3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 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
14 * [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
18 * This file is the lexer for Perl. It's closely linked to the
21 * The main routine is yylex(), which returns the next token.
25 #define PERL_IN_TOKE_C
28 #define new_constant(a,b,c,d,e,f,g) \
29 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
31 #define pl_yylval (PL_parser->yylval)
33 /* YYINITDEPTH -- initial size of the parser's stacks. */
34 #define YYINITDEPTH 200
36 /* XXX temporary backwards compatibility */
37 #define PL_lex_brackets (PL_parser->lex_brackets)
38 #define PL_lex_brackstack (PL_parser->lex_brackstack)
39 #define PL_lex_casemods (PL_parser->lex_casemods)
40 #define PL_lex_casestack (PL_parser->lex_casestack)
41 #define PL_lex_defer (PL_parser->lex_defer)
42 #define PL_lex_dojoin (PL_parser->lex_dojoin)
43 #define PL_lex_expect (PL_parser->lex_expect)
44 #define PL_lex_formbrack (PL_parser->lex_formbrack)
45 #define PL_lex_inpat (PL_parser->lex_inpat)
46 #define PL_lex_inwhat (PL_parser->lex_inwhat)
47 #define PL_lex_op (PL_parser->lex_op)
48 #define PL_lex_repl (PL_parser->lex_repl)
49 #define PL_lex_starts (PL_parser->lex_starts)
50 #define PL_lex_stuff (PL_parser->lex_stuff)
51 #define PL_multi_start (PL_parser->multi_start)
52 #define PL_multi_open (PL_parser->multi_open)
53 #define PL_multi_close (PL_parser->multi_close)
54 #define PL_pending_ident (PL_parser->pending_ident)
55 #define PL_preambled (PL_parser->preambled)
56 #define PL_sublex_info (PL_parser->sublex_info)
57 #define PL_linestr (PL_parser->linestr)
58 #define PL_expect (PL_parser->expect)
59 #define PL_copline (PL_parser->copline)
60 #define PL_bufptr (PL_parser->bufptr)
61 #define PL_oldbufptr (PL_parser->oldbufptr)
62 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
63 #define PL_linestart (PL_parser->linestart)
64 #define PL_bufend (PL_parser->bufend)
65 #define PL_last_uni (PL_parser->last_uni)
66 #define PL_last_lop (PL_parser->last_lop)
67 #define PL_last_lop_op (PL_parser->last_lop_op)
68 #define PL_lex_state (PL_parser->lex_state)
69 #define PL_rsfp (PL_parser->rsfp)
70 #define PL_rsfp_filters (PL_parser->rsfp_filters)
71 #define PL_in_my (PL_parser->in_my)
72 #define PL_in_my_stash (PL_parser->in_my_stash)
73 #define PL_tokenbuf (PL_parser->tokenbuf)
74 #define PL_multi_end (PL_parser->multi_end)
75 #define PL_error_count (PL_parser->error_count)
78 # define PL_endwhite (PL_parser->endwhite)
79 # define PL_faketokens (PL_parser->faketokens)
80 # define PL_lasttoke (PL_parser->lasttoke)
81 # define PL_nextwhite (PL_parser->nextwhite)
82 # define PL_realtokenstart (PL_parser->realtokenstart)
83 # define PL_skipwhite (PL_parser->skipwhite)
84 # define PL_thisclose (PL_parser->thisclose)
85 # define PL_thismad (PL_parser->thismad)
86 # define PL_thisopen (PL_parser->thisopen)
87 # define PL_thisstuff (PL_parser->thisstuff)
88 # define PL_thistoken (PL_parser->thistoken)
89 # define PL_thiswhite (PL_parser->thiswhite)
90 # define PL_thiswhite (PL_parser->thiswhite)
91 # define PL_nexttoke (PL_parser->nexttoke)
92 # define PL_curforce (PL_parser->curforce)
94 # define PL_nexttoke (PL_parser->nexttoke)
95 # define PL_nexttype (PL_parser->nexttype)
96 # define PL_nextval (PL_parser->nextval)
100 S_pending_ident(pTHX);
102 static const char ident_too_long[] = "Identifier too long";
104 #ifndef PERL_NO_UTF16_FILTER
105 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
106 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
110 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
111 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
113 # define CURMAD(slot,sv)
114 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
117 #define XFAKEBRACK 128
118 #define XENUMMASK 127
120 #ifdef USE_UTF8_SCRIPTS
121 # define UTF (!IN_BYTES)
123 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
126 /* The maximum number of characters preceding the unrecognized one to display */
127 #define UNRECOGNIZED_PRECEDE_COUNT 10
129 /* In variables named $^X, these are the legal values for X.
130 * 1999-02-27 mjd-perl-patch@plover.com */
131 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
133 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
135 /* LEX_* are values for PL_lex_state, the state of the lexer.
136 * They are arranged oddly so that the guard on the switch statement
137 * can get by with a single comparison (if the compiler is smart enough).
140 /* #define LEX_NOTPARSING 11 is done in perl.h. */
142 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
143 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
144 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
145 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
146 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
148 /* at end of code, eg "$x" followed by: */
149 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
150 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
152 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
153 string or after \E, $foo, etc */
154 #define LEX_INTERPCONST 2 /* NOT USED */
155 #define LEX_FORMLINE 1 /* expecting a format line */
156 #define LEX_KNOWNEXT 0 /* next token known; just return it */
160 static const char* const lex_state_names[] = {
179 #include "keywords.h"
181 /* CLINE is a macro that ensures PL_copline has a sane value */
186 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
189 # define SKIPSPACE0(s) skipspace0(s)
190 # define SKIPSPACE1(s) skipspace1(s)
191 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
192 # define PEEKSPACE(s) skipspace2(s,0)
194 # define SKIPSPACE0(s) skipspace(s)
195 # define SKIPSPACE1(s) skipspace(s)
196 # define SKIPSPACE2(s,tsv) skipspace(s)
197 # define PEEKSPACE(s) skipspace(s)
201 * Convenience functions to return different tokens and prime the
202 * lexer for the next token. They all take an argument.
204 * TOKEN : generic token (used for '(', DOLSHARP, etc)
205 * OPERATOR : generic operator
206 * AOPERATOR : assignment operator
207 * PREBLOCK : beginning the block after an if, while, foreach, ...
208 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
209 * PREREF : *EXPR where EXPR is not a simple identifier
210 * TERM : expression term
211 * LOOPX : loop exiting command (goto, last, dump, etc)
212 * FTST : file test operator
213 * FUN0 : zero-argument function
214 * FUN1 : not used, except for not, which isn't a UNIOP
215 * BOop : bitwise or or xor
217 * SHop : shift operator
218 * PWop : power operator
219 * PMop : pattern-matching operator
220 * Aop : addition-level operator
221 * Mop : multiplication-level operator
222 * Eop : equality-testing operator
223 * Rop : relational operator <= != gt
225 * Also see LOP and lop() below.
228 #ifdef DEBUGGING /* Serve -DT. */
229 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
231 # define REPORT(retval) (retval)
234 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
235 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
236 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
237 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
238 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
239 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
240 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
241 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
242 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
243 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
244 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
245 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
246 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
247 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
248 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
249 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
250 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
251 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
252 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
253 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
255 /* This bit of chicanery makes a unary function followed by
256 * a parenthesis into a function with one argument, highest precedence.
257 * The UNIDOR macro is for unary functions that can be followed by the //
258 * operator (such as C<shift // 0>).
260 #define UNI2(f,x) { \
261 pl_yylval.ival = f; \
264 PL_last_uni = PL_oldbufptr; \
265 PL_last_lop_op = f; \
267 return REPORT( (int)FUNC1 ); \
269 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
271 #define UNI(f) UNI2(f,XTERM)
272 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
274 #define UNIBRACK(f) { \
275 pl_yylval.ival = f; \
277 PL_last_uni = PL_oldbufptr; \
279 return REPORT( (int)FUNC1 ); \
281 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
284 /* grandfather return to old style */
285 #define OLDLOP(f) return(pl_yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
289 /* how to interpret the pl_yylval associated with the token */
293 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
299 static struct debug_tokens {
301 enum token_type type;
303 } const debug_tokens[] =
305 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
306 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
307 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
308 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
309 { ARROW, TOKENTYPE_NONE, "ARROW" },
310 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
311 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
312 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
313 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
314 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
315 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
316 { DO, TOKENTYPE_NONE, "DO" },
317 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
318 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
319 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
320 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
321 { ELSE, TOKENTYPE_NONE, "ELSE" },
322 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
323 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
324 { FOR, TOKENTYPE_IVAL, "FOR" },
325 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
326 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
327 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
328 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
329 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
330 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
331 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
332 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
333 { IF, TOKENTYPE_IVAL, "IF" },
334 { LABEL, TOKENTYPE_PVAL, "LABEL" },
335 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
336 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
337 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
338 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
339 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
340 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
341 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
342 { MY, TOKENTYPE_IVAL, "MY" },
343 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
344 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
345 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
346 { OROP, TOKENTYPE_IVAL, "OROP" },
347 { OROR, TOKENTYPE_NONE, "OROR" },
348 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
349 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
350 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
351 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
352 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
353 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
354 { PREINC, TOKENTYPE_NONE, "PREINC" },
355 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
356 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
357 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
358 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
359 { SUB, TOKENTYPE_NONE, "SUB" },
360 { THING, TOKENTYPE_OPVAL, "THING" },
361 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
362 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
363 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
364 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
365 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
366 { USE, TOKENTYPE_IVAL, "USE" },
367 { WHEN, TOKENTYPE_IVAL, "WHEN" },
368 { WHILE, TOKENTYPE_IVAL, "WHILE" },
369 { WORD, TOKENTYPE_OPVAL, "WORD" },
370 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
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 PERL_ARGS_ASSERT_TOKEREPORT;
384 const char *name = NULL;
385 enum token_type type = TOKENTYPE_NONE;
386 const struct debug_tokens *p;
387 SV* const report = newSVpvs("<== ");
389 for (p = debug_tokens; p->token; p++) {
390 if (p->token == (int)rv) {
397 Perl_sv_catpv(aTHX_ report, name);
398 else if ((char)rv > ' ' && (char)rv < '~')
399 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
401 sv_catpvs(report, "EOF");
403 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
406 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
409 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
411 case TOKENTYPE_OPNUM:
412 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
413 PL_op_name[lvalp->ival]);
416 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
418 case TOKENTYPE_OPVAL:
420 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
421 PL_op_name[lvalp->opval->op_type]);
422 if (lvalp->opval->op_type == OP_CONST) {
423 Perl_sv_catpvf(aTHX_ report, " %s",
424 SvPEEK(cSVOPx_sv(lvalp->opval)));
429 sv_catpvs(report, "(opval=null)");
432 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
438 /* print the buffer with suitable escapes */
441 S_printbuf(pTHX_ const char *const fmt, const char *const s)
443 SV* const tmp = newSVpvs("");
445 PERL_ARGS_ASSERT_PRINTBUF;
447 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
454 S_deprecate_commaless_var_list(pTHX) {
456 deprecate("comma-less variable list");
457 return REPORT(','); /* grandfather non-comma-format format */
463 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
464 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
468 S_ao(pTHX_ int toketype)
471 if (*PL_bufptr == '=') {
473 if (toketype == ANDAND)
474 pl_yylval.ival = OP_ANDASSIGN;
475 else if (toketype == OROR)
476 pl_yylval.ival = OP_ORASSIGN;
477 else if (toketype == DORDOR)
478 pl_yylval.ival = OP_DORASSIGN;
486 * When Perl expects an operator and finds something else, no_op
487 * prints the warning. It always prints "<something> found where
488 * operator expected. It prints "Missing semicolon on previous line?"
489 * if the surprise occurs at the start of the line. "do you need to
490 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
491 * where the compiler doesn't know if foo is a method call or a function.
492 * It prints "Missing operator before end of line" if there's nothing
493 * after the missing operator, or "... before <...>" if there is something
494 * after the missing operator.
498 S_no_op(pTHX_ const char *const what, char *s)
501 char * const oldbp = PL_bufptr;
502 const bool is_first = (PL_oldbufptr == PL_linestart);
504 PERL_ARGS_ASSERT_NO_OP;
510 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
511 if (ckWARN_d(WARN_SYNTAX)) {
513 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
514 "\t(Missing semicolon on previous line?)\n");
515 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
517 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
519 if (t < PL_bufptr && isSPACE(*t))
520 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
521 "\t(Do you need to predeclare %.*s?)\n",
522 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
526 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
527 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
535 * Complain about missing quote/regexp/heredoc terminator.
536 * If it's called with NULL then it cauterizes the line buffer.
537 * If we're in a delimited string and the delimiter is a control
538 * character, it's reformatted into a two-char sequence like ^C.
543 S_missingterm(pTHX_ char *s)
549 char * const nl = strrchr(s,'\n');
553 else if (isCNTRL(PL_multi_close)) {
555 tmpbuf[1] = (char)toCTRL(PL_multi_close);
560 *tmpbuf = (char)PL_multi_close;
564 q = strchr(s,'"') ? '\'' : '"';
565 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
568 #define FEATURE_IS_ENABLED(name) \
569 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
570 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
571 /* The longest string we pass in. */
572 #define MAX_FEATURE_LEN (sizeof("switch")-1)
575 * S_feature_is_enabled
576 * Check whether the named feature is enabled.
579 S_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
582 HV * const hinthv = GvHV(PL_hintgv);
583 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
585 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
587 assert(namelen <= MAX_FEATURE_LEN);
588 memcpy(&he_name[8], name, namelen);
590 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
594 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
595 * utf16-to-utf8-reversed.
598 #ifdef PERL_CR_FILTER
602 register const char *s = SvPVX_const(sv);
603 register const char * const e = s + SvCUR(sv);
605 PERL_ARGS_ASSERT_STRIP_RETURN;
607 /* outer loop optimized to do nothing if there are no CR-LFs */
609 if (*s++ == '\r' && *s == '\n') {
610 /* hit a CR-LF, need to copy the rest */
611 register char *d = s - 1;
614 if (*s == '\r' && s[1] == '\n')
625 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
627 const I32 count = FILTER_READ(idx+1, sv, maxlen);
628 if (count > 0 && !maxlen)
639 * Create a parser object and initialise its parser and lexer fields
641 * rsfp is the opened file handle to read from (if any),
643 * line holds any initial content already read from the file (or in
644 * the case of no file, such as an eval, the whole contents);
646 * new_filter indicates that this is a new file and it shouldn't inherit
647 * the filters from the current parser (ie require).
651 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
654 const char *s = NULL;
656 yy_parser *parser, *oparser;
658 /* create and initialise a parser */
660 Newxz(parser, 1, yy_parser);
661 parser->old_parser = oparser = PL_parser;
664 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
665 parser->ps = parser->stack;
666 parser->stack_size = YYINITDEPTH;
668 parser->stack->state = 0;
669 parser->yyerrstatus = 0;
670 parser->yychar = YYEMPTY; /* Cause a token to be read. */
672 /* on scope exit, free this parser and restore any outer one */
674 parser->saved_curcop = PL_curcop;
676 /* initialise lexer state */
679 parser->curforce = -1;
681 parser->nexttoke = 0;
683 parser->error_count = oparser ? oparser->error_count : 0;
684 parser->copline = NOLINE;
685 parser->lex_state = LEX_NORMAL;
686 parser->expect = XSTATE;
688 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
689 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
691 Newx(parser->lex_brackstack, 120, char);
692 Newx(parser->lex_casestack, 12, char);
693 *parser->lex_casestack = '\0';
696 s = SvPV_const(line, len);
702 parser->linestr = newSVpvs("\n;");
703 } else if (SvREADONLY(line) || s[len-1] != ';') {
704 parser->linestr = newSVsv(line);
706 sv_catpvs(parser->linestr, "\n;");
709 SvREFCNT_inc_simple_void_NN(line);
710 parser->linestr = line;
712 parser->oldoldbufptr =
715 parser->linestart = SvPVX(parser->linestr);
716 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
717 parser->last_lop = parser->last_uni = NULL;
721 /* delete a parser object */
724 Perl_parser_free(pTHX_ const yy_parser *parser)
726 PERL_ARGS_ASSERT_PARSER_FREE;
728 PL_curcop = parser->saved_curcop;
729 SvREFCNT_dec(parser->linestr);
731 if (parser->rsfp == PerlIO_stdin())
732 PerlIO_clearerr(parser->rsfp);
733 else if (parser->rsfp && (!parser->old_parser ||
734 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
735 PerlIO_close(parser->rsfp);
736 SvREFCNT_dec(parser->rsfp_filters);
738 Safefree(parser->stack);
739 Safefree(parser->lex_brackstack);
740 Safefree(parser->lex_casestack);
741 PL_parser = parser->old_parser;
748 * Finalizer for lexing operations. Must be called when the parser is
749 * done with the lexer.
756 PL_doextract = FALSE;
761 * This subroutine has nothing to do with tilting, whether at windmills
762 * or pinball tables. Its name is short for "increment line". It
763 * increments the current line number in CopLINE(PL_curcop) and checks
764 * to see whether the line starts with a comment of the form
765 * # line 500 "foo.pm"
766 * If so, it sets the current line number and file to the values in the comment.
770 S_incline(pTHX_ const char *s)
777 PERL_ARGS_ASSERT_INCLINE;
779 CopLINE_inc(PL_curcop);
782 while (SPACE_OR_TAB(*s))
784 if (strnEQ(s, "line", 4))
788 if (SPACE_OR_TAB(*s))
792 while (SPACE_OR_TAB(*s))
800 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
802 while (SPACE_OR_TAB(*s))
804 if (*s == '"' && (t = strchr(s+1, '"'))) {
814 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
816 if (*e != '\n' && *e != '\0')
817 return; /* false alarm */
820 const STRLEN len = t - s;
822 SV *const temp_sv = CopFILESV(PL_curcop);
828 tmplen = SvCUR(temp_sv);
834 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
835 /* must copy *{"::_<(eval N)[oldfilename:L]"}
836 * to *{"::_<newfilename"} */
837 /* However, the long form of evals is only turned on by the
838 debugger - usually they're "(eval %lu)" */
842 STRLEN tmplen2 = len;
843 if (tmplen + 2 <= sizeof smallbuf)
846 Newx(tmpbuf, tmplen + 2, char);
849 memcpy(tmpbuf + 2, cf, tmplen);
851 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
856 if (tmplen2 + 2 <= sizeof smallbuf)
859 Newx(tmpbuf2, tmplen2 + 2, char);
861 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
862 /* Either they malloc'd it, or we malloc'd it,
863 so no prefix is present in ours. */
868 memcpy(tmpbuf2 + 2, s, tmplen2);
871 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
873 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
874 /* adjust ${"::_<newfilename"} to store the new file name */
875 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
876 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
877 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
880 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
882 if (tmpbuf != smallbuf) Safefree(tmpbuf);
885 CopFILE_free(PL_curcop);
886 CopFILE_setn(PL_curcop, s, len);
888 CopLINE_set(PL_curcop, atoi(n)-1);
892 /* skip space before PL_thistoken */
895 S_skipspace0(pTHX_ register char *s)
897 PERL_ARGS_ASSERT_SKIPSPACE0;
904 PL_thiswhite = newSVpvs("");
905 sv_catsv(PL_thiswhite, PL_skipwhite);
906 sv_free(PL_skipwhite);
909 PL_realtokenstart = s - SvPVX(PL_linestr);
913 /* skip space after PL_thistoken */
916 S_skipspace1(pTHX_ register char *s)
918 const char *start = s;
919 I32 startoff = start - SvPVX(PL_linestr);
921 PERL_ARGS_ASSERT_SKIPSPACE1;
926 start = SvPVX(PL_linestr) + startoff;
927 if (!PL_thistoken && PL_realtokenstart >= 0) {
928 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
929 PL_thistoken = newSVpvn(tstart, start - tstart);
931 PL_realtokenstart = -1;
934 PL_nextwhite = newSVpvs("");
935 sv_catsv(PL_nextwhite, PL_skipwhite);
936 sv_free(PL_skipwhite);
943 S_skipspace2(pTHX_ register char *s, SV **svp)
946 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
947 const I32 startoff = s - SvPVX(PL_linestr);
949 PERL_ARGS_ASSERT_SKIPSPACE2;
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);
1003 PERL_ARGS_ASSERT_SKIPSPACE;
1006 sv_free(PL_skipwhite);
1010 PERL_ARGS_ASSERT_SKIPSPACE;
1012 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1013 while (s < PL_bufend && SPACE_OR_TAB(*s))
1023 SSize_t oldprevlen, oldoldprevlen;
1024 SSize_t oldloplen = 0, oldunilen = 0;
1025 while (s < PL_bufend && isSPACE(*s)) {
1026 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
1031 if (s < PL_bufend && *s == '#') {
1032 while (s < PL_bufend && *s != '\n')
1034 if (s < PL_bufend) {
1036 if (PL_in_eval && !PL_rsfp) {
1043 /* only continue to recharge the buffer if we're at the end
1044 * of the buffer, we're not reading from a source filter, and
1045 * we're in normal lexing mode
1047 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1048 PL_lex_state == LEX_FORMLINE)
1055 /* try to recharge the buffer */
1057 curoff = s - SvPVX(PL_linestr);
1060 if ((s = filter_gets(PL_linestr, PL_rsfp,
1061 (prevlen = SvCUR(PL_linestr)))) == NULL)
1064 if (PL_madskills && curoff != startoff) {
1066 PL_skipwhite = newSVpvs("");
1067 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1071 /* mustn't throw out old stuff yet if madpropping */
1072 SvCUR(PL_linestr) = curoff;
1073 s = SvPVX(PL_linestr) + curoff;
1075 if (curoff && s[-1] == '\n')
1079 /* end of file. Add on the -p or -n magic */
1080 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1083 sv_catpvs(PL_linestr,
1084 ";}continue{print or die qq(-p destination: $!\\n);}");
1086 sv_setpvs(PL_linestr,
1087 ";}continue{print or die qq(-p destination: $!\\n);}");
1089 PL_minus_n = PL_minus_p = 0;
1091 else if (PL_minus_n) {
1093 sv_catpvs(PL_linestr, ";}");
1095 sv_setpvs(PL_linestr, ";}");
1101 sv_catpvs(PL_linestr,";");
1103 sv_setpvs(PL_linestr,";");
1106 /* reset variables for next time we lex */
1107 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1113 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1114 PL_last_lop = PL_last_uni = NULL;
1116 /* Close the filehandle. Could be from
1117 * STDIN, or a regular file. If we were reading code from
1118 * STDIN (because the commandline held no -e or filename)
1119 * then we don't close it, we reset it so the code can
1120 * read from STDIN too.
1123 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1124 PerlIO_clearerr(PL_rsfp);
1126 (void)PerlIO_close(PL_rsfp);
1131 /* not at end of file, so we only read another line */
1132 /* make corresponding updates to old pointers, for yyerror() */
1133 oldprevlen = PL_oldbufptr - PL_bufend;
1134 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1136 oldunilen = PL_last_uni - PL_bufend;
1138 oldloplen = PL_last_lop - PL_bufend;
1139 PL_linestart = PL_bufptr = s + prevlen;
1140 PL_bufend = s + SvCUR(PL_linestr);
1142 PL_oldbufptr = s + oldprevlen;
1143 PL_oldoldbufptr = s + oldoldprevlen;
1145 PL_last_uni = s + oldunilen;
1147 PL_last_lop = s + oldloplen;
1150 /* debugger active and we're not compiling the debugger code,
1151 * so store the line into the debugger's array of lines
1153 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
1154 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1161 PL_skipwhite = newSVpvs("");
1162 curoff = s - SvPVX(PL_linestr);
1163 if (curoff - startoff)
1164 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1173 * Check the unary operators to ensure there's no ambiguity in how they're
1174 * used. An ambiguous piece of code would be:
1176 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1177 * the +5 is its argument.
1187 if (PL_oldoldbufptr != PL_last_uni)
1189 while (isSPACE(*PL_last_uni))
1192 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1194 if ((t = strchr(s, '(')) && t < PL_bufptr)
1197 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1198 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1199 (int)(s - PL_last_uni), PL_last_uni);
1203 * LOP : macro to build a list operator. Its behaviour has been replaced
1204 * with a subroutine, S_lop() for which LOP is just another name.
1207 #define LOP(f,x) return lop(f,x,s)
1211 * Build a list operator (or something that might be one). The rules:
1212 * - if we have a next token, then it's a list operator [why?]
1213 * - if the next thing is an opening paren, then it's a function
1214 * - else it's a list operator
1218 S_lop(pTHX_ I32 f, int x, char *s)
1222 PERL_ARGS_ASSERT_LOP;
1228 PL_last_lop = PL_oldbufptr;
1229 PL_last_lop_op = (OPCODE)f;
1232 return REPORT(LSTOP);
1235 return REPORT(LSTOP);
1238 return REPORT(FUNC);
1241 return REPORT(FUNC);
1243 return REPORT(LSTOP);
1249 * Sets up for an eventual force_next(). start_force(0) basically does
1250 * an unshift, while start_force(-1) does a push. yylex removes items
1255 S_start_force(pTHX_ int where)
1259 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1260 where = PL_lasttoke;
1261 assert(PL_curforce < 0 || PL_curforce == where);
1262 if (PL_curforce != where) {
1263 for (i = PL_lasttoke; i > where; --i) {
1264 PL_nexttoke[i] = PL_nexttoke[i-1];
1268 if (PL_curforce < 0) /* in case of duplicate start_force() */
1269 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1270 PL_curforce = where;
1273 curmad('^', newSVpvs(""));
1274 CURMAD('_', PL_nextwhite);
1279 S_curmad(pTHX_ char slot, SV *sv)
1285 if (PL_curforce < 0)
1286 where = &PL_thismad;
1288 where = &PL_nexttoke[PL_curforce].next_mad;
1294 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1296 else if (PL_encoding) {
1297 sv_recode_to_utf8(sv, PL_encoding);
1302 /* keep a slot open for the head of the list? */
1303 if (slot != '_' && *where && (*where)->mad_key == '^') {
1304 (*where)->mad_key = slot;
1305 sv_free(MUTABLE_SV(((*where)->mad_val)));
1306 (*where)->mad_val = (void*)sv;
1309 addmad(newMADsv(slot, sv), where, 0);
1312 # define start_force(where) NOOP
1313 # define curmad(slot, sv) NOOP
1318 * When the lexer realizes it knows the next token (for instance,
1319 * it is reordering tokens for the parser) then it can call S_force_next
1320 * to know what token to return the next time the lexer is called. Caller
1321 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1322 * and possibly PL_expect to ensure the lexer handles the token correctly.
1326 S_force_next(pTHX_ I32 type)
1331 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1332 tokereport(type, &NEXTVAL_NEXTTOKE);
1336 if (PL_curforce < 0)
1337 start_force(PL_lasttoke);
1338 PL_nexttoke[PL_curforce].next_type = type;
1339 if (PL_lex_state != LEX_KNOWNEXT)
1340 PL_lex_defer = PL_lex_state;
1341 PL_lex_state = LEX_KNOWNEXT;
1342 PL_lex_expect = PL_expect;
1345 PL_nexttype[PL_nexttoke] = type;
1347 if (PL_lex_state != LEX_KNOWNEXT) {
1348 PL_lex_defer = PL_lex_state;
1349 PL_lex_expect = PL_expect;
1350 PL_lex_state = LEX_KNOWNEXT;
1356 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1359 SV * const sv = newSVpvn_utf8(start, len,
1362 && !is_ascii_string((const U8*)start, len)
1363 && is_utf8_string((const U8*)start, len));
1369 * When the lexer knows the next thing is a word (for instance, it has
1370 * just seen -> and it knows that the next char is a word char, then
1371 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1375 * char *start : buffer position (must be within PL_linestr)
1376 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1377 * int check_keyword : if true, Perl checks to make sure the word isn't
1378 * a keyword (do this if the word is a label, e.g. goto FOO)
1379 * int allow_pack : if true, : characters will also be allowed (require,
1380 * use, etc. do this)
1381 * int allow_initial_tick : used by the "sub" lexer only.
1385 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1391 PERL_ARGS_ASSERT_FORCE_WORD;
1393 start = SKIPSPACE1(start);
1395 if (isIDFIRST_lazy_if(s,UTF) ||
1396 (allow_pack && *s == ':') ||
1397 (allow_initial_tick && *s == '\'') )
1399 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1400 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1402 start_force(PL_curforce);
1404 curmad('X', newSVpvn(start,s-start));
1405 if (token == METHOD) {
1410 PL_expect = XOPERATOR;
1414 curmad('g', newSVpvs( "forced" ));
1415 NEXTVAL_NEXTTOKE.opval
1416 = (OP*)newSVOP(OP_CONST,0,
1417 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1418 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1426 * Called when the lexer wants $foo *foo &foo etc, but the program
1427 * text only contains the "foo" portion. The first argument is a pointer
1428 * to the "foo", and the second argument is the type symbol to prefix.
1429 * Forces the next token to be a "WORD".
1430 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1434 S_force_ident(pTHX_ register const char *s, int kind)
1438 PERL_ARGS_ASSERT_FORCE_IDENT;
1441 const STRLEN len = strlen(s);
1442 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1443 start_force(PL_curforce);
1444 NEXTVAL_NEXTTOKE.opval = o;
1447 o->op_private = OPpCONST_ENTERED;
1448 /* XXX see note in pp_entereval() for why we forgo typo
1449 warnings if the symbol must be introduced in an eval.
1451 gv_fetchpvn_flags(s, len,
1452 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1454 kind == '$' ? SVt_PV :
1455 kind == '@' ? SVt_PVAV :
1456 kind == '%' ? SVt_PVHV :
1464 Perl_str_to_version(pTHX_ SV *sv)
1469 const char *start = SvPV_const(sv,len);
1470 const char * const end = start + len;
1471 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1473 PERL_ARGS_ASSERT_STR_TO_VERSION;
1475 while (start < end) {
1479 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1484 retval += ((NV)n)/nshift;
1493 * Forces the next token to be a version number.
1494 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1495 * and if "guessing" is TRUE, then no new token is created (and the caller
1496 * must use an alternative parsing method).
1500 S_force_version(pTHX_ char *s, int guessing)
1506 I32 startoff = s - SvPVX(PL_linestr);
1509 PERL_ARGS_ASSERT_FORCE_VERSION;
1517 while (isDIGIT(*d) || *d == '_' || *d == '.')
1521 start_force(PL_curforce);
1522 curmad('X', newSVpvn(s,d-s));
1525 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1527 s = scan_num(s, &pl_yylval);
1528 version = pl_yylval.opval;
1529 ver = cSVOPx(version)->op_sv;
1530 if (SvPOK(ver) && !SvNIOK(ver)) {
1531 SvUPGRADE(ver, SVt_PVNV);
1532 SvNV_set(ver, str_to_version(ver));
1533 SvNOK_on(ver); /* hint that it is a version */
1536 else if (guessing) {
1539 sv_free(PL_nextwhite); /* let next token collect whitespace */
1541 s = SvPVX(PL_linestr) + startoff;
1549 if (PL_madskills && !version) {
1550 sv_free(PL_nextwhite); /* let next token collect whitespace */
1552 s = SvPVX(PL_linestr) + startoff;
1555 /* NOTE: The parser sees the package name and the VERSION swapped */
1556 start_force(PL_curforce);
1557 NEXTVAL_NEXTTOKE.opval = version;
1565 * Tokenize a quoted string passed in as an SV. It finds the next
1566 * chunk, up to end of string or a backslash. It may make a new
1567 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1572 S_tokeq(pTHX_ SV *sv)
1576 register char *send;
1581 PERL_ARGS_ASSERT_TOKEQ;
1586 s = SvPV_force(sv, len);
1587 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1590 while (s < send && *s != '\\')
1595 if ( PL_hints & HINT_NEW_STRING ) {
1596 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
1600 if (s + 1 < send && (s[1] == '\\'))
1601 s++; /* all that, just for this */
1606 SvCUR_set(sv, d - SvPVX_const(sv));
1608 if ( PL_hints & HINT_NEW_STRING )
1609 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
1614 * Now come three functions related to double-quote context,
1615 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1616 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1617 * interact with PL_lex_state, and create fake ( ... ) argument lists
1618 * to handle functions and concatenation.
1619 * They assume that whoever calls them will be setting up a fake
1620 * join call, because each subthing puts a ',' after it. This lets
1623 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1625 * (I'm not sure whether the spurious commas at the end of lcfirst's
1626 * arguments and join's arguments are created or not).
1631 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1633 * Pattern matching will set PL_lex_op to the pattern-matching op to
1634 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
1636 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1638 * Everything else becomes a FUNC.
1640 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1641 * had an OP_CONST or OP_READLINE). This just sets us up for a
1642 * call to S_sublex_push().
1646 S_sublex_start(pTHX)
1649 register const I32 op_type = pl_yylval.ival;
1651 if (op_type == OP_NULL) {
1652 pl_yylval.opval = PL_lex_op;
1656 if (op_type == OP_CONST || op_type == OP_READLINE) {
1657 SV *sv = tokeq(PL_lex_stuff);
1659 if (SvTYPE(sv) == SVt_PVIV) {
1660 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1662 const char * const p = SvPV_const(sv, len);
1663 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
1667 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1668 PL_lex_stuff = NULL;
1669 /* Allow <FH> // "foo" */
1670 if (op_type == OP_READLINE)
1671 PL_expect = XTERMORDORDOR;
1674 else if (op_type == OP_BACKTICK && PL_lex_op) {
1675 /* readpipe() vas overriden */
1676 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1677 pl_yylval.opval = PL_lex_op;
1679 PL_lex_stuff = NULL;
1683 PL_sublex_info.super_state = PL_lex_state;
1684 PL_sublex_info.sub_inwhat = (U16)op_type;
1685 PL_sublex_info.sub_op = PL_lex_op;
1686 PL_lex_state = LEX_INTERPPUSH;
1690 pl_yylval.opval = PL_lex_op;
1700 * Create a new scope to save the lexing state. The scope will be
1701 * ended in S_sublex_done. Returns a '(', starting the function arguments
1702 * to the uc, lc, etc. found before.
1703 * Sets PL_lex_state to LEX_INTERPCONCAT.
1712 PL_lex_state = PL_sublex_info.super_state;
1713 SAVEBOOL(PL_lex_dojoin);
1714 SAVEI32(PL_lex_brackets);
1715 SAVEI32(PL_lex_casemods);
1716 SAVEI32(PL_lex_starts);
1717 SAVEI8(PL_lex_state);
1718 SAVEVPTR(PL_lex_inpat);
1719 SAVEI16(PL_lex_inwhat);
1720 SAVECOPLINE(PL_curcop);
1721 SAVEPPTR(PL_bufptr);
1722 SAVEPPTR(PL_bufend);
1723 SAVEPPTR(PL_oldbufptr);
1724 SAVEPPTR(PL_oldoldbufptr);
1725 SAVEPPTR(PL_last_lop);
1726 SAVEPPTR(PL_last_uni);
1727 SAVEPPTR(PL_linestart);
1728 SAVESPTR(PL_linestr);
1729 SAVEGENERICPV(PL_lex_brackstack);
1730 SAVEGENERICPV(PL_lex_casestack);
1732 PL_linestr = PL_lex_stuff;
1733 PL_lex_stuff = NULL;
1735 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1736 = SvPVX(PL_linestr);
1737 PL_bufend += SvCUR(PL_linestr);
1738 PL_last_lop = PL_last_uni = NULL;
1739 SAVEFREESV(PL_linestr);
1741 PL_lex_dojoin = FALSE;
1742 PL_lex_brackets = 0;
1743 Newx(PL_lex_brackstack, 120, char);
1744 Newx(PL_lex_casestack, 12, char);
1745 PL_lex_casemods = 0;
1746 *PL_lex_casestack = '\0';
1748 PL_lex_state = LEX_INTERPCONCAT;
1749 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1751 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1752 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1753 PL_lex_inpat = PL_sublex_info.sub_op;
1755 PL_lex_inpat = NULL;
1762 * Restores lexer state after a S_sublex_push.
1769 if (!PL_lex_starts++) {
1770 SV * const sv = newSVpvs("");
1771 if (SvUTF8(PL_linestr))
1773 PL_expect = XOPERATOR;
1774 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1778 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1779 PL_lex_state = LEX_INTERPCASEMOD;
1783 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1784 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1785 PL_linestr = PL_lex_repl;
1787 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1788 PL_bufend += SvCUR(PL_linestr);
1789 PL_last_lop = PL_last_uni = NULL;
1790 SAVEFREESV(PL_linestr);
1791 PL_lex_dojoin = FALSE;
1792 PL_lex_brackets = 0;
1793 PL_lex_casemods = 0;
1794 *PL_lex_casestack = '\0';
1796 if (SvEVALED(PL_lex_repl)) {
1797 PL_lex_state = LEX_INTERPNORMAL;
1799 /* we don't clear PL_lex_repl here, so that we can check later
1800 whether this is an evalled subst; that means we rely on the
1801 logic to ensure sublex_done() is called again only via the
1802 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1805 PL_lex_state = LEX_INTERPCONCAT;
1815 PL_endwhite = newSVpvs("");
1816 sv_catsv(PL_endwhite, PL_thiswhite);
1820 sv_setpvs(PL_thistoken,"");
1822 PL_realtokenstart = -1;
1826 PL_bufend = SvPVX(PL_linestr);
1827 PL_bufend += SvCUR(PL_linestr);
1828 PL_expect = XOPERATOR;
1829 PL_sublex_info.sub_inwhat = 0;
1837 Extracts a pattern, double-quoted string, or transliteration. This
1840 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1841 processing a pattern (PL_lex_inpat is true), a transliteration
1842 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1844 Returns a pointer to the character scanned up to. If this is
1845 advanced from the start pointer supplied (i.e. if anything was
1846 successfully parsed), will leave an OP for the substring scanned
1847 in pl_yylval. Caller must intuit reason for not parsing further
1848 by looking at the next characters herself.
1852 double-quoted style: \r and \n
1853 regexp special ones: \D \s
1856 case and quoting: \U \Q \E
1857 stops on @ and $, but not for $ as tail anchor
1859 In transliterations:
1860 characters are VERY literal, except for - not at the start or end
1861 of the string, which indicates a range. If the range is in bytes,
1862 scan_const expands the range to the full set of intermediate
1863 characters. If the range is in utf8, the hyphen is replaced with
1864 a certain range mark which will be handled by pmtrans() in op.c.
1866 In double-quoted strings:
1868 double-quoted style: \r and \n
1870 deprecated backrefs: \1 (in substitution replacements)
1871 case and quoting: \U \Q \E
1874 scan_const does *not* construct ops to handle interpolated strings.
1875 It stops processing as soon as it finds an embedded $ or @ variable
1876 and leaves it to the caller to work out what's going on.
1878 embedded arrays (whether in pattern or not) could be:
1879 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1881 $ in double-quoted strings must be the symbol of an embedded scalar.
1883 $ in pattern could be $foo or could be tail anchor. Assumption:
1884 it's a tail anchor if $ is the last thing in the string, or if it's
1885 followed by one of "()| \r\n\t"
1887 \1 (backreferences) are turned into $1
1889 The structure of the code is
1890 while (there's a character to process) {
1891 handle transliteration ranges
1892 skip regexp comments /(?#comment)/ and codes /(?{code})/
1893 skip #-initiated comments in //x patterns
1894 check for embedded arrays
1895 check for embedded scalars
1897 leave intact backslashes from leaveit (below)
1898 deprecate \1 in substitution replacements
1899 handle string-changing backslashes \l \U \Q \E, etc.
1900 switch (what was escaped) {
1901 handle \- in a transliteration (becomes a literal -)
1902 handle \132 (octal characters)
1903 handle \x15 and \x{1234} (hex characters)
1904 handle \N{name} (named characters)
1905 handle \cV (control characters)
1906 handle printf-style backslashes (\f, \r, \n, etc)
1909 } (end if backslash)
1910 handle regular character
1911 } (end while character to read)
1916 S_scan_const(pTHX_ char *start)
1919 register char *send = PL_bufend; /* end of the constant */
1920 SV *sv = newSV(send - start); /* sv for the constant. See
1921 note below on sizing. */
1922 register char *s = start; /* start of the constant */
1923 register char *d = SvPVX(sv); /* destination for copies */
1924 bool dorange = FALSE; /* are we in a translit range? */
1925 bool didrange = FALSE; /* did we just finish a range? */
1926 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1927 I32 this_utf8 = UTF; /* Is the source string assumed
1928 to be UTF8? But, this can
1929 show as true when the source
1930 isn't utf8, as for example
1931 when it is entirely composed
1934 /* Note on sizing: The scanned constant is placed into sv, which is
1935 * initialized by newSV() assuming one byte of output for every byte of
1936 * input. This routine expects newSV() to allocate an extra byte for a
1937 * trailing NUL, which this routine will append if it gets to the end of
1938 * the input. There may be more bytes of input than output (eg., \N{LATIN
1939 * CAPITAL LETTER A}), or more output than input if the constant ends up
1940 * recoded to utf8, but each time a construct is found that might increase
1941 * the needed size, SvGROW() is called. Its size parameter each time is
1942 * based on the best guess estimate at the time, namely the length used so
1943 * far, plus the length the current construct will occupy, plus room for
1944 * the trailing NUL, plus one byte for every input byte still unscanned */
1948 UV literal_endpoint = 0;
1949 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1952 PERL_ARGS_ASSERT_SCAN_CONST;
1954 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1955 /* If we are doing a trans and we know we want UTF8 set expectation */
1956 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1957 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1961 while (s < send || dorange) {
1962 /* get transliterations out of the way (they're most literal) */
1963 if (PL_lex_inwhat == OP_TRANS) {
1964 /* expand a range A-Z to the full set of characters. AIE! */
1966 I32 i; /* current expanded character */
1967 I32 min; /* first character in range */
1968 I32 max; /* last character in range */
1979 char * const c = (char*)utf8_hop((U8*)d, -1);
1983 *c = (char)UTF_TO_NATIVE(0xff);
1984 /* mark the range as done, and continue */
1990 i = d - SvPVX_const(sv); /* remember current offset */
1993 SvLEN(sv) + (has_utf8 ?
1994 (512 - UTF_CONTINUATION_MARK +
1997 /* How many two-byte within 0..255: 128 in UTF-8,
1998 * 96 in UTF-8-mod. */
2000 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2002 d = SvPVX(sv) + i; /* refresh d after realloc */
2006 for (j = 0; j <= 1; j++) {
2007 char * const c = (char*)utf8_hop((U8*)d, -1);
2008 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2014 max = (U8)0xff; /* only to \xff */
2015 uvmax = uv; /* \x{100} to uvmax */
2017 d = c; /* eat endpoint chars */
2022 d -= 2; /* eat the first char and the - */
2023 min = (U8)*d; /* first char in range */
2024 max = (U8)d[1]; /* last char in range */
2031 "Invalid range \"%c-%c\" in transliteration operator",
2032 (char)min, (char)max);
2036 if (literal_endpoint == 2 &&
2037 ((isLOWER(min) && isLOWER(max)) ||
2038 (isUPPER(min) && isUPPER(max)))) {
2040 for (i = min; i <= max; i++)
2042 *d++ = NATIVE_TO_NEED(has_utf8,i);
2044 for (i = min; i <= max; i++)
2046 *d++ = NATIVE_TO_NEED(has_utf8,i);
2051 for (i = min; i <= max; i++)
2054 const U8 ch = (U8)NATIVE_TO_UTF(i);
2055 if (UNI_IS_INVARIANT(ch))
2058 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2059 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2068 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2070 *d++ = (char)UTF_TO_NATIVE(0xff);
2072 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2076 /* mark the range as done, and continue */
2080 literal_endpoint = 0;
2085 /* range begins (ignore - as first or last char) */
2086 else if (*s == '-' && s+1 < send && s != start) {
2088 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2095 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2105 literal_endpoint = 0;
2106 native_range = TRUE;
2111 /* if we get here, we're not doing a transliteration */
2113 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2114 except for the last char, which will be done separately. */
2115 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2117 while (s+1 < send && *s != ')')
2118 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2120 else if (s[2] == '{' /* This should match regcomp.c */
2121 || (s[2] == '?' && s[3] == '{'))
2124 char *regparse = s + (s[2] == '{' ? 3 : 4);
2127 while (count && (c = *regparse)) {
2128 if (c == '\\' && regparse[1])
2136 if (*regparse != ')')
2137 regparse--; /* Leave one char for continuation. */
2138 while (s < regparse)
2139 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2143 /* likewise skip #-initiated comments in //x patterns */
2144 else if (*s == '#' && PL_lex_inpat &&
2145 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2146 while (s+1 < send && *s != '\n')
2147 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2150 /* check for embedded arrays
2151 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2153 else if (*s == '@' && s[1]) {
2154 if (isALNUM_lazy_if(s+1,UTF))
2156 if (strchr(":'{$", s[1]))
2158 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2159 break; /* in regexp, neither @+ nor @- are interpolated */
2162 /* check for embedded scalars. only stop if we're sure it's a
2165 else if (*s == '$') {
2166 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2168 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2170 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2171 "Possible unintended interpolation of $\\ in regex");
2173 break; /* in regexp, $ might be tail anchor */
2177 /* End of else if chain - OP_TRANS rejoin rest */
2180 if (*s == '\\' && s+1 < send) {
2183 /* deprecate \1 in strings and substitution replacements */
2184 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2185 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2187 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2192 /* string-change backslash escapes */
2193 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2197 /* skip any other backslash escapes in a pattern */
2198 else if (PL_lex_inpat) {
2199 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2200 goto default_action;
2203 /* if we get here, it's either a quoted -, or a digit */
2206 /* quoted - in transliterations */
2208 if (PL_lex_inwhat == OP_TRANS) {
2215 if ((isALPHA(*s) || isDIGIT(*s)))
2216 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2217 "Unrecognized escape \\%c passed through",
2219 /* default action is to copy the quoted character */
2220 goto default_action;
2223 /* eg. \132 indicates the octal constant 0x132 */
2224 case '0': case '1': case '2': case '3':
2225 case '4': case '5': case '6': case '7':
2229 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2232 goto NUM_ESCAPE_INSERT;
2234 /* eg. \x24 indicates the hex constant 0x24 */
2238 char* const e = strchr(s, '}');
2239 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2240 PERL_SCAN_DISALLOW_PREFIX;
2245 yyerror("Missing right brace on \\x{}");
2249 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2255 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2256 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2262 /* Insert oct, hex, or \N{U+...} escaped character. There will
2263 * always be enough room in sv since such escapes will be
2264 * longer than any UTF-8 sequence they can end up as, except if
2265 * they force us to recode the rest of the string into utf8 */
2267 /* Here uv is the ordinal of the next character being added in
2268 * unicode (converted from native). (It has to be done before
2269 * here because \N is interpreted as unicode, and oct and hex
2271 if (!UNI_IS_INVARIANT(uv)) {
2272 if (!has_utf8 && uv > 255) {
2273 /* Might need to recode whatever we have accumulated so
2274 * far if it contains any chars variant in utf8 or
2277 SvCUR_set(sv, d - SvPVX_const(sv));
2280 /* See Note on sizing above. */
2281 sv_utf8_upgrade_flags_grow(sv,
2282 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2283 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2284 d = SvPVX(sv) + SvCUR(sv);
2289 d = (char*)uvuni_to_utf8((U8*)d, uv);
2290 if (PL_lex_inwhat == OP_TRANS &&
2291 PL_sublex_info.sub_op) {
2292 PL_sublex_info.sub_op->op_private |=
2293 (PL_lex_repl ? OPpTRANS_FROM_UTF
2297 if (uv > 255 && !dorange)
2298 native_range = FALSE;
2310 /* \N{LATIN SMALL LETTER A} is a named character, and so is
2315 char* e = strchr(s, '}');
2321 yyerror("Missing right brace on \\N{}");
2325 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2326 /* \N{U+...} The ... is a unicode value even on EBCDIC
2328 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2329 PERL_SCAN_DISALLOW_PREFIX;
2332 uv = grok_hex(s, &len, &flags, NULL);
2333 if ( e > s && len != (STRLEN)(e - s) ) {
2337 goto NUM_ESCAPE_INSERT;
2339 res = newSVpvn(s + 1, e - s - 1);
2340 res = new_constant( NULL, 0, "charnames",
2341 res, NULL, s - 2, e - s + 3 );
2343 sv_utf8_upgrade(res);
2344 str = SvPV_const(res,len);
2345 #ifdef EBCDIC_NEVER_MIND
2346 /* charnames uses pack U and that has been
2347 * recently changed to do the below uni->native
2348 * mapping, so this would be redundant (and wrong,
2349 * the code point would be doubly converted).
2350 * But leave this in just in case the pack U change
2351 * gets revoked, but the semantics is still
2352 * desireable for charnames. --jhi */
2354 UV uv = utf8_to_uvchr((const U8*)str, 0);
2357 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2359 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2360 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2361 str = SvPV_const(res, len);
2365 /* If destination is not in utf8 but this new character is,
2366 * recode the dest to utf8 */
2367 if (!has_utf8 && SvUTF8(res)) {
2368 SvCUR_set(sv, d - SvPVX_const(sv));
2371 /* See Note on sizing above. */
2372 sv_utf8_upgrade_flags_grow(sv,
2373 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2374 len + (STRLEN)(send - s) + 1);
2375 d = SvPVX(sv) + SvCUR(sv);
2377 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2379 /* See Note on sizing above. (NOTE: SvCUR() is not set
2380 * correctly here). */
2381 const STRLEN off = d - SvPVX_const(sv);
2382 d = SvGROW(sv, off + len + (STRLEN)(send - s) + 1) + off;
2386 native_range = FALSE; /* \N{} is guessed to be Unicode */
2388 Copy(str, d, len, char);
2395 yyerror("Missing braces on \\N{}");
2398 /* \c is a control character */
2407 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2410 yyerror("Missing control char name in \\c");
2414 /* printf-style backslashes, formfeeds, newlines, etc */
2416 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2419 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2422 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2425 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2428 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2431 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2434 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2440 } /* end if (backslash) */
2447 /* If we started with encoded form, or already know we want it,
2448 then encode the next character */
2449 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
2453 /* One might think that it is wasted effort in the case of the
2454 * source being utf8 (this_utf8 == TRUE) to take the next character
2455 * in the source, convert it to an unsigned value, and then convert
2456 * it back again. But the source has not been validated here. The
2457 * routine that does the conversion checks for errors like
2460 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2461 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2463 SvCUR_set(sv, d - SvPVX_const(sv));
2466 /* See Note on sizing above. */
2467 sv_utf8_upgrade_flags_grow(sv,
2468 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2469 need + (STRLEN)(send - s) + 1);
2470 d = SvPVX(sv) + SvCUR(sv);
2472 } else if (need > len) {
2473 /* encoded value larger than old, may need extra space (NOTE:
2474 * SvCUR() is not set correctly here). See Note on sizing
2476 const STRLEN off = d - SvPVX_const(sv);
2477 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
2481 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2483 if (uv > 255 && !dorange)
2484 native_range = FALSE;
2488 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2490 } /* while loop to process each character */
2492 /* terminate the string and set up the sv */
2494 SvCUR_set(sv, d - SvPVX_const(sv));
2495 if (SvCUR(sv) >= SvLEN(sv))
2496 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2499 if (PL_encoding && !has_utf8) {
2500 sv_recode_to_utf8(sv, PL_encoding);
2506 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2507 PL_sublex_info.sub_op->op_private |=
2508 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2512 /* shrink the sv if we allocated more than we used */
2513 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2514 SvPV_shrink_to_cur(sv);
2517 /* return the substring (via pl_yylval) only if we parsed anything */
2518 if (s > PL_bufptr) {
2519 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
2520 const char *const key = PL_lex_inpat ? "qr" : "q";
2521 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
2525 if (PL_lex_inwhat == OP_TRANS) {
2528 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
2536 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
2539 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2546 * Returns TRUE if there's more to the expression (e.g., a subscript),
2549 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2551 * ->[ and ->{ return TRUE
2552 * { and [ outside a pattern are always subscripts, so return TRUE
2553 * if we're outside a pattern and it's not { or [, then return FALSE
2554 * if we're in a pattern and the first char is a {
2555 * {4,5} (any digits around the comma) returns FALSE
2556 * if we're in a pattern and the first char is a [
2558 * [SOMETHING] has a funky algorithm to decide whether it's a
2559 * character class or not. It has to deal with things like
2560 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2561 * anything else returns TRUE
2564 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2567 S_intuit_more(pTHX_ register char *s)
2571 PERL_ARGS_ASSERT_INTUIT_MORE;
2573 if (PL_lex_brackets)
2575 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2577 if (*s != '{' && *s != '[')
2582 /* In a pattern, so maybe we have {n,m}. */
2599 /* On the other hand, maybe we have a character class */
2602 if (*s == ']' || *s == '^')
2605 /* this is terrifying, and it works */
2606 int weight = 2; /* let's weigh the evidence */
2608 unsigned char un_char = 255, last_un_char;
2609 const char * const send = strchr(s,']');
2610 char tmpbuf[sizeof PL_tokenbuf * 4];
2612 if (!send) /* has to be an expression */
2615 Zero(seen,256,char);
2618 else if (isDIGIT(*s)) {
2620 if (isDIGIT(s[1]) && s[2] == ']')
2626 for (; s < send; s++) {
2627 last_un_char = un_char;
2628 un_char = (unsigned char)*s;
2633 weight -= seen[un_char] * 10;
2634 if (isALNUM_lazy_if(s+1,UTF)) {
2636 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2637 len = (int)strlen(tmpbuf);
2638 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2643 else if (*s == '$' && s[1] &&
2644 strchr("[#!%*<>()-=",s[1])) {
2645 if (/*{*/ strchr("])} =",s[2]))
2654 if (strchr("wds]",s[1]))
2656 else if (seen[(U8)'\''] || seen[(U8)'"'])
2658 else if (strchr("rnftbxcav",s[1]))
2660 else if (isDIGIT(s[1])) {
2662 while (s[1] && isDIGIT(s[1]))
2672 if (strchr("aA01! ",last_un_char))
2674 if (strchr("zZ79~",s[1]))
2676 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2677 weight -= 5; /* cope with negative subscript */
2680 if (!isALNUM(last_un_char)
2681 && !(last_un_char == '$' || last_un_char == '@'
2682 || last_un_char == '&')
2683 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2688 if (keyword(tmpbuf, d - tmpbuf, 0))
2691 if (un_char == last_un_char + 1)
2693 weight -= seen[un_char];
2698 if (weight >= 0) /* probably a character class */
2708 * Does all the checking to disambiguate
2710 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2711 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2713 * First argument is the stuff after the first token, e.g. "bar".
2715 * Not a method if bar is a filehandle.
2716 * Not a method if foo is a subroutine prototyped to take a filehandle.
2717 * Not a method if it's really "Foo $bar"
2718 * Method if it's "foo $bar"
2719 * Not a method if it's really "print foo $bar"
2720 * Method if it's really "foo package::" (interpreted as package->foo)
2721 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2722 * Not a method if bar is a filehandle or package, but is quoted with
2727 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2730 char *s = start + (*start == '$');
2731 char tmpbuf[sizeof PL_tokenbuf];
2738 PERL_ARGS_ASSERT_INTUIT_METHOD;
2741 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2745 const char *proto = SvPVX_const(cv);
2756 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2757 /* start is the beginning of the possible filehandle/object,
2758 * and s is the end of it
2759 * tmpbuf is a copy of it
2762 if (*start == '$') {
2763 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
2764 isUPPER(*PL_tokenbuf))
2767 len = start - SvPVX(PL_linestr);
2771 start = SvPVX(PL_linestr) + len;
2775 return *s == '(' ? FUNCMETH : METHOD;
2777 if (!keyword(tmpbuf, len, 0)) {
2778 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2782 soff = s - SvPVX(PL_linestr);
2786 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2787 if (indirgv && GvCVu(indirgv))
2789 /* filehandle or package name makes it a method */
2790 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2792 soff = s - SvPVX(PL_linestr);
2795 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2796 return 0; /* no assumptions -- "=>" quotes bearword */
2798 start_force(PL_curforce);
2799 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2800 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
2801 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2803 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2808 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2810 return *s == '(' ? FUNCMETH : METHOD;
2816 /* Encoded script support. filter_add() effectively inserts a
2817 * 'pre-processing' function into the current source input stream.
2818 * Note that the filter function only applies to the current source file
2819 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2821 * The datasv parameter (which may be NULL) can be used to pass
2822 * private data to this instance of the filter. The filter function
2823 * can recover the SV using the FILTER_DATA macro and use it to
2824 * store private buffers and state information.
2826 * The supplied datasv parameter is upgraded to a PVIO type
2827 * and the IoDIRP/IoANY field is used to store the function pointer,
2828 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2829 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2830 * private use must be set using malloc'd pointers.
2834 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2843 if (!PL_rsfp_filters)
2844 PL_rsfp_filters = newAV();
2847 SvUPGRADE(datasv, SVt_PVIO);
2848 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2849 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2850 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2851 FPTR2DPTR(void *, IoANY(datasv)),
2852 SvPV_nolen(datasv)));
2853 av_unshift(PL_rsfp_filters, 1);
2854 av_store(PL_rsfp_filters, 0, datasv) ;
2859 /* Delete most recently added instance of this filter function. */
2861 Perl_filter_del(pTHX_ filter_t funcp)
2866 PERL_ARGS_ASSERT_FILTER_DEL;
2869 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2870 FPTR2DPTR(void*, funcp)));
2872 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2874 /* if filter is on top of stack (usual case) just pop it off */
2875 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2876 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2877 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2878 IoANY(datasv) = (void *)NULL;
2879 sv_free(av_pop(PL_rsfp_filters));
2883 /* we need to search for the correct entry and clear it */
2884 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2888 /* Invoke the idxth filter function for the current rsfp. */
2889 /* maxlen 0 = read one text line */
2891 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2896 /* This API is bad. It should have been using unsigned int for maxlen.
2897 Not sure if we want to change the API, but if not we should sanity
2898 check the value here. */
2899 const unsigned int correct_length
2908 PERL_ARGS_ASSERT_FILTER_READ;
2910 if (!PL_parser || !PL_rsfp_filters)
2912 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2913 /* Provide a default input filter to make life easy. */
2914 /* Note that we append to the line. This is handy. */
2915 DEBUG_P(PerlIO_printf(Perl_debug_log,
2916 "filter_read %d: from rsfp\n", idx));
2917 if (correct_length) {
2920 const int old_len = SvCUR(buf_sv);
2922 /* ensure buf_sv is large enough */
2923 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2924 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2925 correct_length)) <= 0) {
2926 if (PerlIO_error(PL_rsfp))
2927 return -1; /* error */
2929 return 0 ; /* end of file */
2931 SvCUR_set(buf_sv, old_len + len) ;
2934 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2935 if (PerlIO_error(PL_rsfp))
2936 return -1; /* error */
2938 return 0 ; /* end of file */
2941 return SvCUR(buf_sv);
2943 /* Skip this filter slot if filter has been deleted */
2944 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2945 DEBUG_P(PerlIO_printf(Perl_debug_log,
2946 "filter_read %d: skipped (filter deleted)\n",
2948 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2950 /* Get function pointer hidden within datasv */
2951 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2952 DEBUG_P(PerlIO_printf(Perl_debug_log,
2953 "filter_read %d: via function %p (%s)\n",
2954 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2955 /* Call function. The function is expected to */
2956 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2957 /* Return: <0:error, =0:eof, >0:not eof */
2958 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2962 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2966 PERL_ARGS_ASSERT_FILTER_GETS;
2968 #ifdef PERL_CR_FILTER
2969 if (!PL_rsfp_filters) {
2970 filter_add(S_cr_textfilter,NULL);
2973 if (PL_rsfp_filters) {
2975 SvCUR_set(sv, 0); /* start with empty line */
2976 if (FILTER_READ(0, sv, 0) > 0)
2977 return ( SvPVX(sv) ) ;
2982 return (sv_gets(sv, fp, append));
2986 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
2991 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
2993 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2997 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2998 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3000 return GvHV(gv); /* Foo:: */
3003 /* use constant CLASS => 'MyClass' */
3004 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3005 if (gv && GvCV(gv)) {
3006 SV * const sv = cv_const_sv(GvCV(gv));
3008 pkgname = SvPV_const(sv, len);
3011 return gv_stashpvn(pkgname, len, 0);
3015 * S_readpipe_override
3016 * Check whether readpipe() is overriden, and generates the appropriate
3017 * optree, provided sublex_start() is called afterwards.
3020 S_readpipe_override(pTHX)
3023 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3024 pl_yylval.ival = OP_BACKTICK;
3026 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3028 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3029 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3030 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3032 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3033 append_elem(OP_LIST,
3034 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3035 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3042 * The intent of this yylex wrapper is to minimize the changes to the
3043 * tokener when we aren't interested in collecting madprops. It remains
3044 * to be seen how successful this strategy will be...
3051 char *s = PL_bufptr;
3053 /* make sure PL_thiswhite is initialized */
3057 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3058 if (PL_pending_ident)
3059 return S_pending_ident(aTHX);
3061 /* previous token ate up our whitespace? */
3062 if (!PL_lasttoke && PL_nextwhite) {
3063 PL_thiswhite = PL_nextwhite;
3067 /* isolate the token, and figure out where it is without whitespace */
3068 PL_realtokenstart = -1;
3072 assert(PL_curforce < 0);
3074 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3075 if (!PL_thistoken) {
3076 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3077 PL_thistoken = newSVpvs("");
3079 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3080 PL_thistoken = newSVpvn(tstart, s - tstart);
3083 if (PL_thismad) /* install head */
3084 CURMAD('X', PL_thistoken);
3087 /* last whitespace of a sublex? */
3088 if (optype == ')' && PL_endwhite) {
3089 CURMAD('X', PL_endwhite);
3094 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3095 if (!PL_thiswhite && !PL_endwhite && !optype) {
3096 sv_free(PL_thistoken);
3101 /* put off final whitespace till peg */
3102 if (optype == ';' && !PL_rsfp) {
3103 PL_nextwhite = PL_thiswhite;
3106 else if (PL_thisopen) {
3107 CURMAD('q', PL_thisopen);
3109 sv_free(PL_thistoken);
3113 /* Store actual token text as madprop X */
3114 CURMAD('X', PL_thistoken);
3118 /* add preceding whitespace as madprop _ */
3119 CURMAD('_', PL_thiswhite);
3123 /* add quoted material as madprop = */
3124 CURMAD('=', PL_thisstuff);
3128 /* add terminating quote as madprop Q */
3129 CURMAD('Q', PL_thisclose);
3133 /* special processing based on optype */
3137 /* opval doesn't need a TOKEN since it can already store mp */
3147 if (pl_yylval.opval)
3148 append_madprops(PL_thismad, pl_yylval.opval, 0);
3156 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3165 /* remember any fake bracket that lexer is about to discard */
3166 if (PL_lex_brackets == 1 &&
3167 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3170 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3173 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3174 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3177 break; /* don't bother looking for trailing comment */
3186 /* attach a trailing comment to its statement instead of next token */
3190 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3192 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3194 if (*s == '\n' || *s == '#') {
3195 while (s < PL_bufend && *s != '\n')
3199 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3200 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3217 /* Create new token struct. Note: opvals return early above. */
3218 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
3225 S_tokenize_use(pTHX_ int is_use, char *s) {
3228 PERL_ARGS_ASSERT_TOKENIZE_USE;
3230 if (PL_expect != XSTATE)
3231 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3232 is_use ? "use" : "no"));
3234 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3235 s = force_version(s, TRUE);
3236 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3237 start_force(PL_curforce);
3238 NEXTVAL_NEXTTOKE.opval = NULL;
3241 else if (*s == 'v') {
3242 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3243 s = force_version(s, FALSE);
3247 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3248 s = force_version(s, FALSE);
3250 pl_yylval.ival = is_use;
3254 static const char* const exp_name[] =
3255 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3256 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3263 Works out what to call the token just pulled out of the input
3264 stream. The yacc parser takes care of taking the ops we return and
3265 stitching them into a tree.
3271 if read an identifier
3272 if we're in a my declaration
3273 croak if they tried to say my($foo::bar)
3274 build the ops for a my() declaration
3275 if it's an access to a my() variable
3276 are we in a sort block?
3277 croak if my($a); $a <=> $b
3278 build ops for access to a my() variable
3279 if in a dq string, and they've said @foo and we can't find @foo
3281 build ops for a bareword
3282 if we already built the token before, use it.
3287 #pragma segment Perl_yylex
3293 register char *s = PL_bufptr;
3298 /* orig_keyword, gvp, and gv are initialized here because
3299 * jump to the label just_a_word_zero can bypass their
3300 * initialization later. */
3301 I32 orig_keyword = 0;
3306 SV* tmp = newSVpvs("");
3307 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3308 (IV)CopLINE(PL_curcop),
3309 lex_state_names[PL_lex_state],
3310 exp_name[PL_expect],
3311 pv_display(tmp, s, strlen(s), 0, 60));
3314 /* check if there's an identifier for us to look at */
3315 if (PL_pending_ident)
3316 return REPORT(S_pending_ident(aTHX));
3318 /* no identifier pending identification */
3320 switch (PL_lex_state) {
3322 case LEX_NORMAL: /* Some compilers will produce faster */
3323 case LEX_INTERPNORMAL: /* code if we comment these out. */
3327 /* when we've already built the next token, just pull it out of the queue */
3331 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
3333 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3334 PL_nexttoke[PL_lasttoke].next_mad = 0;
3335 if (PL_thismad && PL_thismad->mad_key == '_') {
3336 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
3337 PL_thismad->mad_val = 0;
3338 mad_free(PL_thismad);
3343 PL_lex_state = PL_lex_defer;
3344 PL_expect = PL_lex_expect;
3345 PL_lex_defer = LEX_NORMAL;
3346 if (!PL_nexttoke[PL_lasttoke].next_type)
3351 pl_yylval = PL_nextval[PL_nexttoke];
3353 PL_lex_state = PL_lex_defer;
3354 PL_expect = PL_lex_expect;
3355 PL_lex_defer = LEX_NORMAL;
3359 /* FIXME - can these be merged? */
3360 return(PL_nexttoke[PL_lasttoke].next_type);
3362 return REPORT(PL_nexttype[PL_nexttoke]);
3365 /* interpolated case modifiers like \L \U, including \Q and \E.
3366 when we get here, PL_bufptr is at the \
3368 case LEX_INTERPCASEMOD:
3370 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3371 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3373 /* handle \E or end of string */
3374 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3376 if (PL_lex_casemods) {
3377 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3378 PL_lex_casestack[PL_lex_casemods] = '\0';
3380 if (PL_bufptr != PL_bufend
3381 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3383 PL_lex_state = LEX_INTERPCONCAT;
3386 PL_thistoken = newSVpvs("\\E");
3392 while (PL_bufptr != PL_bufend &&
3393 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3395 PL_thiswhite = newSVpvs("");
3396 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3400 if (PL_bufptr != PL_bufend)
3403 PL_lex_state = LEX_INTERPCONCAT;
3407 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3408 "### Saw case modifier\n"); });
3410 if (s[1] == '\\' && s[2] == 'E') {
3413 PL_thiswhite = newSVpvs("");
3414 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3417 PL_lex_state = LEX_INTERPCONCAT;
3422 if (!PL_madskills) /* when just compiling don't need correct */
3423 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3424 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3425 if ((*s == 'L' || *s == 'U') &&
3426 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3427 PL_lex_casestack[--PL_lex_casemods] = '\0';
3430 if (PL_lex_casemods > 10)
3431 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3432 PL_lex_casestack[PL_lex_casemods++] = *s;
3433 PL_lex_casestack[PL_lex_casemods] = '\0';
3434 PL_lex_state = LEX_INTERPCONCAT;
3435 start_force(PL_curforce);
3436 NEXTVAL_NEXTTOKE.ival = 0;
3438 start_force(PL_curforce);
3440 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3442 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3444 NEXTVAL_NEXTTOKE.ival = OP_LC;
3446 NEXTVAL_NEXTTOKE.ival = OP_UC;
3448 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3450 Perl_croak(aTHX_ "panic: yylex");
3452 SV* const tmpsv = newSVpvs("\\ ");
3453 /* replace the space with the character we want to escape
3455 SvPVX(tmpsv)[1] = *s;
3461 if (PL_lex_starts) {
3467 sv_free(PL_thistoken);
3468 PL_thistoken = newSVpvs("");
3471 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3472 if (PL_lex_casemods == 1 && PL_lex_inpat)
3481 case LEX_INTERPPUSH:
3482 return REPORT(sublex_push());
3484 case LEX_INTERPSTART:
3485 if (PL_bufptr == PL_bufend)
3486 return REPORT(sublex_done());
3487 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3488 "### Interpolated variable\n"); });
3490 PL_lex_dojoin = (*PL_bufptr == '@');
3491 PL_lex_state = LEX_INTERPNORMAL;
3492 if (PL_lex_dojoin) {
3493 start_force(PL_curforce);
3494 NEXTVAL_NEXTTOKE.ival = 0;
3496 start_force(PL_curforce);
3497 force_ident("\"", '$');
3498 start_force(PL_curforce);
3499 NEXTVAL_NEXTTOKE.ival = 0;
3501 start_force(PL_curforce);
3502 NEXTVAL_NEXTTOKE.ival = 0;
3504 start_force(PL_curforce);
3505 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3508 if (PL_lex_starts++) {
3513 sv_free(PL_thistoken);
3514 PL_thistoken = newSVpvs("");
3517 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3518 if (!PL_lex_casemods && PL_lex_inpat)
3525 case LEX_INTERPENDMAYBE:
3526 if (intuit_more(PL_bufptr)) {
3527 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3533 if (PL_lex_dojoin) {
3534 PL_lex_dojoin = FALSE;
3535 PL_lex_state = LEX_INTERPCONCAT;
3539 sv_free(PL_thistoken);
3540 PL_thistoken = newSVpvs("");
3545 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3546 && SvEVALED(PL_lex_repl))
3548 if (PL_bufptr != PL_bufend)
3549 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3553 case LEX_INTERPCONCAT:
3555 if (PL_lex_brackets)
3556 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3558 if (PL_bufptr == PL_bufend)
3559 return REPORT(sublex_done());
3561 if (SvIVX(PL_linestr) == '\'') {
3562 SV *sv = newSVsv(PL_linestr);
3565 else if ( PL_hints & HINT_NEW_RE )
3566 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
3567 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3571 s = scan_const(PL_bufptr);
3573 PL_lex_state = LEX_INTERPCASEMOD;
3575 PL_lex_state = LEX_INTERPSTART;
3578 if (s != PL_bufptr) {
3579 start_force(PL_curforce);
3581 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3583 NEXTVAL_NEXTTOKE = pl_yylval;
3586 if (PL_lex_starts++) {
3590 sv_free(PL_thistoken);
3591 PL_thistoken = newSVpvs("");
3594 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3595 if (!PL_lex_casemods && PL_lex_inpat)
3608 PL_lex_state = LEX_NORMAL;
3609 s = scan_formline(PL_bufptr);
3610 if (!PL_lex_formbrack)
3616 PL_oldoldbufptr = PL_oldbufptr;
3622 sv_free(PL_thistoken);
3625 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3629 if (isIDFIRST_lazy_if(s,UTF))
3632 unsigned char c = *s;
3633 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
3634 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
3635 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
3640 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
3644 goto fake_eof; /* emulate EOF on ^D or ^Z */
3653 if (PL_lex_brackets) {
3654 yyerror((const char *)
3656 ? "Format not terminated"
3657 : "Missing right curly or square bracket"));
3659 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3660 "### Tokener got EOF\n");
3664 if (s++ < PL_bufend)
3665 goto retry; /* ignore stray nulls */
3668 if (!PL_in_eval && !PL_preambled) {
3669 PL_preambled = TRUE;
3675 /* Generate a string of Perl code to load the debugger.
3676 * If PERL5DB is set, it will return the contents of that,
3677 * otherwise a compile-time require of perl5db.pl. */
3679 const char * const pdb = PerlEnv_getenv("PERL5DB");
3682 sv_setpv(PL_linestr, pdb);
3683 sv_catpvs(PL_linestr,";");
3685 SETERRNO(0,SS_NORMAL);
3686 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
3689 sv_setpvs(PL_linestr,"");
3690 if (PL_preambleav) {
3691 SV **svp = AvARRAY(PL_preambleav);
3692 SV **const end = svp + AvFILLp(PL_preambleav);
3694 sv_catsv(PL_linestr, *svp);
3696 sv_catpvs(PL_linestr, ";");
3698 sv_free(MUTABLE_SV(PL_preambleav));
3699 PL_preambleav = NULL;
3702 sv_catpvs(PL_linestr,
3703 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
3704 if (PL_minus_n || PL_minus_p) {
3705 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3707 sv_catpvs(PL_linestr,"chomp;");
3710 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3711 || *PL_splitstr == '"')
3712 && strchr(PL_splitstr + 1, *PL_splitstr))
3713 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3715 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3716 bytes can be used as quoting characters. :-) */
3717 const char *splits = PL_splitstr;
3718 sv_catpvs(PL_linestr, "our @F=split(q\0");
3721 if (*splits == '\\')
3722 sv_catpvn(PL_linestr, splits, 1);
3723 sv_catpvn(PL_linestr, splits, 1);
3724 } while (*splits++);
3725 /* This loop will embed the trailing NUL of
3726 PL_linestr as the last thing it does before
3728 sv_catpvs(PL_linestr, ");");
3732 sv_catpvs(PL_linestr,"our @F=split(' ');");
3735 sv_catpvs(PL_linestr, "\n");
3736 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3737 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3738 PL_last_lop = PL_last_uni = NULL;
3739 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
3740 update_debugger_info(PL_linestr, NULL, 0);
3744 bof = PL_rsfp ? TRUE : FALSE;
3745 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3748 PL_realtokenstart = -1;
3751 if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3752 PerlIO_clearerr(PL_rsfp);
3754 (void)PerlIO_close(PL_rsfp);
3756 PL_doextract = FALSE;
3758 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3764 sv_setpvs(PL_linestr, ";}continue{print;}");
3766 sv_setpvs(PL_linestr, ";}");
3767 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3768 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3769 PL_last_lop = PL_last_uni = NULL;
3770 PL_minus_n = PL_minus_p = 0;
3773 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3774 PL_last_lop = PL_last_uni = NULL;
3775 sv_setpvs(PL_linestr,"");
3776 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3778 /* If it looks like the start of a BOM or raw UTF-16,
3779 * check if it in fact is. */
3785 #ifdef PERLIO_IS_STDIO
3786 # ifdef __GNU_LIBRARY__
3787 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3788 # define FTELL_FOR_PIPE_IS_BROKEN
3792 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3793 # define FTELL_FOR_PIPE_IS_BROKEN
3798 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3800 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3801 s = swallow_bom((U8*)s);
3805 /* Incest with pod. */
3808 sv_catsv(PL_thiswhite, PL_linestr);
3810 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3811 sv_setpvs(PL_linestr, "");
3812 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3813 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3814 PL_last_lop = PL_last_uni = NULL;
3815 PL_doextract = FALSE;
3819 } while (PL_doextract);
3820 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3821 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
3822 update_debugger_info(PL_linestr, NULL, 0);
3823 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3824 PL_last_lop = PL_last_uni = NULL;
3825 if (CopLINE(PL_curcop) == 1) {
3826 while (s < PL_bufend && isSPACE(*s))
3828 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3832 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3836 if (*s == '#' && *(s+1) == '!')
3838 #ifdef ALTERNATE_SHEBANG
3840 static char const as[] = ALTERNATE_SHEBANG;
3841 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3842 d = s + (sizeof(as) - 1);
3844 #endif /* ALTERNATE_SHEBANG */
3853 while (*d && !isSPACE(*d))
3857 #ifdef ARG_ZERO_IS_SCRIPT
3858 if (ipathend > ipath) {
3860 * HP-UX (at least) sets argv[0] to the script name,
3861 * which makes $^X incorrect. And Digital UNIX and Linux,
3862 * at least, set argv[0] to the basename of the Perl
3863 * interpreter. So, having found "#!", we'll set it right.
3865 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3867 assert(SvPOK(x) || SvGMAGICAL(x));
3868 if (sv_eq(x, CopFILESV(PL_curcop))) {
3869 sv_setpvn(x, ipath, ipathend - ipath);
3875 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3876 const char * const lstart = SvPV_const(x,llen);
3878 bstart += blen - llen;
3879 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3880 sv_setpvn(x, ipath, ipathend - ipath);
3885 TAINT_NOT; /* $^X is always tainted, but that's OK */
3887 #endif /* ARG_ZERO_IS_SCRIPT */
3892 d = instr(s,"perl -");
3894 d = instr(s,"perl");
3896 /* avoid getting into infinite loops when shebang
3897 * line contains "Perl" rather than "perl" */
3899 for (d = ipathend-4; d >= ipath; --d) {
3900 if ((*d == 'p' || *d == 'P')
3901 && !ibcmp(d, "perl", 4))
3911 #ifdef ALTERNATE_SHEBANG
3913 * If the ALTERNATE_SHEBANG on this system starts with a
3914 * character that can be part of a Perl expression, then if
3915 * we see it but not "perl", we're probably looking at the
3916 * start of Perl code, not a request to hand off to some
3917 * other interpreter. Similarly, if "perl" is there, but
3918 * not in the first 'word' of the line, we assume the line
3919 * contains the start of the Perl program.
3921 if (d && *s != '#') {
3922 const char *c = ipath;
3923 while (*c && !strchr("; \t\r\n\f\v#", *c))
3926 d = NULL; /* "perl" not in first word; ignore */
3928 *s = '#'; /* Don't try to parse shebang line */
3930 #endif /* ALTERNATE_SHEBANG */
3935 !instr(s,"indir") &&
3936 instr(PL_origargv[0],"perl"))
3943 while (s < PL_bufend && isSPACE(*s))
3945 if (s < PL_bufend) {
3946 Newx(newargv,PL_origargc+3,char*);
3948 while (s < PL_bufend && !isSPACE(*s))
3951 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3954 newargv = PL_origargv;
3957 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3959 Perl_croak(aTHX_ "Can't exec %s", ipath);
3962 while (*d && !isSPACE(*d))
3964 while (SPACE_OR_TAB(*d))
3968 const bool switches_done = PL_doswitches;
3969 const U32 oldpdb = PL_perldb;
3970 const bool oldn = PL_minus_n;
3971 const bool oldp = PL_minus_p;
3975 bool baduni = FALSE;
3977 const char *d2 = d1 + 1;
3978 if (parse_unicode_opts((const char **)&d2)
3982 if (baduni || *d1 == 'M' || *d1 == 'm') {
3983 const char * const m = d1;
3984 while (*d1 && !isSPACE(*d1))
3986 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3989 d1 = moreswitches(d1);
3991 if (PL_doswitches && !switches_done) {
3992 int argc = PL_origargc;
3993 char **argv = PL_origargv;
3996 } while (argc && argv[0][0] == '-' && argv[0][1]);
3997 init_argv_symbols(argc,argv);
3999 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4000 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4001 /* if we have already added "LINE: while (<>) {",
4002 we must not do it again */
4004 sv_setpvs(PL_linestr, "");
4005 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4006 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4007 PL_last_lop = PL_last_uni = NULL;
4008 PL_preambled = FALSE;
4009 if (PERLDB_LINE || PERLDB_SAVESRC)
4010 (void)gv_fetchfile(PL_origfilename);
4017 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4019 PL_lex_state = LEX_FORMLINE;
4024 #ifdef PERL_STRICT_CR
4025 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4027 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4029 case ' ': case '\t': case '\f': case 013:
4031 PL_realtokenstart = -1;
4033 PL_thiswhite = newSVpvs("");
4034 sv_catpvn(PL_thiswhite, s, 1);
4041 PL_realtokenstart = -1;
4045 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4046 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4047 /* handle eval qq[#line 1 "foo"\n ...] */
4048 CopLINE_dec(PL_curcop);
4051 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4053 if (!PL_in_eval || PL_rsfp)
4058 while (d < PL_bufend && *d != '\n')
4062 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4063 Perl_croak(aTHX_ "panic: input overflow");
4066 PL_thiswhite = newSVpvn(s, d - s);
4071 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4073 PL_lex_state = LEX_FORMLINE;
4079 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4080 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4083 TOKEN(PEG); /* make sure any #! line is accessible */
4088 /* if (PL_madskills && PL_lex_formbrack) { */
4090 while (d < PL_bufend && *d != '\n')
4094 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4095 Perl_croak(aTHX_ "panic: input overflow");
4096 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4098 PL_thiswhite = newSVpvs("");
4099 if (CopLINE(PL_curcop) == 1) {
4100 sv_setpvs(PL_thiswhite, "");
4103 sv_catpvn(PL_thiswhite, s, d - s);
4117 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4125 while (s < PL_bufend && SPACE_OR_TAB(*s))
4128 if (strnEQ(s,"=>",2)) {
4129 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4130 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4131 OPERATOR('-'); /* unary minus */
4133 PL_last_uni = PL_oldbufptr;
4135 case 'r': ftst = OP_FTEREAD; break;
4136 case 'w': ftst = OP_FTEWRITE; break;
4137 case 'x': ftst = OP_FTEEXEC; break;
4138 case 'o': ftst = OP_FTEOWNED; break;
4139 case 'R': ftst = OP_FTRREAD; break;
4140 case 'W': ftst = OP_FTRWRITE; break;
4141 case 'X': ftst = OP_FTREXEC; break;
4142 case 'O': ftst = OP_FTROWNED; break;
4143 case 'e': ftst = OP_FTIS; break;
4144 case 'z': ftst = OP_FTZERO; break;
4145 case 's': ftst = OP_FTSIZE; break;
4146 case 'f': ftst = OP_FTFILE; break;
4147 case 'd': ftst = OP_FTDIR; break;
4148 case 'l': ftst = OP_FTLINK; break;
4149 case 'p': ftst = OP_FTPIPE; break;
4150 case 'S': ftst = OP_FTSOCK; break;
4151 case 'u': ftst = OP_FTSUID; break;
4152 case 'g': ftst = OP_FTSGID; break;
4153 case 'k': ftst = OP_FTSVTX; break;
4154 case 'b': ftst = OP_FTBLK; break;
4155 case 'c': ftst = OP_FTCHR; break;
4156 case 't': ftst = OP_FTTTY; break;
4157 case 'T': ftst = OP_FTTEXT; break;
4158 case 'B': ftst = OP_FTBINARY; break;
4159 case 'M': case 'A': case 'C':
4160 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4162 case 'M': ftst = OP_FTMTIME; break;
4163 case 'A': ftst = OP_FTATIME; break;
4164 case 'C': ftst = OP_FTCTIME; break;
4172 PL_last_lop_op = (OPCODE)ftst;
4173 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4174 "### Saw file test %c\n", (int)tmp);
4179 /* Assume it was a minus followed by a one-letter named
4180 * subroutine call (or a -bareword), then. */
4181 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4182 "### '-%c' looked like a file test but was not\n",
4189 const char tmp = *s++;
4192 if (PL_expect == XOPERATOR)
4197 else if (*s == '>') {
4200 if (isIDFIRST_lazy_if(s,UTF)) {
4201 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4209 if (PL_expect == XOPERATOR)
4212 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4214 OPERATOR('-'); /* unary minus */
4220 const char tmp = *s++;
4223 if (PL_expect == XOPERATOR)
4228 if (PL_expect == XOPERATOR)
4231 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4238 if (PL_expect != XOPERATOR) {
4239 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4240 PL_expect = XOPERATOR;
4241 force_ident(PL_tokenbuf, '*');
4254 if (PL_expect == XOPERATOR) {
4258 PL_tokenbuf[0] = '%';
4259 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4260 sizeof PL_tokenbuf - 1, FALSE);
4261 if (!PL_tokenbuf[1]) {
4264 PL_pending_ident = '%';
4273 const char tmp = *s++;
4278 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4285 const char tmp = *s++;
4291 goto just_a_word_zero_gv;
4294 switch (PL_expect) {
4300 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4302 PL_bufptr = s; /* update in case we back off */
4308 PL_expect = XTERMBLOCK;
4311 stuffstart = s - SvPVX(PL_linestr) - 1;
4315 while (isIDFIRST_lazy_if(s,UTF)) {
4318 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4319 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4320 if (tmp < 0) tmp = -tmp;
4335 sv = newSVpvn(s, len);
4337 d = scan_str(d,TRUE,TRUE);
4339 /* MUST advance bufptr here to avoid bogus
4340 "at end of line" context messages from yyerror().
4342 PL_bufptr = s + len;
4343 yyerror("Unterminated attribute parameter in attribute list");
4347 return REPORT(0); /* EOF indicator */
4351 sv_catsv(sv, PL_lex_stuff);
4352 attrs = append_elem(OP_LIST, attrs,
4353 newSVOP(OP_CONST, 0, sv));
4354 SvREFCNT_dec(PL_lex_stuff);
4355 PL_lex_stuff = NULL;
4358 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4360 if (PL_in_my == KEY_our) {
4361 deprecate(":unique");
4364 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4367 /* NOTE: any CV attrs applied here need to be part of
4368 the CVf_BUILTIN_ATTRS define in cv.h! */
4369 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4371 CvLVALUE_on(PL_compcv);
4373 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4375 deprecate(":locked");
4377 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4379 CvMETHOD_on(PL_compcv);
4381 /* After we've set the flags, it could be argued that
4382 we don't need to do the attributes.pm-based setting
4383 process, and shouldn't bother appending recognized
4384 flags. To experiment with that, uncomment the
4385 following "else". (Note that's already been
4386 uncommented. That keeps the above-applied built-in
4387 attributes from being intercepted (and possibly
4388 rejected) by a package's attribute routines, but is
4389 justified by the performance win for the common case
4390 of applying only built-in attributes.) */
4392 attrs = append_elem(OP_LIST, attrs,
4393 newSVOP(OP_CONST, 0,
4397 if (*s == ':' && s[1] != ':')
4400 break; /* require real whitespace or :'s */
4401 /* XXX losing whitespace on sequential attributes here */
4405 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4406 if (*s != ';' && *s != '}' && *s != tmp
4407 && (tmp != '=' || *s != ')')) {
4408 const char q = ((*s == '\'') ? '"' : '\'');
4409 /* If here for an expression, and parsed no attrs, back
4411 if (tmp == '=' && !attrs) {
4415 /* MUST advance bufptr here to avoid bogus "at end of line"
4416 context messages from yyerror().
4419 yyerror( (const char *)
4421 ? Perl_form(aTHX_ "Invalid separator character "
4422 "%c%c%c in attribute list", q, *s, q)
4423 : "Unterminated attribute list" ) );
4431 start_force(PL_curforce);
4432 NEXTVAL_NEXTTOKE.opval = attrs;
4433 CURMAD('_', PL_nextwhite);
4438 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4439 (s - SvPVX(PL_linestr)) - stuffstart);
4447 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4448 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4456 const char tmp = *s++;
4461 const char tmp = *s++;
4469 if (PL_lex_brackets <= 0)
4470 yyerror("Unmatched right square bracket");
4473 if (PL_lex_state == LEX_INTERPNORMAL) {
4474 if (PL_lex_brackets == 0) {
4475 if (*s == '-' && s[1] == '>')
4476 PL_lex_state = LEX_INTERPENDMAYBE;
4477 else if (*s != '[' && *s != '{')
4478 PL_lex_state = LEX_INTERPEND;
4485 if (PL_lex_brackets > 100) {
4486 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4488 switch (PL_expect) {
4490 if (PL_lex_formbrack) {
4494 if (PL_oldoldbufptr == PL_last_lop)
4495 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4497 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4498 OPERATOR(HASHBRACK);
4500 while (s < PL_bufend && SPACE_OR_TAB(*s))
4503 PL_tokenbuf[0] = '\0';
4504 if (d < PL_bufend && *d == '-') {
4505 PL_tokenbuf[0] = '-';
4507 while (d < PL_bufend && SPACE_OR_TAB(*d))
4510 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4511 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4513 while (d < PL_bufend && SPACE_OR_TAB(*d))
4516 const char minus = (PL_tokenbuf[0] == '-');
4517 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4525 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4530 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4535 if (PL_oldoldbufptr == PL_last_lop)
4536 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4538 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4541 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4543 /* This hack is to get the ${} in the message. */
4545 yyerror("syntax error");
4548 OPERATOR(HASHBRACK);
4550 /* This hack serves to disambiguate a pair of curlies
4551 * as being a block or an anon hash. Normally, expectation
4552 * determines that, but in cases where we're not in a
4553 * position to expect anything in particular (like inside
4554 * eval"") we have to resolve the ambiguity. This code
4555 * covers the case where the first term in the curlies is a
4556 * quoted string. Most other cases need to be explicitly
4557 * disambiguated by prepending a "+" before the opening
4558 * curly in order to force resolution as an anon hash.
4560 * XXX should probably propagate the outer expectation
4561 * into eval"" to rely less on this hack, but that could
4562 * potentially break current behavior of eval"".
4566 if (*s == '\'' || *s == '"' || *s == '`') {
4567 /* common case: get past first string, handling escapes */
4568 for (t++; t < PL_bufend && *t != *s;)
4569 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4573 else if (*s == 'q') {
4576 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4579 /* skip q//-like construct */
4581 char open, close, term;
4584 while (t < PL_bufend && isSPACE(*t))
4586 /* check for q => */
4587 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4588 OPERATOR(HASHBRACK);
4592 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4596 for (t++; t < PL_bufend; t++) {
4597 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4599 else if (*t == open)
4603 for (t++; t < PL_bufend; t++) {
4604 if (*t == '\\' && t+1 < PL_bufend)
4606 else if (*t == close && --brackets <= 0)
4608 else if (*t == open)
4615 /* skip plain q word */
4616 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4619 else if (isALNUM_lazy_if(t,UTF)) {
4621 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4624 while (t < PL_bufend && isSPACE(*t))
4626 /* if comma follows first term, call it an anon hash */
4627 /* XXX it could be a comma expression with loop modifiers */
4628 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4629 || (*t == '=' && t[1] == '>')))
4630 OPERATOR(HASHBRACK);
4631 if (PL_expect == XREF)
4634 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4640 pl_yylval.ival = CopLINE(PL_curcop);
4641 if (isSPACE(*s) || *s == '#')
4642 PL_copline = NOLINE; /* invalidate current command line number */
4647 if (PL_lex_brackets <= 0)
4648 yyerror("Unmatched right curly bracket");
4650 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4651 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4652 PL_lex_formbrack = 0;
4653 if (PL_lex_state == LEX_INTERPNORMAL) {
4654 if (PL_lex_brackets == 0) {
4655 if (PL_expect & XFAKEBRACK) {
4656 PL_expect &= XENUMMASK;
4657 PL_lex_state = LEX_INTERPEND;
4662 PL_thiswhite = newSVpvs("");
4663 sv_catpvs(PL_thiswhite,"}");
4666 return yylex(); /* ignore fake brackets */
4668 if (*s == '-' && s[1] == '>')
4669 PL_lex_state = LEX_INTERPENDMAYBE;
4670 else if (*s != '[' && *s != '{')
4671 PL_lex_state = LEX_INTERPEND;
4674 if (PL_expect & XFAKEBRACK) {
4675 PL_expect &= XENUMMASK;
4677 return yylex(); /* ignore fake brackets */
4679 start_force(PL_curforce);
4681 curmad('X', newSVpvn(s-1,1));
4682 CURMAD('_', PL_thiswhite);
4687 PL_thistoken = newSVpvs("");
4695 if (PL_expect == XOPERATOR) {
4696 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4697 && isIDFIRST_lazy_if(s,UTF))
4699 CopLINE_dec(PL_curcop);
4700 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
4701 CopLINE_inc(PL_curcop);
4706 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4708 PL_expect = XOPERATOR;
4709 force_ident(PL_tokenbuf, '&');
4713 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
4725 const char tmp = *s++;
4732 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4733 && strchr("+-*/%.^&|<",tmp))
4734 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4735 "Reversed %c= operator",(int)tmp);
4737 if (PL_expect == XSTATE && isALPHA(tmp) &&
4738 (s == PL_linestart+1 || s[-2] == '\n') )
4740 if (PL_in_eval && !PL_rsfp) {
4745 if (strnEQ(s,"=cut",4)) {
4761 PL_thiswhite = newSVpvs("");
4762 sv_catpvn(PL_thiswhite, PL_linestart,
4763 PL_bufend - PL_linestart);
4767 PL_doextract = TRUE;
4771 if (PL_lex_brackets < PL_lex_formbrack) {
4773 #ifdef PERL_STRICT_CR
4774 while (SPACE_OR_TAB(*t))
4776 while (SPACE_OR_TAB(*t) || *t == '\r')
4779 if (*t == '\n' || *t == '#') {
4790 const char tmp = *s++;
4792 /* was this !=~ where !~ was meant?
4793 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4795 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4796 const char *t = s+1;
4798 while (t < PL_bufend && isSPACE(*t))
4801 if (*t == '/' || *t == '?' ||
4802 ((*t == 'm' || *t == 's' || *t == 'y')
4803 && !isALNUM(t[1])) ||
4804 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4805 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4806 "!=~ should be !~");
4816 if (PL_expect != XOPERATOR) {
4817 if (s[1] != '<' && !strchr(s,'>'))
4820 s = scan_heredoc(s);
4822 s = scan_inputsymbol(s);
4823 TERM(sublex_start());
4829 SHop(OP_LEFT_SHIFT);
4843 const char tmp = *s++;
4845 SHop(OP_RIGHT_SHIFT);
4846 else if (tmp == '=')
4855 if (PL_expect == XOPERATOR) {
4856 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4857 return deprecate_commaless_var_list();
4861 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4862 PL_tokenbuf[0] = '@';
4863 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4864 sizeof PL_tokenbuf - 1, FALSE);
4865 if (PL_expect == XOPERATOR)
4866 no_op("Array length", s);
4867 if (!PL_tokenbuf[1])
4869 PL_expect = XOPERATOR;
4870 PL_pending_ident = '#';
4874 PL_tokenbuf[0] = '$';
4875 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4876 sizeof PL_tokenbuf - 1, FALSE);
4877 if (PL_expect == XOPERATOR)
4879 if (!PL_tokenbuf[1]) {
4881 yyerror("Final $ should be \\$ or $name");
4885 /* This kludge not intended to be bulletproof. */
4886 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4887 pl_yylval.opval = newSVOP(OP_CONST, 0,
4888 newSViv(CopARYBASE_get(&PL_compiling)));
4889 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
4895 const char tmp = *s;
4896 if (PL_lex_state == LEX_NORMAL)
4899 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4900 && intuit_more(s)) {
4902 PL_tokenbuf[0] = '@';
4903 if (ckWARN(WARN_SYNTAX)) {
4906 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4909 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4910 while (t < PL_bufend && *t != ']')
4912 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4913 "Multidimensional syntax %.*s not supported",
4914 (int)((t - PL_bufptr) + 1), PL_bufptr);
4918 else if (*s == '{') {
4920 PL_tokenbuf[0] = '%';
4921 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4922 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4924 char tmpbuf[sizeof PL_tokenbuf];
4927 } while (isSPACE(*t));
4928 if (isIDFIRST_lazy_if(t,UTF)) {
4930 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4934 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4935 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4936 "You need to quote \"%s\"",
4943 PL_expect = XOPERATOR;
4944 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4945 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4946 if (!islop || PL_last_lop_op == OP_GREPSTART)
4947 PL_expect = XOPERATOR;
4948 else if (strchr("$@\"'`q", *s))
4949 PL_expect = XTERM; /* e.g. print $fh "foo" */
4950 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4951 PL_expect = XTERM; /* e.g. print $fh &sub */
4952 else if (isIDFIRST_lazy_if(s,UTF)) {
4953 char tmpbuf[sizeof PL_tokenbuf];
4955 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4956 if ((t2 = keyword(tmpbuf, len, 0))) {
4957 /* binary operators exclude handle interpretations */
4969 PL_expect = XTERM; /* e.g. print $fh length() */
4974 PL_expect = XTERM; /* e.g. print $fh subr() */
4977 else if (isDIGIT(*s))
4978 PL_expect = XTERM; /* e.g. print $fh 3 */
4979 else if (*s == '.' && isDIGIT(s[1]))
4980 PL_expect = XTERM; /* e.g. print $fh .3 */
4981 else if ((*s == '?' || *s == '-' || *s == '+')
4982 && !isSPACE(s[1]) && s[1] != '=')
4983 PL_expect = XTERM; /* e.g. print $fh -1 */
4984 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4986 PL_expect = XTERM; /* e.g. print $fh /.../
4987 XXX except DORDOR operator
4989 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4991 PL_expect = XTERM; /* print $fh <<"EOF" */
4994 PL_pending_ident = '$';
4998 if (PL_expect == XOPERATOR)
5000 PL_tokenbuf[0] = '@';
5001 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5002 if (!PL_tokenbuf[1]) {
5005 if (PL_lex_state == LEX_NORMAL)
5007 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5009 PL_tokenbuf[0] = '%';
5011 /* Warn about @ where they meant $. */
5012 if (*s == '[' || *s == '{') {
5013 if (ckWARN(WARN_SYNTAX)) {
5014 const char *t = s + 1;
5015 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5017 if (*t == '}' || *t == ']') {
5019 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5020 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5021 "Scalar value %.*s better written as $%.*s",
5022 (int)(t-PL_bufptr), PL_bufptr,
5023 (int)(t-PL_bufptr-1), PL_bufptr+1);
5028 PL_pending_ident = '@';
5031 case '/': /* may be division, defined-or, or pattern */
5032 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5036 case '?': /* may either be conditional or pattern */
5037 if (PL_expect == XOPERATOR) {
5045 /* A // operator. */
5055 /* Disable warning on "study /blah/" */
5056 if (PL_oldoldbufptr == PL_last_uni
5057 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5058 || memNE(PL_last_uni, "study", 5)
5059 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5062 s = scan_pat(s,OP_MATCH);
5063 TERM(sublex_start());
5067 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5068 #ifdef PERL_STRICT_CR
5071 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5073 && (s == PL_linestart || s[-1] == '\n') )
5075 PL_lex_formbrack = 0;
5079 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5083 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5089 pl_yylval.ival = OPf_SPECIAL;
5095 if (PL_expect != XOPERATOR)
5100 case '0': case '1': case '2': case '3': case '4':
5101 case '5': case '6': case '7': case '8': case '9':
5102 s = scan_num(s, &pl_yylval);
5103 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5104 if (PL_expect == XOPERATOR)
5109 s = scan_str(s,!!PL_madskills,FALSE);
5110 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5111 if (PL_expect == XOPERATOR) {
5112 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5113 return deprecate_commaless_var_list();
5120 pl_yylval.ival = OP_CONST;
5121 TERM(sublex_start());
5124 s = scan_str(s,!!PL_madskills,FALSE);
5125 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5126 if (PL_expect == XOPERATOR) {
5127 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5128 return deprecate_commaless_var_list();
5135 pl_yylval.ival = OP_CONST;
5136 /* FIXME. I think that this can be const if char *d is replaced by
5137 more localised variables. */
5138 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5139 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5140 pl_yylval.ival = OP_STRINGIFY;
5144 TERM(sublex_start());
5147 s = scan_str(s,!!PL_madskills,FALSE);
5148 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5149 if (PL_expect == XOPERATOR)
5150 no_op("Backticks",s);
5153 readpipe_override();
5154 TERM(sublex_start());
5158 if (PL_lex_inwhat && isDIGIT(*s))
5159 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5161 if (PL_expect == XOPERATOR)
5162 no_op("Backslash",s);
5166 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5167 char *start = s + 2;
5168 while (isDIGIT(*start) || *start == '_')
5170 if (*start == '.' && isDIGIT(start[1])) {
5171 s = scan_num(s, &pl_yylval);
5174 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5175 else if (!isALPHA(*start) && (PL_expect == XTERM
5176 || PL_expect == XREF || PL_expect == XSTATE
5177 || PL_expect == XTERMORDORDOR)) {
5178 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5180 s = scan_num(s, &pl_yylval);
5187 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5229 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5231 /* Some keywords can be followed by any delimiter, including ':' */
5232 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5233 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5234 (PL_tokenbuf[0] == 'q' &&
5235 strchr("qwxr", PL_tokenbuf[1])))));
5237 /* x::* is just a word, unless x is "CORE" */
5238 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5242 while (d < PL_bufend && isSPACE(*d))
5243 d++; /* no comments skipped here, or s### is misparsed */
5245 /* Is this a label? */
5246 if (!tmp && PL_expect == XSTATE
5247 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5248 tmp = keyword(PL_tokenbuf, len, 0);
5250 Perl_croak(aTHX_ "Can't use keyword '%s' as a label", PL_tokenbuf);
5252 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5257 /* Check for keywords */
5258 tmp = keyword(PL_tokenbuf, len, 0);
5260 /* Is this a word before a => operator? */
5261 if (*d == '=' && d[1] == '>') {
5264 = (OP*)newSVOP(OP_CONST, 0,
5265 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5266 pl_yylval.opval->op_private = OPpCONST_BARE;
5270 if (tmp < 0) { /* second-class keyword? */
5271 GV *ogv = NULL; /* override (winner) */
5272 GV *hgv = NULL; /* hidden (loser) */
5273 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5275 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5278 if (GvIMPORTED_CV(gv))
5280 else if (! CvMETHOD(cv))
5284 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5285 (gv = *gvp) && isGV_with_GP(gv) &&
5286 GvCVu(gv) && GvIMPORTED_CV(gv))
5293 tmp = 0; /* overridden by import or by GLOBAL */
5296 && -tmp==KEY_lock /* XXX generalizable kludge */
5299 tmp = 0; /* any sub overrides "weak" keyword */
5301 else { /* no override */
5303 if (tmp == KEY_dump) {
5304 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
5305 "dump() better written as CORE::dump()");
5309 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
5310 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5311 "Ambiguous call resolved as CORE::%s(), %s",
5312 GvENAME(hgv), "qualify as such or use &");
5319 default: /* not a keyword */
5320 /* Trade off - by using this evil construction we can pull the
5321 variable gv into the block labelled keylookup. If not, then
5322 we have to give it function scope so that the goto from the
5323 earlier ':' case doesn't bypass the initialisation. */
5325 just_a_word_zero_gv:
5333 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5336 SV *nextPL_nextwhite = 0;
5340 /* Get the rest if it looks like a package qualifier */
5342 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5344 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5347 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5348 *s == '\'' ? "'" : "::");
5353 if (PL_expect == XOPERATOR) {
5354 if (PL_bufptr == PL_linestart) {
5355 CopLINE_dec(PL_curcop);
5356 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5357 CopLINE_inc(PL_curcop);
5360 no_op("Bareword",s);
5363 /* Look for a subroutine with this name in current package,
5364 unless name is "Foo::", in which case Foo is a bearword
5365 (and a package name). */
5367 if (len > 2 && !PL_madskills &&
5368 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5370 if (ckWARN(WARN_BAREWORD)
5371 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5372 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5373 "Bareword \"%s\" refers to nonexistent package",
5376 PL_tokenbuf[len] = '\0';
5382 /* Mustn't actually add anything to a symbol table.
5383 But also don't want to "initialise" any placeholder
5384 constants that might already be there into full
5385 blown PVGVs with attached PVCV. */
5386 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5387 GV_NOADD_NOINIT, SVt_PVCV);
5392 /* if we saw a global override before, get the right name */
5395 sv = newSVpvs("CORE::GLOBAL::");
5396 sv_catpv(sv,PL_tokenbuf);
5399 /* If len is 0, newSVpv does strlen(), which is correct.
5400 If len is non-zero, then it will be the true length,
5401 and so the scalar will be created correctly. */
5402 sv = newSVpv(PL_tokenbuf,len);
5405 if (PL_madskills && !PL_thistoken) {
5406 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5407 PL_thistoken = newSVpvn(start,s - start);
5408 PL_realtokenstart = s - SvPVX(PL_linestr);
5412 /* Presume this is going to be a bareword of some sort. */
5415 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5416 pl_yylval.opval->op_private = OPpCONST_BARE;
5417 /* UTF-8 package name? */
5418 if (UTF && !IN_BYTES &&
5419 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5422 /* And if "Foo::", then that's what it certainly is. */
5427 /* Do the explicit type check so that we don't need to force
5428 the initialisation of the symbol table to have a real GV.
5429 Beware - gv may not really be a PVGV, cv may not really be
5430 a PVCV, (because of the space optimisations that gv_init
5431 understands) But they're true if for this symbol there is
5432 respectively a typeglob and a subroutine.
5434 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5435 /* Real typeglob, so get the real subroutine: */
5437 /* A proxy for a subroutine in this package? */
5438 : SvOK(gv) ? MUTABLE_CV(gv) : NULL)
5441 /* See if it's the indirect object for a list operator. */
5443 if (PL_oldoldbufptr &&
5444 PL_oldoldbufptr < PL_bufptr &&
5445 (PL_oldoldbufptr == PL_last_lop
5446 || PL_oldoldbufptr == PL_last_uni) &&
5447 /* NO SKIPSPACE BEFORE HERE! */
5448 (PL_expect == XREF ||
5449 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5451 bool immediate_paren = *s == '(';
5453 /* (Now we can afford to cross potential line boundary.) */
5454 s = SKIPSPACE2(s,nextPL_nextwhite);
5456 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5459 /* Two barewords in a row may indicate method call. */
5461 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5462 (tmp = intuit_method(s, gv, cv)))
5465 /* If not a declared subroutine, it's an indirect object. */
5466 /* (But it's an indir obj regardless for sort.) */
5467 /* Also, if "_" follows a filetest operator, it's a bareword */
5470 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5472 (PL_last_lop_op != OP_MAPSTART &&
5473 PL_last_lop_op != OP_GREPSTART))))
5474 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5475 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5478 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5483 PL_expect = XOPERATOR;
5486 s = SKIPSPACE2(s,nextPL_nextwhite);
5487 PL_nextwhite = nextPL_nextwhite;
5492 /* Is this a word before a => operator? */
5493 if (*s == '=' && s[1] == '>' && !pkgname) {
5495 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
5496 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5497 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
5501 /* If followed by a paren, it's certainly a subroutine. */
5506 while (SPACE_OR_TAB(*d))
5508 if (*d == ')' && (sv = gv_const_sv(gv))) {
5515 PL_nextwhite = PL_thiswhite;
5518 start_force(PL_curforce);
5520 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5521 PL_expect = XOPERATOR;
5524 PL_nextwhite = nextPL_nextwhite;
5525 curmad('X', PL_thistoken);
5526 PL_thistoken = newSVpvs("");
5534 /* If followed by var or block, call it a method (unless sub) */
5536 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5537 PL_last_lop = PL_oldbufptr;
5538 PL_last_lop_op = OP_METHOD;
5542 /* If followed by a bareword, see if it looks like indir obj. */
5545 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5546 && (tmp = intuit_method(s, gv, cv)))
5549 /* Not a method, so call it a subroutine (if defined) */
5552 if (lastchar == '-')
5553 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
5554 "Ambiguous use of -%s resolved as -&%s()",
5555 PL_tokenbuf, PL_tokenbuf);
5556 /* Check for a constant sub */
5557 if ((sv = gv_const_sv(gv))) {
5559 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
5560 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5561 pl_yylval.opval->op_private = 0;
5565 /* Resolve to GV now. */
5566 if (SvTYPE(gv) != SVt_PVGV) {
5567 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5568 assert (SvTYPE(gv) == SVt_PVGV);
5569 /* cv must have been some sort of placeholder, so
5570 now needs replacing with a real code reference. */
5574 op_free(pl_yylval.opval);
5575 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5576 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5577 PL_last_lop = PL_oldbufptr;
5578 PL_last_lop_op = OP_ENTERSUB;
5579 /* Is there a prototype? */
5587 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
5590 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5592 while (*proto == ';')
5594 if (*proto == '&' && *s == '{') {
5596 sv_setpvs(PL_subname, "__ANON__");
5598 sv_setpvs(PL_subname, "__ANON__::__ANON__");
5605 PL_nextwhite = PL_thiswhite;
5608 start_force(PL_curforce);
5609 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5612 PL_nextwhite = nextPL_nextwhite;
5613 curmad('X', PL_thistoken);
5614 PL_thistoken = newSVpvs("");
5621 /* Guess harder when madskills require "best effort". */
5622 if (PL_madskills && (!gv || !GvCVu(gv))) {
5623 int probable_sub = 0;
5624 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5626 else if (isALPHA(*s)) {
5630 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5631 if (!keyword(tmpbuf, tmplen, 0))
5634 while (d < PL_bufend && isSPACE(*d))
5636 if (*d == '=' && d[1] == '>')
5641 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5642 op_free(pl_yylval.opval);
5643 pl_yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5644 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5645 PL_last_lop = PL_oldbufptr;
5646 PL_last_lop_op = OP_ENTERSUB;
5647 PL_nextwhite = PL_thiswhite;
5649 start_force(PL_curforce);
5650 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5652 PL_nextwhite = nextPL_nextwhite;
5653 curmad('X', PL_thistoken);
5654 PL_thistoken = newSVpvs("");
5659 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
5666 /* Call it a bare word */
5668 if (PL_hints & HINT_STRICT_SUBS)
5669 pl_yylval.opval->op_private |= OPpCONST_STRICT;
5672 /* after "print" and similar functions (corresponding to
5673 * "F? L" in opcode.pl), whatever wasn't already parsed as
5674 * a filehandle should be subject to "strict subs".
5675 * Likewise for the optional indirect-object argument to system
5676 * or exec, which can't be a bareword */
5677 if ((PL_last_lop_op == OP_PRINT
5678 || PL_last_lop_op == OP_PRTF
5679 || PL_last_lop_op == OP_SAY
5680 || PL_last_lop_op == OP_SYSTEM
5681 || PL_last_lop_op == OP_EXEC)
5682 && (PL_hints & HINT_STRICT_SUBS))
5683 pl_yylval.opval->op_private |= OPpCONST_STRICT;
5684 if (lastchar != '-') {
5685 if (ckWARN(WARN_RESERVED)) {
5689 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5690 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5697 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
5698 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
5699 "Operator or semicolon missing before %c%s",
5700 lastchar, PL_tokenbuf);
5701 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
5702 "Ambiguous use of %c resolved as operator %c",
5703 lastchar, lastchar);
5709 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5710 newSVpv(CopFILE(PL_curcop),0));
5714 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5715 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5718 case KEY___PACKAGE__:
5719 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5721 ? newSVhek(HvNAME_HEK(PL_curstash))
5728 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5729 const char *pname = "main";
5730 if (PL_tokenbuf[2] == 'D')
5731 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5732 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5736 GvIOp(gv) = newIO();
5737 IoIFP(GvIOp(gv)) = PL_rsfp;
5738 #if defined(HAS_FCNTL) && defined(F_SETFD)
5740 const int fd = PerlIO_fileno(PL_rsfp);
5741 fcntl(fd,F_SETFD,fd >= 3);
5744 /* Mark this internal pseudo-handle as clean */
5745 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5746 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5747 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5749 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5750 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5751 /* if the script was opened in binmode, we need to revert
5752 * it to text mode for compatibility; but only iff it has CRs
5753 * XXX this is a questionable hack at best. */
5754 if (PL_bufend-PL_bufptr > 2
5755 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5758 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5759 loc = PerlIO_tell(PL_rsfp);
5760 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5763 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5765 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5766 #endif /* NETWARE */
5767 #ifdef PERLIO_IS_STDIO /* really? */
5768 # if defined(__BORLANDC__)
5769 /* XXX see note in do_binmode() */
5770 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5774 PerlIO_seek(PL_rsfp, loc, 0);
5778 #ifdef PERLIO_LAYERS
5781 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5782 else if (PL_encoding) {
5789 XPUSHs(PL_encoding);
5791 call_method("name", G_SCALAR);
5795 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5796 Perl_form(aTHX_ ":encoding(%"SVf")",
5805 if (PL_realtokenstart >= 0) {
5806 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5808 PL_endwhite = newSVpvs("");
5809 sv_catsv(PL_endwhite, PL_thiswhite);
5811 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5812 PL_realtokenstart = -1;
5814 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5815 SvCUR(PL_endwhite))) != NULL) ;
5830 if (PL_expect == XSTATE) {
5837 if (*s == ':' && s[1] == ':') {
5840 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5841 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5842 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5845 else if (tmp == KEY_require || tmp == KEY_do)
5846 /* that's a way to remember we saw "CORE::" */
5859 LOP(OP_ACCEPT,XTERM);
5865 LOP(OP_ATAN2,XTERM);
5871 LOP(OP_BINMODE,XTERM);
5874 LOP(OP_BLESS,XTERM);
5883 /* When 'use switch' is in effect, continue has a dual
5884 life as a control operator. */
5886 if (!FEATURE_IS_ENABLED("switch"))
5889 /* We have to disambiguate the two senses of
5890 "continue". If the next token is a '{' then
5891 treat it as the start of a continue block;
5892 otherwise treat it as a control operator.
5904 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5921 if (!PL_cryptseen) {
5922 PL_cryptseen = TRUE;
5926 LOP(OP_CRYPT,XTERM);
5929 LOP(OP_CHMOD,XTERM);
5932 LOP(OP_CHOWN,XTERM);
5935 LOP(OP_CONNECT,XTERM);
5954 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5955 if (orig_keyword == KEY_do) {
5964 PL_hints |= HINT_BLOCK_SCOPE;
5974 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5975 LOP(OP_DBMOPEN,XTERM);
5981 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5988 pl_yylval.ival = CopLINE(PL_curcop);
6004 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
6005 UNIBRACK(OP_ENTEREVAL);
6019 case KEY_endhostent:
6025 case KEY_endservent:
6028 case KEY_endprotoent:
6039 pl_yylval.ival = CopLINE(PL_curcop);
6041 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6044 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6047 if ((PL_bufend - p) >= 3 &&
6048 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6050 else if ((PL_bufend - p) >= 4 &&
6051 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6054 if (isIDFIRST_lazy_if(p,UTF)) {
6055 p = scan_ident(p, PL_bufend,
6056 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6060 Perl_croak(aTHX_ "Missing $ on loop variable");
6062 s = SvPVX(PL_linestr) + soff;
6068 LOP(OP_FORMLINE,XTERM);
6074 LOP(OP_FCNTL,XTERM);
6080 LOP(OP_FLOCK,XTERM);
6089 LOP(OP_GREPSTART, XREF);
6092 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6107 case KEY_getpriority:
6108 LOP(OP_GETPRIORITY,XTERM);
6110 case KEY_getprotobyname:
6113 case KEY_getprotobynumber:
6114 LOP(OP_GPBYNUMBER,XTERM);
6116 case KEY_getprotoent:
6128 case KEY_getpeername:
6129 UNI(OP_GETPEERNAME);
6131 case KEY_gethostbyname:
6134 case KEY_gethostbyaddr:
6135 LOP(OP_GHBYADDR,XTERM);
6137 case KEY_gethostent:
6140 case KEY_getnetbyname:
6143 case KEY_getnetbyaddr:
6144 LOP(OP_GNBYADDR,XTERM);
6149 case KEY_getservbyname:
6150 LOP(OP_GSBYNAME,XTERM);
6152 case KEY_getservbyport:
6153 LOP(OP_GSBYPORT,XTERM);
6155 case KEY_getservent:
6158 case KEY_getsockname:
6159 UNI(OP_GETSOCKNAME);
6161 case KEY_getsockopt:
6162 LOP(OP_GSOCKOPT,XTERM);
6177 pl_yylval.ival = CopLINE(PL_curcop);
6187 pl_yylval.ival = CopLINE(PL_curcop);
6191 LOP(OP_INDEX,XTERM);
6197 LOP(OP_IOCTL,XTERM);
6209 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6241 LOP(OP_LISTEN,XTERM);
6250 s = scan_pat(s,OP_MATCH);
6251 TERM(sublex_start());
6254 LOP(OP_MAPSTART, XREF);
6257 LOP(OP_MKDIR,XTERM);
6260 LOP(OP_MSGCTL,XTERM);
6263 LOP(OP_MSGGET,XTERM);
6266 LOP(OP_MSGRCV,XTERM);
6269 LOP(OP_MSGSND,XTERM);
6274 PL_in_my = (U16)tmp;
6276 if (isIDFIRST_lazy_if(s,UTF)) {
6280 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6281 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6283 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6284 if (!PL_in_my_stash) {
6287 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6291 if (PL_madskills) { /* just add type to declarator token */
6292 sv_catsv(PL_thistoken, PL_nextwhite);
6294 sv_catpvn(PL_thistoken, start, s - start);
6302 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6309 s = tokenize_use(0, s);
6313 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6320 if (isIDFIRST_lazy_if(s,UTF)) {
6322 for (d = s; isALNUM_lazy_if(d,UTF);)
6324 for (t=d; isSPACE(*t);)
6326 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6328 && !(t[0] == '=' && t[1] == '>')
6330 int parms_len = (int)(d-s);
6331 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6332 "Precedence problem: open %.*s should be open(%.*s)",
6333 parms_len, s, parms_len, s);
6339 pl_yylval.ival = OP_OR;
6349 LOP(OP_OPEN_DIR,XTERM);
6352 checkcomma(s,PL_tokenbuf,"filehandle");
6356 checkcomma(s,PL_tokenbuf,"filehandle");
6375 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6376 s = force_version(s, FALSE);
6380 LOP(OP_PIPE_OP,XTERM);
6383 s = scan_str(s,!!PL_madskills,FALSE);
6386 pl_yylval.ival = OP_CONST;
6387 TERM(sublex_start());
6393 s = scan_str(s,!!PL_madskills,FALSE);
6396 PL_expect = XOPERATOR;
6398 if (SvCUR(PL_lex_stuff)) {
6401 d = SvPV_force(PL_lex_stuff, len);
6403 for (; isSPACE(*d) && len; --len, ++d)
6408 if (!warned && ckWARN(WARN_QW)) {
6409 for (; !isSPACE(*d) && len; --len, ++d) {
6411 Perl_warner(aTHX_ packWARN(WARN_QW),
6412 "Possible attempt to separate words with commas");
6415 else if (*d == '#') {
6416 Perl_warner(aTHX_ packWARN(WARN_QW),
6417 "Possible attempt to put comments in qw() list");
6423 for (; !isSPACE(*d) && len; --len, ++d)
6426 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
6427 words = append_elem(OP_LIST, words,
6428 newSVOP(OP_CONST, 0, tokeq(sv)));
6432 start_force(PL_curforce);
6433 NEXTVAL_NEXTTOKE.opval = words;
6438 SvREFCNT_dec(PL_lex_stuff);
6439 PL_lex_stuff = NULL;
6445 s = scan_str(s,!!PL_madskills,FALSE);
6448 pl_yylval.ival = OP_STRINGIFY;
6449 if (SvIVX(PL_lex_stuff) == '\'')
6450 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6451 TERM(sublex_start());
6454 s = scan_pat(s,OP_QR);
6455 TERM(sublex_start());
6458 s = scan_str(s,!!PL_madskills,FALSE);
6461 readpipe_override();
6462 TERM(sublex_start());
6470 s = force_version(s, FALSE);
6472 else if (*s != 'v' || !isDIGIT(s[1])
6473 || (s = force_version(s, TRUE), *s == 'v'))
6475 *PL_tokenbuf = '\0';
6476 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6477 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6478 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6480 yyerror("<> should be quotes");
6482 if (orig_keyword == KEY_require) {
6490 PL_last_uni = PL_oldbufptr;
6491 PL_last_lop_op = OP_REQUIRE;
6493 return REPORT( (int)REQUIRE );
6499 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6503 LOP(OP_RENAME,XTERM);
6512 LOP(OP_RINDEX,XTERM);
6521 UNIDOR(OP_READLINE);
6524 UNIDOR(OP_BACKTICK);
6533 LOP(OP_REVERSE,XTERM);
6536 UNIDOR(OP_READLINK);
6543 if (pl_yylval.opval)
6544 TERM(sublex_start());
6546 TOKEN(1); /* force error */
6549 checkcomma(s,PL_tokenbuf,"filehandle");
6559 LOP(OP_SELECT,XTERM);
6565 LOP(OP_SEMCTL,XTERM);
6568 LOP(OP_SEMGET,XTERM);
6571 LOP(OP_SEMOP,XTERM);
6577 LOP(OP_SETPGRP,XTERM);
6579 case KEY_setpriority:
6580 LOP(OP_SETPRIORITY,XTERM);
6582 case KEY_sethostent:
6588 case KEY_setservent:
6591 case KEY_setprotoent:
6601 LOP(OP_SEEKDIR,XTERM);
6603 case KEY_setsockopt:
6604 LOP(OP_SSOCKOPT,XTERM);
6610 LOP(OP_SHMCTL,XTERM);
6613 LOP(OP_SHMGET,XTERM);
6616 LOP(OP_SHMREAD,XTERM);
6619 LOP(OP_SHMWRITE,XTERM);
6622 LOP(OP_SHUTDOWN,XTERM);
6631 LOP(OP_SOCKET,XTERM);
6633 case KEY_socketpair:
6634 LOP(OP_SOCKPAIR,XTERM);
6637 checkcomma(s,PL_tokenbuf,"subroutine name");
6639 if (*s == ';' || *s == ')') /* probably a close */
6640 Perl_croak(aTHX_ "sort is now a reserved word");
6642 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6646 LOP(OP_SPLIT,XTERM);
6649 LOP(OP_SPRINTF,XTERM);
6652 LOP(OP_SPLICE,XTERM);
6667 LOP(OP_SUBSTR,XTERM);
6673 char tmpbuf[sizeof PL_tokenbuf];
6674 SSize_t tboffset = 0;
6675 expectation attrful;
6676 bool have_name, have_proto;
6677 const int key = tmp;
6682 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6683 SV *subtoken = newSVpvn(tstart, s - tstart);
6687 s = SKIPSPACE2(s,tmpwhite);
6692 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6693 (*s == ':' && s[1] == ':'))
6696 SV *nametoke = NULL;
6700 attrful = XATTRBLOCK;
6701 /* remember buffer pos'n for later force_word */
6702 tboffset = s - PL_oldbufptr;
6703 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6706 nametoke = newSVpvn(s, d - s);
6708 if (memchr(tmpbuf, ':', len))
6709 sv_setpvn(PL_subname, tmpbuf, len);
6711 sv_setsv(PL_subname,PL_curstname);
6712 sv_catpvs(PL_subname,"::");
6713 sv_catpvn(PL_subname,tmpbuf,len);
6720 CURMAD('X', nametoke);
6721 CURMAD('_', tmpwhite);
6722 (void) force_word(PL_oldbufptr + tboffset, WORD,
6725 s = SKIPSPACE2(d,tmpwhite);
6732 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6733 PL_expect = XTERMBLOCK;
6734 attrful = XATTRTERM;
6735 sv_setpvs(PL_subname,"?");
6739 if (key == KEY_format) {
6741 PL_lex_formbrack = PL_lex_brackets + 1;
6743 PL_thistoken = subtoken;
6747 (void) force_word(PL_oldbufptr + tboffset, WORD,
6753 /* Look for a prototype */
6756 bool bad_proto = FALSE;
6757 bool in_brackets = FALSE;
6758 char greedy_proto = ' ';
6759 bool proto_after_greedy_proto = FALSE;
6760 bool must_be_last = FALSE;
6761 bool underscore = FALSE;
6762 bool seen_underscore = FALSE;
6763 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6765 s = scan_str(s,!!PL_madskills,FALSE);
6767 Perl_croak(aTHX_ "Prototype not terminated");
6768 /* strip spaces and check for bad characters */
6769 d = SvPVX(PL_lex_stuff);
6771 for (p = d; *p; ++p) {
6777 proto_after_greedy_proto = TRUE;
6778 if (!strchr("$@%*;[]&\\_", *p)) {
6790 else if ( *p == ']' ) {
6791 in_brackets = FALSE;
6793 else if ( (*p == '@' || *p == '%') &&
6794 ( tmp < 2 || d[tmp-2] != '\\' ) &&
6796 must_be_last = TRUE;
6799 else if ( *p == '_' ) {
6800 underscore = seen_underscore = TRUE;
6807 if (proto_after_greedy_proto)
6808 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6809 "Prototype after '%c' for %"SVf" : %s",
6810 greedy_proto, SVfARG(PL_subname), d);
6812 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6813 "Illegal character %sin prototype for %"SVf" : %s",
6814 seen_underscore ? "after '_' " : "",
6815 SVfARG(PL_subname), d);
6816 SvCUR_set(PL_lex_stuff, tmp);
6821 CURMAD('q', PL_thisopen);
6822 CURMAD('_', tmpwhite);
6823 CURMAD('=', PL_thisstuff);
6824 CURMAD('Q', PL_thisclose);
6825 NEXTVAL_NEXTTOKE.opval =
6826 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6827 PL_lex_stuff = NULL;
6830 s = SKIPSPACE2(s,tmpwhite);
6838 if (*s == ':' && s[1] != ':')
6839 PL_expect = attrful;
6840 else if (*s != '{' && key == KEY_sub) {
6842 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6844 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6851 curmad('^', newSVpvs(""));
6852 CURMAD('_', tmpwhite);
6856 PL_thistoken = subtoken;
6859 NEXTVAL_NEXTTOKE.opval =
6860 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6861 PL_lex_stuff = NULL;
6867 sv_setpvs(PL_subname, "__ANON__");
6869 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6873 (void) force_word(PL_oldbufptr + tboffset, WORD,
6882 LOP(OP_SYSTEM,XREF);
6885 LOP(OP_SYMLINK,XTERM);
6888 LOP(OP_SYSCALL,XTERM);
6891 LOP(OP_SYSOPEN,XTERM);
6894 LOP(OP_SYSSEEK,XTERM);
6897 LOP(OP_SYSREAD,XTERM);
6900 LOP(OP_SYSWRITE,XTERM);
6904 TERM(sublex_start());
6925 LOP(OP_TRUNCATE,XTERM);
6937 pl_yylval.ival = CopLINE(PL_curcop);
6941 pl_yylval.ival = CopLINE(PL_curcop);
6945 LOP(OP_UNLINK,XTERM);
6951 LOP(OP_UNPACK,XTERM);
6954 LOP(OP_UTIME,XTERM);
6960 LOP(OP_UNSHIFT,XTERM);
6963 s = tokenize_use(1, s);
6973 pl_yylval.ival = CopLINE(PL_curcop);
6977 pl_yylval.ival = CopLINE(PL_curcop);
6981 PL_hints |= HINT_BLOCK_SCOPE;
6988 LOP(OP_WAITPID,XTERM);
6997 ctl_l[0] = toCTRL('L');
6999 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7002 /* Make sure $^L is defined */
7003 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7008 if (PL_expect == XOPERATOR)
7014 pl_yylval.ival = OP_XOR;
7019 TERM(sublex_start());
7024 #pragma segment Main
7028 S_pending_ident(pTHX)
7033 /* pit holds the identifier we read and pending_ident is reset */
7034 char pit = PL_pending_ident;
7035 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7036 /* All routes through this function want to know if there is a colon. */
7037 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7038 PL_pending_ident = 0;
7040 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7041 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7042 "### Pending identifier '%s'\n", PL_tokenbuf); });
7044 /* if we're in a my(), we can't allow dynamics here.
7045 $foo'bar has already been turned into $foo::bar, so
7046 just check for colons.
7048 if it's a legal name, the OP is a PADANY.
7051 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7053 yyerror(Perl_form(aTHX_ "No package name allowed for "
7054 "variable %s in \"our\"",
7056 tmp = allocmy(PL_tokenbuf);
7060 yyerror(Perl_form(aTHX_ PL_no_myglob,
7061 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7063 pl_yylval.opval = newOP(OP_PADANY, 0);
7064 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf);
7070 build the ops for accesses to a my() variable.
7072 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7073 then used in a comparison. This catches most, but not
7074 all cases. For instance, it catches
7075 sort { my($a); $a <=> $b }
7077 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7078 (although why you'd do that is anyone's guess).
7083 tmp = pad_findmy(PL_tokenbuf);
7084 if (tmp != NOT_IN_PAD) {
7085 /* might be an "our" variable" */
7086 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7087 /* build ops for a bareword */
7088 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7089 HEK * const stashname = HvNAME_HEK(stash);
7090 SV * const sym = newSVhek(stashname);
7091 sv_catpvs(sym, "::");
7092 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7093 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7094 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7097 ? (GV_ADDMULTI | GV_ADDINEVAL)
7100 ((PL_tokenbuf[0] == '$') ? SVt_PV
7101 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7106 /* if it's a sort block and they're naming $a or $b */
7107 if (PL_last_lop_op == OP_SORT &&
7108 PL_tokenbuf[0] == '$' &&
7109 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
7112 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
7113 d < PL_bufend && *d != '\n';
7116 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7117 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7123 pl_yylval.opval = newOP(OP_PADANY, 0);
7124 pl_yylval.opval->op_targ = tmp;
7130 Whine if they've said @foo in a doublequoted string,
7131 and @foo isn't a variable we can find in the symbol
7134 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7135 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7137 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7138 /* DO NOT warn for @- and @+ */
7139 && !( PL_tokenbuf[2] == '\0' &&
7140 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7143 /* Downgraded from fatal to warning 20000522 mjd */
7144 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7145 "Possible unintended interpolation of %s in string",
7150 /* build ops for a bareword */
7151 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7153 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7155 PL_tokenbuf + 1, tokenbuf_len - 1,
7156 /* If the identifier refers to a stash, don't autovivify it.
7157 * Change 24660 had the side effect of causing symbol table
7158 * hashes to always be defined, even if they were freshly
7159 * created and the only reference in the entire program was
7160 * the single statement with the defined %foo::bar:: test.
7161 * It appears that all code in the wild doing this actually
7162 * wants to know whether sub-packages have been loaded, so
7163 * by avoiding auto-vivifying symbol tables, we ensure that
7164 * defined %foo::bar:: continues to be false, and the existing
7165 * tests still give the expected answers, even though what
7166 * they're actually testing has now changed subtly.
7168 (*PL_tokenbuf == '%'
7169 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7172 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7173 ((PL_tokenbuf[0] == '$') ? SVt_PV
7174 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7180 * The following code was generated by perl_keyword.pl.
7184 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7188 PERL_ARGS_ASSERT_KEYWORD;
7192 case 1: /* 5 tokens of length 1 */
7224 case 2: /* 18 tokens of length 2 */
7370 case 3: /* 29 tokens of length 3 */
7374 if (name[1] == 'N' &&
7437 if (name[1] == 'i' &&
7469 if (name[1] == 'o' &&
7478 if (name[1] == 'e' &&
7487 if (name[1] == 'n' &&
7496 if (name[1] == 'o' &&
7505 if (name[1] == 'a' &&
7514 if (name[1] == 'o' &&
7576 if (name[1] == 'e' &&
7590 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7616 if (name[1] == 'i' &&
7625 if (name[1] == 's' &&
7634 if (name[1] == 'e' &&
7643 if (name[1] == 'o' &&
7655 case 4: /* 41 tokens of length 4 */
7659 if (name[1] == 'O' &&
7669 if (name[1] == 'N' &&
7679 if (name[1] == 'i' &&
7689 if (name[1] == 'h' &&
7699 if (name[1] == 'u' &&
7712 if (name[2] == 'c' &&
7721 if (name[2] == 's' &&
7730 if (name[2] == 'a' &&
7766 if (name[1] == 'o' &&
7779 if (name[2] == 't' &&
7788 if (name[2] == 'o' &&
7797 if (name[2] == 't' &&
7806 if (name[2] == 'e' &&
7819 if (name[1] == 'o' &&
7832 if (name[2] == 'y' &&
7841 if (name[2] == 'l' &&
7857 if (name[2] == 's' &&
7866 if (name[2] == 'n' &&
7875 if (name[2] == 'c' &&
7888 if (name[1] == 'e' &&
7898 if (name[1] == 'p' &&
7911 if (name[2] == 'c' &&
7920 if (name[2] == 'p' &&
7929 if (name[2] == 's' &&
7945 if (name[2] == 'n' &&
8015 if (name[2] == 'r' &&
8024 if (name[2] == 'r' &&
8033 if (name[2] == 'a' &&
8049 if (name[2] == 'l' &&
8111 if (name[2] == 'e' &&
8114 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8127 case 5: /* 39 tokens of length 5 */
8131 if (name[1] == 'E' &&
8142 if (name[1] == 'H' &&
8156 if (name[2] == 'a' &&
8166 if (name[2] == 'a' &&
8183 if (name[2] == 'e' &&
8193 if (name[2] == 'e' &&
8197 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8213 if (name[3] == 'i' &&
8222 if (name[3] == 'o' &&
8258 if (name[2] == 'o' &&
8268 if (name[2] == 'y' &&
8282 if (name[1] == 'l' &&
8296 if (name[2] == 'n' &&
8306 if (name[2] == 'o' &&
8320 if (name[1] == 'i' &&
8325 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8334 if (name[2] == 'd' &&
8344 if (name[2] == 'c' &&
8361 if (name[2] == 'c' &&
8371 if (name[2] == 't' &&
8385 if (name[1] == 'k' &&
8396 if (name[1] == 'r' &&
8410 if (name[2] == 's' &&
8420 if (name[2] == 'd' &&
8437 if (name[2] == 'm' &&
8447 if (name[2] == 'i' &&
8457 if (name[2] == 'e' &&
8467 if (name[2] == 'l' &&
8477 if (name[2] == 'a' &&
8490 if (name[3] == 't' &&
8493 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8499 if (name[3] == 'd' &&
8516 if (name[1] == 'i' &&
8530 if (name[2] == 'a' &&
8543 if (name[3] == 'e' &&
8578 if (name[2] == 'i' &&
8595 if (name[2] == 'i' &&
8605 if (name[2] == 'i' &&
8622 case 6: /* 33 tokens of length 6 */
8626 if (name[1] == 'c' &&
8641 if (name[2] == 'l' &&
8652 if (name[2] == 'r' &&
8667 if (name[1] == 'e' &&
8682 if (name[2] == 's' &&
8687 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8693 if (name[2] == 'i' &&
8711 if (name[2] == 'l' &&
8722 if (name[2] == 'r' &&
8737 if (name[1] == 'm' &&
8752 if (name[2] == 'n' &&
8763 if (name[2] == 's' &&
8778 if (name[1] == 's' &&
8784 if (name[4] == 't' &&
8793 if (name[4] == 'e' &&
8802 if (name[4] == 'c' &&
8811 if (name[4] == 'n' &&
8827 if (name[1] == 'r' &&
8845 if (name[3] == 'a' &&
8855 if (name[3] == 'u' &&
8869 if (name[2] == 'n' &&
8887 if (name[2] == 'a' &&
8901 if (name[3] == 'e' &&
8914 if (name[4] == 't' &&
8923 if (name[4] == 'e' &&
8945 if (name[4] == 't' &&
8954 if (name[4] == 'e' &&
8970 if (name[2] == 'c' &&
8981 if (name[2] == 'l' &&
8992 if (name[2] == 'b' &&
9003 if (name[2] == 's' &&
9026 if (name[4] == 's' &&
9035 if (name[4] == 'n' &&
9048 if (name[3] == 'a' &&
9065 if (name[1] == 'a' &&
9080 case 7: /* 29 tokens of length 7 */
9084 if (name[1] == 'E' &&
9097 if (name[1] == '_' &&
9110 if (name[1] == 'i' &&
9117 return -KEY_binmode;
9123 if (name[1] == 'o' &&
9130 return -KEY_connect;
9139 if (name[2] == 'm' &&
9145 return -KEY_dbmopen;
9156 if (name[4] == 'u' &&
9160 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9166 if (name[4] == 'n' &&
9187 if (name[1] == 'o' &&
9200 if (name[1] == 'e' &&
9207 if (name[5] == 'r' &&
9210 return -KEY_getpgrp;
9216 if (name[5] == 'i' &&
9219 return -KEY_getppid;
9232 if (name[1] == 'c' &&
9239 return -KEY_lcfirst;
9245 if (name[1] == 'p' &&
9252 return -KEY_opendir;
9258 if (name[1] == 'a' &&
9276 if (name[3] == 'd' &&
9281 return -KEY_readdir;
9287 if (name[3] == 'u' &&
9298 if (name[3] == 'e' &&
9303 return -KEY_reverse;
9322 if (name[3] == 'k' &&
9327 return -KEY_seekdir;
9333 if (name[3] == 'p' &&
9338 return -KEY_setpgrp;
9348 if (name[2] == 'm' &&
9354 return -KEY_shmread;
9360 if (name[2] == 'r' &&
9366 return -KEY_sprintf;
9375 if (name[3] == 'l' &&
9380 return -KEY_symlink;
9389 if (name[4] == 'a' &&
9393 return -KEY_syscall;
9399 if (name[4] == 'p' &&
9403 return -KEY_sysopen;
9409 if (name[4] == 'e' &&
9413 return -KEY_sysread;
9419 if (name[4] == 'e' &&
9423 return -KEY_sysseek;
9441 if (name[1] == 'e' &&
9448 return -KEY_telldir;
9457 if (name[2] == 'f' &&
9463 return -KEY_ucfirst;
9469 if (name[2] == 's' &&
9475 return -KEY_unshift;
9485 if (name[1] == 'a' &&
9492 return -KEY_waitpid;
9501 case 8: /* 26 tokens of length 8 */
9505 if (name[1] == 'U' &&
9513 return KEY_AUTOLOAD;
9524 if (name[3] == 'A' &&
9530 return KEY___DATA__;
9536 if (name[3] == 'I' &&
9542 return -KEY___FILE__;
9548 if (name[3] == 'I' &&
9554 return -KEY___LINE__;
9570 if (name[2] == 'o' &&
9577 return -KEY_closedir;
9583 if (name[2] == 'n' &&
9590 return -KEY_continue;
9600 if (name[1] == 'b' &&
9608 return -KEY_dbmclose;
9614 if (name[1] == 'n' &&
9620 if (name[4] == 'r' &&
9625 return -KEY_endgrent;
9631 if (name[4] == 'w' &&
9636 return -KEY_endpwent;
9649 if (name[1] == 'o' &&
9657 return -KEY_formline;
9663 if (name[1] == 'e' &&
9674 if (name[6] == 'n' &&
9677 return -KEY_getgrent;
9683 if (name[6] == 'i' &&
9686 return -KEY_getgrgid;
9692 if (name[6] == 'a' &&
9695 return -KEY_getgrnam;
9708 if (name[4] == 'o' &&
9713 return -KEY_getlogin;
9724 if (name[6] == 'n' &&
9727 return -KEY_getpwent;
9733 if (name[6] == 'a' &&
9736 return -KEY_getpwnam;
9742 if (name[6] == 'i' &&
9745 return -KEY_getpwuid;
9765 if (name[1] == 'e' &&
9772 if (name[5] == 'i' &&
9779 return -KEY_readline;
9784 return -KEY_readlink;
9795 if (name[5] == 'i' &&
9799 return -KEY_readpipe;
9820 if (name[4] == 'r' &&
9825 return -KEY_setgrent;
9831 if (name[4] == 'w' &&
9836 return -KEY_setpwent;
9852 if (name[3] == 'w' &&
9858 return -KEY_shmwrite;
9864 if (name[3] == 't' &&
9870 return -KEY_shutdown;
9880 if (name[2] == 's' &&
9887 return -KEY_syswrite;
9897 if (name[1] == 'r' &&
9905 return -KEY_truncate;
9914 case 9: /* 9 tokens of length 9 */
9918 if (name[1] == 'N' &&
9927 return KEY_UNITCHECK;
9933 if (name[1] == 'n' &&
9942 return -KEY_endnetent;
9948 if (name[1] == 'e' &&
9957 return -KEY_getnetent;
9963 if (name[1] == 'o' &&
9972 return -KEY_localtime;
9978 if (name[1] == 'r' &&
9987 return KEY_prototype;
9993 if (name[1] == 'u' &&
10002 return -KEY_quotemeta;
10008 if (name[1] == 'e' &&
10017 return -KEY_rewinddir;
10023 if (name[1] == 'e' &&
10032 return -KEY_setnetent;
10038 if (name[1] == 'a' &&
10047 return -KEY_wantarray;
10056 case 10: /* 9 tokens of length 10 */
10060 if (name[1] == 'n' &&
10066 if (name[4] == 'o' &&
10073 return -KEY_endhostent;
10079 if (name[4] == 'e' &&
10086 return -KEY_endservent;
10099 if (name[1] == 'e' &&
10105 if (name[4] == 'o' &&
10112 return -KEY_gethostent;
10121 if (name[5] == 'r' &&
10127 return -KEY_getservent;
10133 if (name[5] == 'c' &&
10139 return -KEY_getsockopt;
10159 if (name[2] == 't')
10164 if (name[4] == 'o' &&
10171 return -KEY_sethostent;
10180 if (name[5] == 'r' &&
10186 return -KEY_setservent;
10192 if (name[5] == 'c' &&
10198 return -KEY_setsockopt;
10215 if (name[2] == 'c' &&
10224 return -KEY_socketpair;
10237 case 11: /* 8 tokens of length 11 */
10241 if (name[1] == '_' &&
10251 { /* __PACKAGE__ */
10252 return -KEY___PACKAGE__;
10258 if (name[1] == 'n' &&
10268 { /* endprotoent */
10269 return -KEY_endprotoent;
10275 if (name[1] == 'e' &&
10284 if (name[5] == 'e' &&
10290 { /* getpeername */
10291 return -KEY_getpeername;
10300 if (name[6] == 'o' &&
10305 { /* getpriority */
10306 return -KEY_getpriority;
10312 if (name[6] == 't' &&
10317 { /* getprotoent */
10318 return -KEY_getprotoent;
10332 if (name[4] == 'o' &&
10339 { /* getsockname */
10340 return -KEY_getsockname;
10353 if (name[1] == 'e' &&
10361 if (name[6] == 'o' &&
10366 { /* setpriority */
10367 return -KEY_setpriority;
10373 if (name[6] == 't' &&
10378 { /* setprotoent */
10379 return -KEY_setprotoent;
10395 case 12: /* 2 tokens of length 12 */
10396 if (name[0] == 'g' &&
10408 if (name[9] == 'd' &&
10411 { /* getnetbyaddr */
10412 return -KEY_getnetbyaddr;
10418 if (name[9] == 'a' &&
10421 { /* getnetbyname */
10422 return -KEY_getnetbyname;
10434 case 13: /* 4 tokens of length 13 */
10435 if (name[0] == 'g' &&
10442 if (name[4] == 'o' &&
10451 if (name[10] == 'd' &&
10454 { /* gethostbyaddr */
10455 return -KEY_gethostbyaddr;
10461 if (name[10] == 'a' &&
10464 { /* gethostbyname */
10465 return -KEY_gethostbyname;
10478 if (name[4] == 'e' &&
10487 if (name[10] == 'a' &&
10490 { /* getservbyname */
10491 return -KEY_getservbyname;
10497 if (name[10] == 'o' &&
10500 { /* getservbyport */
10501 return -KEY_getservbyport;
10520 case 14: /* 1 tokens of length 14 */
10521 if (name[0] == 'g' &&
10535 { /* getprotobyname */
10536 return -KEY_getprotobyname;
10541 case 16: /* 1 tokens of length 16 */
10542 if (name[0] == 'g' &&
10558 { /* getprotobynumber */
10559 return -KEY_getprotobynumber;
10573 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10577 PERL_ARGS_ASSERT_CHECKCOMMA;
10579 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10580 if (ckWARN(WARN_SYNTAX)) {
10583 for (w = s+2; *w && level; w++) {
10586 else if (*w == ')')
10589 while (isSPACE(*w))
10591 /* the list of chars below is for end of statements or
10592 * block / parens, boolean operators (&&, ||, //) and branch
10593 * constructs (or, and, if, until, unless, while, err, for).
10594 * Not a very solid hack... */
10595 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10596 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10597 "%s (...) interpreted as function",name);
10600 while (s < PL_bufend && isSPACE(*s))
10604 while (s < PL_bufend && isSPACE(*s))
10606 if (isIDFIRST_lazy_if(s,UTF)) {
10607 const char * const w = s++;
10608 while (isALNUM_lazy_if(s,UTF))
10610 while (s < PL_bufend && isSPACE(*s))
10614 if (keyword(w, s - w, 0))
10617 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10618 if (gv && GvCVu(gv))
10620 Perl_croak(aTHX_ "No comma allowed after %s", what);
10625 /* Either returns sv, or mortalizes sv and returns a new SV*.
10626 Best used as sv=new_constant(..., sv, ...).
10627 If s, pv are NULL, calls subroutine with one argument,
10628 and type is used with error messages only. */
10631 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
10632 SV *sv, SV *pv, const char *type, STRLEN typelen)
10635 HV * const table = GvHV(PL_hintgv); /* ^H */
10639 const char *why1 = "", *why2 = "", *why3 = "";
10641 PERL_ARGS_ASSERT_NEW_CONSTANT;
10643 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10646 why2 = (const char *)
10647 (strEQ(key,"charnames")
10648 ? "(possibly a missing \"use charnames ...\")"
10650 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10651 (type ? type: "undef"), why2);
10653 /* This is convoluted and evil ("goto considered harmful")
10654 * but I do not understand the intricacies of all the different
10655 * failure modes of %^H in here. The goal here is to make
10656 * the most probable error message user-friendly. --jhi */
10661 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10662 (type ? type: "undef"), why1, why2, why3);
10664 yyerror(SvPVX_const(msg));
10668 cvp = hv_fetch(table, key, keylen, FALSE);
10669 if (!cvp || !SvOK(*cvp)) {
10672 why3 = "} is not defined";
10675 sv_2mortal(sv); /* Parent created it permanently */
10678 pv = newSVpvn_flags(s, len, SVs_TEMP);
10680 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
10682 typesv = &PL_sv_undef;
10684 PUSHSTACKi(PERLSI_OVERLOAD);
10696 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10700 /* Check the eval first */
10701 if (!PL_in_eval && SvTRUE(ERRSV)) {
10702 sv_catpvs(ERRSV, "Propagated");
10703 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10705 res = SvREFCNT_inc_simple(sv);
10709 SvREFCNT_inc_simple_void(res);
10718 why1 = "Call to &{$^H{";
10720 why3 = "}} did not return a defined value";
10728 /* Returns a NUL terminated string, with the length of the string written to
10732 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10735 register char *d = dest;
10736 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10738 PERL_ARGS_ASSERT_SCAN_WORD;
10742 Perl_croak(aTHX_ ident_too_long);
10743 if (isALNUM(*s)) /* UTF handled below */
10745 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10750 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10754 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10755 char *t = s + UTF8SKIP(s);
10757 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10761 Perl_croak(aTHX_ ident_too_long);
10762 Copy(s, d, len, char);
10775 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10778 char *bracket = NULL;
10780 register char *d = dest;
10781 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10783 PERL_ARGS_ASSERT_SCAN_IDENT;
10788 while (isDIGIT(*s)) {
10790 Perl_croak(aTHX_ ident_too_long);
10797 Perl_croak(aTHX_ ident_too_long);
10798 if (isALNUM(*s)) /* UTF handled below */
10800 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10805 else if (*s == ':' && s[1] == ':') {
10809 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10810 char *t = s + UTF8SKIP(s);
10811 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10813 if (d + (t - s) > e)
10814 Perl_croak(aTHX_ ident_too_long);
10815 Copy(s, d, t - s, char);
10826 if (PL_lex_state != LEX_NORMAL)
10827 PL_lex_state = LEX_INTERPENDMAYBE;
10830 if (*s == '$' && s[1] &&
10831 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10844 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10849 if (isSPACE(s[-1])) {
10851 const char ch = *s++;
10852 if (!SPACE_OR_TAB(ch)) {
10858 if (isIDFIRST_lazy_if(d,UTF)) {
10862 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10863 end += UTF8SKIP(end);
10864 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10865 end += UTF8SKIP(end);
10867 Copy(s, d, end - s, char);
10872 while ((isALNUM(*s) || *s == ':') && d < e)
10875 Perl_croak(aTHX_ ident_too_long);
10878 while (s < send && SPACE_OR_TAB(*s))
10880 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10881 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10882 const char * const brack =
10884 ((*s == '[') ? "[...]" : "{...}");
10885 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10886 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10887 funny, dest, brack, funny, dest, brack);
10890 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10894 /* Handle extended ${^Foo} variables
10895 * 1999-02-27 mjd-perl-patch@plover.com */
10896 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10900 while (isALNUM(*s) && d < e) {
10904 Perl_croak(aTHX_ ident_too_long);
10909 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10910 PL_lex_state = LEX_INTERPEND;
10913 if (PL_lex_state == LEX_NORMAL) {
10914 if (ckWARN(WARN_AMBIGUOUS) &&
10915 (keyword(dest, d - dest, 0)
10916 || get_cvn_flags(dest, d - dest, 0)))
10920 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10921 "Ambiguous use of %c{%s} resolved to %c%s",
10922 funny, dest, funny, dest);
10927 s = bracket; /* let the parser handle it */
10931 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10932 PL_lex_state = LEX_INTERPEND;
10937 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10939 PERL_ARGS_ASSERT_PMFLAG;
10941 PERL_UNUSED_CONTEXT;
10943 const char c = (char)ch;
10945 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10946 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10947 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10948 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10949 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10955 S_scan_pat(pTHX_ char *start, I32 type)
10959 char *s = scan_str(start,!!PL_madskills,FALSE);
10960 const char * const valid_flags =
10961 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10966 PERL_ARGS_ASSERT_SCAN_PAT;
10969 const char * const delimiter = skipspace(start);
10973 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10974 : "Search pattern not terminated" ));
10977 pm = (PMOP*)newPMOP(type, 0);
10978 if (PL_multi_open == '?') {
10979 /* This is the only point in the code that sets PMf_ONCE: */
10980 pm->op_pmflags |= PMf_ONCE;
10982 /* Hence it's safe to do this bit of PMOP book-keeping here, which
10983 allows us to restrict the list needed by reset to just the ??
10985 assert(type != OP_TRANS);
10987 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
10990 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
10993 elements = mg->mg_len / sizeof(PMOP**);
10994 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
10995 ((PMOP**)mg->mg_ptr) [elements++] = pm;
10996 mg->mg_len = elements * sizeof(PMOP**);
10997 PmopSTASH_set(pm,PL_curstash);
11003 while (*s && strchr(valid_flags, *s))
11004 pmflag(&pm->op_pmflags,*s++);
11006 if (PL_madskills && modstart != s) {
11007 SV* tmptoken = newSVpvn(modstart, s - modstart);
11008 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11011 /* issue a warning if /c is specified,but /g is not */
11012 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11014 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11015 "Use of /c modifier is meaningless without /g" );
11018 PL_lex_op = (OP*)pm;
11019 pl_yylval.ival = OP_MATCH;
11024 S_scan_subst(pTHX_ char *start)
11035 PERL_ARGS_ASSERT_SCAN_SUBST;
11037 pl_yylval.ival = OP_NULL;
11039 s = scan_str(start,!!PL_madskills,FALSE);
11042 Perl_croak(aTHX_ "Substitution pattern not terminated");
11044 if (s[-1] == PL_multi_open)
11047 if (PL_madskills) {
11048 CURMAD('q', PL_thisopen);
11049 CURMAD('_', PL_thiswhite);
11050 CURMAD('E', PL_thisstuff);
11051 CURMAD('Q', PL_thisclose);
11052 PL_realtokenstart = s - SvPVX(PL_linestr);
11056 first_start = PL_multi_start;
11057 s = scan_str(s,!!PL_madskills,FALSE);
11059 if (PL_lex_stuff) {
11060 SvREFCNT_dec(PL_lex_stuff);
11061 PL_lex_stuff = NULL;
11063 Perl_croak(aTHX_ "Substitution replacement not terminated");
11065 PL_multi_start = first_start; /* so whole substitution is taken together */
11067 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11070 if (PL_madskills) {
11071 CURMAD('z', PL_thisopen);
11072 CURMAD('R', PL_thisstuff);
11073 CURMAD('Z', PL_thisclose);
11079 if (*s == EXEC_PAT_MOD) {
11083 else if (strchr(S_PAT_MODS, *s))
11084 pmflag(&pm->op_pmflags,*s++);
11090 if (PL_madskills) {
11092 curmad('m', newSVpvn(modstart, s - modstart));
11093 append_madprops(PL_thismad, (OP*)pm, 0);
11097 if ((pm->op_pmflags & PMf_CONTINUE)) {
11098 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
11102 SV * const repl = newSVpvs("");
11104 PL_sublex_info.super_bufptr = s;
11105 PL_sublex_info.super_bufend = PL_bufend;
11107 pm->op_pmflags |= PMf_EVAL;
11110 sv_catpvs(repl, "eval ");
11112 sv_catpvs(repl, "do ");
11114 sv_catpvs(repl, "{");
11115 sv_catsv(repl, PL_lex_repl);
11116 if (strchr(SvPVX(PL_lex_repl), '#'))
11117 sv_catpvs(repl, "\n");
11118 sv_catpvs(repl, "}");
11120 SvREFCNT_dec(PL_lex_repl);
11121 PL_lex_repl = repl;
11124 PL_lex_op = (OP*)pm;
11125 pl_yylval.ival = OP_SUBST;
11130 S_scan_trans(pTHX_ char *start)
11143 PERL_ARGS_ASSERT_SCAN_TRANS;
11145 pl_yylval.ival = OP_NULL;
11147 s = scan_str(start,!!PL_madskills,FALSE);
11149 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11151 if (s[-1] == PL_multi_open)
11154 if (PL_madskills) {
11155 CURMAD('q', PL_thisopen);
11156 CURMAD('_', PL_thiswhite);
11157 CURMAD('E', PL_thisstuff);
11158 CURMAD('Q', PL_thisclose);
11159 PL_realtokenstart = s - SvPVX(PL_linestr);
11163 s = scan_str(s,!!PL_madskills,FALSE);
11165 if (PL_lex_stuff) {
11166 SvREFCNT_dec(PL_lex_stuff);
11167 PL_lex_stuff = NULL;
11169 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11171 if (PL_madskills) {
11172 CURMAD('z', PL_thisopen);
11173 CURMAD('R', PL_thisstuff);
11174 CURMAD('Z', PL_thisclose);
11177 complement = del = squash = 0;
11184 complement = OPpTRANS_COMPLEMENT;
11187 del = OPpTRANS_DELETE;
11190 squash = OPpTRANS_SQUASH;
11199 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11200 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11201 o->op_private &= ~OPpTRANS_ALL;
11202 o->op_private |= del|squash|complement|
11203 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11204 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11207 pl_yylval.ival = OP_TRANS;
11210 if (PL_madskills) {
11212 curmad('m', newSVpvn(modstart, s - modstart));
11213 append_madprops(PL_thismad, o, 0);
11222 S_scan_heredoc(pTHX_ register char *s)
11226 I32 op_type = OP_SCALAR;
11230 const char *found_newline;
11234 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11236 I32 stuffstart = s - SvPVX(PL_linestr);
11239 PL_realtokenstart = -1;
11242 PERL_ARGS_ASSERT_SCAN_HEREDOC;
11246 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11250 while (SPACE_OR_TAB(*peek))
11252 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11255 s = delimcpy(d, e, s, PL_bufend, term, &len);
11265 if (!isALNUM_lazy_if(s,UTF))
11266 deprecate("bare << to mean <<\"\"");
11267 for (; isALNUM_lazy_if(s,UTF); s++) {
11272 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11273 Perl_croak(aTHX_ "Delimiter for here document is too long");
11276 len = d - PL_tokenbuf;
11279 if (PL_madskills) {
11280 tstart = PL_tokenbuf + !outer;
11281 PL_thisclose = newSVpvn(tstart, len - !outer);
11282 tstart = SvPVX(PL_linestr) + stuffstart;
11283 PL_thisopen = newSVpvn(tstart, s - tstart);
11284 stuffstart = s - SvPVX(PL_linestr);
11287 #ifndef PERL_STRICT_CR
11288 d = strchr(s, '\r');
11290 char * const olds = s;
11292 while (s < PL_bufend) {
11298 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11307 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11314 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11315 herewas = newSVpvn(s,PL_bufend-s);
11319 herewas = newSVpvn(s-1,found_newline-s+1);
11322 herewas = newSVpvn(s,found_newline-s);
11326 if (PL_madskills) {
11327 tstart = SvPVX(PL_linestr) + stuffstart;
11329 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11331 PL_thisstuff = newSVpvn(tstart, s - tstart);
11334 s += SvCUR(herewas);
11337 stuffstart = s - SvPVX(PL_linestr);
11343 tmpstr = newSV_type(SVt_PVIV);
11344 SvGROW(tmpstr, 80);
11345 if (term == '\'') {
11346 op_type = OP_CONST;
11347 SvIV_set(tmpstr, -1);
11349 else if (term == '`') {
11350 op_type = OP_BACKTICK;
11351 SvIV_set(tmpstr, '\\');
11355 PL_multi_start = CopLINE(PL_curcop);
11356 PL_multi_open = PL_multi_close = '<';
11357 term = *PL_tokenbuf;
11358 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11359 char * const bufptr = PL_sublex_info.super_bufptr;
11360 char * const bufend = PL_sublex_info.super_bufend;
11361 char * const olds = s - SvCUR(herewas);
11362 s = strchr(bufptr, '\n');
11366 while (s < bufend &&
11367 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11369 CopLINE_inc(PL_curcop);
11372 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11373 missingterm(PL_tokenbuf);
11375 sv_setpvn(herewas,bufptr,d-bufptr+1);
11376 sv_setpvn(tmpstr,d+1,s-d);
11378 sv_catpvn(herewas,s,bufend-s);
11379 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11386 while (s < PL_bufend &&
11387 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11389 CopLINE_inc(PL_curcop);
11391 if (s >= PL_bufend) {
11392 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11393 missingterm(PL_tokenbuf);
11395 sv_setpvn(tmpstr,d+1,s-d);
11397 if (PL_madskills) {
11399 sv_catpvn(PL_thisstuff, d + 1, s - d);
11401 PL_thisstuff = newSVpvn(d + 1, s - d);
11402 stuffstart = s - SvPVX(PL_linestr);
11406 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11408 sv_catpvn(herewas,s,PL_bufend-s);
11409 sv_setsv(PL_linestr,herewas);
11410 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11411 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11412 PL_last_lop = PL_last_uni = NULL;
11415 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
11416 while (s >= PL_bufend) { /* multiple line string? */
11418 if (PL_madskills) {
11419 tstart = SvPVX(PL_linestr) + stuffstart;
11421 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11423 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11427 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11428 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11429 missingterm(PL_tokenbuf);
11432 stuffstart = s - SvPVX(PL_linestr);
11434 CopLINE_inc(PL_curcop);
11435 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11436 PL_last_lop = PL_last_uni = NULL;
11437 #ifndef PERL_STRICT_CR
11438 if (PL_bufend - PL_linestart >= 2) {
11439 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11440 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11442 PL_bufend[-2] = '\n';
11444 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11446 else if (PL_bufend[-1] == '\r')
11447 PL_bufend[-1] = '\n';
11449 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11450 PL_bufend[-1] = '\n';
11452 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
11453 update_debugger_info(PL_linestr, NULL, 0);
11454 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11455 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11456 *(SvPVX(PL_linestr) + off ) = ' ';
11457 sv_catsv(PL_linestr,herewas);
11458 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11459 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11463 sv_catsv(tmpstr,PL_linestr);
11468 PL_multi_end = CopLINE(PL_curcop);
11469 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11470 SvPV_shrink_to_cur(tmpstr);
11472 SvREFCNT_dec(herewas);
11474 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11476 else if (PL_encoding)
11477 sv_recode_to_utf8(tmpstr, PL_encoding);
11479 PL_lex_stuff = tmpstr;
11480 pl_yylval.ival = op_type;
11484 /* scan_inputsymbol
11485 takes: current position in input buffer
11486 returns: new position in input buffer
11487 side-effects: pl_yylval and lex_op are set.
11492 <FH> read from filehandle
11493 <pkg::FH> read from package qualified filehandle
11494 <pkg'FH> read from package qualified filehandle
11495 <$fh> read from filehandle in $fh
11496 <*.h> filename glob
11501 S_scan_inputsymbol(pTHX_ char *start)
11504 register char *s = start; /* current position in buffer */
11507 char *d = PL_tokenbuf; /* start of temp holding space */
11508 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11510 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
11512 end = strchr(s, '\n');
11515 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11517 /* die if we didn't have space for the contents of the <>,
11518 or if it didn't end, or if we see a newline
11521 if (len >= (I32)sizeof PL_tokenbuf)
11522 Perl_croak(aTHX_ "Excessively long <> operator");
11524 Perl_croak(aTHX_ "Unterminated <> operator");
11529 Remember, only scalar variables are interpreted as filehandles by
11530 this code. Anything more complex (e.g., <$fh{$num}>) will be
11531 treated as a glob() call.
11532 This code makes use of the fact that except for the $ at the front,
11533 a scalar variable and a filehandle look the same.
11535 if (*d == '$' && d[1]) d++;
11537 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11538 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11541 /* If we've tried to read what we allow filehandles to look like, and
11542 there's still text left, then it must be a glob() and not a getline.
11543 Use scan_str to pull out the stuff between the <> and treat it
11544 as nothing more than a string.
11547 if (d - PL_tokenbuf != len) {
11548 pl_yylval.ival = OP_GLOB;
11549 s = scan_str(start,!!PL_madskills,FALSE);
11551 Perl_croak(aTHX_ "Glob not terminated");
11555 bool readline_overriden = FALSE;
11558 /* we're in a filehandle read situation */
11561 /* turn <> into <ARGV> */
11563 Copy("ARGV",d,5,char);
11565 /* Check whether readline() is overriden */
11566 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11568 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11570 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11571 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11572 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11573 readline_overriden = TRUE;
11575 /* if <$fh>, create the ops to turn the variable into a
11579 /* try to find it in the pad for this block, otherwise find
11580 add symbol table ops
11582 const PADOFFSET tmp = pad_findmy(d);
11583 if (tmp != NOT_IN_PAD) {
11584 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11585 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11586 HEK * const stashname = HvNAME_HEK(stash);
11587 SV * const sym = sv_2mortal(newSVhek(stashname));
11588 sv_catpvs(sym, "::");
11589 sv_catpv(sym, d+1);
11594 OP * const o = newOP(OP_PADSV, 0);
11596 PL_lex_op = readline_overriden
11597 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11598 append_elem(OP_LIST, o,
11599 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11600 : (OP*)newUNOP(OP_READLINE, 0, o);
11609 ? (GV_ADDMULTI | GV_ADDINEVAL)
11612 PL_lex_op = readline_overriden
11613 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11614 append_elem(OP_LIST,
11615 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11616 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11617 : (OP*)newUNOP(OP_READLINE, 0,
11618 newUNOP(OP_RV2SV, 0,
11619 newGVOP(OP_GV, 0, gv)));
11621 if (!readline_overriden)
11622 PL_lex_op->op_flags |= OPf_SPECIAL;
11623 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
11624 pl_yylval.ival = OP_NULL;
11627 /* If it's none of the above, it must be a literal filehandle
11628 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11630 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11631 PL_lex_op = readline_overriden
11632 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11633 append_elem(OP_LIST,
11634 newGVOP(OP_GV, 0, gv),
11635 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11636 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11637 pl_yylval.ival = OP_NULL;
11646 takes: start position in buffer
11647 keep_quoted preserve \ on the embedded delimiter(s)
11648 keep_delims preserve the delimiters around the string
11649 returns: position to continue reading from buffer
11650 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11651 updates the read buffer.
11653 This subroutine pulls a string out of the input. It is called for:
11654 q single quotes q(literal text)
11655 ' single quotes 'literal text'
11656 qq double quotes qq(interpolate $here please)
11657 " double quotes "interpolate $here please"
11658 qx backticks qx(/bin/ls -l)
11659 ` backticks `/bin/ls -l`
11660 qw quote words @EXPORT_OK = qw( func() $spam )
11661 m// regexp match m/this/
11662 s/// regexp substitute s/this/that/
11663 tr/// string transliterate tr/this/that/
11664 y/// string transliterate y/this/that/
11665 ($*@) sub prototypes sub foo ($)
11666 (stuff) sub attr parameters sub foo : attr(stuff)
11667 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11669 In most of these cases (all but <>, patterns and transliterate)
11670 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11671 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11672 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11675 It skips whitespace before the string starts, and treats the first
11676 character as the delimiter. If the delimiter is one of ([{< then
11677 the corresponding "close" character )]}> is used as the closing
11678 delimiter. It allows quoting of delimiters, and if the string has
11679 balanced delimiters ([{<>}]) it allows nesting.
11681 On success, the SV with the resulting string is put into lex_stuff or,
11682 if that is already non-NULL, into lex_repl. The second case occurs only
11683 when parsing the RHS of the special constructs s/// and tr/// (y///).
11684 For convenience, the terminating delimiter character is stuffed into
11689 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11692 SV *sv; /* scalar value: string */
11693 const char *tmps; /* temp string, used for delimiter matching */
11694 register char *s = start; /* current position in the buffer */
11695 register char term; /* terminating character */
11696 register char *to; /* current position in the sv's data */
11697 I32 brackets = 1; /* bracket nesting level */
11698 bool has_utf8 = FALSE; /* is there any utf8 content? */
11699 I32 termcode; /* terminating char. code */
11700 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11701 STRLEN termlen; /* length of terminating string */
11702 int last_off = 0; /* last position for nesting bracket */
11708 PERL_ARGS_ASSERT_SCAN_STR;
11710 /* skip space before the delimiter */
11716 if (PL_realtokenstart >= 0) {
11717 stuffstart = PL_realtokenstart;
11718 PL_realtokenstart = -1;
11721 stuffstart = start - SvPVX(PL_linestr);
11723 /* mark where we are, in case we need to report errors */
11726 /* after skipping whitespace, the next character is the terminator */
11729 termcode = termstr[0] = term;
11733 termcode = utf8_to_uvchr((U8*)s, &termlen);
11734 Copy(s, termstr, termlen, U8);
11735 if (!UTF8_IS_INVARIANT(term))
11739 /* mark where we are */
11740 PL_multi_start = CopLINE(PL_curcop);
11741 PL_multi_open = term;
11743 /* find corresponding closing delimiter */
11744 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11745 termcode = termstr[0] = term = tmps[5];
11747 PL_multi_close = term;
11749 /* create a new SV to hold the contents. 79 is the SV's initial length.
11750 What a random number. */
11751 sv = newSV_type(SVt_PVIV);
11753 SvIV_set(sv, termcode);
11754 (void)SvPOK_only(sv); /* validate pointer */
11756 /* move past delimiter and try to read a complete string */
11758 sv_catpvn(sv, s, termlen);
11761 tstart = SvPVX(PL_linestr) + stuffstart;
11762 if (!PL_thisopen && !keep_delims) {
11763 PL_thisopen = newSVpvn(tstart, s - tstart);
11764 stuffstart = s - SvPVX(PL_linestr);
11768 if (PL_encoding && !UTF) {
11772 int offset = s - SvPVX_const(PL_linestr);
11773 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11774 &offset, (char*)termstr, termlen);
11775 const char * const ns = SvPVX_const(PL_linestr) + offset;
11776 char * const svlast = SvEND(sv) - 1;
11778 for (; s < ns; s++) {
11779 if (*s == '\n' && !PL_rsfp)
11780 CopLINE_inc(PL_curcop);
11783 goto read_more_line;
11785 /* handle quoted delimiters */
11786 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11788 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11790 if ((svlast-1 - t) % 2) {
11791 if (!keep_quoted) {
11792 *(svlast-1) = term;
11794 SvCUR_set(sv, SvCUR(sv) - 1);
11799 if (PL_multi_open == PL_multi_close) {
11805 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11806 /* At here, all closes are "was quoted" one,
11807 so we don't check PL_multi_close. */
11809 if (!keep_quoted && *(t+1) == PL_multi_open)
11814 else if (*t == PL_multi_open)
11822 SvCUR_set(sv, w - SvPVX_const(sv));
11824 last_off = w - SvPVX(sv);
11825 if (--brackets <= 0)
11830 if (!keep_delims) {
11831 SvCUR_set(sv, SvCUR(sv) - 1);
11837 /* extend sv if need be */
11838 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11839 /* set 'to' to the next character in the sv's string */
11840 to = SvPVX(sv)+SvCUR(sv);
11842 /* if open delimiter is the close delimiter read unbridle */
11843 if (PL_multi_open == PL_multi_close) {
11844 for (; s < PL_bufend; s++,to++) {
11845 /* embedded newlines increment the current line number */
11846 if (*s == '\n' && !PL_rsfp)
11847 CopLINE_inc(PL_curcop);
11848 /* handle quoted delimiters */
11849 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11850 if (!keep_quoted && s[1] == term)
11852 /* any other quotes are simply copied straight through */
11856 /* terminate when run out of buffer (the for() condition), or
11857 have found the terminator */
11858 else if (*s == term) {
11861 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11864 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11870 /* if the terminator isn't the same as the start character (e.g.,
11871 matched brackets), we have to allow more in the quoting, and
11872 be prepared for nested brackets.
11875 /* read until we run out of string, or we find the terminator */
11876 for (; s < PL_bufend; s++,to++) {
11877 /* embedded newlines increment the line count */
11878 if (*s == '\n' && !PL_rsfp)
11879 CopLINE_inc(PL_curcop);
11880 /* backslashes can escape the open or closing characters */
11881 if (*s == '\\' && s+1 < PL_bufend) {
11882 if (!keep_quoted &&
11883 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11888 /* allow nested opens and closes */
11889 else if (*s == PL_multi_close && --brackets <= 0)
11891 else if (*s == PL_multi_open)
11893 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11898 /* terminate the copied string and update the sv's end-of-string */
11900 SvCUR_set(sv, to - SvPVX_const(sv));
11903 * this next chunk reads more into the buffer if we're not done yet
11907 break; /* handle case where we are done yet :-) */
11909 #ifndef PERL_STRICT_CR
11910 if (to - SvPVX_const(sv) >= 2) {
11911 if ((to[-2] == '\r' && to[-1] == '\n') ||
11912 (to[-2] == '\n' && to[-1] == '\r'))
11916 SvCUR_set(sv, to - SvPVX_const(sv));
11918 else if (to[-1] == '\r')
11921 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11926 /* if we're out of file, or a read fails, bail and reset the current
11927 line marker so we can report where the unterminated string began
11930 if (PL_madskills) {
11931 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11933 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11935 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11939 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11941 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11947 /* we read a line, so increment our line counter */
11948 CopLINE_inc(PL_curcop);
11950 /* update debugger info */
11951 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
11952 update_debugger_info(PL_linestr, NULL, 0);
11954 /* having changed the buffer, we must update PL_bufend */
11955 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11956 PL_last_lop = PL_last_uni = NULL;
11959 /* at this point, we have successfully read the delimited string */
11961 if (!PL_encoding || UTF) {
11963 if (PL_madskills) {
11964 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11965 const int len = s - tstart;
11967 sv_catpvn(PL_thisstuff, tstart, len);
11969 PL_thisstuff = newSVpvn(tstart, len);
11970 if (!PL_thisclose && !keep_delims)
11971 PL_thisclose = newSVpvn(s,termlen);
11976 sv_catpvn(sv, s, termlen);
11981 if (PL_madskills) {
11982 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11983 const int len = s - tstart - termlen;
11985 sv_catpvn(PL_thisstuff, tstart, len);
11987 PL_thisstuff = newSVpvn(tstart, len);
11988 if (!PL_thisclose && !keep_delims)
11989 PL_thisclose = newSVpvn(s - termlen,termlen);
11993 if (has_utf8 || PL_encoding)
11996 PL_multi_end = CopLINE(PL_curcop);
11998 /* if we allocated too much space, give some back */
11999 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12000 SvLEN_set(sv, SvCUR(sv) + 1);
12001 SvPV_renew(sv, SvLEN(sv));
12004 /* decide whether this is the first or second quoted string we've read
12017 takes: pointer to position in buffer
12018 returns: pointer to new position in buffer
12019 side-effects: builds ops for the constant in pl_yylval.op
12021 Read a number in any of the formats that Perl accepts:
12023 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12024 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12027 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12029 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12032 If it reads a number without a decimal point or an exponent, it will
12033 try converting the number to an integer and see if it can do so
12034 without loss of precision.
12038 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12041 register const char *s = start; /* current position in buffer */
12042 register char *d; /* destination in temp buffer */
12043 register char *e; /* end of temp buffer */
12044 NV nv; /* number read, as a double */
12045 SV *sv = NULL; /* place to put the converted number */
12046 bool floatit; /* boolean: int or float? */
12047 const char *lastub = NULL; /* position of last underbar */
12048 static char const number_too_long[] = "Number too long";
12050 PERL_ARGS_ASSERT_SCAN_NUM;
12052 /* We use the first character to decide what type of number this is */
12056 Perl_croak(aTHX_ "panic: scan_num");
12058 /* if it starts with a 0, it could be an octal number, a decimal in
12059 0.13 disguise, or a hexadecimal number, or a binary number. */
12063 u holds the "number so far"
12064 shift the power of 2 of the base
12065 (hex == 4, octal == 3, binary == 1)
12066 overflowed was the number more than we can hold?
12068 Shift is used when we add a digit. It also serves as an "are
12069 we in octal/hex/binary?" indicator to disallow hex characters
12070 when in octal mode.
12075 bool overflowed = FALSE;
12076 bool just_zero = TRUE; /* just plain 0 or binary number? */
12077 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12078 static const char* const bases[5] =
12079 { "", "binary", "", "octal", "hexadecimal" };
12080 static const char* const Bases[5] =
12081 { "", "Binary", "", "Octal", "Hexadecimal" };
12082 static const char* const maxima[5] =
12084 "0b11111111111111111111111111111111",
12088 const char *base, *Base, *max;
12090 /* check for hex */
12095 } else if (s[1] == 'b') {
12100 /* check for a decimal in disguise */
12101 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
12103 /* so it must be octal */
12110 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12111 "Misplaced _ in number");
12115 base = bases[shift];
12116 Base = Bases[shift];
12117 max = maxima[shift];
12119 /* read the rest of the number */
12121 /* x is used in the overflow test,
12122 b is the digit we're adding on. */
12127 /* if we don't mention it, we're done */
12131 /* _ are ignored -- but warned about if consecutive */
12133 if (lastub && s == lastub + 1)
12134 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12135 "Misplaced _ in number");
12139 /* 8 and 9 are not octal */
12140 case '8': case '9':
12142 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12146 case '2': case '3': case '4':
12147 case '5': case '6': case '7':
12149 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12152 case '0': case '1':
12153 b = *s++ & 15; /* ASCII digit -> value of digit */
12157 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12158 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12159 /* make sure they said 0x */
12162 b = (*s++ & 7) + 9;
12164 /* Prepare to put the digit we have onto the end
12165 of the number so far. We check for overflows.
12171 x = u << shift; /* make room for the digit */
12173 if ((x >> shift) != u
12174 && !(PL_hints & HINT_NEW_BINARY)) {
12177 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
12178 "Integer overflow in %s number",
12181 u = x | b; /* add the digit to the end */
12184 n *= nvshift[shift];
12185 /* If an NV has not enough bits in its
12186 * mantissa to represent an UV this summing of
12187 * small low-order numbers is a waste of time
12188 * (because the NV cannot preserve the
12189 * low-order bits anyway): we could just
12190 * remember when did we overflow and in the
12191 * end just multiply n by the right
12199 /* if we get here, we had success: make a scalar value from
12204 /* final misplaced underbar check */
12205 if (s[-1] == '_') {
12206 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12211 if (n > 4294967295.0)
12212 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12213 "%s number > %s non-portable",
12219 if (u > 0xffffffff)
12220 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12221 "%s number > %s non-portable",
12226 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12227 sv = new_constant(start, s - start, "integer",
12228 sv, NULL, NULL, 0);
12229 else if (PL_hints & HINT_NEW_BINARY)
12230 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12235 handle decimal numbers.
12236 we're also sent here when we read a 0 as the first digit
12238 case '1': case '2': case '3': case '4': case '5':
12239 case '6': case '7': case '8': case '9': case '.':
12242 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12245 /* read next group of digits and _ and copy into d */
12246 while (isDIGIT(*s) || *s == '_') {
12247 /* skip underscores, checking for misplaced ones
12251 if (lastub && s == lastub + 1)
12252 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12253 "Misplaced _ in number");
12257 /* check for end of fixed-length buffer */
12259 Perl_croak(aTHX_ number_too_long);
12260 /* if we're ok, copy the character */
12265 /* final misplaced underbar check */
12266 if (lastub && s == lastub + 1) {
12267 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12270 /* read a decimal portion if there is one. avoid
12271 3..5 being interpreted as the number 3. followed
12274 if (*s == '.' && s[1] != '.') {
12279 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12280 "Misplaced _ in number");
12284 /* copy, ignoring underbars, until we run out of digits.
12286 for (; isDIGIT(*s) || *s == '_'; s++) {
12287 /* fixed length buffer check */
12289 Perl_croak(aTHX_ number_too_long);
12291 if (lastub && s == lastub + 1)
12292 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12293 "Misplaced _ in number");
12299 /* fractional part ending in underbar? */
12300 if (s[-1] == '_') {
12301 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12302 "Misplaced _ in number");
12304 if (*s == '.' && isDIGIT(s[1])) {
12305 /* oops, it's really a v-string, but without the "v" */
12311 /* read exponent part, if present */
12312 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12316 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12317 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12319 /* stray preinitial _ */
12321 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12322 "Misplaced _ in number");
12326 /* allow positive or negative exponent */
12327 if (*s == '+' || *s == '-')
12330 /* stray initial _ */
12332 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12333 "Misplaced _ in number");
12337 /* read digits of exponent */
12338 while (isDIGIT(*s) || *s == '_') {
12341 Perl_croak(aTHX_ number_too_long);
12345 if (((lastub && s == lastub + 1) ||
12346 (!isDIGIT(s[1]) && s[1] != '_')))
12347 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12348 "Misplaced _ in number");
12355 /* make an sv from the string */
12359 We try to do an integer conversion first if no characters
12360 indicating "float" have been found.
12365 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12367 if (flags == IS_NUMBER_IN_UV) {
12369 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12372 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12373 if (uv <= (UV) IV_MIN)
12374 sv_setiv(sv, -(IV)uv);
12381 /* terminate the string */
12383 nv = Atof(PL_tokenbuf);
12388 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12389 const char *const key = floatit ? "float" : "integer";
12390 const STRLEN keylen = floatit ? 5 : 7;
12391 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12392 key, keylen, sv, NULL, NULL, 0);
12396 /* if it starts with a v, it could be a v-string */
12399 sv = newSV(5); /* preallocate storage space */
12400 s = scan_vstring(s, PL_bufend, sv);
12404 /* make the op for the constant and return */
12407 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12409 lvalp->opval = NULL;
12415 S_scan_formline(pTHX_ register char *s)
12418 register char *eol;
12420 SV * const stuff = newSVpvs("");
12421 bool needargs = FALSE;
12422 bool eofmt = FALSE;
12424 char *tokenstart = s;
12425 SV* savewhite = NULL;
12427 if (PL_madskills) {
12428 savewhite = PL_thiswhite;
12433 PERL_ARGS_ASSERT_SCAN_FORMLINE;
12435 while (!needargs) {
12438 #ifdef PERL_STRICT_CR
12439 while (SPACE_OR_TAB(*t))
12442 while (SPACE_OR_TAB(*t) || *t == '\r')
12445 if (*t == '\n' || t == PL_bufend) {
12450 if (PL_in_eval && !PL_rsfp) {
12451 eol = (char *) memchr(s,'\n',PL_bufend-s);
12456 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12458 for (t = s; t < eol; t++) {
12459 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12461 goto enough; /* ~~ must be first line in formline */
12463 if (*t == '@' || *t == '^')
12467 sv_catpvn(stuff, s, eol-s);
12468 #ifndef PERL_STRICT_CR
12469 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12470 char *end = SvPVX(stuff) + SvCUR(stuff);
12473 SvCUR_set(stuff, SvCUR(stuff) - 1);
12483 if (PL_madskills) {
12485 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12487 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12490 s = filter_gets(PL_linestr, PL_rsfp, 0);
12492 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12494 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12496 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12497 PL_last_lop = PL_last_uni = NULL;
12506 if (SvCUR(stuff)) {
12509 PL_lex_state = LEX_NORMAL;
12510 start_force(PL_curforce);
12511 NEXTVAL_NEXTTOKE.ival = 0;
12515 PL_lex_state = LEX_FORMLINE;
12517 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12519 else if (PL_encoding)
12520 sv_recode_to_utf8(stuff, PL_encoding);
12522 start_force(PL_curforce);
12523 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12525 start_force(PL_curforce);
12526 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12530 SvREFCNT_dec(stuff);
12532 PL_lex_formbrack = 0;
12536 if (PL_madskills) {
12538 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12540 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12541 PL_thiswhite = savewhite;
12548 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12551 const I32 oldsavestack_ix = PL_savestack_ix;
12552 CV* const outsidecv = PL_compcv;
12555 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12557 SAVEI32(PL_subline);
12558 save_item(PL_subname);
12559 SAVESPTR(PL_compcv);
12561 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
12562 CvFLAGS(PL_compcv) |= flags;
12564 PL_subline = CopLINE(PL_curcop);
12565 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12566 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
12567 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12569 return oldsavestack_ix;
12573 #pragma segment Perl_yylex
12576 S_yywarn(pTHX_ const char *const s)
12580 PERL_ARGS_ASSERT_YYWARN;
12582 PL_in_eval |= EVAL_WARNONLY;
12584 PL_in_eval &= ~EVAL_WARNONLY;
12589 Perl_yyerror(pTHX_ const char *const s)
12592 const char *where = NULL;
12593 const char *context = NULL;
12596 int yychar = PL_parser->yychar;
12598 PERL_ARGS_ASSERT_YYERROR;
12600 if (!yychar || (yychar == ';' && !PL_rsfp))
12602 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12603 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12604 PL_oldbufptr != PL_bufptr) {
12607 The code below is removed for NetWare because it abends/crashes on NetWare
12608 when the script has error such as not having the closing quotes like:
12609 if ($var eq "value)
12610 Checking of white spaces is anyway done in NetWare code.
12613 while (isSPACE(*PL_oldoldbufptr))
12616 context = PL_oldoldbufptr;
12617 contlen = PL_bufptr - PL_oldoldbufptr;
12619 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12620 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12623 The code below is removed for NetWare because it abends/crashes on NetWare
12624 when the script has error such as not having the closing quotes like:
12625 if ($var eq "value)
12626 Checking of white spaces is anyway done in NetWare code.
12629 while (isSPACE(*PL_oldbufptr))
12632 context = PL_oldbufptr;
12633 contlen = PL_bufptr - PL_oldbufptr;
12635 else if (yychar > 255)
12636 where = "next token ???";
12637 else if (yychar == -2) { /* YYEMPTY */
12638 if (PL_lex_state == LEX_NORMAL ||
12639 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12640 where = "at end of line";
12641 else if (PL_lex_inpat)
12642 where = "within pattern";
12644 where = "within string";
12647 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
12649 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12650 else if (isPRINT_LC(yychar)) {
12651 const char string = yychar;
12652 sv_catpvn(where_sv, &string, 1);
12655 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12656 where = SvPVX_const(where_sv);
12658 msg = sv_2mortal(newSVpv(s, 0));
12659 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12660 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12662 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12664 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12665 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12666 Perl_sv_catpvf(aTHX_ msg,
12667 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12668 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12671 if (PL_in_eval & EVAL_WARNONLY) {
12672 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12676 if (PL_error_count >= 10) {
12677 if (PL_in_eval && SvCUR(ERRSV))
12678 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12679 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12681 Perl_croak(aTHX_ "%s has too many errors.\n",
12682 OutCopFILE(PL_curcop));
12685 PL_in_my_stash = NULL;
12689 #pragma segment Main
12693 S_swallow_bom(pTHX_ U8 *s)
12696 const STRLEN slen = SvCUR(PL_linestr);
12698 PERL_ARGS_ASSERT_SWALLOW_BOM;
12702 if (s[1] == 0xFE) {
12703 /* UTF-16 little-endian? (or UTF32-LE?) */
12704 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12705 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12706 #ifndef PERL_NO_UTF16_FILTER
12707 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12710 if (PL_bufend > (char*)s) {
12714 filter_add(utf16rev_textfilter, NULL);
12715 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12716 utf16_to_utf8_reversed(s, news,
12717 PL_bufend - (char*)s - 1,
12719 sv_setpvn(PL_linestr, (const char*)news, newlen);
12721 s = (U8*)SvPVX(PL_linestr);
12722 Copy(news, s, newlen, U8);
12726 SvUTF8_on(PL_linestr);
12727 s = (U8*)SvPVX(PL_linestr);
12729 /* FIXME - is this a general bug fix? */
12732 PL_bufend = SvPVX(PL_linestr) + newlen;
12735 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12740 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12741 #ifndef PERL_NO_UTF16_FILTER
12742 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12745 if (PL_bufend > (char *)s) {
12749 filter_add(utf16_textfilter, NULL);
12750 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12751 utf16_to_utf8(s, news,
12752 PL_bufend - (char*)s,
12754 sv_setpvn(PL_linestr, (const char*)news, newlen);
12756 SvUTF8_on(PL_linestr);
12757 s = (U8*)SvPVX(PL_linestr);
12758 PL_bufend = SvPVX(PL_linestr) + newlen;
12761 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12766 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12767 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12768 s += 3; /* UTF-8 */
12774 if (s[2] == 0xFE && s[3] == 0xFF) {
12775 /* UTF-32 big-endian */
12776 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12779 else if (s[2] == 0 && s[3] != 0) {
12782 * are a good indicator of UTF-16BE. */
12783 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12789 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12790 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12791 s += 4; /* UTF-8 */
12797 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12800 * are a good indicator of UTF-16LE. */
12801 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12809 #ifndef PERL_NO_UTF16_FILTER
12811 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12814 const STRLEN old = SvCUR(sv);
12815 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12816 DEBUG_P(PerlIO_printf(Perl_debug_log,
12817 "utf16_textfilter(%p): %d %d (%d)\n",
12818 FPTR2DPTR(void *, utf16_textfilter),
12819 idx, maxlen, (int) count));
12823 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12824 Copy(SvPVX_const(sv), tmps, old, char);
12825 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12826 SvCUR(sv) - old, &newlen);
12827 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12829 DEBUG_P({sv_dump(sv);});
12834 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12837 const STRLEN old = SvCUR(sv);
12838 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12839 DEBUG_P(PerlIO_printf(Perl_debug_log,
12840 "utf16rev_textfilter(%p): %d %d (%d)\n",
12841 FPTR2DPTR(void *, utf16rev_textfilter),
12842 idx, maxlen, (int) count));
12846 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12847 Copy(SvPVX_const(sv), tmps, old, char);
12848 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12849 SvCUR(sv) - old, &newlen);
12850 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12852 DEBUG_P({ sv_dump(sv); });
12858 Returns a pointer to the next character after the parsed
12859 vstring, as well as updating the passed in sv.
12861 Function must be called like
12864 s = scan_vstring(s,e,sv);
12866 where s and e are the start and end of the string.
12867 The sv should already be large enough to store the vstring
12868 passed in, for performance reasons.
12873 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
12876 const char *pos = s;
12877 const char *start = s;
12879 PERL_ARGS_ASSERT_SCAN_VSTRING;
12881 if (*pos == 'v') pos++; /* get past 'v' */
12882 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12884 if ( *pos != '.') {
12885 /* this may not be a v-string if followed by => */
12886 const char *next = pos;
12887 while (next < e && isSPACE(*next))
12889 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12890 /* return string not v-string */
12891 sv_setpvn(sv,(char *)s,pos-s);
12892 return (char *)pos;
12896 if (!isALPHA(*pos)) {
12897 U8 tmpbuf[UTF8_MAXBYTES+1];
12900 s++; /* get past 'v' */
12905 /* this is atoi() that tolerates underscores */
12908 const char *end = pos;
12910 while (--end >= s) {
12912 const UV orev = rev;
12913 rev += (*end - '0') * mult;
12916 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
12917 "Integer overflow in decimal number");
12921 if (rev > 0x7FFFFFFF)
12922 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12924 /* Append native character for the rev point */
12925 tmpend = uvchr_to_utf8(tmpbuf, rev);
12926 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12927 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12929 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12935 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12939 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12947 * c-indentation-style: bsd
12948 * c-basic-offset: 4
12949 * indent-tabs-mode: t
12952 * ex: set ts=8 sts=4 sw=4 noet: