, (I32)CopLINE(PL_curcop), sv);
991 * Called to gobble the appropriate amount and type of whitespace.
992 * Skips comments as well.
996 S_skipspace(pTHX_ register char *s)
1001 int startoff = s - SvPVX(PL_linestr);
1004 sv_free(PL_skipwhite);
1009 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1010 while (s < PL_bufend && SPACE_OR_TAB(*s))
1020 SSize_t oldprevlen, oldoldprevlen;
1021 SSize_t oldloplen = 0, oldunilen = 0;
1022 while (s < PL_bufend && isSPACE(*s)) {
1023 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
1028 if (s < PL_bufend && *s == '#') {
1029 while (s < PL_bufend && *s != '\n')
1031 if (s < PL_bufend) {
1033 if (PL_in_eval && !PL_rsfp) {
1040 /* only continue to recharge the buffer if we're at the end
1041 * of the buffer, we're not reading from a source filter, and
1042 * we're in normal lexing mode
1044 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1045 PL_lex_state == LEX_FORMLINE)
1052 /* try to recharge the buffer */
1054 curoff = s - SvPVX(PL_linestr);
1057 if ((s = filter_gets(PL_linestr, PL_rsfp,
1058 (prevlen = SvCUR(PL_linestr)))) == NULL)
1061 if (PL_madskills && curoff != startoff) {
1063 PL_skipwhite = newSVpvs("");
1064 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1068 /* mustn't throw out old stuff yet if madpropping */
1069 SvCUR(PL_linestr) = curoff;
1070 s = SvPVX(PL_linestr) + curoff;
1072 if (curoff && s[-1] == '\n')
1076 /* end of file. Add on the -p or -n magic */
1077 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1080 sv_catpvs(PL_linestr,
1081 ";}continue{print or die qq(-p destination: $!\\n);}");
1083 sv_setpvs(PL_linestr,
1084 ";}continue{print or die qq(-p destination: $!\\n);}");
1086 PL_minus_n = PL_minus_p = 0;
1088 else if (PL_minus_n) {
1090 sv_catpvn(PL_linestr, ";}", 2);
1092 sv_setpvn(PL_linestr, ";}", 2);
1098 sv_catpvn(PL_linestr,";", 1);
1100 sv_setpvn(PL_linestr,";", 1);
1103 /* reset variables for next time we lex */
1104 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1110 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1111 PL_last_lop = PL_last_uni = NULL;
1113 /* Close the filehandle. Could be from
1114 * STDIN, or a regular file. If we were reading code from
1115 * STDIN (because the commandline held no -e or filename)
1116 * then we don't close it, we reset it so the code can
1117 * read from STDIN too.
1120 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1121 PerlIO_clearerr(PL_rsfp);
1123 (void)PerlIO_close(PL_rsfp);
1128 /* not at end of file, so we only read another line */
1129 /* make corresponding updates to old pointers, for yyerror() */
1130 oldprevlen = PL_oldbufptr - PL_bufend;
1131 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1133 oldunilen = PL_last_uni - PL_bufend;
1135 oldloplen = PL_last_lop - PL_bufend;
1136 PL_linestart = PL_bufptr = s + prevlen;
1137 PL_bufend = s + SvCUR(PL_linestr);
1139 PL_oldbufptr = s + oldprevlen;
1140 PL_oldoldbufptr = s + oldoldprevlen;
1142 PL_last_uni = s + oldunilen;
1144 PL_last_lop = s + oldloplen;
1147 /* debugger active and we're not compiling the debugger code,
1148 * so store the line into the debugger's array of lines
1150 if (PERLDB_LINE && PL_curstash != PL_debstash)
1151 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1158 PL_skipwhite = newSVpvs("");
1159 curoff = s - SvPVX(PL_linestr);
1160 if (curoff - startoff)
1161 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1170 * Check the unary operators to ensure there's no ambiguity in how they're
1171 * used. An ambiguous piece of code would be:
1173 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1174 * the +5 is its argument.
1184 if (PL_oldoldbufptr != PL_last_uni)
1186 while (isSPACE(*PL_last_uni))
1189 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1191 if ((t = strchr(s, '(')) && t < PL_bufptr)
1194 if (ckWARN_d(WARN_AMBIGUOUS)){
1195 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1196 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1197 (int)(s - PL_last_uni), PL_last_uni);
1202 * LOP : macro to build a list operator. Its behaviour has been replaced
1203 * with a subroutine, S_lop() for which LOP is just another name.
1206 #define LOP(f,x) return lop(f,x,s)
1210 * Build a list operator (or something that might be one). The rules:
1211 * - if we have a next token, then it's a list operator [why?]
1212 * - if the next thing is an opening paren, then it's a function
1213 * - else it's a list operator
1217 S_lop(pTHX_ I32 f, int x, char *s)
1224 PL_last_lop = PL_oldbufptr;
1225 PL_last_lop_op = (OPCODE)f;
1228 return REPORT(LSTOP);
1231 return REPORT(LSTOP);
1234 return REPORT(FUNC);
1237 return REPORT(FUNC);
1239 return REPORT(LSTOP);
1245 * Sets up for an eventual force_next(). start_force(0) basically does
1246 * an unshift, while start_force(-1) does a push. yylex removes items
1251 S_start_force(pTHX_ int where)
1255 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1256 where = PL_lasttoke;
1257 assert(PL_curforce < 0 || PL_curforce == where);
1258 if (PL_curforce != where) {
1259 for (i = PL_lasttoke; i > where; --i) {
1260 PL_nexttoke[i] = PL_nexttoke[i-1];
1264 if (PL_curforce < 0) /* in case of duplicate start_force() */
1265 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1266 PL_curforce = where;
1269 curmad('^', newSVpvs(""));
1270 CURMAD('_', PL_nextwhite);
1275 S_curmad(pTHX_ char slot, SV *sv)
1281 if (PL_curforce < 0)
1282 where = &PL_thismad;
1284 where = &PL_nexttoke[PL_curforce].next_mad;
1287 sv_setpvn(sv, "", 0);
1290 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1292 else if (PL_encoding) {
1293 sv_recode_to_utf8(sv, PL_encoding);
1298 /* keep a slot open for the head of the list? */
1299 if (slot != '_' && *where && (*where)->mad_key == '^') {
1300 (*where)->mad_key = slot;
1301 sv_free((SV*)((*where)->mad_val));
1302 (*where)->mad_val = (void*)sv;
1305 addmad(newMADsv(slot, sv), where, 0);
1308 # define start_force(where) NOOP
1309 # define curmad(slot, sv) NOOP
1314 * When the lexer realizes it knows the next token (for instance,
1315 * it is reordering tokens for the parser) then it can call S_force_next
1316 * to know what token to return the next time the lexer is called. Caller
1317 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1318 * and possibly PL_expect to ensure the lexer handles the token correctly.
1322 S_force_next(pTHX_ I32 type)
1327 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1328 tokereport(THING, &NEXTVAL_NEXTTOKE);
1332 if (PL_curforce < 0)
1333 start_force(PL_lasttoke);
1334 PL_nexttoke[PL_curforce].next_type = type;
1335 if (PL_lex_state != LEX_KNOWNEXT)
1336 PL_lex_defer = PL_lex_state;
1337 PL_lex_state = LEX_KNOWNEXT;
1338 PL_lex_expect = PL_expect;
1341 PL_nexttype[PL_nexttoke] = type;
1343 if (PL_lex_state != LEX_KNOWNEXT) {
1344 PL_lex_defer = PL_lex_state;
1345 PL_lex_expect = PL_expect;
1346 PL_lex_state = LEX_KNOWNEXT;
1352 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1355 SV * const sv = newSVpvn_utf8(start, len,
1357 && is_utf8_string((const U8*)start, len));
1363 * When the lexer knows the next thing is a word (for instance, it has
1364 * just seen -> and it knows that the next char is a word char, then
1365 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1369 * char *start : buffer position (must be within PL_linestr)
1370 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1371 * int check_keyword : if true, Perl checks to make sure the word isn't
1372 * a keyword (do this if the word is a label, e.g. goto FOO)
1373 * int allow_pack : if true, : characters will also be allowed (require,
1374 * use, etc. do this)
1375 * int allow_initial_tick : used by the "sub" lexer only.
1379 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1385 start = SKIPSPACE1(start);
1387 if (isIDFIRST_lazy_if(s,UTF) ||
1388 (allow_pack && *s == ':') ||
1389 (allow_initial_tick && *s == '\'') )
1391 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1392 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1394 start_force(PL_curforce);
1396 curmad('X', newSVpvn(start,s-start));
1397 if (token == METHOD) {
1402 PL_expect = XOPERATOR;
1406 curmad('g', newSVpvs( "forced" ));
1407 NEXTVAL_NEXTTOKE.opval
1408 = (OP*)newSVOP(OP_CONST,0,
1409 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1410 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1418 * Called when the lexer wants $foo *foo &foo etc, but the program
1419 * text only contains the "foo" portion. The first argument is a pointer
1420 * to the "foo", and the second argument is the type symbol to prefix.
1421 * Forces the next token to be a "WORD".
1422 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1426 S_force_ident(pTHX_ register const char *s, int kind)
1430 const STRLEN len = strlen(s);
1431 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1432 start_force(PL_curforce);
1433 NEXTVAL_NEXTTOKE.opval = o;
1436 o->op_private = OPpCONST_ENTERED;
1437 /* XXX see note in pp_entereval() for why we forgo typo
1438 warnings if the symbol must be introduced in an eval.
1440 gv_fetchpvn_flags(s, len,
1441 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1443 kind == '$' ? SVt_PV :
1444 kind == '@' ? SVt_PVAV :
1445 kind == '%' ? SVt_PVHV :
1453 Perl_str_to_version(pTHX_ SV *sv)
1458 const char *start = SvPV_const(sv,len);
1459 const char * const end = start + len;
1460 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1461 while (start < end) {
1465 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1470 retval += ((NV)n)/nshift;
1479 * Forces the next token to be a version number.
1480 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1481 * and if "guessing" is TRUE, then no new token is created (and the caller
1482 * must use an alternative parsing method).
1486 S_force_version(pTHX_ char *s, int guessing)
1492 I32 startoff = s - SvPVX(PL_linestr);
1501 while (isDIGIT(*d) || *d == '_' || *d == '.')
1505 start_force(PL_curforce);
1506 curmad('X', newSVpvn(s,d-s));
1509 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1511 s = scan_num(s, &pl_yylval);
1512 version = pl_yylval.opval;
1513 ver = cSVOPx(version)->op_sv;
1514 if (SvPOK(ver) && !SvNIOK(ver)) {
1515 SvUPGRADE(ver, SVt_PVNV);
1516 SvNV_set(ver, str_to_version(ver));
1517 SvNOK_on(ver); /* hint that it is a version */
1520 else if (guessing) {
1523 sv_free(PL_nextwhite); /* let next token collect whitespace */
1525 s = SvPVX(PL_linestr) + startoff;
1533 if (PL_madskills && !version) {
1534 sv_free(PL_nextwhite); /* let next token collect whitespace */
1536 s = SvPVX(PL_linestr) + startoff;
1539 /* NOTE: The parser sees the package name and the VERSION swapped */
1540 start_force(PL_curforce);
1541 NEXTVAL_NEXTTOKE.opval = version;
1549 * Tokenize a quoted string passed in as an SV. It finds the next
1550 * chunk, up to end of string or a backslash. It may make a new
1551 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1556 S_tokeq(pTHX_ SV *sv)
1560 register char *send;
1568 s = SvPV_force(sv, len);
1569 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1572 while (s < send && *s != '\\')
1577 if ( PL_hints & HINT_NEW_STRING ) {
1578 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
1582 if (s + 1 < send && (s[1] == '\\'))
1583 s++; /* all that, just for this */
1588 SvCUR_set(sv, d - SvPVX_const(sv));
1590 if ( PL_hints & HINT_NEW_STRING )
1591 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
1596 * Now come three functions related to double-quote context,
1597 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1598 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1599 * interact with PL_lex_state, and create fake ( ... ) argument lists
1600 * to handle functions and concatenation.
1601 * They assume that whoever calls them will be setting up a fake
1602 * join call, because each subthing puts a ',' after it. This lets
1605 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1607 * (I'm not sure whether the spurious commas at the end of lcfirst's
1608 * arguments and join's arguments are created or not).
1613 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1615 * Pattern matching will set PL_lex_op to the pattern-matching op to
1616 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
1618 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1620 * Everything else becomes a FUNC.
1622 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1623 * had an OP_CONST or OP_READLINE). This just sets us up for a
1624 * call to S_sublex_push().
1628 S_sublex_start(pTHX)
1631 register const I32 op_type = pl_yylval.ival;
1633 if (op_type == OP_NULL) {
1634 pl_yylval.opval = PL_lex_op;
1638 if (op_type == OP_CONST || op_type == OP_READLINE) {
1639 SV *sv = tokeq(PL_lex_stuff);
1641 if (SvTYPE(sv) == SVt_PVIV) {
1642 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1644 const char * const p = SvPV_const(sv, len);
1645 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
1649 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1650 PL_lex_stuff = NULL;
1651 /* Allow <FH> // "foo" */
1652 if (op_type == OP_READLINE)
1653 PL_expect = XTERMORDORDOR;
1656 else if (op_type == OP_BACKTICK && PL_lex_op) {
1657 /* readpipe() vas overriden */
1658 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1659 pl_yylval.opval = PL_lex_op;
1661 PL_lex_stuff = NULL;
1665 PL_sublex_info.super_state = PL_lex_state;
1666 PL_sublex_info.sub_inwhat = (U16)op_type;
1667 PL_sublex_info.sub_op = PL_lex_op;
1668 PL_lex_state = LEX_INTERPPUSH;
1672 pl_yylval.opval = PL_lex_op;
1682 * Create a new scope to save the lexing state. The scope will be
1683 * ended in S_sublex_done. Returns a '(', starting the function arguments
1684 * to the uc, lc, etc. found before.
1685 * Sets PL_lex_state to LEX_INTERPCONCAT.
1694 PL_lex_state = PL_sublex_info.super_state;
1695 SAVEBOOL(PL_lex_dojoin);
1696 SAVEI32(PL_lex_brackets);
1697 SAVEI32(PL_lex_casemods);
1698 SAVEI32(PL_lex_starts);
1699 SAVEI8(PL_lex_state);
1700 SAVEVPTR(PL_lex_inpat);
1701 SAVEI16(PL_lex_inwhat);
1702 SAVECOPLINE(PL_curcop);
1703 SAVEPPTR(PL_bufptr);
1704 SAVEPPTR(PL_bufend);
1705 SAVEPPTR(PL_oldbufptr);
1706 SAVEPPTR(PL_oldoldbufptr);
1707 SAVEPPTR(PL_last_lop);
1708 SAVEPPTR(PL_last_uni);
1709 SAVEPPTR(PL_linestart);
1710 SAVESPTR(PL_linestr);
1711 SAVEGENERICPV(PL_lex_brackstack);
1712 SAVEGENERICPV(PL_lex_casestack);
1714 PL_linestr = PL_lex_stuff;
1715 PL_lex_stuff = NULL;
1717 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1718 = SvPVX(PL_linestr);
1719 PL_bufend += SvCUR(PL_linestr);
1720 PL_last_lop = PL_last_uni = NULL;
1721 SAVEFREESV(PL_linestr);
1723 PL_lex_dojoin = FALSE;
1724 PL_lex_brackets = 0;
1725 Newx(PL_lex_brackstack, 120, char);
1726 Newx(PL_lex_casestack, 12, char);
1727 PL_lex_casemods = 0;
1728 *PL_lex_casestack = '\0';
1730 PL_lex_state = LEX_INTERPCONCAT;
1731 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1733 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1734 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1735 PL_lex_inpat = PL_sublex_info.sub_op;
1737 PL_lex_inpat = NULL;
1744 * Restores lexer state after a S_sublex_push.
1751 if (!PL_lex_starts++) {
1752 SV * const sv = newSVpvs("");
1753 if (SvUTF8(PL_linestr))
1755 PL_expect = XOPERATOR;
1756 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1760 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1761 PL_lex_state = LEX_INTERPCASEMOD;
1765 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1766 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1767 PL_linestr = PL_lex_repl;
1769 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1770 PL_bufend += SvCUR(PL_linestr);
1771 PL_last_lop = PL_last_uni = NULL;
1772 SAVEFREESV(PL_linestr);
1773 PL_lex_dojoin = FALSE;
1774 PL_lex_brackets = 0;
1775 PL_lex_casemods = 0;
1776 *PL_lex_casestack = '\0';
1778 if (SvEVALED(PL_lex_repl)) {
1779 PL_lex_state = LEX_INTERPNORMAL;
1781 /* we don't clear PL_lex_repl here, so that we can check later
1782 whether this is an evalled subst; that means we rely on the
1783 logic to ensure sublex_done() is called again only via the
1784 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1787 PL_lex_state = LEX_INTERPCONCAT;
1797 PL_endwhite = newSVpvs("");
1798 sv_catsv(PL_endwhite, PL_thiswhite);
1802 sv_setpvn(PL_thistoken,"",0);
1804 PL_realtokenstart = -1;
1808 PL_bufend = SvPVX(PL_linestr);
1809 PL_bufend += SvCUR(PL_linestr);
1810 PL_expect = XOPERATOR;
1811 PL_sublex_info.sub_inwhat = 0;
1819 Extracts a pattern, double-quoted string, or transliteration. This
1822 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1823 processing a pattern (PL_lex_inpat is true), a transliteration
1824 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1826 Returns a pointer to the character scanned up to. If this is
1827 advanced from the start pointer supplied (i.e. if anything was
1828 successfully parsed), will leave an OP for the substring scanned
1829 in pl_yylval. Caller must intuit reason for not parsing further
1830 by looking at the next characters herself.
1834 double-quoted style: \r and \n
1835 regexp special ones: \D \s
1838 case and quoting: \U \Q \E
1839 stops on @ and $, but not for $ as tail anchor
1841 In transliterations:
1842 characters are VERY literal, except for - not at the start or end
1843 of the string, which indicates a range. If the range is in bytes,
1844 scan_const expands the range to the full set of intermediate
1845 characters. If the range is in utf8, the hyphen is replaced with
1846 a certain range mark which will be handled by pmtrans() in op.c.
1848 In double-quoted strings:
1850 double-quoted style: \r and \n
1852 deprecated backrefs: \1 (in substitution replacements)
1853 case and quoting: \U \Q \E
1856 scan_const does *not* construct ops to handle interpolated strings.
1857 It stops processing as soon as it finds an embedded $ or @ variable
1858 and leaves it to the caller to work out what's going on.
1860 embedded arrays (whether in pattern or not) could be:
1861 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1863 $ in double-quoted strings must be the symbol of an embedded scalar.
1865 $ in pattern could be $foo or could be tail anchor. Assumption:
1866 it's a tail anchor if $ is the last thing in the string, or if it's
1867 followed by one of "()| \r\n\t"
1869 \1 (backreferences) are turned into $1
1871 The structure of the code is
1872 while (there's a character to process) {
1873 handle transliteration ranges
1874 skip regexp comments /(?#comment)/ and codes /(?{code})/
1875 skip #-initiated comments in //x patterns
1876 check for embedded arrays
1877 check for embedded scalars
1879 leave intact backslashes from leaveit (below)
1880 deprecate \1 in substitution replacements
1881 handle string-changing backslashes \l \U \Q \E, etc.
1882 switch (what was escaped) {
1883 handle \- in a transliteration (becomes a literal -)
1884 handle \132 (octal characters)
1885 handle \x15 and \x{1234} (hex characters)
1886 handle \N{name} (named characters)
1887 handle \cV (control characters)
1888 handle printf-style backslashes (\f, \r, \n, etc)
1890 } (end if backslash)
1891 } (end while character to read)
1896 S_scan_const(pTHX_ char *start)
1899 register char *send = PL_bufend; /* end of the constant */
1900 SV *sv = newSV(send - start); /* sv for the constant */
1901 register char *s = start; /* start of the constant */
1902 register char *d = SvPVX(sv); /* destination for copies */
1903 bool dorange = FALSE; /* are we in a translit range? */
1904 bool didrange = FALSE; /* did we just finish a range? */
1905 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1906 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1909 UV literal_endpoint = 0;
1910 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1913 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1914 /* If we are doing a trans and we know we want UTF8 set expectation */
1915 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1916 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1920 while (s < send || dorange) {
1921 /* get transliterations out of the way (they're most literal) */
1922 if (PL_lex_inwhat == OP_TRANS) {
1923 /* expand a range A-Z to the full set of characters. AIE! */
1925 I32 i; /* current expanded character */
1926 I32 min; /* first character in range */
1927 I32 max; /* last character in range */
1938 char * const c = (char*)utf8_hop((U8*)d, -1);
1942 *c = (char)UTF_TO_NATIVE(0xff);
1943 /* mark the range as done, and continue */
1949 i = d - SvPVX_const(sv); /* remember current offset */
1952 SvLEN(sv) + (has_utf8 ?
1953 (512 - UTF_CONTINUATION_MARK +
1956 /* How many two-byte within 0..255: 128 in UTF-8,
1957 * 96 in UTF-8-mod. */
1959 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1961 d = SvPVX(sv) + i; /* refresh d after realloc */
1965 for (j = 0; j <= 1; j++) {
1966 char * const c = (char*)utf8_hop((U8*)d, -1);
1967 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1973 max = (U8)0xff; /* only to \xff */
1974 uvmax = uv; /* \x{100} to uvmax */
1976 d = c; /* eat endpoint chars */
1981 d -= 2; /* eat the first char and the - */
1982 min = (U8)*d; /* first char in range */
1983 max = (U8)d[1]; /* last char in range */
1990 "Invalid range \"%c-%c\" in transliteration operator",
1991 (char)min, (char)max);
1995 if (literal_endpoint == 2 &&
1996 ((isLOWER(min) && isLOWER(max)) ||
1997 (isUPPER(min) && isUPPER(max)))) {
1999 for (i = min; i <= max; i++)
2001 *d++ = NATIVE_TO_NEED(has_utf8,i);
2003 for (i = min; i <= max; i++)
2005 *d++ = NATIVE_TO_NEED(has_utf8,i);
2010 for (i = min; i <= max; i++)
2013 const U8 ch = (U8)NATIVE_TO_UTF(i);
2014 if (UNI_IS_INVARIANT(ch))
2017 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2018 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2027 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2029 *d++ = (char)UTF_TO_NATIVE(0xff);
2031 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2035 /* mark the range as done, and continue */
2039 literal_endpoint = 0;
2044 /* range begins (ignore - as first or last char) */
2045 else if (*s == '-' && s+1 < send && s != start) {
2047 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2054 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2064 literal_endpoint = 0;
2065 native_range = TRUE;
2070 /* if we get here, we're not doing a transliteration */
2072 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2073 except for the last char, which will be done separately. */
2074 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2076 while (s+1 < send && *s != ')')
2077 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2079 else if (s[2] == '{' /* This should match regcomp.c */
2080 || (s[2] == '?' && s[3] == '{'))
2083 char *regparse = s + (s[2] == '{' ? 3 : 4);
2086 while (count && (c = *regparse)) {
2087 if (c == '\\' && regparse[1])
2095 if (*regparse != ')')
2096 regparse--; /* Leave one char for continuation. */
2097 while (s < regparse)
2098 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2102 /* likewise skip #-initiated comments in //x patterns */
2103 else if (*s == '#' && PL_lex_inpat &&
2104 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2105 while (s+1 < send && *s != '\n')
2106 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2109 /* check for embedded arrays
2110 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2112 else if (*s == '@' && s[1]) {
2113 if (isALNUM_lazy_if(s+1,UTF))
2115 if (strchr(":'{$", s[1]))
2117 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2118 break; /* in regexp, neither @+ nor @- are interpolated */
2121 /* check for embedded scalars. only stop if we're sure it's a
2124 else if (*s == '$') {
2125 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2127 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2128 break; /* in regexp, $ might be tail anchor */
2131 /* End of else if chain - OP_TRANS rejoin rest */
2134 if (*s == '\\' && s+1 < send) {
2137 /* deprecate \1 in strings and substitution replacements */
2138 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2139 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2141 if (ckWARN(WARN_SYNTAX))
2142 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2147 /* string-change backslash escapes */
2148 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2152 /* skip any other backslash escapes in a pattern */
2153 else if (PL_lex_inpat) {
2154 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2155 goto default_action;
2158 /* if we get here, it's either a quoted -, or a digit */
2161 /* quoted - in transliterations */
2163 if (PL_lex_inwhat == OP_TRANS) {
2170 if ((isALPHA(*s) || isDIGIT(*s)) &&
2172 Perl_warner(aTHX_ packWARN(WARN_MISC),
2173 "Unrecognized escape \\%c passed through",
2175 /* default action is to copy the quoted character */
2176 goto default_action;
2179 /* \132 indicates an octal constant */
2180 case '0': case '1': case '2': case '3':
2181 case '4': case '5': case '6': case '7':
2185 uv = grok_oct(s, &len, &flags, NULL);
2188 goto NUM_ESCAPE_INSERT;
2190 /* \x24 indicates a hex constant */
2194 char* const e = strchr(s, '}');
2195 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2196 PERL_SCAN_DISALLOW_PREFIX;
2201 yyerror("Missing right brace on \\x{}");
2205 uv = grok_hex(s, &len, &flags, NULL);
2211 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2212 uv = grok_hex(s, &len, &flags, NULL);
2218 /* Insert oct or hex escaped character.
2219 * There will always enough room in sv since such
2220 * escapes will be longer than any UTF-8 sequence
2221 * they can end up as. */
2223 /* We need to map to chars to ASCII before doing the tests
2226 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2227 if (!has_utf8 && uv > 255) {
2228 /* Might need to recode whatever we have
2229 * accumulated so far if it contains any
2232 * (Can't we keep track of that and avoid
2233 * this rescan? --jhi)
2237 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2238 if (!NATIVE_IS_INVARIANT(*c)) {
2243 const STRLEN offset = d - SvPVX_const(sv);
2245 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2249 while (src >= (const U8 *)SvPVX_const(sv)) {
2250 if (!NATIVE_IS_INVARIANT(*src)) {
2251 const U8 ch = NATIVE_TO_ASCII(*src);
2252 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2253 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2263 if (has_utf8 || uv > 255) {
2264 d = (char*)uvchr_to_utf8((U8*)d, uv);
2266 if (PL_lex_inwhat == OP_TRANS &&
2267 PL_sublex_info.sub_op) {
2268 PL_sublex_info.sub_op->op_private |=
2269 (PL_lex_repl ? OPpTRANS_FROM_UTF
2273 if (uv > 255 && !dorange)
2274 native_range = FALSE;
2286 /* \N{LATIN SMALL LETTER A} is a named character */
2290 char* e = strchr(s, '}');
2296 yyerror("Missing right brace on \\N{}");
2300 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2302 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2303 PERL_SCAN_DISALLOW_PREFIX;
2306 uv = grok_hex(s, &len, &flags, NULL);
2307 if ( e > s && len != (STRLEN)(e - s) ) {
2311 goto NUM_ESCAPE_INSERT;
2313 res = newSVpvn(s + 1, e - s - 1);
2314 res = new_constant( NULL, 0, "charnames",
2315 res, NULL, s - 2, e - s + 3 );
2317 sv_utf8_upgrade(res);
2318 str = SvPV_const(res,len);
2319 #ifdef EBCDIC_NEVER_MIND
2320 /* charnames uses pack U and that has been
2321 * recently changed to do the below uni->native
2322 * mapping, so this would be redundant (and wrong,
2323 * the code point would be doubly converted).
2324 * But leave this in just in case the pack U change
2325 * gets revoked, but the semantics is still
2326 * desireable for charnames. --jhi */
2328 UV uv = utf8_to_uvchr((const U8*)str, 0);
2331 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2333 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2334 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2335 str = SvPV_const(res, len);
2339 if (!has_utf8 && SvUTF8(res)) {
2340 const char * const ostart = SvPVX_const(sv);
2341 SvCUR_set(sv, d - ostart);
2344 sv_utf8_upgrade(sv);
2345 /* this just broke our allocation above... */
2346 SvGROW(sv, (STRLEN)(send - start));
2347 d = SvPVX(sv) + SvCUR(sv);
2350 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2351 const char * const odest = SvPVX_const(sv);
2353 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2354 d = SvPVX(sv) + (d - odest);
2358 native_range = FALSE; /* \N{} is guessed to be Unicode */
2360 Copy(str, d, len, char);
2367 yyerror("Missing braces on \\N{}");
2370 /* \c is a control character */
2379 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2382 yyerror("Missing control char name in \\c");
2386 /* printf-style backslashes, formfeeds, newlines, etc */
2388 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2391 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2394 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2397 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2400 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2403 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2406 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2412 } /* end if (backslash) */
2419 /* If we started with encoded form, or already know we want it
2420 and then encode the next character */
2421 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2423 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2424 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2427 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2428 const STRLEN off = d - SvPVX_const(sv);
2429 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2431 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2434 if (uv > 255 && !dorange)
2435 native_range = FALSE;
2439 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2441 } /* while loop to process each character */
2443 /* terminate the string and set up the sv */
2445 SvCUR_set(sv, d - SvPVX_const(sv));
2446 if (SvCUR(sv) >= SvLEN(sv))
2447 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2450 if (PL_encoding && !has_utf8) {
2451 sv_recode_to_utf8(sv, PL_encoding);
2457 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2458 PL_sublex_info.sub_op->op_private |=
2459 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2463 /* shrink the sv if we allocated more than we used */
2464 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2465 SvPV_shrink_to_cur(sv);
2468 /* return the substring (via pl_yylval) only if we parsed anything */
2469 if (s > PL_bufptr) {
2470 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
2471 const char *const key = PL_lex_inpat ? "qr" : "q";
2472 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
2476 if (PL_lex_inwhat == OP_TRANS) {
2479 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
2487 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
2490 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2497 * Returns TRUE if there's more to the expression (e.g., a subscript),
2500 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2502 * ->[ and ->{ return TRUE
2503 * { and [ outside a pattern are always subscripts, so return TRUE
2504 * if we're outside a pattern and it's not { or [, then return FALSE
2505 * if we're in a pattern and the first char is a {
2506 * {4,5} (any digits around the comma) returns FALSE
2507 * if we're in a pattern and the first char is a [
2509 * [SOMETHING] has a funky algorithm to decide whether it's a
2510 * character class or not. It has to deal with things like
2511 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2512 * anything else returns TRUE
2515 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2518 S_intuit_more(pTHX_ register char *s)
2521 if (PL_lex_brackets)
2523 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2525 if (*s != '{' && *s != '[')
2530 /* In a pattern, so maybe we have {n,m}. */
2547 /* On the other hand, maybe we have a character class */
2550 if (*s == ']' || *s == '^')
2553 /* this is terrifying, and it works */
2554 int weight = 2; /* let's weigh the evidence */
2556 unsigned char un_char = 255, last_un_char;
2557 const char * const send = strchr(s,']');
2558 char tmpbuf[sizeof PL_tokenbuf * 4];
2560 if (!send) /* has to be an expression */
2563 Zero(seen,256,char);
2566 else if (isDIGIT(*s)) {
2568 if (isDIGIT(s[1]) && s[2] == ']')
2574 for (; s < send; s++) {
2575 last_un_char = un_char;
2576 un_char = (unsigned char)*s;
2581 weight -= seen[un_char] * 10;
2582 if (isALNUM_lazy_if(s+1,UTF)) {
2584 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2585 len = (int)strlen(tmpbuf);
2586 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2591 else if (*s == '$' && s[1] &&
2592 strchr("[#!%*<>()-=",s[1])) {
2593 if (/*{*/ strchr("])} =",s[2]))
2602 if (strchr("wds]",s[1]))
2604 else if (seen[(U8)'\''] || seen[(U8)'"'])
2606 else if (strchr("rnftbxcav",s[1]))
2608 else if (isDIGIT(s[1])) {
2610 while (s[1] && isDIGIT(s[1]))
2620 if (strchr("aA01! ",last_un_char))
2622 if (strchr("zZ79~",s[1]))
2624 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2625 weight -= 5; /* cope with negative subscript */
2628 if (!isALNUM(last_un_char)
2629 && !(last_un_char == '$' || last_un_char == '@'
2630 || last_un_char == '&')
2631 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2636 if (keyword(tmpbuf, d - tmpbuf, 0))
2639 if (un_char == last_un_char + 1)
2641 weight -= seen[un_char];
2646 if (weight >= 0) /* probably a character class */
2656 * Does all the checking to disambiguate
2658 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2659 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2661 * First argument is the stuff after the first token, e.g. "bar".
2663 * Not a method if bar is a filehandle.
2664 * Not a method if foo is a subroutine prototyped to take a filehandle.
2665 * Not a method if it's really "Foo $bar"
2666 * Method if it's "foo $bar"
2667 * Not a method if it's really "print foo $bar"
2668 * Method if it's really "foo package::" (interpreted as package->foo)
2669 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2670 * Not a method if bar is a filehandle or package, but is quoted with
2675 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2678 char *s = start + (*start == '$');
2679 char tmpbuf[sizeof PL_tokenbuf];
2687 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2691 const char *proto = SvPVX_const(cv);
2702 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2703 /* start is the beginning of the possible filehandle/object,
2704 * and s is the end of it
2705 * tmpbuf is a copy of it
2708 if (*start == '$') {
2709 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
2710 isUPPER(*PL_tokenbuf))
2713 len = start - SvPVX(PL_linestr);
2717 start = SvPVX(PL_linestr) + len;
2721 return *s == '(' ? FUNCMETH : METHOD;
2723 if (!keyword(tmpbuf, len, 0)) {
2724 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2728 soff = s - SvPVX(PL_linestr);
2732 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2733 if (indirgv && GvCVu(indirgv))
2735 /* filehandle or package name makes it a method */
2736 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2738 soff = s - SvPVX(PL_linestr);
2741 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2742 return 0; /* no assumptions -- "=>" quotes bearword */
2744 start_force(PL_curforce);
2745 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2746 newSVpvn(tmpbuf,len));
2747 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2749 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2754 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2756 return *s == '(' ? FUNCMETH : METHOD;
2762 /* Encoded script support. filter_add() effectively inserts a
2763 * 'pre-processing' function into the current source input stream.
2764 * Note that the filter function only applies to the current source file
2765 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2767 * The datasv parameter (which may be NULL) can be used to pass
2768 * private data to this instance of the filter. The filter function
2769 * can recover the SV using the FILTER_DATA macro and use it to
2770 * store private buffers and state information.
2772 * The supplied datasv parameter is upgraded to a PVIO type
2773 * and the IoDIRP/IoANY field is used to store the function pointer,
2774 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2775 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2776 * private use must be set using malloc'd pointers.
2780 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2789 if (!PL_rsfp_filters)
2790 PL_rsfp_filters = newAV();
2793 SvUPGRADE(datasv, SVt_PVIO);
2794 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2795 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2796 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2797 FPTR2DPTR(void *, IoANY(datasv)),
2798 SvPV_nolen(datasv)));
2799 av_unshift(PL_rsfp_filters, 1);
2800 av_store(PL_rsfp_filters, 0, datasv) ;
2805 /* Delete most recently added instance of this filter function. */
2807 Perl_filter_del(pTHX_ filter_t funcp)
2813 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2814 FPTR2DPTR(void*, funcp)));
2816 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2818 /* if filter is on top of stack (usual case) just pop it off */
2819 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2820 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2821 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2822 IoANY(datasv) = (void *)NULL;
2823 sv_free(av_pop(PL_rsfp_filters));
2827 /* we need to search for the correct entry and clear it */
2828 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2832 /* Invoke the idxth filter function for the current rsfp. */
2833 /* maxlen 0 = read one text line */
2835 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2840 /* This API is bad. It should have been using unsigned int for maxlen.
2841 Not sure if we want to change the API, but if not we should sanity
2842 check the value here. */
2843 const unsigned int correct_length
2852 if (!PL_parser || !PL_rsfp_filters)
2854 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2855 /* Provide a default input filter to make life easy. */
2856 /* Note that we append to the line. This is handy. */
2857 DEBUG_P(PerlIO_printf(Perl_debug_log,
2858 "filter_read %d: from rsfp\n", idx));
2859 if (correct_length) {
2862 const int old_len = SvCUR(buf_sv);
2864 /* ensure buf_sv is large enough */
2865 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2866 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2867 correct_length)) <= 0) {
2868 if (PerlIO_error(PL_rsfp))
2869 return -1; /* error */
2871 return 0 ; /* end of file */
2873 SvCUR_set(buf_sv, old_len + len) ;
2876 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2877 if (PerlIO_error(PL_rsfp))
2878 return -1; /* error */
2880 return 0 ; /* end of file */
2883 return SvCUR(buf_sv);
2885 /* Skip this filter slot if filter has been deleted */
2886 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2887 DEBUG_P(PerlIO_printf(Perl_debug_log,
2888 "filter_read %d: skipped (filter deleted)\n",
2890 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2892 /* Get function pointer hidden within datasv */
2893 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2894 DEBUG_P(PerlIO_printf(Perl_debug_log,
2895 "filter_read %d: via function %p (%s)\n",
2896 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2897 /* Call function. The function is expected to */
2898 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2899 /* Return: <0:error, =0:eof, >0:not eof */
2900 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2904 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2907 #ifdef PERL_CR_FILTER
2908 if (!PL_rsfp_filters) {
2909 filter_add(S_cr_textfilter,NULL);
2912 if (PL_rsfp_filters) {
2914 SvCUR_set(sv, 0); /* start with empty line */
2915 if (FILTER_READ(0, sv, 0) > 0)
2916 return ( SvPVX(sv) ) ;
2921 return (sv_gets(sv, fp, append));
2925 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
2930 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2934 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2935 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2937 return GvHV(gv); /* Foo:: */
2940 /* use constant CLASS => 'MyClass' */
2941 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2942 if (gv && GvCV(gv)) {
2943 SV * const sv = cv_const_sv(GvCV(gv));
2945 pkgname = SvPV_const(sv, len);
2948 return gv_stashpvn(pkgname, len, 0);
2952 * S_readpipe_override
2953 * Check whether readpipe() is overriden, and generates the appropriate
2954 * optree, provided sublex_start() is called afterwards.
2957 S_readpipe_override(pTHX)
2960 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2961 pl_yylval.ival = OP_BACKTICK;
2963 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2965 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2966 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
2967 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2969 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2970 append_elem(OP_LIST,
2971 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2972 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2979 * The intent of this yylex wrapper is to minimize the changes to the
2980 * tokener when we aren't interested in collecting madprops. It remains
2981 * to be seen how successful this strategy will be...
2988 char *s = PL_bufptr;
2990 /* make sure PL_thiswhite is initialized */
2994 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2995 if (PL_pending_ident)
2996 return S_pending_ident(aTHX);
2998 /* previous token ate up our whitespace? */
2999 if (!PL_lasttoke && PL_nextwhite) {
3000 PL_thiswhite = PL_nextwhite;
3004 /* isolate the token, and figure out where it is without whitespace */
3005 PL_realtokenstart = -1;
3009 assert(PL_curforce < 0);
3011 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3012 if (!PL_thistoken) {
3013 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3014 PL_thistoken = newSVpvs("");
3016 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3017 PL_thistoken = newSVpvn(tstart, s - tstart);
3020 if (PL_thismad) /* install head */
3021 CURMAD('X', PL_thistoken);
3024 /* last whitespace of a sublex? */
3025 if (optype == ')' && PL_endwhite) {
3026 CURMAD('X', PL_endwhite);
3031 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3032 if (!PL_thiswhite && !PL_endwhite && !optype) {
3033 sv_free(PL_thistoken);
3038 /* put off final whitespace till peg */
3039 if (optype == ';' && !PL_rsfp) {
3040 PL_nextwhite = PL_thiswhite;
3043 else if (PL_thisopen) {
3044 CURMAD('q', PL_thisopen);
3046 sv_free(PL_thistoken);
3050 /* Store actual token text as madprop X */
3051 CURMAD('X', PL_thistoken);
3055 /* add preceding whitespace as madprop _ */
3056 CURMAD('_', PL_thiswhite);
3060 /* add quoted material as madprop = */
3061 CURMAD('=', PL_thisstuff);
3065 /* add terminating quote as madprop Q */
3066 CURMAD('Q', PL_thisclose);
3070 /* special processing based on optype */
3074 /* opval doesn't need a TOKEN since it can already store mp */
3084 if (pl_yylval.opval)
3085 append_madprops(PL_thismad, pl_yylval.opval, 0);
3093 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3102 /* remember any fake bracket that lexer is about to discard */
3103 if (PL_lex_brackets == 1 &&
3104 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3107 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3110 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3111 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3114 break; /* don't bother looking for trailing comment */
3123 /* attach a trailing comment to its statement instead of next token */
3127 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3129 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3131 if (*s == '\n' || *s == '#') {
3132 while (s < PL_bufend && *s != '\n')
3136 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3137 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3154 /* Create new token struct. Note: opvals return early above. */
3155 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
3162 S_tokenize_use(pTHX_ int is_use, char *s) {
3164 if (PL_expect != XSTATE)
3165 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3166 is_use ? "use" : "no"));
3168 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3169 s = force_version(s, TRUE);
3170 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3171 start_force(PL_curforce);
3172 NEXTVAL_NEXTTOKE.opval = NULL;
3175 else if (*s == 'v') {
3176 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3177 s = force_version(s, FALSE);
3181 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3182 s = force_version(s, FALSE);
3184 pl_yylval.ival = is_use;
3188 static const char* const exp_name[] =
3189 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3190 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3197 Works out what to call the token just pulled out of the input
3198 stream. The yacc parser takes care of taking the ops we return and
3199 stitching them into a tree.
3205 if read an identifier
3206 if we're in a my declaration
3207 croak if they tried to say my($foo::bar)
3208 build the ops for a my() declaration
3209 if it's an access to a my() variable
3210 are we in a sort block?
3211 croak if my($a); $a <=> $b
3212 build ops for access to a my() variable
3213 if in a dq string, and they've said @foo and we can't find @foo
3215 build ops for a bareword
3216 if we already built the token before, use it.
3221 #pragma segment Perl_yylex
3227 register char *s = PL_bufptr;
3232 /* orig_keyword, gvp, and gv are initialized here because
3233 * jump to the label just_a_word_zero can bypass their
3234 * initialization later. */
3235 I32 orig_keyword = 0;
3240 SV* tmp = newSVpvs("");
3241 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3242 (IV)CopLINE(PL_curcop),
3243 lex_state_names[PL_lex_state],
3244 exp_name[PL_expect],
3245 pv_display(tmp, s, strlen(s), 0, 60));
3248 /* check if there's an identifier for us to look at */
3249 if (PL_pending_ident)
3250 return REPORT(S_pending_ident(aTHX));
3252 /* no identifier pending identification */
3254 switch (PL_lex_state) {
3256 case LEX_NORMAL: /* Some compilers will produce faster */
3257 case LEX_INTERPNORMAL: /* code if we comment these out. */
3261 /* when we've already built the next token, just pull it out of the queue */
3265 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
3267 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3268 PL_nexttoke[PL_lasttoke].next_mad = 0;
3269 if (PL_thismad && PL_thismad->mad_key == '_') {
3270 PL_thiswhite = (SV*)PL_thismad->mad_val;
3271 PL_thismad->mad_val = 0;
3272 mad_free(PL_thismad);
3277 PL_lex_state = PL_lex_defer;
3278 PL_expect = PL_lex_expect;
3279 PL_lex_defer = LEX_NORMAL;
3280 if (!PL_nexttoke[PL_lasttoke].next_type)
3285 pl_yylval = PL_nextval[PL_nexttoke];
3287 PL_lex_state = PL_lex_defer;
3288 PL_expect = PL_lex_expect;
3289 PL_lex_defer = LEX_NORMAL;
3293 /* FIXME - can these be merged? */
3294 return(PL_nexttoke[PL_lasttoke].next_type);
3296 return REPORT(PL_nexttype[PL_nexttoke]);
3299 /* interpolated case modifiers like \L \U, including \Q and \E.
3300 when we get here, PL_bufptr is at the \
3302 case LEX_INTERPCASEMOD:
3304 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3305 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3307 /* handle \E or end of string */
3308 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3310 if (PL_lex_casemods) {
3311 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3312 PL_lex_casestack[PL_lex_casemods] = '\0';
3314 if (PL_bufptr != PL_bufend
3315 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3317 PL_lex_state = LEX_INTERPCONCAT;
3320 PL_thistoken = newSVpvs("\\E");
3326 while (PL_bufptr != PL_bufend &&
3327 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3329 PL_thiswhite = newSVpvs("");
3330 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3334 if (PL_bufptr != PL_bufend)
3337 PL_lex_state = LEX_INTERPCONCAT;
3341 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3342 "### Saw case modifier\n"); });
3344 if (s[1] == '\\' && s[2] == 'E') {
3347 PL_thiswhite = newSVpvs("");
3348 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3351 PL_lex_state = LEX_INTERPCONCAT;
3356 if (!PL_madskills) /* when just compiling don't need correct */
3357 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3358 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3359 if ((*s == 'L' || *s == 'U') &&
3360 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3361 PL_lex_casestack[--PL_lex_casemods] = '\0';
3364 if (PL_lex_casemods > 10)
3365 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3366 PL_lex_casestack[PL_lex_casemods++] = *s;
3367 PL_lex_casestack[PL_lex_casemods] = '\0';
3368 PL_lex_state = LEX_INTERPCONCAT;
3369 start_force(PL_curforce);
3370 NEXTVAL_NEXTTOKE.ival = 0;
3372 start_force(PL_curforce);
3374 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3376 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3378 NEXTVAL_NEXTTOKE.ival = OP_LC;
3380 NEXTVAL_NEXTTOKE.ival = OP_UC;
3382 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3384 Perl_croak(aTHX_ "panic: yylex");
3386 SV* const tmpsv = newSVpvs("\\ ");
3387 /* replace the space with the character we want to escape
3389 SvPVX(tmpsv)[1] = *s;
3395 if (PL_lex_starts) {
3401 sv_free(PL_thistoken);
3402 PL_thistoken = newSVpvs("");
3405 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3406 if (PL_lex_casemods == 1 && PL_lex_inpat)
3415 case LEX_INTERPPUSH:
3416 return REPORT(sublex_push());
3418 case LEX_INTERPSTART:
3419 if (PL_bufptr == PL_bufend)
3420 return REPORT(sublex_done());
3421 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3422 "### Interpolated variable\n"); });
3424 PL_lex_dojoin = (*PL_bufptr == '@');
3425 PL_lex_state = LEX_INTERPNORMAL;
3426 if (PL_lex_dojoin) {
3427 start_force(PL_curforce);
3428 NEXTVAL_NEXTTOKE.ival = 0;
3430 start_force(PL_curforce);
3431 force_ident("\"", '$');
3432 start_force(PL_curforce);
3433 NEXTVAL_NEXTTOKE.ival = 0;
3435 start_force(PL_curforce);
3436 NEXTVAL_NEXTTOKE.ival = 0;
3438 start_force(PL_curforce);
3439 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3442 if (PL_lex_starts++) {
3447 sv_free(PL_thistoken);
3448 PL_thistoken = newSVpvs("");
3451 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3452 if (!PL_lex_casemods && PL_lex_inpat)
3459 case LEX_INTERPENDMAYBE:
3460 if (intuit_more(PL_bufptr)) {
3461 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3467 if (PL_lex_dojoin) {
3468 PL_lex_dojoin = FALSE;
3469 PL_lex_state = LEX_INTERPCONCAT;
3473 sv_free(PL_thistoken);
3474 PL_thistoken = newSVpvs("");
3479 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3480 && SvEVALED(PL_lex_repl))
3482 if (PL_bufptr != PL_bufend)
3483 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3487 case LEX_INTERPCONCAT:
3489 if (PL_lex_brackets)
3490 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3492 if (PL_bufptr == PL_bufend)
3493 return REPORT(sublex_done());
3495 if (SvIVX(PL_linestr) == '\'') {
3496 SV *sv = newSVsv(PL_linestr);
3499 else if ( PL_hints & HINT_NEW_RE )
3500 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
3501 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3505 s = scan_const(PL_bufptr);
3507 PL_lex_state = LEX_INTERPCASEMOD;
3509 PL_lex_state = LEX_INTERPSTART;
3512 if (s != PL_bufptr) {
3513 start_force(PL_curforce);
3515 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3517 NEXTVAL_NEXTTOKE = pl_yylval;
3520 if (PL_lex_starts++) {
3524 sv_free(PL_thistoken);
3525 PL_thistoken = newSVpvs("");
3528 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3529 if (!PL_lex_casemods && PL_lex_inpat)
3542 PL_lex_state = LEX_NORMAL;
3543 s = scan_formline(PL_bufptr);
3544 if (!PL_lex_formbrack)
3550 PL_oldoldbufptr = PL_oldbufptr;
3556 sv_free(PL_thistoken);
3559 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3563 if (isIDFIRST_lazy_if(s,UTF))
3565 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3566 Perl_croak(aTHX_ "Unrecognized character \\x%02X in column %d", *s & 255, (int) len + 1);
3569 goto fake_eof; /* emulate EOF on ^D or ^Z */
3578 if (PL_lex_brackets) {
3579 yyerror((const char *)
3581 ? "Format not terminated"
3582 : "Missing right curly or square bracket"));
3584 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3585 "### Tokener got EOF\n");
3589 if (s++ < PL_bufend)
3590 goto retry; /* ignore stray nulls */
3593 if (!PL_in_eval && !PL_preambled) {
3594 PL_preambled = TRUE;
3600 /* Generate a string of Perl code to load the debugger.
3601 * If PERL5DB is set, it will return the contents of that,
3602 * otherwise a compile-time require of perl5db.pl. */
3604 const char * const pdb = PerlEnv_getenv("PERL5DB");
3607 sv_setpv(PL_linestr, pdb);
3608 sv_catpvs(PL_linestr,";");
3610 SETERRNO(0,SS_NORMAL);
3611 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
3614 sv_setpvs(PL_linestr,"");
3615 if (PL_preambleav) {
3616 SV **svp = AvARRAY(PL_preambleav);
3617 SV **const end = svp + AvFILLp(PL_preambleav);
3619 sv_catsv(PL_linestr, *svp);
3621 sv_catpvs(PL_linestr, ";");
3623 sv_free((SV*)PL_preambleav);
3624 PL_preambleav = NULL;
3626 if (PL_minus_n || PL_minus_p) {
3627 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3629 sv_catpvs(PL_linestr,"chomp;");
3632 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3633 || *PL_splitstr == '"')
3634 && strchr(PL_splitstr + 1, *PL_splitstr))
3635 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3637 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3638 bytes can be used as quoting characters. :-) */
3639 const char *splits = PL_splitstr;
3640 sv_catpvs(PL_linestr, "our @F=split(q\0");
3643 if (*splits == '\\')
3644 sv_catpvn(PL_linestr, splits, 1);
3645 sv_catpvn(PL_linestr, splits, 1);
3646 } while (*splits++);
3647 /* This loop will embed the trailing NUL of
3648 PL_linestr as the last thing it does before
3650 sv_catpvs(PL_linestr, ");");
3654 sv_catpvs(PL_linestr,"our @F=split(' ');");
3658 sv_catpvs(PL_linestr,
3659 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
3660 sv_catpvs(PL_linestr, "\n");
3661 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3662 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3663 PL_last_lop = PL_last_uni = NULL;
3664 if (PERLDB_LINE && PL_curstash != PL_debstash)
3665 update_debugger_info(PL_linestr, NULL, 0);
3669 bof = PL_rsfp ? TRUE : FALSE;
3670 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3673 PL_realtokenstart = -1;
3676 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 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3725 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3726 s = swallow_bom((U8*)s);
3730 /* Incest with pod. */
3733 sv_catsv(PL_thiswhite, PL_linestr);
3735 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3736 sv_setpvn(PL_linestr, "", 0);
3737 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3738 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3739 PL_last_lop = PL_last_uni = NULL;
3740 PL_doextract = FALSE;
3744 } while (PL_doextract);
3745 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3746 if (PERLDB_LINE && PL_curstash != PL_debstash)
3747 update_debugger_info(PL_linestr, NULL, 0);
3748 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3749 PL_last_lop = PL_last_uni = NULL;
3750 if (CopLINE(PL_curcop) == 1) {
3751 while (s < PL_bufend && isSPACE(*s))
3753 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3757 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3761 if (*s == '#' && *(s+1) == '!')
3763 #ifdef ALTERNATE_SHEBANG
3765 static char const as[] = ALTERNATE_SHEBANG;
3766 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3767 d = s + (sizeof(as) - 1);
3769 #endif /* ALTERNATE_SHEBANG */
3778 while (*d && !isSPACE(*d))
3782 #ifdef ARG_ZERO_IS_SCRIPT
3783 if (ipathend > ipath) {
3785 * HP-UX (at least) sets argv[0] to the script name,
3786 * which makes $^X incorrect. And Digital UNIX and Linux,
3787 * at least, set argv[0] to the basename of the Perl
3788 * interpreter. So, having found "#!", we'll set it right.
3790 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3792 assert(SvPOK(x) || SvGMAGICAL(x));
3793 if (sv_eq(x, CopFILESV(PL_curcop))) {
3794 sv_setpvn(x, ipath, ipathend - ipath);
3800 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3801 const char * const lstart = SvPV_const(x,llen);
3803 bstart += blen - llen;
3804 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3805 sv_setpvn(x, ipath, ipathend - ipath);
3810 TAINT_NOT; /* $^X is always tainted, but that's OK */
3812 #endif /* ARG_ZERO_IS_SCRIPT */
3817 d = instr(s,"perl -");
3819 d = instr(s,"perl");
3821 /* avoid getting into infinite loops when shebang
3822 * line contains "Perl" rather than "perl" */
3824 for (d = ipathend-4; d >= ipath; --d) {
3825 if ((*d == 'p' || *d == 'P')
3826 && !ibcmp(d, "perl", 4))
3836 #ifdef ALTERNATE_SHEBANG
3838 * If the ALTERNATE_SHEBANG on this system starts with a
3839 * character that can be part of a Perl expression, then if
3840 * we see it but not "perl", we're probably looking at the
3841 * start of Perl code, not a request to hand off to some
3842 * other interpreter. Similarly, if "perl" is there, but
3843 * not in the first 'word' of the line, we assume the line
3844 * contains the start of the Perl program.
3846 if (d && *s != '#') {
3847 const char *c = ipath;
3848 while (*c && !strchr("; \t\r\n\f\v#", *c))
3851 d = NULL; /* "perl" not in first word; ignore */
3853 *s = '#'; /* Don't try to parse shebang line */
3855 #endif /* ALTERNATE_SHEBANG */
3856 #ifndef MACOS_TRADITIONAL
3861 !instr(s,"indir") &&
3862 instr(PL_origargv[0],"perl"))
3869 while (s < PL_bufend && isSPACE(*s))
3871 if (s < PL_bufend) {
3872 Newxz(newargv,PL_origargc+3,char*);
3874 while (s < PL_bufend && !isSPACE(*s))
3877 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3880 newargv = PL_origargv;
3883 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3885 Perl_croak(aTHX_ "Can't exec %s", ipath);
3889 while (*d && !isSPACE(*d))
3891 while (SPACE_OR_TAB(*d))
3895 const bool switches_done = PL_doswitches;
3896 const U32 oldpdb = PL_perldb;
3897 const bool oldn = PL_minus_n;
3898 const bool oldp = PL_minus_p;
3902 if (*d1 == 'M' || *d1 == 'm' || *d1 == 'C') {
3903 const char * const m = d1;
3904 while (*d1 && !isSPACE(*d1))
3906 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3909 d1 = moreswitches(d1);
3911 if (PL_doswitches && !switches_done) {
3912 int argc = PL_origargc;
3913 char **argv = PL_origargv;
3916 } while (argc && argv[0][0] == '-' && argv[0][1]);
3917 init_argv_symbols(argc,argv);
3919 if ((PERLDB_LINE && !oldpdb) ||
3920 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3921 /* if we have already added "LINE: while (<>) {",
3922 we must not do it again */
3924 sv_setpvn(PL_linestr, "", 0);
3925 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3926 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3927 PL_last_lop = PL_last_uni = NULL;
3928 PL_preambled = FALSE;
3930 (void)gv_fetchfile(PL_origfilename);
3937 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3939 PL_lex_state = LEX_FORMLINE;
3944 #ifdef PERL_STRICT_CR
3945 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3947 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3949 case ' ': case '\t': case '\f': case 013:
3950 #ifdef MACOS_TRADITIONAL
3954 PL_realtokenstart = -1;
3956 PL_thiswhite = newSVpvs("");
3957 sv_catpvn(PL_thiswhite, s, 1);
3964 PL_realtokenstart = -1;
3968 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3969 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3970 /* handle eval qq[#line 1 "foo"\n ...] */
3971 CopLINE_dec(PL_curcop);
3974 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3976 if (!PL_in_eval || PL_rsfp)
3981 while (d < PL_bufend && *d != '\n')
3985 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3986 Perl_croak(aTHX_ "panic: input overflow");
3989 PL_thiswhite = newSVpvn(s, d - s);
3994 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3996 PL_lex_state = LEX_FORMLINE;
4002 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4003 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4006 TOKEN(PEG); /* make sure any #! line is accessible */
4011 /* if (PL_madskills && PL_lex_formbrack) { */
4013 while (d < PL_bufend && *d != '\n')
4017 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4018 Perl_croak(aTHX_ "panic: input overflow");
4019 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4021 PL_thiswhite = newSVpvs("");
4022 if (CopLINE(PL_curcop) == 1) {
4023 sv_setpvn(PL_thiswhite, "", 0);
4026 sv_catpvn(PL_thiswhite, s, d - s);
4040 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4048 while (s < PL_bufend && SPACE_OR_TAB(*s))
4051 if (strnEQ(s,"=>",2)) {
4052 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4053 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4054 OPERATOR('-'); /* unary minus */
4056 PL_last_uni = PL_oldbufptr;
4058 case 'r': ftst = OP_FTEREAD; break;
4059 case 'w': ftst = OP_FTEWRITE; break;
4060 case 'x': ftst = OP_FTEEXEC; break;
4061 case 'o': ftst = OP_FTEOWNED; break;
4062 case 'R': ftst = OP_FTRREAD; break;
4063 case 'W': ftst = OP_FTRWRITE; break;
4064 case 'X': ftst = OP_FTREXEC; break;
4065 case 'O': ftst = OP_FTROWNED; break;
4066 case 'e': ftst = OP_FTIS; break;
4067 case 'z': ftst = OP_FTZERO; break;
4068 case 's': ftst = OP_FTSIZE; break;
4069 case 'f': ftst = OP_FTFILE; break;
4070 case 'd': ftst = OP_FTDIR; break;
4071 case 'l': ftst = OP_FTLINK; break;
4072 case 'p': ftst = OP_FTPIPE; break;
4073 case 'S': ftst = OP_FTSOCK; break;
4074 case 'u': ftst = OP_FTSUID; break;
4075 case 'g': ftst = OP_FTSGID; break;
4076 case 'k': ftst = OP_FTSVTX; break;
4077 case 'b': ftst = OP_FTBLK; break;
4078 case 'c': ftst = OP_FTCHR; break;
4079 case 't': ftst = OP_FTTTY; break;
4080 case 'T': ftst = OP_FTTEXT; break;
4081 case 'B': ftst = OP_FTBINARY; break;
4082 case 'M': case 'A': case 'C':
4083 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4085 case 'M': ftst = OP_FTMTIME; break;
4086 case 'A': ftst = OP_FTATIME; break;
4087 case 'C': ftst = OP_FTCTIME; break;
4095 PL_last_lop_op = (OPCODE)ftst;
4096 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4097 "### Saw file test %c\n", (int)tmp);
4102 /* Assume it was a minus followed by a one-letter named
4103 * subroutine call (or a -bareword), then. */
4104 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4105 "### '-%c' looked like a file test but was not\n",
4112 const char tmp = *s++;
4115 if (PL_expect == XOPERATOR)
4120 else if (*s == '>') {
4123 if (isIDFIRST_lazy_if(s,UTF)) {
4124 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4132 if (PL_expect == XOPERATOR)
4135 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4137 OPERATOR('-'); /* unary minus */
4143 const char tmp = *s++;
4146 if (PL_expect == XOPERATOR)
4151 if (PL_expect == XOPERATOR)
4154 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4161 if (PL_expect != XOPERATOR) {
4162 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4163 PL_expect = XOPERATOR;
4164 force_ident(PL_tokenbuf, '*');
4177 if (PL_expect == XOPERATOR) {
4181 PL_tokenbuf[0] = '%';
4182 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4183 sizeof PL_tokenbuf - 1, FALSE);
4184 if (!PL_tokenbuf[1]) {
4187 PL_pending_ident = '%';
4198 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4205 const char tmp = *s++;
4211 goto just_a_word_zero_gv;
4214 switch (PL_expect) {
4220 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4222 PL_bufptr = s; /* update in case we back off */
4228 PL_expect = XTERMBLOCK;
4231 stuffstart = s - SvPVX(PL_linestr) - 1;
4235 while (isIDFIRST_lazy_if(s,UTF)) {
4238 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4239 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4240 if (tmp < 0) tmp = -tmp;
4254 sv = newSVpvn(s, len);
4256 d = scan_str(d,TRUE,TRUE);
4258 /* MUST advance bufptr here to avoid bogus
4259 "at end of line" context messages from yyerror().
4261 PL_bufptr = s + len;
4262 yyerror("Unterminated attribute parameter in attribute list");
4266 return REPORT(0); /* EOF indicator */
4270 sv_catsv(sv, PL_lex_stuff);
4271 attrs = append_elem(OP_LIST, attrs,
4272 newSVOP(OP_CONST, 0, sv));
4273 SvREFCNT_dec(PL_lex_stuff);
4274 PL_lex_stuff = NULL;
4277 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4279 if (PL_in_my == KEY_our) {
4281 GvUNIQUE_on(cGVOPx_gv(pl_yylval.opval));
4283 /* skip to avoid loading attributes.pm */
4285 deprecate(":unique");
4288 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4291 /* NOTE: any CV attrs applied here need to be part of
4292 the CVf_BUILTIN_ATTRS define in cv.h! */
4293 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4295 CvLVALUE_on(PL_compcv);
4297 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4299 CvLOCKED_on(PL_compcv);
4301 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4303 CvMETHOD_on(PL_compcv);
4305 /* After we've set the flags, it could be argued that
4306 we don't need to do the attributes.pm-based setting
4307 process, and shouldn't bother appending recognized
4308 flags. To experiment with that, uncomment the
4309 following "else". (Note that's already been
4310 uncommented. That keeps the above-applied built-in
4311 attributes from being intercepted (and possibly
4312 rejected) by a package's attribute routines, but is
4313 justified by the performance win for the common case
4314 of applying only built-in attributes.) */
4316 attrs = append_elem(OP_LIST, attrs,
4317 newSVOP(OP_CONST, 0,
4321 if (*s == ':' && s[1] != ':')
4324 break; /* require real whitespace or :'s */
4325 /* XXX losing whitespace on sequential attributes here */
4329 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4330 if (*s != ';' && *s != '}' && *s != tmp
4331 && (tmp != '=' || *s != ')')) {
4332 const char q = ((*s == '\'') ? '"' : '\'');
4333 /* If here for an expression, and parsed no attrs, back
4335 if (tmp == '=' && !attrs) {
4339 /* MUST advance bufptr here to avoid bogus "at end of line"
4340 context messages from yyerror().
4343 yyerror( (const char *)
4345 ? Perl_form(aTHX_ "Invalid separator character "
4346 "%c%c%c in attribute list", q, *s, q)
4347 : "Unterminated attribute list" ) );
4355 start_force(PL_curforce);
4356 NEXTVAL_NEXTTOKE.opval = attrs;
4357 CURMAD('_', PL_nextwhite);
4362 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4363 (s - SvPVX(PL_linestr)) - stuffstart);
4371 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4372 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4380 const char tmp = *s++;
4385 const char tmp = *s++;
4393 if (PL_lex_brackets <= 0)
4394 yyerror("Unmatched right square bracket");
4397 if (PL_lex_state == LEX_INTERPNORMAL) {
4398 if (PL_lex_brackets == 0) {
4399 if (*s == '-' && s[1] == '>')
4400 PL_lex_state = LEX_INTERPENDMAYBE;
4401 else if (*s != '[' && *s != '{')
4402 PL_lex_state = LEX_INTERPEND;
4409 if (PL_lex_brackets > 100) {
4410 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4412 switch (PL_expect) {
4414 if (PL_lex_formbrack) {
4418 if (PL_oldoldbufptr == PL_last_lop)
4419 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4421 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4422 OPERATOR(HASHBRACK);
4424 while (s < PL_bufend && SPACE_OR_TAB(*s))
4427 PL_tokenbuf[0] = '\0';
4428 if (d < PL_bufend && *d == '-') {
4429 PL_tokenbuf[0] = '-';
4431 while (d < PL_bufend && SPACE_OR_TAB(*d))
4434 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4435 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4437 while (d < PL_bufend && SPACE_OR_TAB(*d))
4440 const char minus = (PL_tokenbuf[0] == '-');
4441 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4449 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4454 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4459 if (PL_oldoldbufptr == PL_last_lop)
4460 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4462 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4465 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4467 /* This hack is to get the ${} in the message. */
4469 yyerror("syntax error");
4472 OPERATOR(HASHBRACK);
4474 /* This hack serves to disambiguate a pair of curlies
4475 * as being a block or an anon hash. Normally, expectation
4476 * determines that, but in cases where we're not in a
4477 * position to expect anything in particular (like inside
4478 * eval"") we have to resolve the ambiguity. This code
4479 * covers the case where the first term in the curlies is a
4480 * quoted string. Most other cases need to be explicitly
4481 * disambiguated by prepending a "+" before the opening
4482 * curly in order to force resolution as an anon hash.
4484 * XXX should probably propagate the outer expectation
4485 * into eval"" to rely less on this hack, but that could
4486 * potentially break current behavior of eval"".
4490 if (*s == '\'' || *s == '"' || *s == '`') {
4491 /* common case: get past first string, handling escapes */
4492 for (t++; t < PL_bufend && *t != *s;)
4493 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4497 else if (*s == 'q') {
4500 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4503 /* skip q//-like construct */
4505 char open, close, term;
4508 while (t < PL_bufend && isSPACE(*t))
4510 /* check for q => */
4511 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4512 OPERATOR(HASHBRACK);
4516 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4520 for (t++; t < PL_bufend; t++) {
4521 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4523 else if (*t == open)
4527 for (t++; t < PL_bufend; t++) {
4528 if (*t == '\\' && t+1 < PL_bufend)
4530 else if (*t == close && --brackets <= 0)
4532 else if (*t == open)
4539 /* skip plain q word */
4540 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4543 else if (isALNUM_lazy_if(t,UTF)) {
4545 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4548 while (t < PL_bufend && isSPACE(*t))
4550 /* if comma follows first term, call it an anon hash */
4551 /* XXX it could be a comma expression with loop modifiers */
4552 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4553 || (*t == '=' && t[1] == '>')))
4554 OPERATOR(HASHBRACK);
4555 if (PL_expect == XREF)
4558 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4564 pl_yylval.ival = CopLINE(PL_curcop);
4565 if (isSPACE(*s) || *s == '#')
4566 PL_copline = NOLINE; /* invalidate current command line number */
4571 if (PL_lex_brackets <= 0)
4572 yyerror("Unmatched right curly bracket");
4574 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4575 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4576 PL_lex_formbrack = 0;
4577 if (PL_lex_state == LEX_INTERPNORMAL) {
4578 if (PL_lex_brackets == 0) {
4579 if (PL_expect & XFAKEBRACK) {
4580 PL_expect &= XENUMMASK;
4581 PL_lex_state = LEX_INTERPEND;
4586 PL_thiswhite = newSVpvs("");
4587 sv_catpvn(PL_thiswhite,"}",1);
4590 return yylex(); /* ignore fake brackets */
4592 if (*s == '-' && s[1] == '>')
4593 PL_lex_state = LEX_INTERPENDMAYBE;
4594 else if (*s != '[' && *s != '{')
4595 PL_lex_state = LEX_INTERPEND;
4598 if (PL_expect & XFAKEBRACK) {
4599 PL_expect &= XENUMMASK;
4601 return yylex(); /* ignore fake brackets */
4603 start_force(PL_curforce);
4605 curmad('X', newSVpvn(s-1,1));
4606 CURMAD('_', PL_thiswhite);
4611 PL_thistoken = newSVpvs("");
4619 if (PL_expect == XOPERATOR) {
4620 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4621 && isIDFIRST_lazy_if(s,UTF))
4623 CopLINE_dec(PL_curcop);
4624 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4625 CopLINE_inc(PL_curcop);
4630 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4632 PL_expect = XOPERATOR;
4633 force_ident(PL_tokenbuf, '&');
4637 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
4649 const char tmp = *s++;
4656 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4657 && strchr("+-*/%.^&|<",tmp))
4658 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4659 "Reversed %c= operator",(int)tmp);
4661 if (PL_expect == XSTATE && isALPHA(tmp) &&
4662 (s == PL_linestart+1 || s[-2] == '\n') )
4664 if (PL_in_eval && !PL_rsfp) {
4669 if (strnEQ(s,"=cut",4)) {
4685 PL_thiswhite = newSVpvs("");
4686 sv_catpvn(PL_thiswhite, PL_linestart,
4687 PL_bufend - PL_linestart);
4691 PL_doextract = TRUE;
4695 if (PL_lex_brackets < PL_lex_formbrack) {
4697 #ifdef PERL_STRICT_CR
4698 while (SPACE_OR_TAB(*t))
4700 while (SPACE_OR_TAB(*t) || *t == '\r')
4703 if (*t == '\n' || *t == '#') {
4714 const char tmp = *s++;
4716 /* was this !=~ where !~ was meant?
4717 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4719 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4720 const char *t = s+1;
4722 while (t < PL_bufend && isSPACE(*t))
4725 if (*t == '/' || *t == '?' ||
4726 ((*t == 'm' || *t == 's' || *t == 'y')
4727 && !isALNUM(t[1])) ||
4728 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4729 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4730 "!=~ should be !~");
4740 if (PL_expect != XOPERATOR) {
4741 if (s[1] != '<' && !strchr(s,'>'))
4744 s = scan_heredoc(s);
4746 s = scan_inputsymbol(s);
4747 TERM(sublex_start());
4753 SHop(OP_LEFT_SHIFT);
4767 const char tmp = *s++;
4769 SHop(OP_RIGHT_SHIFT);
4770 else if (tmp == '=')
4779 if (PL_expect == XOPERATOR) {
4780 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4782 deprecate_old(commaless_variable_list);
4783 return REPORT(','); /* grandfather non-comma-format format */
4787 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4788 PL_tokenbuf[0] = '@';
4789 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4790 sizeof PL_tokenbuf - 1, FALSE);
4791 if (PL_expect == XOPERATOR)
4792 no_op("Array length", s);
4793 if (!PL_tokenbuf[1])
4795 PL_expect = XOPERATOR;
4796 PL_pending_ident = '#';
4800 PL_tokenbuf[0] = '$';
4801 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4802 sizeof PL_tokenbuf - 1, FALSE);
4803 if (PL_expect == XOPERATOR)
4805 if (!PL_tokenbuf[1]) {
4807 yyerror("Final $ should be \\$ or $name");
4811 /* This kludge not intended to be bulletproof. */
4812 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4813 pl_yylval.opval = newSVOP(OP_CONST, 0,
4814 newSViv(CopARYBASE_get(&PL_compiling)));
4815 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
4821 const char tmp = *s;
4822 if (PL_lex_state == LEX_NORMAL)
4825 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4826 && intuit_more(s)) {
4828 PL_tokenbuf[0] = '@';
4829 if (ckWARN(WARN_SYNTAX)) {
4832 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4835 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4836 while (t < PL_bufend && *t != ']')
4838 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4839 "Multidimensional syntax %.*s not supported",
4840 (int)((t - PL_bufptr) + 1), PL_bufptr);
4844 else if (*s == '{') {
4846 PL_tokenbuf[0] = '%';
4847 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4848 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4850 char tmpbuf[sizeof PL_tokenbuf];
4853 } while (isSPACE(*t));
4854 if (isIDFIRST_lazy_if(t,UTF)) {
4856 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4860 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4861 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4862 "You need to quote \"%s\"",
4869 PL_expect = XOPERATOR;
4870 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4871 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4872 if (!islop || PL_last_lop_op == OP_GREPSTART)
4873 PL_expect = XOPERATOR;
4874 else if (strchr("$@\"'`q", *s))
4875 PL_expect = XTERM; /* e.g. print $fh "foo" */
4876 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4877 PL_expect = XTERM; /* e.g. print $fh &sub */
4878 else if (isIDFIRST_lazy_if(s,UTF)) {
4879 char tmpbuf[sizeof PL_tokenbuf];
4881 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4882 if ((t2 = keyword(tmpbuf, len, 0))) {
4883 /* binary operators exclude handle interpretations */
4895 PL_expect = XTERM; /* e.g. print $fh length() */
4900 PL_expect = XTERM; /* e.g. print $fh subr() */
4903 else if (isDIGIT(*s))
4904 PL_expect = XTERM; /* e.g. print $fh 3 */
4905 else if (*s == '.' && isDIGIT(s[1]))
4906 PL_expect = XTERM; /* e.g. print $fh .3 */
4907 else if ((*s == '?' || *s == '-' || *s == '+')
4908 && !isSPACE(s[1]) && s[1] != '=')
4909 PL_expect = XTERM; /* e.g. print $fh -1 */
4910 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4912 PL_expect = XTERM; /* e.g. print $fh /.../
4913 XXX except DORDOR operator
4915 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4917 PL_expect = XTERM; /* print $fh <<"EOF" */
4920 PL_pending_ident = '$';
4924 if (PL_expect == XOPERATOR)
4926 PL_tokenbuf[0] = '@';
4927 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4928 if (!PL_tokenbuf[1]) {
4931 if (PL_lex_state == LEX_NORMAL)
4933 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4935 PL_tokenbuf[0] = '%';
4937 /* Warn about @ where they meant $. */
4938 if (*s == '[' || *s == '{') {
4939 if (ckWARN(WARN_SYNTAX)) {
4940 const char *t = s + 1;
4941 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4943 if (*t == '}' || *t == ']') {
4945 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4946 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4947 "Scalar value %.*s better written as $%.*s",
4948 (int)(t-PL_bufptr), PL_bufptr,
4949 (int)(t-PL_bufptr-1), PL_bufptr+1);
4954 PL_pending_ident = '@';
4957 case '/': /* may be division, defined-or, or pattern */
4958 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4962 case '?': /* may either be conditional or pattern */
4963 if(PL_expect == XOPERATOR) {
4971 /* A // operator. */
4981 /* Disable warning on "study /blah/" */
4982 if (PL_oldoldbufptr == PL_last_uni
4983 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4984 || memNE(PL_last_uni, "study", 5)
4985 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4988 s = scan_pat(s,OP_MATCH);
4989 TERM(sublex_start());
4993 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4994 #ifdef PERL_STRICT_CR
4997 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4999 && (s == PL_linestart || s[-1] == '\n') )
5001 PL_lex_formbrack = 0;
5005 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5011 pl_yylval.ival = OPf_SPECIAL;
5017 if (PL_expect != XOPERATOR)
5022 case '0': case '1': case '2': case '3': case '4':
5023 case '5': case '6': case '7': case '8': case '9':
5024 s = scan_num(s, &pl_yylval);
5025 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5026 if (PL_expect == XOPERATOR)
5031 s = scan_str(s,!!PL_madskills,FALSE);
5032 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5033 if (PL_expect == XOPERATOR) {
5034 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5036 deprecate_old(commaless_variable_list);
5037 return REPORT(','); /* grandfather non-comma-format format */
5044 pl_yylval.ival = OP_CONST;
5045 TERM(sublex_start());
5048 s = scan_str(s,!!PL_madskills,FALSE);
5049 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5050 if (PL_expect == XOPERATOR) {
5051 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5053 deprecate_old(commaless_variable_list);
5054 return REPORT(','); /* grandfather non-comma-format format */
5061 pl_yylval.ival = OP_CONST;
5062 /* FIXME. I think that this can be const if char *d is replaced by
5063 more localised variables. */
5064 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5065 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5066 pl_yylval.ival = OP_STRINGIFY;
5070 TERM(sublex_start());
5073 s = scan_str(s,!!PL_madskills,FALSE);
5074 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5075 if (PL_expect == XOPERATOR)
5076 no_op("Backticks",s);
5079 readpipe_override();
5080 TERM(sublex_start());
5084 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5085 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5087 if (PL_expect == XOPERATOR)
5088 no_op("Backslash",s);
5092 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5093 char *start = s + 2;
5094 while (isDIGIT(*start) || *start == '_')
5096 if (*start == '.' && isDIGIT(start[1])) {
5097 s = scan_num(s, &pl_yylval);
5100 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5101 else if (!isALPHA(*start) && (PL_expect == XTERM
5102 || PL_expect == XREF || PL_expect == XSTATE
5103 || PL_expect == XTERMORDORDOR)) {
5104 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5106 s = scan_num(s, &pl_yylval);
5113 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5155 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5157 /* Some keywords can be followed by any delimiter, including ':' */
5158 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5159 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5160 (PL_tokenbuf[0] == 'q' &&
5161 strchr("qwxr", PL_tokenbuf[1])))));
5163 /* x::* is just a word, unless x is "CORE" */
5164 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5168 while (d < PL_bufend && isSPACE(*d))
5169 d++; /* no comments skipped here, or s### is misparsed */
5171 /* Is this a label? */
5172 if (!tmp && PL_expect == XSTATE
5173 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5175 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5180 /* Check for keywords */
5181 tmp = keyword(PL_tokenbuf, len, 0);
5183 /* Is this a word before a => operator? */
5184 if (*d == '=' && d[1] == '>') {
5187 = (OP*)newSVOP(OP_CONST, 0,
5188 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5189 pl_yylval.opval->op_private = OPpCONST_BARE;
5193 if (tmp < 0) { /* second-class keyword? */
5194 GV *ogv = NULL; /* override (winner) */
5195 GV *hgv = NULL; /* hidden (loser) */
5196 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5198 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5201 if (GvIMPORTED_CV(gv))
5203 else if (! CvMETHOD(cv))
5207 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5208 (gv = *gvp) && isGV_with_GP(gv) &&
5209 GvCVu(gv) && GvIMPORTED_CV(gv))
5216 tmp = 0; /* overridden by import or by GLOBAL */
5219 && -tmp==KEY_lock /* XXX generalizable kludge */
5222 tmp = 0; /* any sub overrides "weak" keyword */
5224 else { /* no override */
5226 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5227 Perl_warner(aTHX_ packWARN(WARN_MISC),
5228 "dump() better written as CORE::dump()");
5232 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5233 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5234 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5235 "Ambiguous call resolved as CORE::%s(), %s",
5236 GvENAME(hgv), "qualify as such or use &");
5243 default: /* not a keyword */
5244 /* Trade off - by using this evil construction we can pull the
5245 variable gv into the block labelled keylookup. If not, then
5246 we have to give it function scope so that the goto from the
5247 earlier ':' case doesn't bypass the initialisation. */
5249 just_a_word_zero_gv:
5257 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5260 SV *nextPL_nextwhite = 0;
5264 /* Get the rest if it looks like a package qualifier */
5266 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5268 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5271 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5272 *s == '\'' ? "'" : "::");
5277 if (PL_expect == XOPERATOR) {
5278 if (PL_bufptr == PL_linestart) {
5279 CopLINE_dec(PL_curcop);
5280 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5281 CopLINE_inc(PL_curcop);
5284 no_op("Bareword",s);
5287 /* Look for a subroutine with this name in current package,
5288 unless name is "Foo::", in which case Foo is a bearword
5289 (and a package name). */
5291 if (len > 2 && !PL_madskills &&
5292 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5294 if (ckWARN(WARN_BAREWORD)
5295 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5296 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5297 "Bareword \"%s\" refers to nonexistent package",
5300 PL_tokenbuf[len] = '\0';
5306 /* Mustn't actually add anything to a symbol table.
5307 But also don't want to "initialise" any placeholder
5308 constants that might already be there into full
5309 blown PVGVs with attached PVCV. */
5310 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5311 GV_NOADD_NOINIT, SVt_PVCV);
5316 /* if we saw a global override before, get the right name */
5319 sv = newSVpvs("CORE::GLOBAL::");
5320 sv_catpv(sv,PL_tokenbuf);
5323 /* If len is 0, newSVpv does strlen(), which is correct.
5324 If len is non-zero, then it will be the true length,
5325 and so the scalar will be created correctly. */
5326 sv = newSVpv(PL_tokenbuf,len);
5329 if (PL_madskills && !PL_thistoken) {
5330 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5331 PL_thistoken = newSVpvn(start,s - start);
5332 PL_realtokenstart = s - SvPVX(PL_linestr);
5336 /* Presume this is going to be a bareword of some sort. */
5339 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5340 pl_yylval.opval->op_private = OPpCONST_BARE;
5341 /* UTF-8 package name? */
5342 if (UTF && !IN_BYTES &&
5343 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5346 /* And if "Foo::", then that's what it certainly is. */
5351 /* Do the explicit type check so that we don't need to force
5352 the initialisation of the symbol table to have a real GV.
5353 Beware - gv may not really be a PVGV, cv may not really be
5354 a PVCV, (because of the space optimisations that gv_init
5355 understands) But they're true if for this symbol there is
5356 respectively a typeglob and a subroutine.
5358 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5359 /* Real typeglob, so get the real subroutine: */
5361 /* A proxy for a subroutine in this package? */
5362 : SvOK(gv) ? (CV *) gv : NULL)
5365 /* See if it's the indirect object for a list operator. */
5367 if (PL_oldoldbufptr &&
5368 PL_oldoldbufptr < PL_bufptr &&
5369 (PL_oldoldbufptr == PL_last_lop
5370 || PL_oldoldbufptr == PL_last_uni) &&
5371 /* NO SKIPSPACE BEFORE HERE! */
5372 (PL_expect == XREF ||
5373 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5375 bool immediate_paren = *s == '(';
5377 /* (Now we can afford to cross potential line boundary.) */
5378 s = SKIPSPACE2(s,nextPL_nextwhite);
5380 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5383 /* Two barewords in a row may indicate method call. */
5385 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5386 (tmp = intuit_method(s, gv, cv)))
5389 /* If not a declared subroutine, it's an indirect object. */
5390 /* (But it's an indir obj regardless for sort.) */
5391 /* Also, if "_" follows a filetest operator, it's a bareword */
5394 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5396 (PL_last_lop_op != OP_MAPSTART &&
5397 PL_last_lop_op != OP_GREPSTART))))
5398 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5399 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5402 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5407 PL_expect = XOPERATOR;
5410 s = SKIPSPACE2(s,nextPL_nextwhite);
5411 PL_nextwhite = nextPL_nextwhite;
5416 /* Is this a word before a => operator? */
5417 if (*s == '=' && s[1] == '>' && !pkgname) {
5419 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
5420 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5421 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
5425 /* If followed by a paren, it's certainly a subroutine. */
5430 while (SPACE_OR_TAB(*d))
5432 if (*d == ')' && (sv = gv_const_sv(gv))) {
5439 PL_nextwhite = PL_thiswhite;
5442 start_force(PL_curforce);
5444 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5445 PL_expect = XOPERATOR;
5448 PL_nextwhite = nextPL_nextwhite;
5449 curmad('X', PL_thistoken);
5450 PL_thistoken = newSVpvs("");
5458 /* If followed by var or block, call it a method (unless sub) */
5460 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5461 PL_last_lop = PL_oldbufptr;
5462 PL_last_lop_op = OP_METHOD;
5466 /* If followed by a bareword, see if it looks like indir obj. */
5469 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5470 && (tmp = intuit_method(s, gv, cv)))
5473 /* Not a method, so call it a subroutine (if defined) */
5476 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5477 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5478 "Ambiguous use of -%s resolved as -&%s()",
5479 PL_tokenbuf, PL_tokenbuf);
5480 /* Check for a constant sub */
5481 if ((sv = gv_const_sv(gv))) {
5483 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
5484 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5485 pl_yylval.opval->op_private = 0;
5489 /* Resolve to GV now. */
5490 if (SvTYPE(gv) != SVt_PVGV) {
5491 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5492 assert (SvTYPE(gv) == SVt_PVGV);
5493 /* cv must have been some sort of placeholder, so
5494 now needs replacing with a real code reference. */
5498 op_free(pl_yylval.opval);
5499 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5500 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5501 PL_last_lop = PL_oldbufptr;
5502 PL_last_lop_op = OP_ENTERSUB;
5503 /* Is there a prototype? */
5511 const char *proto = SvPV_const((SV*)cv, protolen);
5514 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5516 while (*proto == ';')
5518 if (*proto == '&' && *s == '{') {
5520 sv_setpvs(PL_subname, "__ANON__");
5522 sv_setpvs(PL_subname, "__ANON__::__ANON__");
5529 PL_nextwhite = PL_thiswhite;
5532 start_force(PL_curforce);
5533 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5536 PL_nextwhite = nextPL_nextwhite;
5537 curmad('X', PL_thistoken);
5538 PL_thistoken = newSVpvs("");
5545 /* Guess harder when madskills require "best effort". */
5546 if (PL_madskills && (!gv || !GvCVu(gv))) {
5547 int probable_sub = 0;
5548 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5550 else if (isALPHA(*s)) {
5554 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5555 if (!keyword(tmpbuf, tmplen, 0))
5558 while (d < PL_bufend && isSPACE(*d))
5560 if (*d == '=' && d[1] == '>')
5565 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5566 op_free(pl_yylval.opval);
5567 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5568 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5569 PL_last_lop = PL_oldbufptr;
5570 PL_last_lop_op = OP_ENTERSUB;
5571 PL_nextwhite = PL_thiswhite;
5573 start_force(PL_curforce);
5574 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5576 PL_nextwhite = nextPL_nextwhite;
5577 curmad('X', PL_thistoken);
5578 PL_thistoken = newSVpvs("");
5583 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5590 /* Call it a bare word */
5592 if (PL_hints & HINT_STRICT_SUBS)
5593 pl_yylval.opval->op_private |= OPpCONST_STRICT;
5596 if (lastchar != '-') {
5597 if (ckWARN(WARN_RESERVED)) {
5601 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5602 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5609 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5610 && ckWARN_d(WARN_AMBIGUOUS)) {
5611 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5612 "Operator or semicolon missing before %c%s",
5613 lastchar, PL_tokenbuf);
5614 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5615 "Ambiguous use of %c resolved as operator %c",
5616 lastchar, lastchar);
5622 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5623 newSVpv(CopFILE(PL_curcop),0));
5627 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5628 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5631 case KEY___PACKAGE__:
5632 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5634 ? newSVhek(HvNAME_HEK(PL_curstash))
5641 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5642 const char *pname = "main";
5643 if (PL_tokenbuf[2] == 'D')
5644 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5645 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5649 GvIOp(gv) = newIO();
5650 IoIFP(GvIOp(gv)) = PL_rsfp;
5651 #if defined(HAS_FCNTL) && defined(F_SETFD)
5653 const int fd = PerlIO_fileno(PL_rsfp);
5654 fcntl(fd,F_SETFD,fd >= 3);
5657 /* Mark this internal pseudo-handle as clean */
5658 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5659 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5660 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5662 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5663 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5664 /* if the script was opened in binmode, we need to revert
5665 * it to text mode for compatibility; but only iff it has CRs
5666 * XXX this is a questionable hack at best. */
5667 if (PL_bufend-PL_bufptr > 2
5668 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5671 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5672 loc = PerlIO_tell(PL_rsfp);
5673 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5676 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5678 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5679 #endif /* NETWARE */
5680 #ifdef PERLIO_IS_STDIO /* really? */
5681 # if defined(__BORLANDC__)
5682 /* XXX see note in do_binmode() */
5683 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5687 PerlIO_seek(PL_rsfp, loc, 0);
5691 #ifdef PERLIO_LAYERS
5694 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5695 else if (PL_encoding) {
5702 XPUSHs(PL_encoding);
5704 call_method("name", G_SCALAR);
5708 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5709 Perl_form(aTHX_ ":encoding(%"SVf")",
5718 if (PL_realtokenstart >= 0) {
5719 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5721 PL_endwhite = newSVpvs("");
5722 sv_catsv(PL_endwhite, PL_thiswhite);
5724 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5725 PL_realtokenstart = -1;
5727 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5728 SvCUR(PL_endwhite))) != NULL) ;
5743 if (PL_expect == XSTATE) {
5750 if (*s == ':' && s[1] == ':') {
5753 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5754 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5755 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5758 else if (tmp == KEY_require || tmp == KEY_do)
5759 /* that's a way to remember we saw "CORE::" */
5772 LOP(OP_ACCEPT,XTERM);
5778 LOP(OP_ATAN2,XTERM);
5784 LOP(OP_BINMODE,XTERM);
5787 LOP(OP_BLESS,XTERM);
5796 /* When 'use switch' is in effect, continue has a dual
5797 life as a control operator. */
5799 if (!FEATURE_IS_ENABLED("switch"))
5802 /* We have to disambiguate the two senses of
5803 "continue". If the next token is a '{' then
5804 treat it as the start of a continue block;
5805 otherwise treat it as a control operator.
5817 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5834 if (!PL_cryptseen) {
5835 PL_cryptseen = TRUE;
5839 LOP(OP_CRYPT,XTERM);
5842 LOP(OP_CHMOD,XTERM);
5845 LOP(OP_CHOWN,XTERM);
5848 LOP(OP_CONNECT,XTERM);
5867 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5868 if (orig_keyword == KEY_do) {
5877 PL_hints |= HINT_BLOCK_SCOPE;
5887 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5888 LOP(OP_DBMOPEN,XTERM);
5894 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5901 pl_yylval.ival = CopLINE(PL_curcop);
5917 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5918 UNIBRACK(OP_ENTEREVAL);
5932 case KEY_endhostent:
5938 case KEY_endservent:
5941 case KEY_endprotoent:
5952 pl_yylval.ival = CopLINE(PL_curcop);
5954 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5957 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5960 if ((PL_bufend - p) >= 3 &&
5961 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5963 else if ((PL_bufend - p) >= 4 &&
5964 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5967 if (isIDFIRST_lazy_if(p,UTF)) {
5968 p = scan_ident(p, PL_bufend,
5969 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5973 Perl_croak(aTHX_ "Missing $ on loop variable");
5975 s = SvPVX(PL_linestr) + soff;
5981 LOP(OP_FORMLINE,XTERM);
5987 LOP(OP_FCNTL,XTERM);
5993 LOP(OP_FLOCK,XTERM);
6002 LOP(OP_GREPSTART, XREF);
6005 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6020 case KEY_getpriority:
6021 LOP(OP_GETPRIORITY,XTERM);
6023 case KEY_getprotobyname:
6026 case KEY_getprotobynumber:
6027 LOP(OP_GPBYNUMBER,XTERM);
6029 case KEY_getprotoent:
6041 case KEY_getpeername:
6042 UNI(OP_GETPEERNAME);
6044 case KEY_gethostbyname:
6047 case KEY_gethostbyaddr:
6048 LOP(OP_GHBYADDR,XTERM);
6050 case KEY_gethostent:
6053 case KEY_getnetbyname:
6056 case KEY_getnetbyaddr:
6057 LOP(OP_GNBYADDR,XTERM);
6062 case KEY_getservbyname:
6063 LOP(OP_GSBYNAME,XTERM);
6065 case KEY_getservbyport:
6066 LOP(OP_GSBYPORT,XTERM);
6068 case KEY_getservent:
6071 case KEY_getsockname:
6072 UNI(OP_GETSOCKNAME);
6074 case KEY_getsockopt:
6075 LOP(OP_GSOCKOPT,XTERM);
6090 pl_yylval.ival = CopLINE(PL_curcop);
6100 pl_yylval.ival = CopLINE(PL_curcop);
6104 LOP(OP_INDEX,XTERM);
6110 LOP(OP_IOCTL,XTERM);
6122 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6154 LOP(OP_LISTEN,XTERM);
6163 s = scan_pat(s,OP_MATCH);
6164 TERM(sublex_start());
6167 LOP(OP_MAPSTART, XREF);
6170 LOP(OP_MKDIR,XTERM);
6173 LOP(OP_MSGCTL,XTERM);
6176 LOP(OP_MSGGET,XTERM);
6179 LOP(OP_MSGRCV,XTERM);
6182 LOP(OP_MSGSND,XTERM);
6187 PL_in_my = (U16)tmp;
6189 if (isIDFIRST_lazy_if(s,UTF)) {
6193 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6194 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6196 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6197 if (!PL_in_my_stash) {
6200 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6204 if (PL_madskills) { /* just add type to declarator token */
6205 sv_catsv(PL_thistoken, PL_nextwhite);
6207 sv_catpvn(PL_thistoken, start, s - start);
6215 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6222 s = tokenize_use(0, s);
6226 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6233 if (isIDFIRST_lazy_if(s,UTF)) {
6235 for (d = s; isALNUM_lazy_if(d,UTF);)
6237 for (t=d; isSPACE(*t);)
6239 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6241 && !(t[0] == '=' && t[1] == '>')
6243 int parms_len = (int)(d-s);
6244 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6245 "Precedence problem: open %.*s should be open(%.*s)",
6246 parms_len, s, parms_len, s);
6252 pl_yylval.ival = OP_OR;
6262 LOP(OP_OPEN_DIR,XTERM);
6265 checkcomma(s,PL_tokenbuf,"filehandle");
6269 checkcomma(s,PL_tokenbuf,"filehandle");
6288 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6292 LOP(OP_PIPE_OP,XTERM);
6295 s = scan_str(s,!!PL_madskills,FALSE);
6298 pl_yylval.ival = OP_CONST;
6299 TERM(sublex_start());
6305 s = scan_str(s,!!PL_madskills,FALSE);
6308 PL_expect = XOPERATOR;
6310 if (SvCUR(PL_lex_stuff)) {
6313 d = SvPV_force(PL_lex_stuff, len);
6315 for (; isSPACE(*d) && len; --len, ++d)
6320 if (!warned && ckWARN(WARN_QW)) {
6321 for (; !isSPACE(*d) && len; --len, ++d) {
6323 Perl_warner(aTHX_ packWARN(WARN_QW),
6324 "Possible attempt to separate words with commas");
6327 else if (*d == '#') {
6328 Perl_warner(aTHX_ packWARN(WARN_QW),
6329 "Possible attempt to put comments in qw() list");
6335 for (; !isSPACE(*d) && len; --len, ++d)
6338 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
6339 words = append_elem(OP_LIST, words,
6340 newSVOP(OP_CONST, 0, tokeq(sv)));
6344 start_force(PL_curforce);
6345 NEXTVAL_NEXTTOKE.opval = words;
6350 SvREFCNT_dec(PL_lex_stuff);
6351 PL_lex_stuff = NULL;
6357 s = scan_str(s,!!PL_madskills,FALSE);
6360 pl_yylval.ival = OP_STRINGIFY;
6361 if (SvIVX(PL_lex_stuff) == '\'')
6362 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6363 TERM(sublex_start());
6366 s = scan_pat(s,OP_QR);
6367 TERM(sublex_start());
6370 s = scan_str(s,!!PL_madskills,FALSE);
6373 readpipe_override();
6374 TERM(sublex_start());
6382 s = force_version(s, FALSE);
6384 else if (*s != 'v' || !isDIGIT(s[1])
6385 || (s = force_version(s, TRUE), *s == 'v'))
6387 *PL_tokenbuf = '\0';
6388 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6389 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6390 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6392 yyerror("<> should be quotes");
6394 if (orig_keyword == KEY_require) {
6402 PL_last_uni = PL_oldbufptr;
6403 PL_last_lop_op = OP_REQUIRE;
6405 return REPORT( (int)REQUIRE );
6411 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6415 LOP(OP_RENAME,XTERM);
6424 LOP(OP_RINDEX,XTERM);
6433 UNIDOR(OP_READLINE);
6436 UNIDOR(OP_BACKTICK);
6445 LOP(OP_REVERSE,XTERM);
6448 UNIDOR(OP_READLINK);
6455 if (pl_yylval.opval)
6456 TERM(sublex_start());
6458 TOKEN(1); /* force error */
6461 checkcomma(s,PL_tokenbuf,"filehandle");
6471 LOP(OP_SELECT,XTERM);
6477 LOP(OP_SEMCTL,XTERM);
6480 LOP(OP_SEMGET,XTERM);
6483 LOP(OP_SEMOP,XTERM);
6489 LOP(OP_SETPGRP,XTERM);
6491 case KEY_setpriority:
6492 LOP(OP_SETPRIORITY,XTERM);
6494 case KEY_sethostent:
6500 case KEY_setservent:
6503 case KEY_setprotoent:
6513 LOP(OP_SEEKDIR,XTERM);
6515 case KEY_setsockopt:
6516 LOP(OP_SSOCKOPT,XTERM);
6522 LOP(OP_SHMCTL,XTERM);
6525 LOP(OP_SHMGET,XTERM);
6528 LOP(OP_SHMREAD,XTERM);
6531 LOP(OP_SHMWRITE,XTERM);
6534 LOP(OP_SHUTDOWN,XTERM);
6543 LOP(OP_SOCKET,XTERM);
6545 case KEY_socketpair:
6546 LOP(OP_SOCKPAIR,XTERM);
6549 checkcomma(s,PL_tokenbuf,"subroutine name");
6551 if (*s == ';' || *s == ')') /* probably a close */
6552 Perl_croak(aTHX_ "sort is now a reserved word");
6554 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6558 LOP(OP_SPLIT,XTERM);
6561 LOP(OP_SPRINTF,XTERM);
6564 LOP(OP_SPLICE,XTERM);
6579 LOP(OP_SUBSTR,XTERM);
6585 char tmpbuf[sizeof PL_tokenbuf];
6586 SSize_t tboffset = 0;
6587 expectation attrful;
6588 bool have_name, have_proto;
6589 const int key = tmp;
6594 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6595 SV *subtoken = newSVpvn(tstart, s - tstart);
6599 s = SKIPSPACE2(s,tmpwhite);
6604 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6605 (*s == ':' && s[1] == ':'))
6612 attrful = XATTRBLOCK;
6613 /* remember buffer pos'n for later force_word */
6614 tboffset = s - PL_oldbufptr;
6615 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6618 nametoke = newSVpvn(s, d - s);
6620 if (memchr(tmpbuf, ':', len))
6621 sv_setpvn(PL_subname, tmpbuf, len);
6623 sv_setsv(PL_subname,PL_curstname);
6624 sv_catpvs(PL_subname,"::");
6625 sv_catpvn(PL_subname,tmpbuf,len);
6632 CURMAD('X', nametoke);
6633 CURMAD('_', tmpwhite);
6634 (void) force_word(PL_oldbufptr + tboffset, WORD,
6637 s = SKIPSPACE2(d,tmpwhite);
6644 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6645 PL_expect = XTERMBLOCK;
6646 attrful = XATTRTERM;
6647 sv_setpvn(PL_subname,"?",1);
6651 if (key == KEY_format) {
6653 PL_lex_formbrack = PL_lex_brackets + 1;
6655 PL_thistoken = subtoken;
6659 (void) force_word(PL_oldbufptr + tboffset, WORD,
6665 /* Look for a prototype */
6668 bool bad_proto = FALSE;
6669 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6671 s = scan_str(s,!!PL_madskills,FALSE);
6673 Perl_croak(aTHX_ "Prototype not terminated");
6674 /* strip spaces and check for bad characters */
6675 d = SvPVX(PL_lex_stuff);
6677 for (p = d; *p; ++p) {
6680 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6686 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6687 "Illegal character in prototype for %"SVf" : %s",
6688 SVfARG(PL_subname), d);
6689 SvCUR_set(PL_lex_stuff, tmp);
6694 CURMAD('q', PL_thisopen);
6695 CURMAD('_', tmpwhite);
6696 CURMAD('=', PL_thisstuff);
6697 CURMAD('Q', PL_thisclose);
6698 NEXTVAL_NEXTTOKE.opval =
6699 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6700 PL_lex_stuff = NULL;
6703 s = SKIPSPACE2(s,tmpwhite);
6711 if (*s == ':' && s[1] != ':')
6712 PL_expect = attrful;
6713 else if (*s != '{' && key == KEY_sub) {
6715 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6717 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6724 curmad('^', newSVpvs(""));
6725 CURMAD('_', tmpwhite);
6729 PL_thistoken = subtoken;
6732 NEXTVAL_NEXTTOKE.opval =
6733 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6734 PL_lex_stuff = NULL;
6740 sv_setpvs(PL_subname, "__ANON__");
6742 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6746 (void) force_word(PL_oldbufptr + tboffset, WORD,
6755 LOP(OP_SYSTEM,XREF);
6758 LOP(OP_SYMLINK,XTERM);
6761 LOP(OP_SYSCALL,XTERM);
6764 LOP(OP_SYSOPEN,XTERM);
6767 LOP(OP_SYSSEEK,XTERM);
6770 LOP(OP_SYSREAD,XTERM);
6773 LOP(OP_SYSWRITE,XTERM);
6777 TERM(sublex_start());
6798 LOP(OP_TRUNCATE,XTERM);
6810 pl_yylval.ival = CopLINE(PL_curcop);
6814 pl_yylval.ival = CopLINE(PL_curcop);
6818 LOP(OP_UNLINK,XTERM);
6824 LOP(OP_UNPACK,XTERM);
6827 LOP(OP_UTIME,XTERM);
6833 LOP(OP_UNSHIFT,XTERM);
6836 s = tokenize_use(1, s);
6846 pl_yylval.ival = CopLINE(PL_curcop);
6850 pl_yylval.ival = CopLINE(PL_curcop);
6854 PL_hints |= HINT_BLOCK_SCOPE;
6861 LOP(OP_WAITPID,XTERM);
6870 ctl_l[0] = toCTRL('L');
6872 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6875 /* Make sure $^L is defined */
6876 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6881 if (PL_expect == XOPERATOR)
6887 pl_yylval.ival = OP_XOR;
6892 TERM(sublex_start());
6897 #pragma segment Main
6901 S_pending_ident(pTHX)
6906 /* pit holds the identifier we read and pending_ident is reset */
6907 char pit = PL_pending_ident;
6908 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
6909 /* All routes through this function want to know if there is a colon. */
6910 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
6911 PL_pending_ident = 0;
6913 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6914 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6915 "### Pending identifier '%s'\n", PL_tokenbuf); });
6917 /* if we're in a my(), we can't allow dynamics here.
6918 $foo'bar has already been turned into $foo::bar, so
6919 just check for colons.
6921 if it's a legal name, the OP is a PADANY.
6924 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6926 yyerror(Perl_form(aTHX_ "No package name allowed for "
6927 "variable %s in \"our\"",
6929 tmp = allocmy(PL_tokenbuf);
6933 yyerror(Perl_form(aTHX_ PL_no_myglob,
6934 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6936 pl_yylval.opval = newOP(OP_PADANY, 0);
6937 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf);
6943 build the ops for accesses to a my() variable.
6945 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6946 then used in a comparison. This catches most, but not
6947 all cases. For instance, it catches
6948 sort { my($a); $a <=> $b }
6950 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6951 (although why you'd do that is anyone's guess).
6956 tmp = pad_findmy(PL_tokenbuf);
6957 if (tmp != NOT_IN_PAD) {
6958 /* might be an "our" variable" */
6959 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6960 /* build ops for a bareword */
6961 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6962 HEK * const stashname = HvNAME_HEK(stash);
6963 SV * const sym = newSVhek(stashname);
6964 sv_catpvs(sym, "::");
6965 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
6966 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6967 pl_yylval.opval->op_private = OPpCONST_ENTERED;
6970 ? (GV_ADDMULTI | GV_ADDINEVAL)
6973 ((PL_tokenbuf[0] == '$') ? SVt_PV
6974 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6979 /* if it's a sort block and they're naming $a or $b */
6980 if (PL_last_lop_op == OP_SORT &&
6981 PL_tokenbuf[0] == '$' &&
6982 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
6985 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
6986 d < PL_bufend && *d != '\n';
6989 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
6990 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
6996 pl_yylval.opval = newOP(OP_PADANY, 0);
6997 pl_yylval.opval->op_targ = tmp;
7003 Whine if they've said @foo in a doublequoted string,
7004 and @foo isn't a variable we can find in the symbol
7007 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7008 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7010 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7011 && ckWARN(WARN_AMBIGUOUS)
7012 /* DO NOT warn for @- and @+ */
7013 && !( PL_tokenbuf[2] == '\0' &&
7014 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7017 /* Downgraded from fatal to warning 20000522 mjd */
7018 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7019 "Possible unintended interpolation of %s in string",
7024 /* build ops for a bareword */
7025 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7027 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7029 PL_tokenbuf + 1, tokenbuf_len - 1,
7030 /* If the identifier refers to a stash, don't autovivify it.
7031 * Change 24660 had the side effect of causing symbol table
7032 * hashes to always be defined, even if they were freshly
7033 * created and the only reference in the entire program was
7034 * the single statement with the defined %foo::bar:: test.
7035 * It appears that all code in the wild doing this actually
7036 * wants to know whether sub-packages have been loaded, so
7037 * by avoiding auto-vivifying symbol tables, we ensure that
7038 * defined %foo::bar:: continues to be false, and the existing
7039 * tests still give the expected answers, even though what
7040 * they're actually testing has now changed subtly.
7042 (*PL_tokenbuf == '%'
7043 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7046 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7047 ((PL_tokenbuf[0] == '$') ? SVt_PV
7048 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7054 * The following code was generated by perl_keyword.pl.
7058 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7063 case 1: /* 5 tokens of length 1 */
7095 case 2: /* 18 tokens of length 2 */
7241 case 3: /* 29 tokens of length 3 */
7245 if (name[1] == 'N' &&
7308 if (name[1] == 'i' &&
7340 if (name[1] == 'o' &&
7349 if (name[1] == 'e' &&
7358 if (name[1] == 'n' &&
7367 if (name[1] == 'o' &&
7376 if (name[1] == 'a' &&
7385 if (name[1] == 'o' &&
7447 if (name[1] == 'e' &&
7461 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7487 if (name[1] == 'i' &&
7496 if (name[1] == 's' &&
7505 if (name[1] == 'e' &&
7514 if (name[1] == 'o' &&
7526 case 4: /* 41 tokens of length 4 */
7530 if (name[1] == 'O' &&
7540 if (name[1] == 'N' &&
7550 if (name[1] == 'i' &&
7560 if (name[1] == 'h' &&
7570 if (name[1] == 'u' &&
7583 if (name[2] == 'c' &&
7592 if (name[2] == 's' &&
7601 if (name[2] == 'a' &&
7637 if (name[1] == 'o' &&
7650 if (name[2] == 't' &&
7659 if (name[2] == 'o' &&
7668 if (name[2] == 't' &&
7677 if (name[2] == 'e' &&
7690 if (name[1] == 'o' &&
7703 if (name[2] == 'y' &&
7712 if (name[2] == 'l' &&
7728 if (name[2] == 's' &&
7737 if (name[2] == 'n' &&
7746 if (name[2] == 'c' &&
7759 if (name[1] == 'e' &&
7769 if (name[1] == 'p' &&
7782 if (name[2] == 'c' &&
7791 if (name[2] == 'p' &&
7800 if (name[2] == 's' &&
7816 if (name[2] == 'n' &&
7886 if (name[2] == 'r' &&
7895 if (name[2] == 'r' &&
7904 if (name[2] == 'a' &&
7920 if (name[2] == 'l' &&
7982 if (name[2] == 'e' &&
7985 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
7998 case 5: /* 39 tokens of length 5 */
8002 if (name[1] == 'E' &&
8013 if (name[1] == 'H' &&
8027 if (name[2] == 'a' &&
8037 if (name[2] == 'a' &&
8054 if (name[2] == 'e' &&
8064 if (name[2] == 'e' &&
8068 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8084 if (name[3] == 'i' &&
8093 if (name[3] == 'o' &&
8129 if (name[2] == 'o' &&
8139 if (name[2] == 'y' &&
8153 if (name[1] == 'l' &&
8167 if (name[2] == 'n' &&
8177 if (name[2] == 'o' &&
8191 if (name[1] == 'i' &&
8196 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8205 if (name[2] == 'd' &&
8215 if (name[2] == 'c' &&
8232 if (name[2] == 'c' &&
8242 if (name[2] == 't' &&
8256 if (name[1] == 'k' &&
8267 if (name[1] == 'r' &&
8281 if (name[2] == 's' &&
8291 if (name[2] == 'd' &&
8308 if (name[2] == 'm' &&
8318 if (name[2] == 'i' &&
8328 if (name[2] == 'e' &&
8338 if (name[2] == 'l' &&
8348 if (name[2] == 'a' &&
8361 if (name[3] == 't' &&
8364 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8370 if (name[3] == 'd' &&
8387 if (name[1] == 'i' &&
8401 if (name[2] == 'a' &&
8414 if (name[3] == 'e' &&
8449 if (name[2] == 'i' &&
8466 if (name[2] == 'i' &&
8476 if (name[2] == 'i' &&
8493 case 6: /* 33 tokens of length 6 */
8497 if (name[1] == 'c' &&
8512 if (name[2] == 'l' &&
8523 if (name[2] == 'r' &&
8538 if (name[1] == 'e' &&
8553 if (name[2] == 's' &&
8558 if(ckWARN_d(WARN_SYNTAX))
8559 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8565 if (name[2] == 'i' &&
8583 if (name[2] == 'l' &&
8594 if (name[2] == 'r' &&
8609 if (name[1] == 'm' &&
8624 if (name[2] == 'n' &&
8635 if (name[2] == 's' &&
8650 if (name[1] == 's' &&
8656 if (name[4] == 't' &&
8665 if (name[4] == 'e' &&
8674 if (name[4] == 'c' &&
8683 if (name[4] == 'n' &&
8699 if (name[1] == 'r' &&
8717 if (name[3] == 'a' &&
8727 if (name[3] == 'u' &&
8741 if (name[2] == 'n' &&
8759 if (name[2] == 'a' &&
8773 if (name[3] == 'e' &&
8786 if (name[4] == 't' &&
8795 if (name[4] == 'e' &&
8817 if (name[4] == 't' &&
8826 if (name[4] == 'e' &&
8842 if (name[2] == 'c' &&
8853 if (name[2] == 'l' &&
8864 if (name[2] == 'b' &&
8875 if (name[2] == 's' &&
8898 if (name[4] == 's' &&
8907 if (name[4] == 'n' &&
8920 if (name[3] == 'a' &&
8937 if (name[1] == 'a' &&
8952 case 7: /* 29 tokens of length 7 */
8956 if (name[1] == 'E' &&
8969 if (name[1] == '_' &&
8982 if (name[1] == 'i' &&
8989 return -KEY_binmode;
8995 if (name[1] == 'o' &&
9002 return -KEY_connect;
9011 if (name[2] == 'm' &&
9017 return -KEY_dbmopen;
9028 if (name[4] == 'u' &&
9032 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9038 if (name[4] == 'n' &&
9059 if (name[1] == 'o' &&
9072 if (name[1] == 'e' &&
9079 if (name[5] == 'r' &&
9082 return -KEY_getpgrp;
9088 if (name[5] == 'i' &&
9091 return -KEY_getppid;
9104 if (name[1] == 'c' &&
9111 return -KEY_lcfirst;
9117 if (name[1] == 'p' &&
9124 return -KEY_opendir;
9130 if (name[1] == 'a' &&
9148 if (name[3] == 'd' &&
9153 return -KEY_readdir;
9159 if (name[3] == 'u' &&
9170 if (name[3] == 'e' &&
9175 return -KEY_reverse;
9194 if (name[3] == 'k' &&
9199 return -KEY_seekdir;
9205 if (name[3] == 'p' &&
9210 return -KEY_setpgrp;
9220 if (name[2] == 'm' &&
9226 return -KEY_shmread;
9232 if (name[2] == 'r' &&
9238 return -KEY_sprintf;
9247 if (name[3] == 'l' &&
9252 return -KEY_symlink;
9261 if (name[4] == 'a' &&
9265 return -KEY_syscall;
9271 if (name[4] == 'p' &&
9275 return -KEY_sysopen;
9281 if (name[4] == 'e' &&
9285 return -KEY_sysread;
9291 if (name[4] == 'e' &&
9295 return -KEY_sysseek;
9313 if (name[1] == 'e' &&
9320 return -KEY_telldir;
9329 if (name[2] == 'f' &&
9335 return -KEY_ucfirst;
9341 if (name[2] == 's' &&
9347 return -KEY_unshift;
9357 if (name[1] == 'a' &&
9364 return -KEY_waitpid;
9373 case 8: /* 26 tokens of length 8 */
9377 if (name[1] == 'U' &&
9385 return KEY_AUTOLOAD;
9396 if (name[3] == 'A' &&
9402 return KEY___DATA__;
9408 if (name[3] == 'I' &&
9414 return -KEY___FILE__;
9420 if (name[3] == 'I' &&
9426 return -KEY___LINE__;
9442 if (name[2] == 'o' &&
9449 return -KEY_closedir;
9455 if (name[2] == 'n' &&
9462 return -KEY_continue;
9472 if (name[1] == 'b' &&
9480 return -KEY_dbmclose;
9486 if (name[1] == 'n' &&
9492 if (name[4] == 'r' &&
9497 return -KEY_endgrent;
9503 if (name[4] == 'w' &&
9508 return -KEY_endpwent;
9521 if (name[1] == 'o' &&
9529 return -KEY_formline;
9535 if (name[1] == 'e' &&
9546 if (name[6] == 'n' &&
9549 return -KEY_getgrent;
9555 if (name[6] == 'i' &&
9558 return -KEY_getgrgid;
9564 if (name[6] == 'a' &&
9567 return -KEY_getgrnam;
9580 if (name[4] == 'o' &&
9585 return -KEY_getlogin;
9596 if (name[6] == 'n' &&
9599 return -KEY_getpwent;
9605 if (name[6] == 'a' &&
9608 return -KEY_getpwnam;
9614 if (name[6] == 'i' &&
9617 return -KEY_getpwuid;
9637 if (name[1] == 'e' &&
9644 if (name[5] == 'i' &&
9651 return -KEY_readline;
9656 return -KEY_readlink;
9667 if (name[5] == 'i' &&
9671 return -KEY_readpipe;
9692 if (name[4] == 'r' &&
9697 return -KEY_setgrent;
9703 if (name[4] == 'w' &&
9708 return -KEY_setpwent;
9724 if (name[3] == 'w' &&
9730 return -KEY_shmwrite;
9736 if (name[3] == 't' &&
9742 return -KEY_shutdown;
9752 if (name[2] == 's' &&
9759 return -KEY_syswrite;
9769 if (name[1] == 'r' &&
9777 return -KEY_truncate;
9786 case 9: /* 9 tokens of length 9 */
9790 if (name[1] == 'N' &&
9799 return KEY_UNITCHECK;
9805 if (name[1] == 'n' &&
9814 return -KEY_endnetent;
9820 if (name[1] == 'e' &&
9829 return -KEY_getnetent;
9835 if (name[1] == 'o' &&
9844 return -KEY_localtime;
9850 if (name[1] == 'r' &&
9859 return KEY_prototype;
9865 if (name[1] == 'u' &&
9874 return -KEY_quotemeta;
9880 if (name[1] == 'e' &&
9889 return -KEY_rewinddir;
9895 if (name[1] == 'e' &&
9904 return -KEY_setnetent;
9910 if (name[1] == 'a' &&
9919 return -KEY_wantarray;
9928 case 10: /* 9 tokens of length 10 */
9932 if (name[1] == 'n' &&
9938 if (name[4] == 'o' &&
9945 return -KEY_endhostent;
9951 if (name[4] == 'e' &&
9958 return -KEY_endservent;
9971 if (name[1] == 'e' &&
9977 if (name[4] == 'o' &&
9984 return -KEY_gethostent;
9993 if (name[5] == 'r' &&
9999 return -KEY_getservent;
10005 if (name[5] == 'c' &&
10011 return -KEY_getsockopt;
10031 if (name[2] == 't')
10036 if (name[4] == 'o' &&
10043 return -KEY_sethostent;
10052 if (name[5] == 'r' &&
10058 return -KEY_setservent;
10064 if (name[5] == 'c' &&
10070 return -KEY_setsockopt;
10087 if (name[2] == 'c' &&
10096 return -KEY_socketpair;
10109 case 11: /* 8 tokens of length 11 */
10113 if (name[1] == '_' &&
10123 { /* __PACKAGE__ */
10124 return -KEY___PACKAGE__;
10130 if (name[1] == 'n' &&
10140 { /* endprotoent */
10141 return -KEY_endprotoent;
10147 if (name[1] == 'e' &&
10156 if (name[5] == 'e' &&
10162 { /* getpeername */
10163 return -KEY_getpeername;
10172 if (name[6] == 'o' &&
10177 { /* getpriority */
10178 return -KEY_getpriority;
10184 if (name[6] == 't' &&
10189 { /* getprotoent */
10190 return -KEY_getprotoent;
10204 if (name[4] == 'o' &&
10211 { /* getsockname */
10212 return -KEY_getsockname;
10225 if (name[1] == 'e' &&
10233 if (name[6] == 'o' &&
10238 { /* setpriority */
10239 return -KEY_setpriority;
10245 if (name[6] == 't' &&
10250 { /* setprotoent */
10251 return -KEY_setprotoent;
10267 case 12: /* 2 tokens of length 12 */
10268 if (name[0] == 'g' &&
10280 if (name[9] == 'd' &&
10283 { /* getnetbyaddr */
10284 return -KEY_getnetbyaddr;
10290 if (name[9] == 'a' &&
10293 { /* getnetbyname */
10294 return -KEY_getnetbyname;
10306 case 13: /* 4 tokens of length 13 */
10307 if (name[0] == 'g' &&
10314 if (name[4] == 'o' &&
10323 if (name[10] == 'd' &&
10326 { /* gethostbyaddr */
10327 return -KEY_gethostbyaddr;
10333 if (name[10] == 'a' &&
10336 { /* gethostbyname */
10337 return -KEY_gethostbyname;
10350 if (name[4] == 'e' &&
10359 if (name[10] == 'a' &&
10362 { /* getservbyname */
10363 return -KEY_getservbyname;
10369 if (name[10] == 'o' &&
10372 { /* getservbyport */
10373 return -KEY_getservbyport;
10392 case 14: /* 1 tokens of length 14 */
10393 if (name[0] == 'g' &&
10407 { /* getprotobyname */
10408 return -KEY_getprotobyname;
10413 case 16: /* 1 tokens of length 16 */
10414 if (name[0] == 'g' &&
10430 { /* getprotobynumber */
10431 return -KEY_getprotobynumber;
10445 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10449 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10450 if (ckWARN(WARN_SYNTAX)) {
10453 for (w = s+2; *w && level; w++) {
10456 else if (*w == ')')
10459 while (isSPACE(*w))
10461 /* the list of chars below is for end of statements or
10462 * block / parens, boolean operators (&&, ||, //) and branch
10463 * constructs (or, and, if, until, unless, while, err, for).
10464 * Not a very solid hack... */
10465 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10466 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10467 "%s (...) interpreted as function",name);
10470 while (s < PL_bufend && isSPACE(*s))
10474 while (s < PL_bufend && isSPACE(*s))
10476 if (isIDFIRST_lazy_if(s,UTF)) {
10477 const char * const w = s++;
10478 while (isALNUM_lazy_if(s,UTF))
10480 while (s < PL_bufend && isSPACE(*s))
10484 if (keyword(w, s - w, 0))
10487 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10488 if (gv && GvCVu(gv))
10490 Perl_croak(aTHX_ "No comma allowed after %s", what);
10495 /* Either returns sv, or mortalizes sv and returns a new SV*.
10496 Best used as sv=new_constant(..., sv, ...).
10497 If s, pv are NULL, calls subroutine with one argument,
10498 and type is used with error messages only. */
10501 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
10502 SV *sv, SV *pv, const char *type, STRLEN typelen)
10505 HV * const table = GvHV(PL_hintgv); /* ^H */
10509 const char *why1 = "", *why2 = "", *why3 = "";
10511 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10514 why2 = (const char *)
10515 (strEQ(key,"charnames")
10516 ? "(possibly a missing \"use charnames ...\")"
10518 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10519 (type ? type: "undef"), why2);
10521 /* This is convoluted and evil ("goto considered harmful")
10522 * but I do not understand the intricacies of all the different
10523 * failure modes of %^H in here. The goal here is to make
10524 * the most probable error message user-friendly. --jhi */
10529 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10530 (type ? type: "undef"), why1, why2, why3);
10532 yyerror(SvPVX_const(msg));
10536 cvp = hv_fetch(table, key, keylen, FALSE);
10537 if (!cvp || !SvOK(*cvp)) {
10540 why3 = "} is not defined";
10543 sv_2mortal(sv); /* Parent created it permanently */
10546 pv = newSVpvn_flags(s, len, SVs_TEMP);
10548 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
10550 typesv = &PL_sv_undef;
10552 PUSHSTACKi(PERLSI_OVERLOAD);
10564 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10568 /* Check the eval first */
10569 if (!PL_in_eval && SvTRUE(ERRSV)) {
10570 sv_catpvs(ERRSV, "Propagated");
10571 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10573 res = SvREFCNT_inc_simple(sv);
10577 SvREFCNT_inc_simple_void(res);
10586 why1 = "Call to &{$^H{";
10588 why3 = "}} did not return a defined value";
10596 /* Returns a NUL terminated string, with the length of the string written to
10600 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10603 register char *d = dest;
10604 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10607 Perl_croak(aTHX_ ident_too_long);
10608 if (isALNUM(*s)) /* UTF handled below */
10610 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10615 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10619 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10620 char *t = s + UTF8SKIP(s);
10622 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10626 Perl_croak(aTHX_ ident_too_long);
10627 Copy(s, d, len, char);
10640 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10643 char *bracket = NULL;
10645 register char *d = dest;
10646 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10651 while (isDIGIT(*s)) {
10653 Perl_croak(aTHX_ ident_too_long);
10660 Perl_croak(aTHX_ ident_too_long);
10661 if (isALNUM(*s)) /* UTF handled below */
10663 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10668 else if (*s == ':' && s[1] == ':') {
10672 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10673 char *t = s + UTF8SKIP(s);
10674 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10676 if (d + (t - s) > e)
10677 Perl_croak(aTHX_ ident_too_long);
10678 Copy(s, d, t - s, char);
10689 if (PL_lex_state != LEX_NORMAL)
10690 PL_lex_state = LEX_INTERPENDMAYBE;
10693 if (*s == '$' && s[1] &&
10694 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10707 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10712 if (isSPACE(s[-1])) {
10714 const char ch = *s++;
10715 if (!SPACE_OR_TAB(ch)) {
10721 if (isIDFIRST_lazy_if(d,UTF)) {
10725 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10726 end += UTF8SKIP(end);
10727 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10728 end += UTF8SKIP(end);
10730 Copy(s, d, end - s, char);
10735 while ((isALNUM(*s) || *s == ':') && d < e)
10738 Perl_croak(aTHX_ ident_too_long);
10741 while (s < send && SPACE_OR_TAB(*s))
10743 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10744 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10745 const char * const brack =
10747 ((*s == '[') ? "[...]" : "{...}");
10748 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10749 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10750 funny, dest, brack, funny, dest, brack);
10753 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10757 /* Handle extended ${^Foo} variables
10758 * 1999-02-27 mjd-perl-patch@plover.com */
10759 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10763 while (isALNUM(*s) && d < e) {
10767 Perl_croak(aTHX_ ident_too_long);
10772 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10773 PL_lex_state = LEX_INTERPEND;
10776 if (PL_lex_state == LEX_NORMAL) {
10777 if (ckWARN(WARN_AMBIGUOUS) &&
10778 (keyword(dest, d - dest, 0)
10779 || get_cvn_flags(dest, d - dest, 0)))
10783 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10784 "Ambiguous use of %c{%s} resolved to %c%s",
10785 funny, dest, funny, dest);
10790 s = bracket; /* let the parser handle it */
10794 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10795 PL_lex_state = LEX_INTERPEND;
10800 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10802 PERL_UNUSED_CONTEXT;
10804 const char c = (char)ch;
10806 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10807 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10808 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10809 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10810 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10816 S_scan_pat(pTHX_ char *start, I32 type)
10820 char *s = scan_str(start,!!PL_madskills,FALSE);
10821 const char * const valid_flags =
10822 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10829 const char * const delimiter = skipspace(start);
10833 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10834 : "Search pattern not terminated" ));
10837 pm = (PMOP*)newPMOP(type, 0);
10838 if (PL_multi_open == '?') {
10839 /* This is the only point in the code that sets PMf_ONCE: */
10840 pm->op_pmflags |= PMf_ONCE;
10842 /* Hence it's safe to do this bit of PMOP book-keeping here, which
10843 allows us to restrict the list needed by reset to just the ??
10845 assert(type != OP_TRANS);
10847 MAGIC *mg = mg_find((SV*)PL_curstash, PERL_MAGIC_symtab);
10850 mg = sv_magicext((SV*)PL_curstash, 0, PERL_MAGIC_symtab, 0, 0,
10853 elements = mg->mg_len / sizeof(PMOP**);
10854 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
10855 ((PMOP**)mg->mg_ptr) [elements++] = pm;
10856 mg->mg_len = elements * sizeof(PMOP**);
10857 PmopSTASH_set(pm,PL_curstash);
10863 while (*s && strchr(valid_flags, *s))
10864 pmflag(&pm->op_pmflags,*s++);
10866 if (PL_madskills && modstart != s) {
10867 SV* tmptoken = newSVpvn(modstart, s - modstart);
10868 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10871 /* issue a warning if /c is specified,but /g is not */
10872 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10873 && ckWARN(WARN_REGEXP))
10875 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
10876 "Use of /c modifier is meaningless without /g" );
10879 PL_lex_op = (OP*)pm;
10880 pl_yylval.ival = OP_MATCH;
10885 S_scan_subst(pTHX_ char *start)
10896 pl_yylval.ival = OP_NULL;
10898 s = scan_str(start,!!PL_madskills,FALSE);
10901 Perl_croak(aTHX_ "Substitution pattern not terminated");
10903 if (s[-1] == PL_multi_open)
10906 if (PL_madskills) {
10907 CURMAD('q', PL_thisopen);
10908 CURMAD('_', PL_thiswhite);
10909 CURMAD('E', PL_thisstuff);
10910 CURMAD('Q', PL_thisclose);
10911 PL_realtokenstart = s - SvPVX(PL_linestr);
10915 first_start = PL_multi_start;
10916 s = scan_str(s,!!PL_madskills,FALSE);
10918 if (PL_lex_stuff) {
10919 SvREFCNT_dec(PL_lex_stuff);
10920 PL_lex_stuff = NULL;
10922 Perl_croak(aTHX_ "Substitution replacement not terminated");
10924 PL_multi_start = first_start; /* so whole substitution is taken together */
10926 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10929 if (PL_madskills) {
10930 CURMAD('z', PL_thisopen);
10931 CURMAD('R', PL_thisstuff);
10932 CURMAD('Z', PL_thisclose);
10938 if (*s == EXEC_PAT_MOD) {
10942 else if (strchr(S_PAT_MODS, *s))
10943 pmflag(&pm->op_pmflags,*s++);
10949 if (PL_madskills) {
10951 curmad('m', newSVpvn(modstart, s - modstart));
10952 append_madprops(PL_thismad, (OP*)pm, 0);
10956 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10957 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10961 SV * const repl = newSVpvs("");
10963 PL_sublex_info.super_bufptr = s;
10964 PL_sublex_info.super_bufend = PL_bufend;
10966 pm->op_pmflags |= PMf_EVAL;
10969 sv_catpvs(repl, "eval ");
10971 sv_catpvs(repl, "do ");
10973 sv_catpvs(repl, "{");
10974 sv_catsv(repl, PL_lex_repl);
10975 if (strchr(SvPVX(PL_lex_repl), '#'))
10976 sv_catpvs(repl, "\n");
10977 sv_catpvs(repl, "}");
10979 SvREFCNT_dec(PL_lex_repl);
10980 PL_lex_repl = repl;
10983 PL_lex_op = (OP*)pm;
10984 pl_yylval.ival = OP_SUBST;
10989 S_scan_trans(pTHX_ char *start)
11002 pl_yylval.ival = OP_NULL;
11004 s = scan_str(start,!!PL_madskills,FALSE);
11006 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11008 if (s[-1] == PL_multi_open)
11011 if (PL_madskills) {
11012 CURMAD('q', PL_thisopen);
11013 CURMAD('_', PL_thiswhite);
11014 CURMAD('E', PL_thisstuff);
11015 CURMAD('Q', PL_thisclose);
11016 PL_realtokenstart = s - SvPVX(PL_linestr);
11020 s = scan_str(s,!!PL_madskills,FALSE);
11022 if (PL_lex_stuff) {
11023 SvREFCNT_dec(PL_lex_stuff);
11024 PL_lex_stuff = NULL;
11026 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11028 if (PL_madskills) {
11029 CURMAD('z', PL_thisopen);
11030 CURMAD('R', PL_thisstuff);
11031 CURMAD('Z', PL_thisclose);
11034 complement = del = squash = 0;
11041 complement = OPpTRANS_COMPLEMENT;
11044 del = OPpTRANS_DELETE;
11047 squash = OPpTRANS_SQUASH;
11056 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11057 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11058 o->op_private &= ~OPpTRANS_ALL;
11059 o->op_private |= del|squash|complement|
11060 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11061 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11064 pl_yylval.ival = OP_TRANS;
11067 if (PL_madskills) {
11069 curmad('m', newSVpvn(modstart, s - modstart));
11070 append_madprops(PL_thismad, o, 0);
11079 S_scan_heredoc(pTHX_ register char *s)
11083 I32 op_type = OP_SCALAR;
11087 const char *found_newline;
11091 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11093 I32 stuffstart = s - SvPVX(PL_linestr);
11096 PL_realtokenstart = -1;
11101 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11105 while (SPACE_OR_TAB(*peek))
11107 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11110 s = delimcpy(d, e, s, PL_bufend, term, &len);
11120 if (!isALNUM_lazy_if(s,UTF))
11121 deprecate_old("bare << to mean <<\"\"");
11122 for (; isALNUM_lazy_if(s,UTF); s++) {
11127 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11128 Perl_croak(aTHX_ "Delimiter for here document is too long");
11131 len = d - PL_tokenbuf;
11134 if (PL_madskills) {
11135 tstart = PL_tokenbuf + !outer;
11136 PL_thisclose = newSVpvn(tstart, len - !outer);
11137 tstart = SvPVX(PL_linestr) + stuffstart;
11138 PL_thisopen = newSVpvn(tstart, s - tstart);
11139 stuffstart = s - SvPVX(PL_linestr);
11142 #ifndef PERL_STRICT_CR
11143 d = strchr(s, '\r');
11145 char * const olds = s;
11147 while (s < PL_bufend) {
11153 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11162 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11169 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11170 herewas = newSVpvn(s,PL_bufend-s);
11174 herewas = newSVpvn(s-1,found_newline-s+1);
11177 herewas = newSVpvn(s,found_newline-s);
11181 if (PL_madskills) {
11182 tstart = SvPVX(PL_linestr) + stuffstart;
11184 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11186 PL_thisstuff = newSVpvn(tstart, s - tstart);
11189 s += SvCUR(herewas);
11192 stuffstart = s - SvPVX(PL_linestr);
11198 tmpstr = newSV_type(SVt_PVIV);
11199 SvGROW(tmpstr, 80);
11200 if (term == '\'') {
11201 op_type = OP_CONST;
11202 SvIV_set(tmpstr, -1);
11204 else if (term == '`') {
11205 op_type = OP_BACKTICK;
11206 SvIV_set(tmpstr, '\\');
11210 PL_multi_start = CopLINE(PL_curcop);
11211 PL_multi_open = PL_multi_close = '<';
11212 term = *PL_tokenbuf;
11213 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11214 char * const bufptr = PL_sublex_info.super_bufptr;
11215 char * const bufend = PL_sublex_info.super_bufend;
11216 char * const olds = s - SvCUR(herewas);
11217 s = strchr(bufptr, '\n');
11221 while (s < bufend &&
11222 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11224 CopLINE_inc(PL_curcop);
11227 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11228 missingterm(PL_tokenbuf);
11230 sv_setpvn(herewas,bufptr,d-bufptr+1);
11231 sv_setpvn(tmpstr,d+1,s-d);
11233 sv_catpvn(herewas,s,bufend-s);
11234 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11241 while (s < PL_bufend &&
11242 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11244 CopLINE_inc(PL_curcop);
11246 if (s >= PL_bufend) {
11247 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11248 missingterm(PL_tokenbuf);
11250 sv_setpvn(tmpstr,d+1,s-d);
11252 if (PL_madskills) {
11254 sv_catpvn(PL_thisstuff, d + 1, s - d);
11256 PL_thisstuff = newSVpvn(d + 1, s - d);
11257 stuffstart = s - SvPVX(PL_linestr);
11261 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11263 sv_catpvn(herewas,s,PL_bufend-s);
11264 sv_setsv(PL_linestr,herewas);
11265 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11266 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11267 PL_last_lop = PL_last_uni = NULL;
11270 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11271 while (s >= PL_bufend) { /* multiple line string? */
11273 if (PL_madskills) {
11274 tstart = SvPVX(PL_linestr) + stuffstart;
11276 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11278 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11282 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11283 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11284 missingterm(PL_tokenbuf);
11287 stuffstart = s - SvPVX(PL_linestr);
11289 CopLINE_inc(PL_curcop);
11290 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11291 PL_last_lop = PL_last_uni = NULL;
11292 #ifndef PERL_STRICT_CR
11293 if (PL_bufend - PL_linestart >= 2) {
11294 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11295 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11297 PL_bufend[-2] = '\n';
11299 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11301 else if (PL_bufend[-1] == '\r')
11302 PL_bufend[-1] = '\n';
11304 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11305 PL_bufend[-1] = '\n';
11307 if (PERLDB_LINE && PL_curstash != PL_debstash)
11308 update_debugger_info(PL_linestr, NULL, 0);
11309 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11310 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11311 *(SvPVX(PL_linestr) + off ) = ' ';
11312 sv_catsv(PL_linestr,herewas);
11313 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11314 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11318 sv_catsv(tmpstr,PL_linestr);
11323 PL_multi_end = CopLINE(PL_curcop);
11324 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11325 SvPV_shrink_to_cur(tmpstr);
11327 SvREFCNT_dec(herewas);
11329 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11331 else if (PL_encoding)
11332 sv_recode_to_utf8(tmpstr, PL_encoding);
11334 PL_lex_stuff = tmpstr;
11335 pl_yylval.ival = op_type;
11339 /* scan_inputsymbol
11340 takes: current position in input buffer
11341 returns: new position in input buffer
11342 side-effects: pl_yylval and lex_op are set.
11347 <FH> read from filehandle
11348 <pkg::FH> read from package qualified filehandle
11349 <pkg'FH> read from package qualified filehandle
11350 <$fh> read from filehandle in $fh
11351 <*.h> filename glob
11356 S_scan_inputsymbol(pTHX_ char *start)
11359 register char *s = start; /* current position in buffer */
11363 char *d = PL_tokenbuf; /* start of temp holding space */
11364 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11366 end = strchr(s, '\n');
11369 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11371 /* die if we didn't have space for the contents of the <>,
11372 or if it didn't end, or if we see a newline
11375 if (len >= (I32)sizeof PL_tokenbuf)
11376 Perl_croak(aTHX_ "Excessively long <> operator");
11378 Perl_croak(aTHX_ "Unterminated <> operator");
11383 Remember, only scalar variables are interpreted as filehandles by
11384 this code. Anything more complex (e.g., <$fh{$num}>) will be
11385 treated as a glob() call.
11386 This code makes use of the fact that except for the $ at the front,
11387 a scalar variable and a filehandle look the same.
11389 if (*d == '$' && d[1]) d++;
11391 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11392 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11395 /* If we've tried to read what we allow filehandles to look like, and
11396 there's still text left, then it must be a glob() and not a getline.
11397 Use scan_str to pull out the stuff between the <> and treat it
11398 as nothing more than a string.
11401 if (d - PL_tokenbuf != len) {
11402 pl_yylval.ival = OP_GLOB;
11403 s = scan_str(start,!!PL_madskills,FALSE);
11405 Perl_croak(aTHX_ "Glob not terminated");
11409 bool readline_overriden = FALSE;
11412 /* we're in a filehandle read situation */
11415 /* turn <> into <ARGV> */
11417 Copy("ARGV",d,5,char);
11419 /* Check whether readline() is overriden */
11420 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11422 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11424 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11425 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11426 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11427 readline_overriden = TRUE;
11429 /* if <$fh>, create the ops to turn the variable into a
11433 /* try to find it in the pad for this block, otherwise find
11434 add symbol table ops
11436 const PADOFFSET tmp = pad_findmy(d);
11437 if (tmp != NOT_IN_PAD) {
11438 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11439 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11440 HEK * const stashname = HvNAME_HEK(stash);
11441 SV * const sym = sv_2mortal(newSVhek(stashname));
11442 sv_catpvs(sym, "::");
11443 sv_catpv(sym, d+1);
11448 OP * const o = newOP(OP_PADSV, 0);
11450 PL_lex_op = readline_overriden
11451 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11452 append_elem(OP_LIST, o,
11453 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11454 : (OP*)newUNOP(OP_READLINE, 0, o);
11463 ? (GV_ADDMULTI | GV_ADDINEVAL)
11466 PL_lex_op = readline_overriden
11467 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11468 append_elem(OP_LIST,
11469 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11470 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11471 : (OP*)newUNOP(OP_READLINE, 0,
11472 newUNOP(OP_RV2SV, 0,
11473 newGVOP(OP_GV, 0, gv)));
11475 if (!readline_overriden)
11476 PL_lex_op->op_flags |= OPf_SPECIAL;
11477 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
11478 pl_yylval.ival = OP_NULL;
11481 /* If it's none of the above, it must be a literal filehandle
11482 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11484 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11485 PL_lex_op = readline_overriden
11486 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11487 append_elem(OP_LIST,
11488 newGVOP(OP_GV, 0, gv),
11489 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11490 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11491 pl_yylval.ival = OP_NULL;
11500 takes: start position in buffer
11501 keep_quoted preserve \ on the embedded delimiter(s)
11502 keep_delims preserve the delimiters around the string
11503 returns: position to continue reading from buffer
11504 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11505 updates the read buffer.
11507 This subroutine pulls a string out of the input. It is called for:
11508 q single quotes q(literal text)
11509 ' single quotes 'literal text'
11510 qq double quotes qq(interpolate $here please)
11511 " double quotes "interpolate $here please"
11512 qx backticks qx(/bin/ls -l)
11513 ` backticks `/bin/ls -l`
11514 qw quote words @EXPORT_OK = qw( func() $spam )
11515 m// regexp match m/this/
11516 s/// regexp substitute s/this/that/
11517 tr/// string transliterate tr/this/that/
11518 y/// string transliterate y/this/that/
11519 ($*@) sub prototypes sub foo ($)
11520 (stuff) sub attr parameters sub foo : attr(stuff)
11521 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11523 In most of these cases (all but <>, patterns and transliterate)
11524 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11525 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11526 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11529 It skips whitespace before the string starts, and treats the first
11530 character as the delimiter. If the delimiter is one of ([{< then
11531 the corresponding "close" character )]}> is used as the closing
11532 delimiter. It allows quoting of delimiters, and if the string has
11533 balanced delimiters ([{<>}]) it allows nesting.
11535 On success, the SV with the resulting string is put into lex_stuff or,
11536 if that is already non-NULL, into lex_repl. The second case occurs only
11537 when parsing the RHS of the special constructs s/// and tr/// (y///).
11538 For convenience, the terminating delimiter character is stuffed into
11543 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11546 SV *sv; /* scalar value: string */
11547 const char *tmps; /* temp string, used for delimiter matching */
11548 register char *s = start; /* current position in the buffer */
11549 register char term; /* terminating character */
11550 register char *to; /* current position in the sv's data */
11551 I32 brackets = 1; /* bracket nesting level */
11552 bool has_utf8 = FALSE; /* is there any utf8 content? */
11553 I32 termcode; /* terminating char. code */
11554 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11555 STRLEN termlen; /* length of terminating string */
11556 int last_off = 0; /* last position for nesting bracket */
11562 /* skip space before the delimiter */
11568 if (PL_realtokenstart >= 0) {
11569 stuffstart = PL_realtokenstart;
11570 PL_realtokenstart = -1;
11573 stuffstart = start - SvPVX(PL_linestr);
11575 /* mark where we are, in case we need to report errors */
11578 /* after skipping whitespace, the next character is the terminator */
11581 termcode = termstr[0] = term;
11585 termcode = utf8_to_uvchr((U8*)s, &termlen);
11586 Copy(s, termstr, termlen, U8);
11587 if (!UTF8_IS_INVARIANT(term))
11591 /* mark where we are */
11592 PL_multi_start = CopLINE(PL_curcop);
11593 PL_multi_open = term;
11595 /* find corresponding closing delimiter */
11596 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11597 termcode = termstr[0] = term = tmps[5];
11599 PL_multi_close = term;
11601 /* create a new SV to hold the contents. 79 is the SV's initial length.
11602 What a random number. */
11603 sv = newSV_type(SVt_PVIV);
11605 SvIV_set(sv, termcode);
11606 (void)SvPOK_only(sv); /* validate pointer */
11608 /* move past delimiter and try to read a complete string */
11610 sv_catpvn(sv, s, termlen);
11613 tstart = SvPVX(PL_linestr) + stuffstart;
11614 if (!PL_thisopen && !keep_delims) {
11615 PL_thisopen = newSVpvn(tstart, s - tstart);
11616 stuffstart = s - SvPVX(PL_linestr);
11620 if (PL_encoding && !UTF) {
11624 int offset = s - SvPVX_const(PL_linestr);
11625 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11626 &offset, (char*)termstr, termlen);
11627 const char * const ns = SvPVX_const(PL_linestr) + offset;
11628 char * const svlast = SvEND(sv) - 1;
11630 for (; s < ns; s++) {
11631 if (*s == '\n' && !PL_rsfp)
11632 CopLINE_inc(PL_curcop);
11635 goto read_more_line;
11637 /* handle quoted delimiters */
11638 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11640 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11642 if ((svlast-1 - t) % 2) {
11643 if (!keep_quoted) {
11644 *(svlast-1) = term;
11646 SvCUR_set(sv, SvCUR(sv) - 1);
11651 if (PL_multi_open == PL_multi_close) {
11657 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11658 /* At here, all closes are "was quoted" one,
11659 so we don't check PL_multi_close. */
11661 if (!keep_quoted && *(t+1) == PL_multi_open)
11666 else if (*t == PL_multi_open)
11674 SvCUR_set(sv, w - SvPVX_const(sv));
11676 last_off = w - SvPVX(sv);
11677 if (--brackets <= 0)
11682 if (!keep_delims) {
11683 SvCUR_set(sv, SvCUR(sv) - 1);
11689 /* extend sv if need be */
11690 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11691 /* set 'to' to the next character in the sv's string */
11692 to = SvPVX(sv)+SvCUR(sv);
11694 /* if open delimiter is the close delimiter read unbridle */
11695 if (PL_multi_open == PL_multi_close) {
11696 for (; s < PL_bufend; s++,to++) {
11697 /* embedded newlines increment the current line number */
11698 if (*s == '\n' && !PL_rsfp)
11699 CopLINE_inc(PL_curcop);
11700 /* handle quoted delimiters */
11701 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11702 if (!keep_quoted && s[1] == term)
11704 /* any other quotes are simply copied straight through */
11708 /* terminate when run out of buffer (the for() condition), or
11709 have found the terminator */
11710 else if (*s == term) {
11713 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11716 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11722 /* if the terminator isn't the same as the start character (e.g.,
11723 matched brackets), we have to allow more in the quoting, and
11724 be prepared for nested brackets.
11727 /* read until we run out of string, or we find the terminator */
11728 for (; s < PL_bufend; s++,to++) {
11729 /* embedded newlines increment the line count */
11730 if (*s == '\n' && !PL_rsfp)
11731 CopLINE_inc(PL_curcop);
11732 /* backslashes can escape the open or closing characters */
11733 if (*s == '\\' && s+1 < PL_bufend) {
11734 if (!keep_quoted &&
11735 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11740 /* allow nested opens and closes */
11741 else if (*s == PL_multi_close && --brackets <= 0)
11743 else if (*s == PL_multi_open)
11745 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11750 /* terminate the copied string and update the sv's end-of-string */
11752 SvCUR_set(sv, to - SvPVX_const(sv));
11755 * this next chunk reads more into the buffer if we're not done yet
11759 break; /* handle case where we are done yet :-) */
11761 #ifndef PERL_STRICT_CR
11762 if (to - SvPVX_const(sv) >= 2) {
11763 if ((to[-2] == '\r' && to[-1] == '\n') ||
11764 (to[-2] == '\n' && to[-1] == '\r'))
11768 SvCUR_set(sv, to - SvPVX_const(sv));
11770 else if (to[-1] == '\r')
11773 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11778 /* if we're out of file, or a read fails, bail and reset the current
11779 line marker so we can report where the unterminated string began
11782 if (PL_madskills) {
11783 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11785 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11787 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11791 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11793 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11799 /* we read a line, so increment our line counter */
11800 CopLINE_inc(PL_curcop);
11802 /* update debugger info */
11803 if (PERLDB_LINE && PL_curstash != PL_debstash)
11804 update_debugger_info(PL_linestr, NULL, 0);
11806 /* having changed the buffer, we must update PL_bufend */
11807 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11808 PL_last_lop = PL_last_uni = NULL;
11811 /* at this point, we have successfully read the delimited string */
11813 if (!PL_encoding || UTF) {
11815 if (PL_madskills) {
11816 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11817 const int len = s - tstart;
11819 sv_catpvn(PL_thisstuff, tstart, len);
11821 PL_thisstuff = newSVpvn(tstart, len);
11822 if (!PL_thisclose && !keep_delims)
11823 PL_thisclose = newSVpvn(s,termlen);
11828 sv_catpvn(sv, s, termlen);
11833 if (PL_madskills) {
11834 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11835 const int len = s - tstart - termlen;
11837 sv_catpvn(PL_thisstuff, tstart, len);
11839 PL_thisstuff = newSVpvn(tstart, len);
11840 if (!PL_thisclose && !keep_delims)
11841 PL_thisclose = newSVpvn(s - termlen,termlen);
11845 if (has_utf8 || PL_encoding)
11848 PL_multi_end = CopLINE(PL_curcop);
11850 /* if we allocated too much space, give some back */
11851 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11852 SvLEN_set(sv, SvCUR(sv) + 1);
11853 SvPV_renew(sv, SvLEN(sv));
11856 /* decide whether this is the first or second quoted string we've read
11869 takes: pointer to position in buffer
11870 returns: pointer to new position in buffer
11871 side-effects: builds ops for the constant in pl_yylval.op
11873 Read a number in any of the formats that Perl accepts:
11875 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11876 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11879 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11881 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11884 If it reads a number without a decimal point or an exponent, it will
11885 try converting the number to an integer and see if it can do so
11886 without loss of precision.
11890 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11893 register const char *s = start; /* current position in buffer */
11894 register char *d; /* destination in temp buffer */
11895 register char *e; /* end of temp buffer */
11896 NV nv; /* number read, as a double */
11897 SV *sv = NULL; /* place to put the converted number */
11898 bool floatit; /* boolean: int or float? */
11899 const char *lastub = NULL; /* position of last underbar */
11900 static char const number_too_long[] = "Number too long";
11902 /* We use the first character to decide what type of number this is */
11906 Perl_croak(aTHX_ "panic: scan_num");
11908 /* if it starts with a 0, it could be an octal number, a decimal in
11909 0.13 disguise, or a hexadecimal number, or a binary number. */
11913 u holds the "number so far"
11914 shift the power of 2 of the base
11915 (hex == 4, octal == 3, binary == 1)
11916 overflowed was the number more than we can hold?
11918 Shift is used when we add a digit. It also serves as an "are
11919 we in octal/hex/binary?" indicator to disallow hex characters
11920 when in octal mode.
11925 bool overflowed = FALSE;
11926 bool just_zero = TRUE; /* just plain 0 or binary number? */
11927 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11928 static const char* const bases[5] =
11929 { "", "binary", "", "octal", "hexadecimal" };
11930 static const char* const Bases[5] =
11931 { "", "Binary", "", "Octal", "Hexadecimal" };
11932 static const char* const maxima[5] =
11934 "0b11111111111111111111111111111111",
11938 const char *base, *Base, *max;
11940 /* check for hex */
11945 } else if (s[1] == 'b') {
11950 /* check for a decimal in disguise */
11951 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11953 /* so it must be octal */
11960 if (ckWARN(WARN_SYNTAX))
11961 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11962 "Misplaced _ in number");
11966 base = bases[shift];
11967 Base = Bases[shift];
11968 max = maxima[shift];
11970 /* read the rest of the number */
11972 /* x is used in the overflow test,
11973 b is the digit we're adding on. */
11978 /* if we don't mention it, we're done */
11982 /* _ are ignored -- but warned about if consecutive */
11984 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
11985 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11986 "Misplaced _ in number");
11990 /* 8 and 9 are not octal */
11991 case '8': case '9':
11993 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
11997 case '2': case '3': case '4':
11998 case '5': case '6': case '7':
12000 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12003 case '0': case '1':
12004 b = *s++ & 15; /* ASCII digit -> value of digit */
12008 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12009 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12010 /* make sure they said 0x */
12013 b = (*s++ & 7) + 9;
12015 /* Prepare to put the digit we have onto the end
12016 of the number so far. We check for overflows.
12022 x = u << shift; /* make room for the digit */
12024 if ((x >> shift) != u
12025 && !(PL_hints & HINT_NEW_BINARY)) {
12028 if (ckWARN_d(WARN_OVERFLOW))
12029 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12030 "Integer overflow in %s number",
12033 u = x | b; /* add the digit to the end */
12036 n *= nvshift[shift];
12037 /* If an NV has not enough bits in its
12038 * mantissa to represent an UV this summing of
12039 * small low-order numbers is a waste of time
12040 * (because the NV cannot preserve the
12041 * low-order bits anyway): we could just
12042 * remember when did we overflow and in the
12043 * end just multiply n by the right
12051 /* if we get here, we had success: make a scalar value from
12056 /* final misplaced underbar check */
12057 if (s[-1] == '_') {
12058 if (ckWARN(WARN_SYNTAX))
12059 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12064 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12065 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12066 "%s number > %s non-portable",
12072 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12073 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12074 "%s number > %s non-portable",
12079 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12080 sv = new_constant(start, s - start, "integer",
12081 sv, NULL, NULL, 0);
12082 else if (PL_hints & HINT_NEW_BINARY)
12083 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12088 handle decimal numbers.
12089 we're also sent here when we read a 0 as the first digit
12091 case '1': case '2': case '3': case '4': case '5':
12092 case '6': case '7': case '8': case '9': case '.':
12095 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12098 /* read next group of digits and _ and copy into d */
12099 while (isDIGIT(*s) || *s == '_') {
12100 /* skip underscores, checking for misplaced ones
12104 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12105 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12106 "Misplaced _ in number");
12110 /* check for end of fixed-length buffer */
12112 Perl_croak(aTHX_ number_too_long);
12113 /* if we're ok, copy the character */
12118 /* final misplaced underbar check */
12119 if (lastub && s == lastub + 1) {
12120 if (ckWARN(WARN_SYNTAX))
12121 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12124 /* read a decimal portion if there is one. avoid
12125 3..5 being interpreted as the number 3. followed
12128 if (*s == '.' && s[1] != '.') {
12133 if (ckWARN(WARN_SYNTAX))
12134 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12135 "Misplaced _ in number");
12139 /* copy, ignoring underbars, until we run out of digits.
12141 for (; isDIGIT(*s) || *s == '_'; s++) {
12142 /* fixed length buffer check */
12144 Perl_croak(aTHX_ number_too_long);
12146 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12147 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12148 "Misplaced _ in number");
12154 /* fractional part ending in underbar? */
12155 if (s[-1] == '_') {
12156 if (ckWARN(WARN_SYNTAX))
12157 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12158 "Misplaced _ in number");
12160 if (*s == '.' && isDIGIT(s[1])) {
12161 /* oops, it's really a v-string, but without the "v" */
12167 /* read exponent part, if present */
12168 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12172 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12173 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12175 /* stray preinitial _ */
12177 if (ckWARN(WARN_SYNTAX))
12178 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12179 "Misplaced _ in number");
12183 /* allow positive or negative exponent */
12184 if (*s == '+' || *s == '-')
12187 /* stray initial _ */
12189 if (ckWARN(WARN_SYNTAX))
12190 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12191 "Misplaced _ in number");
12195 /* read digits of exponent */
12196 while (isDIGIT(*s) || *s == '_') {
12199 Perl_croak(aTHX_ number_too_long);
12203 if (((lastub && s == lastub + 1) ||
12204 (!isDIGIT(s[1]) && s[1] != '_'))
12205 && ckWARN(WARN_SYNTAX))
12206 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12207 "Misplaced _ in number");
12214 /* make an sv from the string */
12218 We try to do an integer conversion first if no characters
12219 indicating "float" have been found.
12224 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12226 if (flags == IS_NUMBER_IN_UV) {
12228 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12231 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12232 if (uv <= (UV) IV_MIN)
12233 sv_setiv(sv, -(IV)uv);
12240 /* terminate the string */
12242 nv = Atof(PL_tokenbuf);
12247 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12248 const char *const key = floatit ? "float" : "integer";
12249 const STRLEN keylen = floatit ? 5 : 7;
12250 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12251 key, keylen, sv, NULL, NULL, 0);
12255 /* if it starts with a v, it could be a v-string */
12258 sv = newSV(5); /* preallocate storage space */
12259 s = scan_vstring(s, PL_bufend, sv);
12263 /* make the op for the constant and return */
12266 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12268 lvalp->opval = NULL;
12274 S_scan_formline(pTHX_ register char *s)
12277 register char *eol;
12279 SV * const stuff = newSVpvs("");
12280 bool needargs = FALSE;
12281 bool eofmt = FALSE;
12283 char *tokenstart = s;
12286 if (PL_madskills) {
12287 savewhite = PL_thiswhite;
12292 while (!needargs) {
12295 #ifdef PERL_STRICT_CR
12296 while (SPACE_OR_TAB(*t))
12299 while (SPACE_OR_TAB(*t) || *t == '\r')
12302 if (*t == '\n' || t == PL_bufend) {
12307 if (PL_in_eval && !PL_rsfp) {
12308 eol = (char *) memchr(s,'\n',PL_bufend-s);
12313 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12315 for (t = s; t < eol; t++) {
12316 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12318 goto enough; /* ~~ must be first line in formline */
12320 if (*t == '@' || *t == '^')
12324 sv_catpvn(stuff, s, eol-s);
12325 #ifndef PERL_STRICT_CR
12326 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12327 char *end = SvPVX(stuff) + SvCUR(stuff);
12330 SvCUR_set(stuff, SvCUR(stuff) - 1);
12340 if (PL_madskills) {
12342 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12344 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12347 s = filter_gets(PL_linestr, PL_rsfp, 0);
12349 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12351 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12353 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12354 PL_last_lop = PL_last_uni = NULL;
12363 if (SvCUR(stuff)) {
12366 PL_lex_state = LEX_NORMAL;
12367 start_force(PL_curforce);
12368 NEXTVAL_NEXTTOKE.ival = 0;
12372 PL_lex_state = LEX_FORMLINE;
12374 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12376 else if (PL_encoding)
12377 sv_recode_to_utf8(stuff, PL_encoding);
12379 start_force(PL_curforce);
12380 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12382 start_force(PL_curforce);
12383 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12387 SvREFCNT_dec(stuff);
12389 PL_lex_formbrack = 0;
12393 if (PL_madskills) {
12395 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12397 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12398 PL_thiswhite = savewhite;
12405 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12408 const I32 oldsavestack_ix = PL_savestack_ix;
12409 CV* const outsidecv = PL_compcv;
12412 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12414 SAVEI32(PL_subline);
12415 save_item(PL_subname);
12416 SAVESPTR(PL_compcv);
12418 PL_compcv = (CV*)newSV_type(is_format ? SVt_PVFM : SVt_PVCV);
12419 CvFLAGS(PL_compcv) |= flags;
12421 PL_subline = CopLINE(PL_curcop);
12422 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12423 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12424 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12426 return oldsavestack_ix;
12430 #pragma segment Perl_yylex
12433 Perl_yywarn(pTHX_ const char *const s)
12436 PL_in_eval |= EVAL_WARNONLY;
12438 PL_in_eval &= ~EVAL_WARNONLY;
12443 Perl_yyerror(pTHX_ const char *const s)
12446 const char *where = NULL;
12447 const char *context = NULL;
12450 int yychar = PL_parser->yychar;
12452 if (!yychar || (yychar == ';' && !PL_rsfp))
12454 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12455 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12456 PL_oldbufptr != PL_bufptr) {
12459 The code below is removed for NetWare because it abends/crashes on NetWare
12460 when the script has error such as not having the closing quotes like:
12461 if ($var eq "value)
12462 Checking of white spaces is anyway done in NetWare code.
12465 while (isSPACE(*PL_oldoldbufptr))
12468 context = PL_oldoldbufptr;
12469 contlen = PL_bufptr - PL_oldoldbufptr;
12471 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12472 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12475 The code below is removed for NetWare because it abends/crashes on NetWare
12476 when the script has error such as not having the closing quotes like:
12477 if ($var eq "value)
12478 Checking of white spaces is anyway done in NetWare code.
12481 while (isSPACE(*PL_oldbufptr))
12484 context = PL_oldbufptr;
12485 contlen = PL_bufptr - PL_oldbufptr;
12487 else if (yychar > 255)
12488 where = "next token ???";
12489 else if (yychar == -2) { /* YYEMPTY */
12490 if (PL_lex_state == LEX_NORMAL ||
12491 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12492 where = "at end of line";
12493 else if (PL_lex_inpat)
12494 where = "within pattern";
12496 where = "within string";
12499 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
12501 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12502 else if (isPRINT_LC(yychar)) {
12503 const char string = yychar;
12504 sv_catpvn(where_sv, &string, 1);
12507 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12508 where = SvPVX_const(where_sv);
12510 msg = sv_2mortal(newSVpv(s, 0));
12511 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12512 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12514 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12516 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12517 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12518 Perl_sv_catpvf(aTHX_ msg,
12519 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12520 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12523 if (PL_in_eval & EVAL_WARNONLY) {
12524 if (ckWARN_d(WARN_SYNTAX))
12525 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12529 if (PL_error_count >= 10) {
12530 if (PL_in_eval && SvCUR(ERRSV))
12531 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12532 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12534 Perl_croak(aTHX_ "%s has too many errors.\n",
12535 OutCopFILE(PL_curcop));
12538 PL_in_my_stash = NULL;
12542 #pragma segment Main
12546 S_swallow_bom(pTHX_ U8 *s)
12549 const STRLEN slen = SvCUR(PL_linestr);
12552 if (s[1] == 0xFE) {
12553 /* UTF-16 little-endian? (or UTF32-LE?) */
12554 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12555 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12556 #ifndef PERL_NO_UTF16_FILTER
12557 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12560 if (PL_bufend > (char*)s) {
12564 filter_add(utf16rev_textfilter, NULL);
12565 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12566 utf16_to_utf8_reversed(s, news,
12567 PL_bufend - (char*)s - 1,
12569 sv_setpvn(PL_linestr, (const char*)news, newlen);
12571 s = (U8*)SvPVX(PL_linestr);
12572 Copy(news, s, newlen, U8);
12576 SvUTF8_on(PL_linestr);
12577 s = (U8*)SvPVX(PL_linestr);
12579 /* FIXME - is this a general bug fix? */
12582 PL_bufend = SvPVX(PL_linestr) + newlen;
12585 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12590 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12591 #ifndef PERL_NO_UTF16_FILTER
12592 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12595 if (PL_bufend > (char *)s) {
12599 filter_add(utf16_textfilter, NULL);
12600 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12601 utf16_to_utf8(s, news,
12602 PL_bufend - (char*)s,
12604 sv_setpvn(PL_linestr, (const char*)news, newlen);
12606 SvUTF8_on(PL_linestr);
12607 s = (U8*)SvPVX(PL_linestr);
12608 PL_bufend = SvPVX(PL_linestr) + newlen;
12611 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12616 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12617 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12618 s += 3; /* UTF-8 */
12624 if (s[2] == 0xFE && s[3] == 0xFF) {
12625 /* UTF-32 big-endian */
12626 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12629 else if (s[2] == 0 && s[3] != 0) {
12632 * are a good indicator of UTF-16BE. */
12633 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12639 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12640 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12641 s += 4; /* UTF-8 */
12647 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12650 * are a good indicator of UTF-16LE. */
12651 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12659 #ifndef PERL_NO_UTF16_FILTER
12661 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12664 const STRLEN old = SvCUR(sv);
12665 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12666 DEBUG_P(PerlIO_printf(Perl_debug_log,
12667 "utf16_textfilter(%p): %d %d (%d)\n",
12668 FPTR2DPTR(void *, utf16_textfilter),
12669 idx, maxlen, (int) count));
12673 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12674 Copy(SvPVX_const(sv), tmps, old, char);
12675 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12676 SvCUR(sv) - old, &newlen);
12677 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12679 DEBUG_P({sv_dump(sv);});
12684 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12687 const STRLEN old = SvCUR(sv);
12688 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12689 DEBUG_P(PerlIO_printf(Perl_debug_log,
12690 "utf16rev_textfilter(%p): %d %d (%d)\n",
12691 FPTR2DPTR(void *, utf16rev_textfilter),
12692 idx, maxlen, (int) count));
12696 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12697 Copy(SvPVX_const(sv), tmps, old, char);
12698 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12699 SvCUR(sv) - old, &newlen);
12700 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12702 DEBUG_P({ sv_dump(sv); });
12708 Returns a pointer to the next character after the parsed
12709 vstring, as well as updating the passed in sv.
12711 Function must be called like
12714 s = scan_vstring(s,e,sv);
12716 where s and e are the start and end of the string.
12717 The sv should already be large enough to store the vstring
12718 passed in, for performance reasons.
12723 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
12726 const char *pos = s;
12727 const char *start = s;
12728 if (*pos == 'v') pos++; /* get past 'v' */
12729 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12731 if ( *pos != '.') {
12732 /* this may not be a v-string if followed by => */
12733 const char *next = pos;
12734 while (next < e && isSPACE(*next))
12736 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12737 /* return string not v-string */
12738 sv_setpvn(sv,(char *)s,pos-s);
12739 return (char *)pos;
12743 if (!isALPHA(*pos)) {
12744 U8 tmpbuf[UTF8_MAXBYTES+1];
12747 s++; /* get past 'v' */
12749 sv_setpvn(sv, "", 0);
12752 /* this is atoi() that tolerates underscores */
12755 const char *end = pos;
12757 while (--end >= s) {
12759 const UV orev = rev;
12760 rev += (*end - '0') * mult;
12762 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12763 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12764 "Integer overflow in decimal number");
12768 if (rev > 0x7FFFFFFF)
12769 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12771 /* Append native character for the rev point */
12772 tmpend = uvchr_to_utf8(tmpbuf, rev);
12773 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12774 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12776 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12782 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12786 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12794 * c-indentation-style: bsd
12795 * c-basic-offset: 4
12796 * indent-tabs-mode: t
12799 * ex: set ts=8 sts=4 sw=4 noet: