3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * "It all comes from here, the stench and the peril." --Frodo
16 * This file is the lexer for Perl. It's closely linked to the
19 * The main routine is yylex(), which returns the next token.
23 #define PERL_IN_TOKE_C
26 #define yylval (PL_parser->yylval)
28 /* YYINITDEPTH -- initial size of the parser's stacks. */
29 #define YYINITDEPTH 200
31 /* XXX temporary backwards compatibility */
32 #define PL_lex_brackets (PL_parser->lex_brackets)
33 #define PL_lex_brackstack (PL_parser->lex_brackstack)
34 #define PL_lex_casemods (PL_parser->lex_casemods)
35 #define PL_lex_casestack (PL_parser->lex_casestack)
36 #define PL_lex_defer (PL_parser->lex_defer)
37 #define PL_lex_dojoin (PL_parser->lex_dojoin)
38 #define PL_lex_expect (PL_parser->lex_expect)
39 #define PL_lex_formbrack (PL_parser->lex_formbrack)
40 #define PL_lex_inpat (PL_parser->lex_inpat)
41 #define PL_lex_inwhat (PL_parser->lex_inwhat)
42 #define PL_lex_op (PL_parser->lex_op)
43 #define PL_lex_repl (PL_parser->lex_repl)
44 #define PL_lex_starts (PL_parser->lex_starts)
45 #define PL_lex_stuff (PL_parser->lex_stuff)
46 #define PL_multi_start (PL_parser->multi_start)
47 #define PL_multi_open (PL_parser->multi_open)
48 #define PL_multi_close (PL_parser->multi_close)
49 #define PL_pending_ident (PL_parser->pending_ident)
50 #define PL_preambled (PL_parser->preambled)
51 #define PL_sublex_info (PL_parser->sublex_info)
52 #define PL_linestr (PL_parser->linestr)
53 #define PL_expect (PL_parser->expect)
54 #define PL_copline (PL_parser->copline)
55 #define PL_bufptr (PL_parser->bufptr)
56 #define PL_oldbufptr (PL_parser->oldbufptr)
57 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
58 #define PL_linestart (PL_parser->linestart)
59 #define PL_bufend (PL_parser->bufend)
60 #define PL_last_uni (PL_parser->last_uni)
61 #define PL_last_lop (PL_parser->last_lop)
62 #define PL_last_lop_op (PL_parser->last_lop_op)
63 #define PL_lex_state (PL_parser->lex_state)
64 #define PL_rsfp (PL_parser->rsfp)
65 #define PL_rsfp_filters (PL_parser->rsfp_filters)
66 #define PL_in_my (PL_parser->in_my)
67 #define PL_in_my_stash (PL_parser->in_my_stash)
68 #define PL_tokenbuf (PL_parser->tokenbuf)
69 #define PL_multi_end (PL_parser->multi_end)
70 #define PL_error_count (PL_parser->error_count)
73 # define PL_endwhite (PL_parser->endwhite)
74 # define PL_faketokens (PL_parser->faketokens)
75 # define PL_lasttoke (PL_parser->lasttoke)
76 # define PL_nextwhite (PL_parser->nextwhite)
77 # define PL_realtokenstart (PL_parser->realtokenstart)
78 # define PL_skipwhite (PL_parser->skipwhite)
79 # define PL_thisclose (PL_parser->thisclose)
80 # define PL_thismad (PL_parser->thismad)
81 # define PL_thisopen (PL_parser->thisopen)
82 # define PL_thisstuff (PL_parser->thisstuff)
83 # define PL_thistoken (PL_parser->thistoken)
84 # define PL_thiswhite (PL_parser->thiswhite)
85 # define PL_thiswhite (PL_parser->thiswhite)
86 # define PL_nexttoke (PL_parser->nexttoke)
87 # define PL_curforce (PL_parser->curforce)
89 # define PL_nexttoke (PL_parser->nexttoke)
90 # define PL_nexttype (PL_parser->nexttype)
91 # define PL_nextval (PL_parser->nextval)
95 S_pending_ident(pTHX);
97 static const char ident_too_long[] = "Identifier too long";
98 static const char commaless_variable_list[] = "comma-less variable list";
100 #ifndef PERL_NO_UTF16_FILTER
101 static I32 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen);
102 static I32 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen);
106 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
107 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
109 # define CURMAD(slot,sv)
110 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
113 #define XFAKEBRACK 128
114 #define XENUMMASK 127
116 #ifdef USE_UTF8_SCRIPTS
117 # define UTF (!IN_BYTES)
119 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
122 /* In variables named $^X, these are the legal values for X.
123 * 1999-02-27 mjd-perl-patch@plover.com */
124 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
126 /* On MacOS, respect nonbreaking spaces */
127 #ifdef MACOS_TRADITIONAL
128 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\312'||(c)=='\t')
130 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
133 /* LEX_* are values for PL_lex_state, the state of the lexer.
134 * They are arranged oddly so that the guard on the switch statement
135 * can get by with a single comparison (if the compiler is smart enough).
138 /* #define LEX_NOTPARSING 11 is done in perl.h. */
140 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
141 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
142 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
143 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
144 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
146 /* at end of code, eg "$x" followed by: */
147 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
148 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
150 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
151 string or after \E, $foo, etc */
152 #define LEX_INTERPCONST 2 /* NOT USED */
153 #define LEX_FORMLINE 1 /* expecting a format line */
154 #define LEX_KNOWNEXT 0 /* next token known; just return it */
158 static const char* const lex_state_names[] = {
177 #include "keywords.h"
179 /* CLINE is a macro that ensures PL_copline has a sane value */
184 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
187 # define SKIPSPACE0(s) skipspace0(s)
188 # define SKIPSPACE1(s) skipspace1(s)
189 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
190 # define PEEKSPACE(s) skipspace2(s,0)
192 # define SKIPSPACE0(s) skipspace(s)
193 # define SKIPSPACE1(s) skipspace(s)
194 # define SKIPSPACE2(s,tsv) skipspace(s)
195 # define PEEKSPACE(s) skipspace(s)
199 * Convenience functions to return different tokens and prime the
200 * lexer for the next token. They all take an argument.
202 * TOKEN : generic token (used for '(', DOLSHARP, etc)
203 * OPERATOR : generic operator
204 * AOPERATOR : assignment operator
205 * PREBLOCK : beginning the block after an if, while, foreach, ...
206 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
207 * PREREF : *EXPR where EXPR is not a simple identifier
208 * TERM : expression term
209 * LOOPX : loop exiting command (goto, last, dump, etc)
210 * FTST : file test operator
211 * FUN0 : zero-argument function
212 * FUN1 : not used, except for not, which isn't a UNIOP
213 * BOop : bitwise or or xor
215 * SHop : shift operator
216 * PWop : power operator
217 * PMop : pattern-matching operator
218 * Aop : addition-level operator
219 * Mop : multiplication-level operator
220 * Eop : equality-testing operator
221 * Rop : relational operator <= != gt
223 * Also see LOP and lop() below.
226 #ifdef DEBUGGING /* Serve -DT. */
227 # define REPORT(retval) tokereport((I32)retval)
229 # define REPORT(retval) (retval)
232 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
233 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
234 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
235 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
236 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
237 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
238 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
239 #define LOOPX(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
240 #define FTST(f) return (yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
241 #define FUN0(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
242 #define FUN1(f) return (yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
243 #define BOop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
244 #define BAop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
245 #define SHop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
246 #define PWop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
247 #define PMop(f) return(yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
248 #define Aop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
249 #define Mop(f) return ao((yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
250 #define Eop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
251 #define Rop(f) return (yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
253 /* This bit of chicanery makes a unary function followed by
254 * a parenthesis into a function with one argument, highest precedence.
255 * The UNIDOR macro is for unary functions that can be followed by the //
256 * operator (such as C<shift // 0>).
258 #define UNI2(f,x) { \
262 PL_last_uni = PL_oldbufptr; \
263 PL_last_lop_op = f; \
265 return REPORT( (int)FUNC1 ); \
267 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
269 #define UNI(f) UNI2(f,XTERM)
270 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
272 #define UNIBRACK(f) { \
275 PL_last_uni = PL_oldbufptr; \
277 return REPORT( (int)FUNC1 ); \
279 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
282 /* grandfather return to old style */
283 #define OLDLOP(f) return(yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
287 /* how to interpret the yylval associated with the token */
291 TOKENTYPE_OPNUM, /* yylval.ival contains an opcode number */
297 static struct debug_tokens {
299 enum token_type type;
301 } const debug_tokens[] =
303 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
304 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
305 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
306 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
307 { ARROW, TOKENTYPE_NONE, "ARROW" },
308 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
309 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
310 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
311 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
312 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
313 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
314 { DO, TOKENTYPE_NONE, "DO" },
315 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
316 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
317 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
318 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
319 { ELSE, TOKENTYPE_NONE, "ELSE" },
320 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
321 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
322 { FOR, TOKENTYPE_IVAL, "FOR" },
323 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
324 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
325 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
326 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
327 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
328 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
329 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
330 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
331 { IF, TOKENTYPE_IVAL, "IF" },
332 { LABEL, TOKENTYPE_PVAL, "LABEL" },
333 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
334 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
335 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
336 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
337 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
338 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
339 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
340 { MY, TOKENTYPE_IVAL, "MY" },
341 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
342 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
343 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
344 { OROP, TOKENTYPE_IVAL, "OROP" },
345 { OROR, TOKENTYPE_NONE, "OROR" },
346 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
347 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
348 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
349 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
350 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
351 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
352 { PREINC, TOKENTYPE_NONE, "PREINC" },
353 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
354 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
355 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
356 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
357 { SUB, TOKENTYPE_NONE, "SUB" },
358 { THING, TOKENTYPE_OPVAL, "THING" },
359 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
360 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
361 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
362 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
363 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
364 { USE, TOKENTYPE_IVAL, "USE" },
365 { WHEN, TOKENTYPE_IVAL, "WHEN" },
366 { WHILE, TOKENTYPE_IVAL, "WHILE" },
367 { WORD, TOKENTYPE_OPVAL, "WORD" },
368 { 0, TOKENTYPE_NONE, NULL }
371 /* dump the returned token in rv, plus any optional arg in yylval */
374 S_tokereport(pTHX_ I32 rv)
378 const char *name = NULL;
379 enum token_type type = TOKENTYPE_NONE;
380 const struct debug_tokens *p;
381 SV* const report = newSVpvs("<== ");
383 for (p = debug_tokens; p->token; p++) {
384 if (p->token == (int)rv) {
391 Perl_sv_catpv(aTHX_ report, name);
392 else if ((char)rv > ' ' && (char)rv < '~')
393 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
395 sv_catpvs(report, "EOF");
397 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
400 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
403 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)yylval.ival);
405 case TOKENTYPE_OPNUM:
406 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
407 PL_op_name[yylval.ival]);
410 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", yylval.pval);
412 case TOKENTYPE_OPVAL:
414 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
415 PL_op_name[yylval.opval->op_type]);
416 if (yylval.opval->op_type == OP_CONST) {
417 Perl_sv_catpvf(aTHX_ report, " %s",
418 SvPEEK(cSVOPx_sv(yylval.opval)));
423 sv_catpvs(report, "(opval=null)");
426 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
432 /* print the buffer with suitable escapes */
435 S_printbuf(pTHX_ const char* fmt, const char* s)
437 SV* const tmp = newSVpvs("");
438 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
447 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
448 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
452 S_ao(pTHX_ int toketype)
455 if (*PL_bufptr == '=') {
457 if (toketype == ANDAND)
458 yylval.ival = OP_ANDASSIGN;
459 else if (toketype == OROR)
460 yylval.ival = OP_ORASSIGN;
461 else if (toketype == DORDOR)
462 yylval.ival = OP_DORASSIGN;
470 * When Perl expects an operator and finds something else, no_op
471 * prints the warning. It always prints "<something> found where
472 * operator expected. It prints "Missing semicolon on previous line?"
473 * if the surprise occurs at the start of the line. "do you need to
474 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
475 * where the compiler doesn't know if foo is a method call or a function.
476 * It prints "Missing operator before end of line" if there's nothing
477 * after the missing operator, or "... before <...>" if there is something
478 * after the missing operator.
482 S_no_op(pTHX_ const char *what, char *s)
485 char * const oldbp = PL_bufptr;
486 const bool is_first = (PL_oldbufptr == PL_linestart);
492 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
493 if (ckWARN_d(WARN_SYNTAX)) {
495 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
496 "\t(Missing semicolon on previous line?)\n");
497 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
499 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
501 if (t < PL_bufptr && isSPACE(*t))
502 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
503 "\t(Do you need to predeclare %.*s?)\n",
504 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
508 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
509 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
517 * Complain about missing quote/regexp/heredoc terminator.
518 * If it's called with NULL then it cauterizes the line buffer.
519 * If we're in a delimited string and the delimiter is a control
520 * character, it's reformatted into a two-char sequence like ^C.
525 S_missingterm(pTHX_ char *s)
531 char * const nl = strrchr(s,'\n');
537 iscntrl(PL_multi_close)
539 PL_multi_close < 32 || PL_multi_close == 127
543 tmpbuf[1] = (char)toCTRL(PL_multi_close);
548 *tmpbuf = (char)PL_multi_close;
552 q = strchr(s,'"') ? '\'' : '"';
553 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
556 #define FEATURE_IS_ENABLED(name) \
557 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
558 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
560 * S_feature_is_enabled
561 * Check whether the named feature is enabled.
564 S_feature_is_enabled(pTHX_ const char *name, STRLEN namelen)
567 HV * const hinthv = GvHV(PL_hintgv);
568 char he_name[32] = "feature_";
569 (void) my_strlcpy(&he_name[8], name, 24);
571 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
579 Perl_deprecate(pTHX_ const char *s)
581 if (ckWARN(WARN_DEPRECATED))
582 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED), "Use of %s is deprecated", s);
586 Perl_deprecate_old(pTHX_ const char *s)
588 /* This function should NOT be called for any new deprecated warnings */
589 /* Use Perl_deprecate instead */
591 /* It is here to maintain backward compatibility with the pre-5.8 */
592 /* warnings category hierarchy. The "deprecated" category used to */
593 /* live under the "syntax" category. It is now a top-level category */
594 /* in its own right. */
596 if (ckWARN2(WARN_DEPRECATED, WARN_SYNTAX))
597 Perl_warner(aTHX_ packWARN2(WARN_DEPRECATED, WARN_SYNTAX),
598 "Use of %s is deprecated", s);
602 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
603 * utf16-to-utf8-reversed.
606 #ifdef PERL_CR_FILTER
610 register const char *s = SvPVX_const(sv);
611 register const char * const e = s + SvCUR(sv);
612 /* outer loop optimized to do nothing if there are no CR-LFs */
614 if (*s++ == '\r' && *s == '\n') {
615 /* hit a CR-LF, need to copy the rest */
616 register char *d = s - 1;
619 if (*s == '\r' && s[1] == '\n')
630 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
632 const I32 count = FILTER_READ(idx+1, sv, maxlen);
633 if (count > 0 && !maxlen)
644 * Create a parser object and initialise its parser and lexer fields
646 * rsfp is the opened file handle to read from (if any),
648 * line holds any initial content already read from the file (or in
649 * the case of no file, such as an eval, the whole contents);
651 * new_filter indicates that this is a new file and it shouldn't inherit
652 * the filters from the current parser (ie require).
656 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
659 const char *s = NULL;
661 yy_parser *parser, *oparser;
663 /* create and initialise a parser */
665 Newxz(parser, 1, yy_parser);
666 parser->old_parser = oparser = PL_parser;
669 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
670 parser->ps = parser->stack;
671 parser->stack_size = YYINITDEPTH;
673 parser->stack->state = 0;
674 parser->yyerrstatus = 0;
675 parser->yychar = YYEMPTY; /* Cause a token to be read. */
677 /* on scope exit, free this parser and restore any outer one */
679 parser->saved_curcop = PL_curcop;
681 /* initialise lexer state */
684 parser->curforce = -1;
686 parser->nexttoke = 0;
688 parser->copline = NOLINE;
689 parser->lex_state = LEX_NORMAL;
690 parser->expect = XSTATE;
692 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
693 : (AV*)SvREFCNT_inc(oparser->rsfp_filters);
695 Newx(parser->lex_brackstack, 120, char);
696 Newx(parser->lex_casestack, 12, char);
697 *parser->lex_casestack = '\0';
700 s = SvPV_const(line, len);
706 parser->linestr = newSVpvs("\n;");
707 } else if (SvREADONLY(line) || s[len-1] != ';') {
708 parser->linestr = newSVsv(line);
710 sv_catpvs(parser->linestr, "\n;");
713 SvREFCNT_inc_simple_void_NN(line);
714 parser->linestr = line;
716 parser->oldoldbufptr =
719 parser->linestart = SvPVX(parser->linestr);
720 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
721 parser->last_lop = parser->last_uni = NULL;
725 /* delete a parser object */
728 Perl_parser_free(pTHX_ const yy_parser *parser)
730 PL_curcop = parser->saved_curcop;
731 SvREFCNT_dec(parser->linestr);
733 if (parser->rsfp == PerlIO_stdin())
734 PerlIO_clearerr(parser->rsfp);
735 else if (parser->rsfp && parser->old_parser
736 && parser->rsfp != parser->old_parser->rsfp)
737 PerlIO_close(parser->rsfp);
738 SvREFCNT_dec(parser->rsfp_filters);
740 Safefree(parser->stack);
741 Safefree(parser->lex_brackstack);
742 Safefree(parser->lex_casestack);
743 PL_parser = parser->old_parser;
750 * Finalizer for lexing operations. Must be called when the parser is
751 * done with the lexer.
758 PL_doextract = FALSE;
763 * This subroutine has nothing to do with tilting, whether at windmills
764 * or pinball tables. Its name is short for "increment line". It
765 * increments the current line number in CopLINE(PL_curcop) and checks
766 * to see whether the line starts with a comment of the form
767 * # line 500 "foo.pm"
768 * If so, it sets the current line number and file to the values in the comment.
772 S_incline(pTHX_ const char *s)
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 while (SPACE_OR_TAB(*s))
802 if (*s == '"' && (t = strchr(s+1, '"'))) {
812 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
814 if (*e != '\n' && *e != '\0')
815 return; /* false alarm */
818 const STRLEN len = t - s;
820 const char * const cf = CopFILE(PL_curcop);
821 STRLEN tmplen = cf ? strlen(cf) : 0;
822 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
823 /* must copy *{"::_<(eval N)[oldfilename:L]"}
824 * to *{"::_<newfilename"} */
825 /* However, the long form of evals is only turned on by the
826 debugger - usually they're "(eval %lu)" */
830 STRLEN tmplen2 = len;
831 if (tmplen + 2 <= sizeof smallbuf)
834 Newx(tmpbuf, tmplen + 2, char);
837 memcpy(tmpbuf + 2, cf, tmplen);
839 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
844 if (tmplen2 + 2 <= sizeof smallbuf)
847 Newx(tmpbuf2, tmplen2 + 2, char);
849 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
850 /* Either they malloc'd it, or we malloc'd it,
851 so no prefix is present in ours. */
856 memcpy(tmpbuf2 + 2, s, tmplen2);
859 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
861 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
862 /* adjust ${"::_<newfilename"} to store the new file name */
863 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
864 GvHV(gv2) = (HV*)SvREFCNT_inc(GvHV(*gvp));
865 GvAV(gv2) = (AV*)SvREFCNT_inc(GvAV(*gvp));
868 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
870 if (tmpbuf != smallbuf) Safefree(tmpbuf);
873 CopFILE_free(PL_curcop);
874 CopFILE_setn(PL_curcop, s, len);
876 CopLINE_set(PL_curcop, atoi(n)-1);
880 /* skip space before PL_thistoken */
883 S_skipspace0(pTHX_ register char *s)
890 PL_thiswhite = newSVpvs("");
891 sv_catsv(PL_thiswhite, PL_skipwhite);
892 sv_free(PL_skipwhite);
895 PL_realtokenstart = s - SvPVX(PL_linestr);
899 /* skip space after PL_thistoken */
902 S_skipspace1(pTHX_ register char *s)
904 const char *start = s;
905 I32 startoff = start - SvPVX(PL_linestr);
910 start = SvPVX(PL_linestr) + startoff;
911 if (!PL_thistoken && PL_realtokenstart >= 0) {
912 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
913 PL_thistoken = newSVpvn(tstart, start - tstart);
915 PL_realtokenstart = -1;
918 PL_nextwhite = newSVpvs("");
919 sv_catsv(PL_nextwhite, PL_skipwhite);
920 sv_free(PL_skipwhite);
927 S_skipspace2(pTHX_ register char *s, SV **svp)
930 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
931 const I32 startoff = s - SvPVX(PL_linestr);
934 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
935 if (!PL_madskills || !svp)
937 start = SvPVX(PL_linestr) + startoff;
938 if (!PL_thistoken && PL_realtokenstart >= 0) {
939 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
940 PL_thistoken = newSVpvn(tstart, start - tstart);
941 PL_realtokenstart = -1;
946 sv_setsv(*svp, PL_skipwhite);
947 sv_free(PL_skipwhite);
956 S_update_debugger_info(pTHX_ SV *orig_sv, const char *buf, STRLEN len)
958 AV *av = CopFILEAVx(PL_curcop);
960 SV * const sv = newSV_type(SVt_PVMG);
962 sv_setsv(sv, orig_sv);
964 sv_setpvn(sv, buf, len);
967 av_store(av, (I32)CopLINE(PL_curcop), sv);
973 * Called to gobble the appropriate amount and type of whitespace.
974 * Skips comments as well.
978 S_skipspace(pTHX_ register char *s)
983 int startoff = s - SvPVX(PL_linestr);
986 sv_free(PL_skipwhite);
991 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
992 while (s < PL_bufend && SPACE_OR_TAB(*s))
1002 SSize_t oldprevlen, oldoldprevlen;
1003 SSize_t oldloplen = 0, oldunilen = 0;
1004 while (s < PL_bufend && isSPACE(*s)) {
1005 if (*s++ == '\n' && PL_in_eval && !PL_rsfp)
1010 if (s < PL_bufend && *s == '#') {
1011 while (s < PL_bufend && *s != '\n')
1013 if (s < PL_bufend) {
1015 if (PL_in_eval && !PL_rsfp) {
1022 /* only continue to recharge the buffer if we're at the end
1023 * of the buffer, we're not reading from a source filter, and
1024 * we're in normal lexing mode
1026 if (s < PL_bufend || !PL_rsfp || PL_sublex_info.sub_inwhat ||
1027 PL_lex_state == LEX_FORMLINE)
1034 /* try to recharge the buffer */
1036 curoff = s - SvPVX(PL_linestr);
1039 if ((s = filter_gets(PL_linestr, PL_rsfp,
1040 (prevlen = SvCUR(PL_linestr)))) == NULL)
1043 if (PL_madskills && curoff != startoff) {
1045 PL_skipwhite = newSVpvs("");
1046 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1050 /* mustn't throw out old stuff yet if madpropping */
1051 SvCUR(PL_linestr) = curoff;
1052 s = SvPVX(PL_linestr) + curoff;
1054 if (curoff && s[-1] == '\n')
1058 /* end of file. Add on the -p or -n magic */
1059 /* XXX these shouldn't really be added here, can't set PL_faketokens */
1062 sv_catpvs(PL_linestr,
1063 ";}continue{print or die qq(-p destination: $!\\n);}");
1065 sv_setpvs(PL_linestr,
1066 ";}continue{print or die qq(-p destination: $!\\n);}");
1068 PL_minus_n = PL_minus_p = 0;
1070 else if (PL_minus_n) {
1072 sv_catpvn(PL_linestr, ";}", 2);
1074 sv_setpvn(PL_linestr, ";}", 2);
1080 sv_catpvn(PL_linestr,";", 1);
1082 sv_setpvn(PL_linestr,";", 1);
1085 /* reset variables for next time we lex */
1086 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart
1092 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
1093 PL_last_lop = PL_last_uni = NULL;
1095 /* Close the filehandle. Could be from -P preprocessor,
1096 * STDIN, or a regular file. If we were reading code from
1097 * STDIN (because the commandline held no -e or filename)
1098 * then we don't close it, we reset it so the code can
1099 * read from STDIN too.
1102 if (PL_preprocess && !PL_in_eval)
1103 (void)PerlProc_pclose(PL_rsfp);
1104 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
1105 PerlIO_clearerr(PL_rsfp);
1107 (void)PerlIO_close(PL_rsfp);
1112 /* not at end of file, so we only read another line */
1113 /* make corresponding updates to old pointers, for yyerror() */
1114 oldprevlen = PL_oldbufptr - PL_bufend;
1115 oldoldprevlen = PL_oldoldbufptr - PL_bufend;
1117 oldunilen = PL_last_uni - PL_bufend;
1119 oldloplen = PL_last_lop - PL_bufend;
1120 PL_linestart = PL_bufptr = s + prevlen;
1121 PL_bufend = s + SvCUR(PL_linestr);
1123 PL_oldbufptr = s + oldprevlen;
1124 PL_oldoldbufptr = s + oldoldprevlen;
1126 PL_last_uni = s + oldunilen;
1128 PL_last_lop = s + oldloplen;
1131 /* debugger active and we're not compiling the debugger code,
1132 * so store the line into the debugger's array of lines
1134 if (PERLDB_LINE && PL_curstash != PL_debstash)
1135 update_debugger_info(NULL, PL_bufptr, PL_bufend - PL_bufptr);
1142 PL_skipwhite = newSVpvs("");
1143 curoff = s - SvPVX(PL_linestr);
1144 if (curoff - startoff)
1145 sv_catpvn(PL_skipwhite, SvPVX(PL_linestr) + startoff,
1154 * Check the unary operators to ensure there's no ambiguity in how they're
1155 * used. An ambiguous piece of code would be:
1157 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1158 * the +5 is its argument.
1168 if (PL_oldoldbufptr != PL_last_uni)
1170 while (isSPACE(*PL_last_uni))
1173 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1175 if ((t = strchr(s, '(')) && t < PL_bufptr)
1178 if (ckWARN_d(WARN_AMBIGUOUS)){
1179 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
1180 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1181 (int)(s - PL_last_uni), PL_last_uni);
1186 * LOP : macro to build a list operator. Its behaviour has been replaced
1187 * with a subroutine, S_lop() for which LOP is just another name.
1190 #define LOP(f,x) return lop(f,x,s)
1194 * Build a list operator (or something that might be one). The rules:
1195 * - if we have a next token, then it's a list operator [why?]
1196 * - if the next thing is an opening paren, then it's a function
1197 * - else it's a list operator
1201 S_lop(pTHX_ I32 f, int x, char *s)
1208 PL_last_lop = PL_oldbufptr;
1209 PL_last_lop_op = (OPCODE)f;
1212 return REPORT(LSTOP);
1215 return REPORT(LSTOP);
1218 return REPORT(FUNC);
1221 return REPORT(FUNC);
1223 return REPORT(LSTOP);
1229 * Sets up for an eventual force_next(). start_force(0) basically does
1230 * an unshift, while start_force(-1) does a push. yylex removes items
1235 S_start_force(pTHX_ int where)
1239 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1240 where = PL_lasttoke;
1241 assert(PL_curforce < 0 || PL_curforce == where);
1242 if (PL_curforce != where) {
1243 for (i = PL_lasttoke; i > where; --i) {
1244 PL_nexttoke[i] = PL_nexttoke[i-1];
1248 if (PL_curforce < 0) /* in case of duplicate start_force() */
1249 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1250 PL_curforce = where;
1253 curmad('^', newSVpvs(""));
1254 CURMAD('_', PL_nextwhite);
1259 S_curmad(pTHX_ char slot, SV *sv)
1265 if (PL_curforce < 0)
1266 where = &PL_thismad;
1268 where = &PL_nexttoke[PL_curforce].next_mad;
1271 sv_setpvn(sv, "", 0);
1274 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1276 else if (PL_encoding) {
1277 sv_recode_to_utf8(sv, PL_encoding);
1282 /* keep a slot open for the head of the list? */
1283 if (slot != '_' && *where && (*where)->mad_key == '^') {
1284 (*where)->mad_key = slot;
1285 sv_free((SV*)((*where)->mad_val));
1286 (*where)->mad_val = (void*)sv;
1289 addmad(newMADsv(slot, sv), where, 0);
1292 # define start_force(where) NOOP
1293 # define curmad(slot, sv) NOOP
1298 * When the lexer realizes it knows the next token (for instance,
1299 * it is reordering tokens for the parser) then it can call S_force_next
1300 * to know what token to return the next time the lexer is called. Caller
1301 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1302 * and possibly PL_expect to ensure the lexer handles the token correctly.
1306 S_force_next(pTHX_ I32 type)
1310 if (PL_curforce < 0)
1311 start_force(PL_lasttoke);
1312 PL_nexttoke[PL_curforce].next_type = type;
1313 if (PL_lex_state != LEX_KNOWNEXT)
1314 PL_lex_defer = PL_lex_state;
1315 PL_lex_state = LEX_KNOWNEXT;
1316 PL_lex_expect = PL_expect;
1319 PL_nexttype[PL_nexttoke] = type;
1321 if (PL_lex_state != LEX_KNOWNEXT) {
1322 PL_lex_defer = PL_lex_state;
1323 PL_lex_expect = PL_expect;
1324 PL_lex_state = LEX_KNOWNEXT;
1330 S_newSV_maybe_utf8(pTHX_ const char *start, STRLEN len)
1333 SV * const sv = newSVpvn(start,len);
1334 if (UTF && !IN_BYTES && is_utf8_string((const U8*)start, len))
1341 * When the lexer knows the next thing is a word (for instance, it has
1342 * just seen -> and it knows that the next char is a word char, then
1343 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1347 * char *start : buffer position (must be within PL_linestr)
1348 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1349 * int check_keyword : if true, Perl checks to make sure the word isn't
1350 * a keyword (do this if the word is a label, e.g. goto FOO)
1351 * int allow_pack : if true, : characters will also be allowed (require,
1352 * use, etc. do this)
1353 * int allow_initial_tick : used by the "sub" lexer only.
1357 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1363 start = SKIPSPACE1(start);
1365 if (isIDFIRST_lazy_if(s,UTF) ||
1366 (allow_pack && *s == ':') ||
1367 (allow_initial_tick && *s == '\'') )
1369 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1370 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1372 start_force(PL_curforce);
1374 curmad('X', newSVpvn(start,s-start));
1375 if (token == METHOD) {
1380 PL_expect = XOPERATOR;
1384 curmad('g', newSVpvs( "forced" ));
1385 NEXTVAL_NEXTTOKE.opval
1386 = (OP*)newSVOP(OP_CONST,0,
1387 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1388 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1396 * Called when the lexer wants $foo *foo &foo etc, but the program
1397 * text only contains the "foo" portion. The first argument is a pointer
1398 * to the "foo", and the second argument is the type symbol to prefix.
1399 * Forces the next token to be a "WORD".
1400 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
1404 S_force_ident(pTHX_ register const char *s, int kind)
1408 const STRLEN len = strlen(s);
1409 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
1410 start_force(PL_curforce);
1411 NEXTVAL_NEXTTOKE.opval = o;
1414 o->op_private = OPpCONST_ENTERED;
1415 /* XXX see note in pp_entereval() for why we forgo typo
1416 warnings if the symbol must be introduced in an eval.
1418 gv_fetchpvn_flags(s, len,
1419 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
1421 kind == '$' ? SVt_PV :
1422 kind == '@' ? SVt_PVAV :
1423 kind == '%' ? SVt_PVHV :
1431 Perl_str_to_version(pTHX_ SV *sv)
1436 const char *start = SvPV_const(sv,len);
1437 const char * const end = start + len;
1438 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
1439 while (start < end) {
1443 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
1448 retval += ((NV)n)/nshift;
1457 * Forces the next token to be a version number.
1458 * If the next token appears to be an invalid version number, (e.g. "v2b"),
1459 * and if "guessing" is TRUE, then no new token is created (and the caller
1460 * must use an alternative parsing method).
1464 S_force_version(pTHX_ char *s, int guessing)
1470 I32 startoff = s - SvPVX(PL_linestr);
1479 while (isDIGIT(*d) || *d == '_' || *d == '.')
1483 start_force(PL_curforce);
1484 curmad('X', newSVpvn(s,d-s));
1487 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
1489 s = scan_num(s, &yylval);
1490 version = yylval.opval;
1491 ver = cSVOPx(version)->op_sv;
1492 if (SvPOK(ver) && !SvNIOK(ver)) {
1493 SvUPGRADE(ver, SVt_PVNV);
1494 SvNV_set(ver, str_to_version(ver));
1495 SvNOK_on(ver); /* hint that it is a version */
1498 else if (guessing) {
1501 sv_free(PL_nextwhite); /* let next token collect whitespace */
1503 s = SvPVX(PL_linestr) + startoff;
1511 if (PL_madskills && !version) {
1512 sv_free(PL_nextwhite); /* let next token collect whitespace */
1514 s = SvPVX(PL_linestr) + startoff;
1517 /* NOTE: The parser sees the package name and the VERSION swapped */
1518 start_force(PL_curforce);
1519 NEXTVAL_NEXTTOKE.opval = version;
1527 * Tokenize a quoted string passed in as an SV. It finds the next
1528 * chunk, up to end of string or a backslash. It may make a new
1529 * SV containing that chunk (if HINT_NEW_STRING is on). It also
1534 S_tokeq(pTHX_ SV *sv)
1538 register char *send;
1546 s = SvPV_force(sv, len);
1547 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
1550 while (s < send && *s != '\\')
1555 if ( PL_hints & HINT_NEW_STRING ) {
1556 pv = sv_2mortal(newSVpvn(SvPVX_const(pv), len));
1562 if (s + 1 < send && (s[1] == '\\'))
1563 s++; /* all that, just for this */
1568 SvCUR_set(sv, d - SvPVX_const(sv));
1570 if ( PL_hints & HINT_NEW_STRING )
1571 return new_constant(NULL, 0, "q", sv, pv, "q");
1576 * Now come three functions related to double-quote context,
1577 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
1578 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
1579 * interact with PL_lex_state, and create fake ( ... ) argument lists
1580 * to handle functions and concatenation.
1581 * They assume that whoever calls them will be setting up a fake
1582 * join call, because each subthing puts a ',' after it. This lets
1585 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
1587 * (I'm not sure whether the spurious commas at the end of lcfirst's
1588 * arguments and join's arguments are created or not).
1593 * Assumes that yylval.ival is the op we're creating (e.g. OP_LCFIRST).
1595 * Pattern matching will set PL_lex_op to the pattern-matching op to
1596 * make (we return THING if yylval.ival is OP_NULL, PMFUNC otherwise).
1598 * OP_CONST and OP_READLINE are easy--just make the new op and return.
1600 * Everything else becomes a FUNC.
1602 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
1603 * had an OP_CONST or OP_READLINE). This just sets us up for a
1604 * call to S_sublex_push().
1608 S_sublex_start(pTHX)
1611 register const I32 op_type = yylval.ival;
1613 if (op_type == OP_NULL) {
1614 yylval.opval = PL_lex_op;
1618 if (op_type == OP_CONST || op_type == OP_READLINE) {
1619 SV *sv = tokeq(PL_lex_stuff);
1621 if (SvTYPE(sv) == SVt_PVIV) {
1622 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
1624 const char * const p = SvPV_const(sv, len);
1625 SV * const nsv = newSVpvn(p, len);
1631 yylval.opval = (OP*)newSVOP(op_type, 0, sv);
1632 PL_lex_stuff = NULL;
1633 /* Allow <FH> // "foo" */
1634 if (op_type == OP_READLINE)
1635 PL_expect = XTERMORDORDOR;
1638 else if (op_type == OP_BACKTICK && PL_lex_op) {
1639 /* readpipe() vas overriden */
1640 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
1641 yylval.opval = PL_lex_op;
1643 PL_lex_stuff = NULL;
1647 PL_sublex_info.super_state = PL_lex_state;
1648 PL_sublex_info.sub_inwhat = (U16)op_type;
1649 PL_sublex_info.sub_op = PL_lex_op;
1650 PL_lex_state = LEX_INTERPPUSH;
1654 yylval.opval = PL_lex_op;
1664 * Create a new scope to save the lexing state. The scope will be
1665 * ended in S_sublex_done. Returns a '(', starting the function arguments
1666 * to the uc, lc, etc. found before.
1667 * Sets PL_lex_state to LEX_INTERPCONCAT.
1676 PL_lex_state = PL_sublex_info.super_state;
1677 SAVEBOOL(PL_lex_dojoin);
1678 SAVEI32(PL_lex_brackets);
1679 SAVEI32(PL_lex_casemods);
1680 SAVEI32(PL_lex_starts);
1681 SAVEI8(PL_lex_state);
1682 SAVEVPTR(PL_lex_inpat);
1683 SAVEI16(PL_lex_inwhat);
1684 SAVECOPLINE(PL_curcop);
1685 SAVEPPTR(PL_bufptr);
1686 SAVEPPTR(PL_bufend);
1687 SAVEPPTR(PL_oldbufptr);
1688 SAVEPPTR(PL_oldoldbufptr);
1689 SAVEPPTR(PL_last_lop);
1690 SAVEPPTR(PL_last_uni);
1691 SAVEPPTR(PL_linestart);
1692 SAVESPTR(PL_linestr);
1693 SAVEGENERICPV(PL_lex_brackstack);
1694 SAVEGENERICPV(PL_lex_casestack);
1696 PL_linestr = PL_lex_stuff;
1697 PL_lex_stuff = NULL;
1699 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
1700 = SvPVX(PL_linestr);
1701 PL_bufend += SvCUR(PL_linestr);
1702 PL_last_lop = PL_last_uni = NULL;
1703 SAVEFREESV(PL_linestr);
1705 PL_lex_dojoin = FALSE;
1706 PL_lex_brackets = 0;
1707 Newx(PL_lex_brackstack, 120, char);
1708 Newx(PL_lex_casestack, 12, char);
1709 PL_lex_casemods = 0;
1710 *PL_lex_casestack = '\0';
1712 PL_lex_state = LEX_INTERPCONCAT;
1713 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
1715 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
1716 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
1717 PL_lex_inpat = PL_sublex_info.sub_op;
1719 PL_lex_inpat = NULL;
1726 * Restores lexer state after a S_sublex_push.
1733 if (!PL_lex_starts++) {
1734 SV * const sv = newSVpvs("");
1735 if (SvUTF8(PL_linestr))
1737 PL_expect = XOPERATOR;
1738 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
1742 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
1743 PL_lex_state = LEX_INTERPCASEMOD;
1747 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
1748 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
1749 PL_linestr = PL_lex_repl;
1751 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
1752 PL_bufend += SvCUR(PL_linestr);
1753 PL_last_lop = PL_last_uni = NULL;
1754 SAVEFREESV(PL_linestr);
1755 PL_lex_dojoin = FALSE;
1756 PL_lex_brackets = 0;
1757 PL_lex_casemods = 0;
1758 *PL_lex_casestack = '\0';
1760 if (SvEVALED(PL_lex_repl)) {
1761 PL_lex_state = LEX_INTERPNORMAL;
1763 /* we don't clear PL_lex_repl here, so that we can check later
1764 whether this is an evalled subst; that means we rely on the
1765 logic to ensure sublex_done() is called again only via the
1766 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
1769 PL_lex_state = LEX_INTERPCONCAT;
1779 PL_endwhite = newSVpvs("");
1780 sv_catsv(PL_endwhite, PL_thiswhite);
1784 sv_setpvn(PL_thistoken,"",0);
1786 PL_realtokenstart = -1;
1790 PL_bufend = SvPVX(PL_linestr);
1791 PL_bufend += SvCUR(PL_linestr);
1792 PL_expect = XOPERATOR;
1793 PL_sublex_info.sub_inwhat = 0;
1801 Extracts a pattern, double-quoted string, or transliteration. This
1804 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
1805 processing a pattern (PL_lex_inpat is true), a transliteration
1806 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
1808 Returns a pointer to the character scanned up to. If this is
1809 advanced from the start pointer supplied (i.e. if anything was
1810 successfully parsed), will leave an OP for the substring scanned
1811 in yylval. Caller must intuit reason for not parsing further
1812 by looking at the next characters herself.
1816 double-quoted style: \r and \n
1817 regexp special ones: \D \s
1820 case and quoting: \U \Q \E
1821 stops on @ and $, but not for $ as tail anchor
1823 In transliterations:
1824 characters are VERY literal, except for - not at the start or end
1825 of the string, which indicates a range. If the range is in bytes,
1826 scan_const expands the range to the full set of intermediate
1827 characters. If the range is in utf8, the hyphen is replaced with
1828 a certain range mark which will be handled by pmtrans() in op.c.
1830 In double-quoted strings:
1832 double-quoted style: \r and \n
1834 deprecated backrefs: \1 (in substitution replacements)
1835 case and quoting: \U \Q \E
1838 scan_const does *not* construct ops to handle interpolated strings.
1839 It stops processing as soon as it finds an embedded $ or @ variable
1840 and leaves it to the caller to work out what's going on.
1842 embedded arrays (whether in pattern or not) could be:
1843 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
1845 $ in double-quoted strings must be the symbol of an embedded scalar.
1847 $ in pattern could be $foo or could be tail anchor. Assumption:
1848 it's a tail anchor if $ is the last thing in the string, or if it's
1849 followed by one of "()| \r\n\t"
1851 \1 (backreferences) are turned into $1
1853 The structure of the code is
1854 while (there's a character to process) {
1855 handle transliteration ranges
1856 skip regexp comments /(?#comment)/ and codes /(?{code})/
1857 skip #-initiated comments in //x patterns
1858 check for embedded arrays
1859 check for embedded scalars
1861 leave intact backslashes from leaveit (below)
1862 deprecate \1 in substitution replacements
1863 handle string-changing backslashes \l \U \Q \E, etc.
1864 switch (what was escaped) {
1865 handle \- in a transliteration (becomes a literal -)
1866 handle \132 (octal characters)
1867 handle \x15 and \x{1234} (hex characters)
1868 handle \N{name} (named characters)
1869 handle \cV (control characters)
1870 handle printf-style backslashes (\f, \r, \n, etc)
1872 } (end if backslash)
1873 } (end while character to read)
1878 S_scan_const(pTHX_ char *start)
1881 register char *send = PL_bufend; /* end of the constant */
1882 SV *sv = newSV(send - start); /* sv for the constant */
1883 register char *s = start; /* start of the constant */
1884 register char *d = SvPVX(sv); /* destination for copies */
1885 bool dorange = FALSE; /* are we in a translit range? */
1886 bool didrange = FALSE; /* did we just finish a range? */
1887 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
1888 I32 this_utf8 = UTF; /* The source string is assumed to be UTF8 */
1891 UV literal_endpoint = 0;
1892 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
1895 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
1896 /* If we are doing a trans and we know we want UTF8 set expectation */
1897 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
1898 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
1902 while (s < send || dorange) {
1903 /* get transliterations out of the way (they're most literal) */
1904 if (PL_lex_inwhat == OP_TRANS) {
1905 /* expand a range A-Z to the full set of characters. AIE! */
1907 I32 i; /* current expanded character */
1908 I32 min; /* first character in range */
1909 I32 max; /* last character in range */
1920 char * const c = (char*)utf8_hop((U8*)d, -1);
1924 *c = (char)UTF_TO_NATIVE(0xff);
1925 /* mark the range as done, and continue */
1931 i = d - SvPVX_const(sv); /* remember current offset */
1934 SvLEN(sv) + (has_utf8 ?
1935 (512 - UTF_CONTINUATION_MARK +
1938 /* How many two-byte within 0..255: 128 in UTF-8,
1939 * 96 in UTF-8-mod. */
1941 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
1943 d = SvPVX(sv) + i; /* refresh d after realloc */
1947 for (j = 0; j <= 1; j++) {
1948 char * const c = (char*)utf8_hop((U8*)d, -1);
1949 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
1955 max = (U8)0xff; /* only to \xff */
1956 uvmax = uv; /* \x{100} to uvmax */
1958 d = c; /* eat endpoint chars */
1963 d -= 2; /* eat the first char and the - */
1964 min = (U8)*d; /* first char in range */
1965 max = (U8)d[1]; /* last char in range */
1972 "Invalid range \"%c-%c\" in transliteration operator",
1973 (char)min, (char)max);
1977 if (literal_endpoint == 2 &&
1978 ((isLOWER(min) && isLOWER(max)) ||
1979 (isUPPER(min) && isUPPER(max)))) {
1981 for (i = min; i <= max; i++)
1983 *d++ = NATIVE_TO_NEED(has_utf8,i);
1985 for (i = min; i <= max; i++)
1987 *d++ = NATIVE_TO_NEED(has_utf8,i);
1992 for (i = min; i <= max; i++)
1995 const U8 ch = (U8)NATIVE_TO_UTF(i);
1996 if (UNI_IS_INVARIANT(ch))
1999 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2000 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2009 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2011 *d++ = (char)UTF_TO_NATIVE(0xff);
2013 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2017 /* mark the range as done, and continue */
2021 literal_endpoint = 0;
2026 /* range begins (ignore - as first or last char) */
2027 else if (*s == '-' && s+1 < send && s != start) {
2029 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2036 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2046 literal_endpoint = 0;
2047 native_range = TRUE;
2052 /* if we get here, we're not doing a transliteration */
2054 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2055 except for the last char, which will be done separately. */
2056 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2058 while (s+1 < send && *s != ')')
2059 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2061 else if (s[2] == '{' /* This should match regcomp.c */
2062 || (s[2] == '?' && s[3] == '{'))
2065 char *regparse = s + (s[2] == '{' ? 3 : 4);
2068 while (count && (c = *regparse)) {
2069 if (c == '\\' && regparse[1])
2077 if (*regparse != ')')
2078 regparse--; /* Leave one char for continuation. */
2079 while (s < regparse)
2080 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2084 /* likewise skip #-initiated comments in //x patterns */
2085 else if (*s == '#' && PL_lex_inpat &&
2086 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2087 while (s+1 < send && *s != '\n')
2088 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2091 /* check for embedded arrays
2092 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2094 else if (*s == '@' && s[1]) {
2095 if (isALNUM_lazy_if(s+1,UTF))
2097 if (strchr(":'{$", s[1]))
2099 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2100 break; /* in regexp, neither @+ nor @- are interpolated */
2103 /* check for embedded scalars. only stop if we're sure it's a
2106 else if (*s == '$') {
2107 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2109 if (s + 1 < send && !strchr("()| \r\n\t", s[1]))
2110 break; /* in regexp, $ might be tail anchor */
2113 /* End of else if chain - OP_TRANS rejoin rest */
2116 if (*s == '\\' && s+1 < send) {
2119 /* deprecate \1 in strings and substitution replacements */
2120 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2121 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2123 if (ckWARN(WARN_SYNTAX))
2124 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2129 /* string-change backslash escapes */
2130 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2134 /* skip any other backslash escapes in a pattern */
2135 else if (PL_lex_inpat) {
2136 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2137 goto default_action;
2140 /* if we get here, it's either a quoted -, or a digit */
2143 /* quoted - in transliterations */
2145 if (PL_lex_inwhat == OP_TRANS) {
2152 if ((isALPHA(*s) || isDIGIT(*s)) &&
2154 Perl_warner(aTHX_ packWARN(WARN_MISC),
2155 "Unrecognized escape \\%c passed through",
2157 /* default action is to copy the quoted character */
2158 goto default_action;
2161 /* \132 indicates an octal constant */
2162 case '0': case '1': case '2': case '3':
2163 case '4': case '5': case '6': case '7':
2167 uv = grok_oct(s, &len, &flags, NULL);
2170 goto NUM_ESCAPE_INSERT;
2172 /* \x24 indicates a hex constant */
2176 char* const e = strchr(s, '}');
2177 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2178 PERL_SCAN_DISALLOW_PREFIX;
2183 yyerror("Missing right brace on \\x{}");
2187 uv = grok_hex(s, &len, &flags, NULL);
2193 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2194 uv = grok_hex(s, &len, &flags, NULL);
2200 /* Insert oct or hex escaped character.
2201 * There will always enough room in sv since such
2202 * escapes will be longer than any UTF-8 sequence
2203 * they can end up as. */
2205 /* We need to map to chars to ASCII before doing the tests
2208 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(uv))) {
2209 if (!has_utf8 && uv > 255) {
2210 /* Might need to recode whatever we have
2211 * accumulated so far if it contains any
2214 * (Can't we keep track of that and avoid
2215 * this rescan? --jhi)
2219 for (c = (U8 *) SvPVX(sv); c < (U8 *)d; c++) {
2220 if (!NATIVE_IS_INVARIANT(*c)) {
2225 const STRLEN offset = d - SvPVX_const(sv);
2227 d = SvGROW(sv, SvLEN(sv) + hicount + 1) + offset;
2231 while (src >= (const U8 *)SvPVX_const(sv)) {
2232 if (!NATIVE_IS_INVARIANT(*src)) {
2233 const U8 ch = NATIVE_TO_ASCII(*src);
2234 *dst-- = (U8)UTF8_EIGHT_BIT_LO(ch);
2235 *dst-- = (U8)UTF8_EIGHT_BIT_HI(ch);
2245 if (has_utf8 || uv > 255) {
2246 d = (char*)uvchr_to_utf8((U8*)d, uv);
2248 if (PL_lex_inwhat == OP_TRANS &&
2249 PL_sublex_info.sub_op) {
2250 PL_sublex_info.sub_op->op_private |=
2251 (PL_lex_repl ? OPpTRANS_FROM_UTF
2255 if (uv > 255 && !dorange)
2256 native_range = FALSE;
2268 /* \N{LATIN SMALL LETTER A} is a named character */
2272 char* e = strchr(s, '}');
2279 yyerror("Missing right brace on \\N{}");
2283 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2285 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2286 PERL_SCAN_DISALLOW_PREFIX;
2289 uv = grok_hex(s, &len, &flags, NULL);
2290 if ( e > s && len != (STRLEN)(e - s) ) {
2294 goto NUM_ESCAPE_INSERT;
2296 res = newSVpvn(s + 1, e - s - 1);
2297 type = newSVpvn(s - 2,e - s + 3);
2298 res = new_constant( NULL, 0, "charnames",
2299 res, NULL, SvPVX(type) );
2302 sv_utf8_upgrade(res);
2303 str = SvPV_const(res,len);
2304 #ifdef EBCDIC_NEVER_MIND
2305 /* charnames uses pack U and that has been
2306 * recently changed to do the below uni->native
2307 * mapping, so this would be redundant (and wrong,
2308 * the code point would be doubly converted).
2309 * But leave this in just in case the pack U change
2310 * gets revoked, but the semantics is still
2311 * desireable for charnames. --jhi */
2313 UV uv = utf8_to_uvchr((const U8*)str, 0);
2316 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2318 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2319 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2320 str = SvPV_const(res, len);
2324 if (!has_utf8 && SvUTF8(res)) {
2325 const char * const ostart = SvPVX_const(sv);
2326 SvCUR_set(sv, d - ostart);
2329 sv_utf8_upgrade(sv);
2330 /* this just broke our allocation above... */
2331 SvGROW(sv, (STRLEN)(send - start));
2332 d = SvPVX(sv) + SvCUR(sv);
2335 if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2336 const char * const odest = SvPVX_const(sv);
2338 SvGROW(sv, (SvLEN(sv) + len - (e - s + 4)));
2339 d = SvPVX(sv) + (d - odest);
2343 native_range = FALSE; /* \N{} is guessed to be Unicode */
2345 Copy(str, d, len, char);
2352 yyerror("Missing braces on \\N{}");
2355 /* \c is a control character */
2364 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2367 yyerror("Missing control char name in \\c");
2371 /* printf-style backslashes, formfeeds, newlines, etc */
2373 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2376 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2379 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2382 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
2385 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
2388 *d++ = ASCII_TO_NEED(has_utf8,'\033');
2391 *d++ = ASCII_TO_NEED(has_utf8,'\007');
2397 } /* end if (backslash) */
2404 /* If we started with encoded form, or already know we want it
2405 and then encode the next character */
2406 if ((has_utf8 || this_utf8) && !NATIVE_IS_INVARIANT((U8)(*s))) {
2408 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
2409 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
2412 /* encoded value larger than old, need extra space (NOTE: SvCUR() not set here) */
2413 const STRLEN off = d - SvPVX_const(sv);
2414 d = SvGROW(sv, SvLEN(sv) + (need-len)) + off;
2416 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
2419 if (uv > 255 && !dorange)
2420 native_range = FALSE;
2424 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2426 } /* while loop to process each character */
2428 /* terminate the string and set up the sv */
2430 SvCUR_set(sv, d - SvPVX_const(sv));
2431 if (SvCUR(sv) >= SvLEN(sv))
2432 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
2435 if (PL_encoding && !has_utf8) {
2436 sv_recode_to_utf8(sv, PL_encoding);
2442 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2443 PL_sublex_info.sub_op->op_private |=
2444 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2448 /* shrink the sv if we allocated more than we used */
2449 if (SvCUR(sv) + 5 < SvLEN(sv)) {
2450 SvPV_shrink_to_cur(sv);
2453 /* return the substring (via yylval) only if we parsed anything */
2454 if (s > PL_bufptr) {
2455 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) )
2456 sv = new_constant(start, s - start,
2457 (const char *)(PL_lex_inpat ? "qr" : "q"),
2460 (( PL_lex_inwhat == OP_TRANS
2462 : ( (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat)
2465 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2472 * Returns TRUE if there's more to the expression (e.g., a subscript),
2475 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
2477 * ->[ and ->{ return TRUE
2478 * { and [ outside a pattern are always subscripts, so return TRUE
2479 * if we're outside a pattern and it's not { or [, then return FALSE
2480 * if we're in a pattern and the first char is a {
2481 * {4,5} (any digits around the comma) returns FALSE
2482 * if we're in a pattern and the first char is a [
2484 * [SOMETHING] has a funky algorithm to decide whether it's a
2485 * character class or not. It has to deal with things like
2486 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
2487 * anything else returns TRUE
2490 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
2493 S_intuit_more(pTHX_ register char *s)
2496 if (PL_lex_brackets)
2498 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
2500 if (*s != '{' && *s != '[')
2505 /* In a pattern, so maybe we have {n,m}. */
2522 /* On the other hand, maybe we have a character class */
2525 if (*s == ']' || *s == '^')
2528 /* this is terrifying, and it works */
2529 int weight = 2; /* let's weigh the evidence */
2531 unsigned char un_char = 255, last_un_char;
2532 const char * const send = strchr(s,']');
2533 char tmpbuf[sizeof PL_tokenbuf * 4];
2535 if (!send) /* has to be an expression */
2538 Zero(seen,256,char);
2541 else if (isDIGIT(*s)) {
2543 if (isDIGIT(s[1]) && s[2] == ']')
2549 for (; s < send; s++) {
2550 last_un_char = un_char;
2551 un_char = (unsigned char)*s;
2556 weight -= seen[un_char] * 10;
2557 if (isALNUM_lazy_if(s+1,UTF)) {
2559 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
2560 len = (int)strlen(tmpbuf);
2561 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
2566 else if (*s == '$' && s[1] &&
2567 strchr("[#!%*<>()-=",s[1])) {
2568 if (/*{*/ strchr("])} =",s[2]))
2577 if (strchr("wds]",s[1]))
2579 else if (seen[(U8)'\''] || seen[(U8)'"'])
2581 else if (strchr("rnftbxcav",s[1]))
2583 else if (isDIGIT(s[1])) {
2585 while (s[1] && isDIGIT(s[1]))
2595 if (strchr("aA01! ",last_un_char))
2597 if (strchr("zZ79~",s[1]))
2599 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
2600 weight -= 5; /* cope with negative subscript */
2603 if (!isALNUM(last_un_char)
2604 && !(last_un_char == '$' || last_un_char == '@'
2605 || last_un_char == '&')
2606 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
2611 if (keyword(tmpbuf, d - tmpbuf, 0))
2614 if (un_char == last_un_char + 1)
2616 weight -= seen[un_char];
2621 if (weight >= 0) /* probably a character class */
2631 * Does all the checking to disambiguate
2633 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
2634 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
2636 * First argument is the stuff after the first token, e.g. "bar".
2638 * Not a method if bar is a filehandle.
2639 * Not a method if foo is a subroutine prototyped to take a filehandle.
2640 * Not a method if it's really "Foo $bar"
2641 * Method if it's "foo $bar"
2642 * Not a method if it's really "print foo $bar"
2643 * Method if it's really "foo package::" (interpreted as package->foo)
2644 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
2645 * Not a method if bar is a filehandle or package, but is quoted with
2650 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
2653 char *s = start + (*start == '$');
2654 char tmpbuf[sizeof PL_tokenbuf];
2662 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
2666 const char *proto = SvPVX_const(cv);
2677 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
2678 /* start is the beginning of the possible filehandle/object,
2679 * and s is the end of it
2680 * tmpbuf is a copy of it
2683 if (*start == '$') {
2684 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
2685 isUPPER(*PL_tokenbuf))
2688 len = start - SvPVX(PL_linestr);
2692 start = SvPVX(PL_linestr) + len;
2696 return *s == '(' ? FUNCMETH : METHOD;
2698 if (!keyword(tmpbuf, len, 0)) {
2699 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
2703 soff = s - SvPVX(PL_linestr);
2707 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
2708 if (indirgv && GvCVu(indirgv))
2710 /* filehandle or package name makes it a method */
2711 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
2713 soff = s - SvPVX(PL_linestr);
2716 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
2717 return 0; /* no assumptions -- "=>" quotes bearword */
2719 start_force(PL_curforce);
2720 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
2721 newSVpvn(tmpbuf,len));
2722 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
2724 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
2729 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
2731 return *s == '(' ? FUNCMETH : METHOD;
2739 * Return a string of Perl code to load the debugger. If PERL5DB
2740 * is set, it will return the contents of that, otherwise a
2741 * compile-time require of perl5db.pl.
2749 const char * const pdb = PerlEnv_getenv("PERL5DB");
2753 SETERRNO(0,SS_NORMAL);
2754 return "BEGIN { require 'perl5db.pl' }";
2760 /* Encoded script support. filter_add() effectively inserts a
2761 * 'pre-processing' function into the current source input stream.
2762 * Note that the filter function only applies to the current source file
2763 * (e.g., it will not affect files 'require'd or 'use'd by this one).
2765 * The datasv parameter (which may be NULL) can be used to pass
2766 * private data to this instance of the filter. The filter function
2767 * can recover the SV using the FILTER_DATA macro and use it to
2768 * store private buffers and state information.
2770 * The supplied datasv parameter is upgraded to a PVIO type
2771 * and the IoDIRP/IoANY field is used to store the function pointer,
2772 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
2773 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
2774 * private use must be set using malloc'd pointers.
2778 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
2787 if (!PL_rsfp_filters)
2788 PL_rsfp_filters = newAV();
2791 SvUPGRADE(datasv, SVt_PVIO);
2792 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
2793 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
2794 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
2795 FPTR2DPTR(void *, IoANY(datasv)),
2796 SvPV_nolen(datasv)));
2797 av_unshift(PL_rsfp_filters, 1);
2798 av_store(PL_rsfp_filters, 0, datasv) ;
2803 /* Delete most recently added instance of this filter function. */
2805 Perl_filter_del(pTHX_ filter_t funcp)
2811 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
2812 FPTR2DPTR(void*, funcp)));
2814 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
2816 /* if filter is on top of stack (usual case) just pop it off */
2817 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
2818 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
2819 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
2820 IoANY(datasv) = (void *)NULL;
2821 sv_free(av_pop(PL_rsfp_filters));
2825 /* we need to search for the correct entry and clear it */
2826 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
2830 /* Invoke the idxth filter function for the current rsfp. */
2831 /* maxlen 0 = read one text line */
2833 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
2838 /* This API is bad. It should have been using unsigned int for maxlen.
2839 Not sure if we want to change the API, but if not we should sanity
2840 check the value here. */
2841 const unsigned int correct_length
2850 if (!PL_parser || !PL_rsfp_filters)
2852 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
2853 /* Provide a default input filter to make life easy. */
2854 /* Note that we append to the line. This is handy. */
2855 DEBUG_P(PerlIO_printf(Perl_debug_log,
2856 "filter_read %d: from rsfp\n", idx));
2857 if (correct_length) {
2860 const int old_len = SvCUR(buf_sv);
2862 /* ensure buf_sv is large enough */
2863 SvGROW(buf_sv, (STRLEN)(old_len + correct_length)) ;
2864 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
2865 correct_length)) <= 0) {
2866 if (PerlIO_error(PL_rsfp))
2867 return -1; /* error */
2869 return 0 ; /* end of file */
2871 SvCUR_set(buf_sv, old_len + len) ;
2874 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
2875 if (PerlIO_error(PL_rsfp))
2876 return -1; /* error */
2878 return 0 ; /* end of file */
2881 return SvCUR(buf_sv);
2883 /* Skip this filter slot if filter has been deleted */
2884 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
2885 DEBUG_P(PerlIO_printf(Perl_debug_log,
2886 "filter_read %d: skipped (filter deleted)\n",
2888 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
2890 /* Get function pointer hidden within datasv */
2891 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
2892 DEBUG_P(PerlIO_printf(Perl_debug_log,
2893 "filter_read %d: via function %p (%s)\n",
2894 idx, (void*)datasv, SvPV_nolen_const(datasv)));
2895 /* Call function. The function is expected to */
2896 /* call "FILTER_READ(idx+1, buf_sv)" first. */
2897 /* Return: <0:error, =0:eof, >0:not eof */
2898 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
2902 S_filter_gets(pTHX_ register SV *sv, register PerlIO *fp, STRLEN append)
2905 #ifdef PERL_CR_FILTER
2906 if (!PL_rsfp_filters) {
2907 filter_add(S_cr_textfilter,NULL);
2910 if (PL_rsfp_filters) {
2912 SvCUR_set(sv, 0); /* start with empty line */
2913 if (FILTER_READ(0, sv, 0) > 0)
2914 return ( SvPVX(sv) ) ;
2919 return (sv_gets(sv, fp, append));
2923 S_find_in_my_stash(pTHX_ const char *pkgname, I32 len)
2928 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
2932 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
2933 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
2935 return GvHV(gv); /* Foo:: */
2938 /* use constant CLASS => 'MyClass' */
2939 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
2940 if (gv && GvCV(gv)) {
2941 SV * const sv = cv_const_sv(GvCV(gv));
2943 pkgname = SvPV_nolen_const(sv);
2946 return gv_stashpv(pkgname, 0);
2950 * S_readpipe_override
2951 * Check whether readpipe() is overriden, and generates the appropriate
2952 * optree, provided sublex_start() is called afterwards.
2955 S_readpipe_override(pTHX)
2958 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
2959 yylval.ival = OP_BACKTICK;
2961 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
2963 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
2964 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
2965 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
2967 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
2968 append_elem(OP_LIST,
2969 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
2970 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
2980 * The intent of this yylex wrapper is to minimize the changes to the
2981 * tokener when we aren't interested in collecting madprops. It remains
2982 * to be seen how successful this strategy will be...
2989 char *s = PL_bufptr;
2991 /* make sure PL_thiswhite is initialized */
2995 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
2996 if (PL_pending_ident)
2997 return S_pending_ident(aTHX);
2999 /* previous token ate up our whitespace? */
3000 if (!PL_lasttoke && PL_nextwhite) {
3001 PL_thiswhite = PL_nextwhite;
3005 /* isolate the token, and figure out where it is without whitespace */
3006 PL_realtokenstart = -1;
3010 assert(PL_curforce < 0);
3012 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3013 if (!PL_thistoken) {
3014 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3015 PL_thistoken = newSVpvs("");
3017 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3018 PL_thistoken = newSVpvn(tstart, s - tstart);
3021 if (PL_thismad) /* install head */
3022 CURMAD('X', PL_thistoken);
3025 /* last whitespace of a sublex? */
3026 if (optype == ')' && PL_endwhite) {
3027 CURMAD('X', PL_endwhite);
3032 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3033 if (!PL_thiswhite && !PL_endwhite && !optype) {
3034 sv_free(PL_thistoken);
3039 /* put off final whitespace till peg */
3040 if (optype == ';' && !PL_rsfp) {
3041 PL_nextwhite = PL_thiswhite;
3044 else if (PL_thisopen) {
3045 CURMAD('q', PL_thisopen);
3047 sv_free(PL_thistoken);
3051 /* Store actual token text as madprop X */
3052 CURMAD('X', PL_thistoken);
3056 /* add preceding whitespace as madprop _ */
3057 CURMAD('_', PL_thiswhite);
3061 /* add quoted material as madprop = */
3062 CURMAD('=', PL_thisstuff);
3066 /* add terminating quote as madprop Q */
3067 CURMAD('Q', PL_thisclose);
3071 /* special processing based on optype */
3075 /* opval doesn't need a TOKEN since it can already store mp */
3086 append_madprops(PL_thismad, yylval.opval, 0);
3094 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3103 /* remember any fake bracket that lexer is about to discard */
3104 if (PL_lex_brackets == 1 &&
3105 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3108 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3111 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3112 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3115 break; /* don't bother looking for trailing comment */
3124 /* attach a trailing comment to its statement instead of next token */
3128 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3130 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3132 if (*s == '\n' || *s == '#') {
3133 while (s < PL_bufend && *s != '\n')
3137 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3138 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3155 /* Create new token struct. Note: opvals return early above. */
3156 yylval.tkval = newTOKEN(optype, yylval, PL_thismad);
3163 S_tokenize_use(pTHX_ int is_use, char *s) {
3165 if (PL_expect != XSTATE)
3166 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3167 is_use ? "use" : "no"));
3169 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3170 s = force_version(s, TRUE);
3171 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3172 start_force(PL_curforce);
3173 NEXTVAL_NEXTTOKE.opval = NULL;
3176 else if (*s == 'v') {
3177 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3178 s = force_version(s, FALSE);
3182 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3183 s = force_version(s, FALSE);
3185 yylval.ival = is_use;
3189 static const char* const exp_name[] =
3190 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3191 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3198 Works out what to call the token just pulled out of the input
3199 stream. The yacc parser takes care of taking the ops we return and
3200 stitching them into a tree.
3206 if read an identifier
3207 if we're in a my declaration
3208 croak if they tried to say my($foo::bar)
3209 build the ops for a my() declaration
3210 if it's an access to a my() variable
3211 are we in a sort block?
3212 croak if my($a); $a <=> $b
3213 build ops for access to a my() variable
3214 if in a dq string, and they've said @foo and we can't find @foo
3216 build ops for a bareword
3217 if we already built the token before, use it.
3222 #pragma segment Perl_yylex
3228 register char *s = PL_bufptr;
3233 /* orig_keyword, gvp, and gv are initialized here because
3234 * jump to the label just_a_word_zero can bypass their
3235 * initialization later. */
3236 I32 orig_keyword = 0;
3241 SV* tmp = newSVpvs("");
3242 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3243 (IV)CopLINE(PL_curcop),
3244 lex_state_names[PL_lex_state],
3245 exp_name[PL_expect],
3246 pv_display(tmp, s, strlen(s), 0, 60));
3249 /* check if there's an identifier for us to look at */
3250 if (PL_pending_ident)
3251 return REPORT(S_pending_ident(aTHX));
3253 /* no identifier pending identification */
3255 switch (PL_lex_state) {
3257 case LEX_NORMAL: /* Some compilers will produce faster */
3258 case LEX_INTERPNORMAL: /* code if we comment these out. */
3262 /* when we've already built the next token, just pull it out of the queue */
3266 yylval = PL_nexttoke[PL_lasttoke].next_val;
3268 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3269 PL_nexttoke[PL_lasttoke].next_mad = 0;
3270 if (PL_thismad && PL_thismad->mad_key == '_') {
3271 PL_thiswhite = (SV*)PL_thismad->mad_val;
3272 PL_thismad->mad_val = 0;
3273 mad_free(PL_thismad);
3278 PL_lex_state = PL_lex_defer;
3279 PL_expect = PL_lex_expect;
3280 PL_lex_defer = LEX_NORMAL;
3281 if (!PL_nexttoke[PL_lasttoke].next_type)
3286 yylval = PL_nextval[PL_nexttoke];
3288 PL_lex_state = PL_lex_defer;
3289 PL_expect = PL_lex_expect;
3290 PL_lex_defer = LEX_NORMAL;
3294 /* FIXME - can these be merged? */
3295 return(PL_nexttoke[PL_lasttoke].next_type);
3297 return REPORT(PL_nexttype[PL_nexttoke]);
3300 /* interpolated case modifiers like \L \U, including \Q and \E.
3301 when we get here, PL_bufptr is at the \
3303 case LEX_INTERPCASEMOD:
3305 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3306 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3308 /* handle \E or end of string */
3309 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3311 if (PL_lex_casemods) {
3312 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3313 PL_lex_casestack[PL_lex_casemods] = '\0';
3315 if (PL_bufptr != PL_bufend
3316 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3318 PL_lex_state = LEX_INTERPCONCAT;
3321 PL_thistoken = newSVpvs("\\E");
3327 while (PL_bufptr != PL_bufend &&
3328 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3330 PL_thiswhite = newSVpvs("");
3331 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3335 if (PL_bufptr != PL_bufend)
3338 PL_lex_state = LEX_INTERPCONCAT;
3342 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3343 "### Saw case modifier\n"); });
3345 if (s[1] == '\\' && s[2] == 'E') {
3348 PL_thiswhite = newSVpvs("");
3349 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3352 PL_lex_state = LEX_INTERPCONCAT;
3357 if (!PL_madskills) /* when just compiling don't need correct */
3358 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3359 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3360 if ((*s == 'L' || *s == 'U') &&
3361 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
3362 PL_lex_casestack[--PL_lex_casemods] = '\0';
3365 if (PL_lex_casemods > 10)
3366 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
3367 PL_lex_casestack[PL_lex_casemods++] = *s;
3368 PL_lex_casestack[PL_lex_casemods] = '\0';
3369 PL_lex_state = LEX_INTERPCONCAT;
3370 start_force(PL_curforce);
3371 NEXTVAL_NEXTTOKE.ival = 0;
3373 start_force(PL_curforce);
3375 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
3377 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
3379 NEXTVAL_NEXTTOKE.ival = OP_LC;
3381 NEXTVAL_NEXTTOKE.ival = OP_UC;
3383 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
3385 Perl_croak(aTHX_ "panic: yylex");
3387 SV* const tmpsv = newSVpvs("");
3388 Perl_sv_catpvf(aTHX_ tmpsv, "\\%c", *s);
3394 if (PL_lex_starts) {
3400 sv_free(PL_thistoken);
3401 PL_thistoken = newSVpvs("");
3404 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3405 if (PL_lex_casemods == 1 && PL_lex_inpat)
3414 case LEX_INTERPPUSH:
3415 return REPORT(sublex_push());
3417 case LEX_INTERPSTART:
3418 if (PL_bufptr == PL_bufend)
3419 return REPORT(sublex_done());
3420 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3421 "### Interpolated variable\n"); });
3423 PL_lex_dojoin = (*PL_bufptr == '@');
3424 PL_lex_state = LEX_INTERPNORMAL;
3425 if (PL_lex_dojoin) {
3426 start_force(PL_curforce);
3427 NEXTVAL_NEXTTOKE.ival = 0;
3429 start_force(PL_curforce);
3430 force_ident("\"", '$');
3431 start_force(PL_curforce);
3432 NEXTVAL_NEXTTOKE.ival = 0;
3434 start_force(PL_curforce);
3435 NEXTVAL_NEXTTOKE.ival = 0;
3437 start_force(PL_curforce);
3438 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
3441 if (PL_lex_starts++) {
3446 sv_free(PL_thistoken);
3447 PL_thistoken = newSVpvs("");
3450 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3451 if (!PL_lex_casemods && PL_lex_inpat)
3458 case LEX_INTERPENDMAYBE:
3459 if (intuit_more(PL_bufptr)) {
3460 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
3466 if (PL_lex_dojoin) {
3467 PL_lex_dojoin = FALSE;
3468 PL_lex_state = LEX_INTERPCONCAT;
3472 sv_free(PL_thistoken);
3473 PL_thistoken = newSVpvs("");
3478 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
3479 && SvEVALED(PL_lex_repl))
3481 if (PL_bufptr != PL_bufend)
3482 Perl_croak(aTHX_ "Bad evalled substitution pattern");
3486 case LEX_INTERPCONCAT:
3488 if (PL_lex_brackets)
3489 Perl_croak(aTHX_ "panic: INTERPCONCAT");
3491 if (PL_bufptr == PL_bufend)
3492 return REPORT(sublex_done());
3494 if (SvIVX(PL_linestr) == '\'') {
3495 SV *sv = newSVsv(PL_linestr);
3498 else if ( PL_hints & HINT_NEW_RE )
3499 sv = new_constant(NULL, 0, "qr", sv, sv, "q");
3500 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3504 s = scan_const(PL_bufptr);
3506 PL_lex_state = LEX_INTERPCASEMOD;
3508 PL_lex_state = LEX_INTERPSTART;
3511 if (s != PL_bufptr) {
3512 start_force(PL_curforce);
3514 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
3516 NEXTVAL_NEXTTOKE = yylval;
3519 if (PL_lex_starts++) {
3523 sv_free(PL_thistoken);
3524 PL_thistoken = newSVpvs("");
3527 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
3528 if (!PL_lex_casemods && PL_lex_inpat)
3541 PL_lex_state = LEX_NORMAL;
3542 s = scan_formline(PL_bufptr);
3543 if (!PL_lex_formbrack)
3549 PL_oldoldbufptr = PL_oldbufptr;
3555 sv_free(PL_thistoken);
3558 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
3562 if (isIDFIRST_lazy_if(s,UTF))
3564 Perl_croak(aTHX_ "Unrecognized character \\x%02X", *s & 255);
3567 goto fake_eof; /* emulate EOF on ^D or ^Z */
3576 if (PL_lex_brackets) {
3577 yyerror((const char *)
3579 ? "Format not terminated"
3580 : "Missing right curly or square bracket"));
3582 DEBUG_T( { PerlIO_printf(Perl_debug_log,
3583 "### Tokener got EOF\n");
3587 if (s++ < PL_bufend)
3588 goto retry; /* ignore stray nulls */
3591 if (!PL_in_eval && !PL_preambled) {
3592 PL_preambled = TRUE;
3597 sv_setpv(PL_linestr,incl_perldb());
3598 if (SvCUR(PL_linestr))
3599 sv_catpvs(PL_linestr,";");
3601 while(AvFILLp(PL_preambleav) >= 0) {
3602 SV *tmpsv = av_shift(PL_preambleav);
3603 sv_catsv(PL_linestr, tmpsv);
3604 sv_catpvs(PL_linestr, ";");
3607 sv_free((SV*)PL_preambleav);
3608 PL_preambleav = NULL;
3610 if (PL_minus_n || PL_minus_p) {
3611 sv_catpvs(PL_linestr, "LINE: while (<>) {");
3613 sv_catpvs(PL_linestr,"chomp;");
3616 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
3617 || *PL_splitstr == '"')
3618 && strchr(PL_splitstr + 1, *PL_splitstr))
3619 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
3621 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
3622 bytes can be used as quoting characters. :-) */
3623 const char *splits = PL_splitstr;
3624 sv_catpvs(PL_linestr, "our @F=split(q\0");
3627 if (*splits == '\\')
3628 sv_catpvn(PL_linestr, splits, 1);
3629 sv_catpvn(PL_linestr, splits, 1);
3630 } while (*splits++);
3631 /* This loop will embed the trailing NUL of
3632 PL_linestr as the last thing it does before
3634 sv_catpvs(PL_linestr, ");");
3638 sv_catpvs(PL_linestr,"our @F=split(' ');");
3642 sv_catpvs(PL_linestr,"use feature ':5.10';");
3643 sv_catpvs(PL_linestr, "\n");
3644 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3645 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3646 PL_last_lop = PL_last_uni = NULL;
3647 if (PERLDB_LINE && PL_curstash != PL_debstash)
3648 update_debugger_info(PL_linestr, NULL, 0);
3652 bof = PL_rsfp ? TRUE : FALSE;
3653 if ((s = filter_gets(PL_linestr, PL_rsfp, 0)) == NULL) {
3656 PL_realtokenstart = -1;
3659 if (PL_preprocess && !PL_in_eval)
3660 (void)PerlProc_pclose(PL_rsfp);
3661 else if ((PerlIO *)PL_rsfp == PerlIO_stdin())
3662 PerlIO_clearerr(PL_rsfp);
3664 (void)PerlIO_close(PL_rsfp);
3666 PL_doextract = FALSE;
3668 if (!PL_in_eval && (PL_minus_n || PL_minus_p)) {
3673 sv_setpv(PL_linestr,
3676 ? ";}continue{print;}" : ";}"));
3677 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3678 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3679 PL_last_lop = PL_last_uni = NULL;
3680 PL_minus_n = PL_minus_p = 0;
3683 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3684 PL_last_lop = PL_last_uni = NULL;
3685 sv_setpvn(PL_linestr,"",0);
3686 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
3688 /* If it looks like the start of a BOM or raw UTF-16,
3689 * check if it in fact is. */
3695 #ifdef PERLIO_IS_STDIO
3696 # ifdef __GNU_LIBRARY__
3697 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
3698 # define FTELL_FOR_PIPE_IS_BROKEN
3702 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
3703 # define FTELL_FOR_PIPE_IS_BROKEN
3708 #ifdef FTELL_FOR_PIPE_IS_BROKEN
3709 /* This loses the possibility to detect the bof
3710 * situation on perl -P when the libc5 is being used.
3711 * Workaround? Maybe attach some extra state to PL_rsfp?
3714 bof = PerlIO_tell(PL_rsfp) == SvCUR(PL_linestr);
3716 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
3719 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3720 s = swallow_bom((U8*)s);
3724 /* Incest with pod. */
3727 sv_catsv(PL_thiswhite, PL_linestr);
3729 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
3730 sv_setpvn(PL_linestr, "", 0);
3731 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3732 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3733 PL_last_lop = PL_last_uni = NULL;
3734 PL_doextract = FALSE;
3738 } while (PL_doextract);
3739 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
3740 if (PERLDB_LINE && PL_curstash != PL_debstash)
3741 update_debugger_info(PL_linestr, NULL, 0);
3742 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3743 PL_last_lop = PL_last_uni = NULL;
3744 if (CopLINE(PL_curcop) == 1) {
3745 while (s < PL_bufend && isSPACE(*s))
3747 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
3751 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
3755 if (*s == '#' && *(s+1) == '!')
3757 #ifdef ALTERNATE_SHEBANG
3759 static char const as[] = ALTERNATE_SHEBANG;
3760 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
3761 d = s + (sizeof(as) - 1);
3763 #endif /* ALTERNATE_SHEBANG */
3772 while (*d && !isSPACE(*d))
3776 #ifdef ARG_ZERO_IS_SCRIPT
3777 if (ipathend > ipath) {
3779 * HP-UX (at least) sets argv[0] to the script name,
3780 * which makes $^X incorrect. And Digital UNIX and Linux,
3781 * at least, set argv[0] to the basename of the Perl
3782 * interpreter. So, having found "#!", we'll set it right.
3784 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
3786 assert(SvPOK(x) || SvGMAGICAL(x));
3787 if (sv_eq(x, CopFILESV(PL_curcop))) {
3788 sv_setpvn(x, ipath, ipathend - ipath);
3794 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
3795 const char * const lstart = SvPV_const(x,llen);
3797 bstart += blen - llen;
3798 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
3799 sv_setpvn(x, ipath, ipathend - ipath);
3804 TAINT_NOT; /* $^X is always tainted, but that's OK */
3806 #endif /* ARG_ZERO_IS_SCRIPT */
3811 d = instr(s,"perl -");
3813 d = instr(s,"perl");
3815 /* avoid getting into infinite loops when shebang
3816 * line contains "Perl" rather than "perl" */
3818 for (d = ipathend-4; d >= ipath; --d) {
3819 if ((*d == 'p' || *d == 'P')
3820 && !ibcmp(d, "perl", 4))
3830 #ifdef ALTERNATE_SHEBANG
3832 * If the ALTERNATE_SHEBANG on this system starts with a
3833 * character that can be part of a Perl expression, then if
3834 * we see it but not "perl", we're probably looking at the
3835 * start of Perl code, not a request to hand off to some
3836 * other interpreter. Similarly, if "perl" is there, but
3837 * not in the first 'word' of the line, we assume the line
3838 * contains the start of the Perl program.
3840 if (d && *s != '#') {
3841 const char *c = ipath;
3842 while (*c && !strchr("; \t\r\n\f\v#", *c))
3845 d = NULL; /* "perl" not in first word; ignore */
3847 *s = '#'; /* Don't try to parse shebang line */
3849 #endif /* ALTERNATE_SHEBANG */
3850 #ifndef MACOS_TRADITIONAL
3855 !instr(s,"indir") &&
3856 instr(PL_origargv[0],"perl"))
3863 while (s < PL_bufend && isSPACE(*s))
3865 if (s < PL_bufend) {
3866 Newxz(newargv,PL_origargc+3,char*);
3868 while (s < PL_bufend && !isSPACE(*s))
3871 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
3874 newargv = PL_origargv;
3877 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
3879 Perl_croak(aTHX_ "Can't exec %s", ipath);
3883 while (*d && !isSPACE(*d))
3885 while (SPACE_OR_TAB(*d))
3889 const bool switches_done = PL_doswitches;
3890 const U32 oldpdb = PL_perldb;
3891 const bool oldn = PL_minus_n;
3892 const bool oldp = PL_minus_p;
3895 if (*d == 'M' || *d == 'm' || *d == 'C') {
3896 const char * const m = d;
3897 while (*d && !isSPACE(*d))
3899 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
3902 d = moreswitches(d);
3904 if (PL_doswitches && !switches_done) {
3905 int argc = PL_origargc;
3906 char **argv = PL_origargv;
3909 } while (argc && argv[0][0] == '-' && argv[0][1]);
3910 init_argv_symbols(argc,argv);
3912 if ((PERLDB_LINE && !oldpdb) ||
3913 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
3914 /* if we have already added "LINE: while (<>) {",
3915 we must not do it again */
3917 sv_setpvn(PL_linestr, "", 0);
3918 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
3919 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
3920 PL_last_lop = PL_last_uni = NULL;
3921 PL_preambled = FALSE;
3923 (void)gv_fetchfile(PL_origfilename);
3930 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3932 PL_lex_state = LEX_FORMLINE;
3937 #ifdef PERL_STRICT_CR
3938 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
3940 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
3942 case ' ': case '\t': case '\f': case 013:
3943 #ifdef MACOS_TRADITIONAL
3947 PL_realtokenstart = -1;
3949 PL_thiswhite = newSVpvs("");
3950 sv_catpvn(PL_thiswhite, s, 1);
3957 PL_realtokenstart = -1;
3961 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
3962 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
3963 /* handle eval qq[#line 1 "foo"\n ...] */
3964 CopLINE_dec(PL_curcop);
3967 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
3969 if (!PL_in_eval || PL_rsfp)
3974 while (d < PL_bufend && *d != '\n')
3978 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
3979 Perl_croak(aTHX_ "panic: input overflow");
3982 PL_thiswhite = newSVpvn(s, d - s);
3987 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
3989 PL_lex_state = LEX_FORMLINE;
3995 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
3996 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
3999 TOKEN(PEG); /* make sure any #! line is accessible */
4004 /* if (PL_madskills && PL_lex_formbrack) { */
4006 while (d < PL_bufend && *d != '\n')
4010 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4011 Perl_croak(aTHX_ "panic: input overflow");
4012 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4014 PL_thiswhite = newSVpvs("");
4015 if (CopLINE(PL_curcop) == 1) {
4016 sv_setpvn(PL_thiswhite, "", 0);
4019 sv_catpvn(PL_thiswhite, s, d - s);
4033 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4041 while (s < PL_bufend && SPACE_OR_TAB(*s))
4044 if (strnEQ(s,"=>",2)) {
4045 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4046 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4047 OPERATOR('-'); /* unary minus */
4049 PL_last_uni = PL_oldbufptr;
4051 case 'r': ftst = OP_FTEREAD; break;
4052 case 'w': ftst = OP_FTEWRITE; break;
4053 case 'x': ftst = OP_FTEEXEC; break;
4054 case 'o': ftst = OP_FTEOWNED; break;
4055 case 'R': ftst = OP_FTRREAD; break;
4056 case 'W': ftst = OP_FTRWRITE; break;
4057 case 'X': ftst = OP_FTREXEC; break;
4058 case 'O': ftst = OP_FTROWNED; break;
4059 case 'e': ftst = OP_FTIS; break;
4060 case 'z': ftst = OP_FTZERO; break;
4061 case 's': ftst = OP_FTSIZE; break;
4062 case 'f': ftst = OP_FTFILE; break;
4063 case 'd': ftst = OP_FTDIR; break;
4064 case 'l': ftst = OP_FTLINK; break;
4065 case 'p': ftst = OP_FTPIPE; break;
4066 case 'S': ftst = OP_FTSOCK; break;
4067 case 'u': ftst = OP_FTSUID; break;
4068 case 'g': ftst = OP_FTSGID; break;
4069 case 'k': ftst = OP_FTSVTX; break;
4070 case 'b': ftst = OP_FTBLK; break;
4071 case 'c': ftst = OP_FTCHR; break;
4072 case 't': ftst = OP_FTTTY; break;
4073 case 'T': ftst = OP_FTTEXT; break;
4074 case 'B': ftst = OP_FTBINARY; break;
4075 case 'M': case 'A': case 'C':
4076 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4078 case 'M': ftst = OP_FTMTIME; break;
4079 case 'A': ftst = OP_FTATIME; break;
4080 case 'C': ftst = OP_FTCTIME; break;
4088 PL_last_lop_op = (OPCODE)ftst;
4089 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4090 "### Saw file test %c\n", (int)tmp);
4095 /* Assume it was a minus followed by a one-letter named
4096 * subroutine call (or a -bareword), then. */
4097 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4098 "### '-%c' looked like a file test but was not\n",
4105 const char tmp = *s++;
4108 if (PL_expect == XOPERATOR)
4113 else if (*s == '>') {
4116 if (isIDFIRST_lazy_if(s,UTF)) {
4117 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4125 if (PL_expect == XOPERATOR)
4128 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4130 OPERATOR('-'); /* unary minus */
4136 const char tmp = *s++;
4139 if (PL_expect == XOPERATOR)
4144 if (PL_expect == XOPERATOR)
4147 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4154 if (PL_expect != XOPERATOR) {
4155 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4156 PL_expect = XOPERATOR;
4157 force_ident(PL_tokenbuf, '*');
4170 if (PL_expect == XOPERATOR) {
4174 PL_tokenbuf[0] = '%';
4175 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4176 sizeof PL_tokenbuf - 1, FALSE);
4177 if (!PL_tokenbuf[1]) {
4180 PL_pending_ident = '%';
4191 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4198 const char tmp = *s++;
4204 goto just_a_word_zero_gv;
4207 switch (PL_expect) {
4213 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4215 PL_bufptr = s; /* update in case we back off */
4221 PL_expect = XTERMBLOCK;
4224 stuffstart = s - SvPVX(PL_linestr) - 1;
4228 while (isIDFIRST_lazy_if(s,UTF)) {
4231 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4232 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4233 if (tmp < 0) tmp = -tmp;
4248 sv = newSVpvn(s, len);
4250 d = scan_str(d,TRUE,TRUE);
4252 /* MUST advance bufptr here to avoid bogus
4253 "at end of line" context messages from yyerror().
4255 PL_bufptr = s + len;
4256 yyerror("Unterminated attribute parameter in attribute list");
4260 return REPORT(0); /* EOF indicator */
4264 sv_catsv(sv, PL_lex_stuff);
4265 attrs = append_elem(OP_LIST, attrs,
4266 newSVOP(OP_CONST, 0, sv));
4267 SvREFCNT_dec(PL_lex_stuff);
4268 PL_lex_stuff = NULL;
4271 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4273 if (PL_in_my == KEY_our) {
4275 GvUNIQUE_on(cGVOPx_gv(yylval.opval));
4277 /* skip to avoid loading attributes.pm */
4279 deprecate(":unique");
4282 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4285 /* NOTE: any CV attrs applied here need to be part of
4286 the CVf_BUILTIN_ATTRS define in cv.h! */
4287 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4289 CvLVALUE_on(PL_compcv);
4291 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4293 CvLOCKED_on(PL_compcv);
4295 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4297 CvMETHOD_on(PL_compcv);
4299 /* After we've set the flags, it could be argued that
4300 we don't need to do the attributes.pm-based setting
4301 process, and shouldn't bother appending recognized
4302 flags. To experiment with that, uncomment the
4303 following "else". (Note that's already been
4304 uncommented. That keeps the above-applied built-in
4305 attributes from being intercepted (and possibly
4306 rejected) by a package's attribute routines, but is
4307 justified by the performance win for the common case
4308 of applying only built-in attributes.) */
4310 attrs = append_elem(OP_LIST, attrs,
4311 newSVOP(OP_CONST, 0,
4315 if (*s == ':' && s[1] != ':')
4318 break; /* require real whitespace or :'s */
4319 /* XXX losing whitespace on sequential attributes here */
4323 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4324 if (*s != ';' && *s != '}' && *s != tmp
4325 && (tmp != '=' || *s != ')')) {
4326 const char q = ((*s == '\'') ? '"' : '\'');
4327 /* If here for an expression, and parsed no attrs, back
4329 if (tmp == '=' && !attrs) {
4333 /* MUST advance bufptr here to avoid bogus "at end of line"
4334 context messages from yyerror().
4337 yyerror( (const char *)
4339 ? Perl_form(aTHX_ "Invalid separator character "
4340 "%c%c%c in attribute list", q, *s, q)
4341 : "Unterminated attribute list" ) );
4349 start_force(PL_curforce);
4350 NEXTVAL_NEXTTOKE.opval = attrs;
4351 CURMAD('_', PL_nextwhite);
4356 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4357 (s - SvPVX(PL_linestr)) - stuffstart);
4365 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
4366 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
4374 const char tmp = *s++;
4379 const char tmp = *s++;
4387 if (PL_lex_brackets <= 0)
4388 yyerror("Unmatched right square bracket");
4391 if (PL_lex_state == LEX_INTERPNORMAL) {
4392 if (PL_lex_brackets == 0) {
4393 if (*s != '[' && *s != '{' && (*s != '-' || s[1] != '>'))
4394 PL_lex_state = LEX_INTERPEND;
4401 if (PL_lex_brackets > 100) {
4402 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4404 switch (PL_expect) {
4406 if (PL_lex_formbrack) {
4410 if (PL_oldoldbufptr == PL_last_lop)
4411 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4413 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4414 OPERATOR(HASHBRACK);
4416 while (s < PL_bufend && SPACE_OR_TAB(*s))
4419 PL_tokenbuf[0] = '\0';
4420 if (d < PL_bufend && *d == '-') {
4421 PL_tokenbuf[0] = '-';
4423 while (d < PL_bufend && SPACE_OR_TAB(*d))
4426 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
4427 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
4429 while (d < PL_bufend && SPACE_OR_TAB(*d))
4432 const char minus = (PL_tokenbuf[0] == '-');
4433 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
4441 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
4446 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4451 if (PL_oldoldbufptr == PL_last_lop)
4452 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
4454 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
4457 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
4459 /* This hack is to get the ${} in the message. */
4461 yyerror("syntax error");
4464 OPERATOR(HASHBRACK);
4466 /* This hack serves to disambiguate a pair of curlies
4467 * as being a block or an anon hash. Normally, expectation
4468 * determines that, but in cases where we're not in a
4469 * position to expect anything in particular (like inside
4470 * eval"") we have to resolve the ambiguity. This code
4471 * covers the case where the first term in the curlies is a
4472 * quoted string. Most other cases need to be explicitly
4473 * disambiguated by prepending a "+" before the opening
4474 * curly in order to force resolution as an anon hash.
4476 * XXX should probably propagate the outer expectation
4477 * into eval"" to rely less on this hack, but that could
4478 * potentially break current behavior of eval"".
4482 if (*s == '\'' || *s == '"' || *s == '`') {
4483 /* common case: get past first string, handling escapes */
4484 for (t++; t < PL_bufend && *t != *s;)
4485 if (*t++ == '\\' && (*t == '\\' || *t == *s))
4489 else if (*s == 'q') {
4492 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
4495 /* skip q//-like construct */
4497 char open, close, term;
4500 while (t < PL_bufend && isSPACE(*t))
4502 /* check for q => */
4503 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
4504 OPERATOR(HASHBRACK);
4508 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
4512 for (t++; t < PL_bufend; t++) {
4513 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
4515 else if (*t == open)
4519 for (t++; t < PL_bufend; t++) {
4520 if (*t == '\\' && t+1 < PL_bufend)
4522 else if (*t == close && --brackets <= 0)
4524 else if (*t == open)
4531 /* skip plain q word */
4532 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4535 else if (isALNUM_lazy_if(t,UTF)) {
4537 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
4540 while (t < PL_bufend && isSPACE(*t))
4542 /* if comma follows first term, call it an anon hash */
4543 /* XXX it could be a comma expression with loop modifiers */
4544 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
4545 || (*t == '=' && t[1] == '>')))
4546 OPERATOR(HASHBRACK);
4547 if (PL_expect == XREF)
4550 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
4556 yylval.ival = CopLINE(PL_curcop);
4557 if (isSPACE(*s) || *s == '#')
4558 PL_copline = NOLINE; /* invalidate current command line number */
4563 if (PL_lex_brackets <= 0)
4564 yyerror("Unmatched right curly bracket");
4566 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
4567 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
4568 PL_lex_formbrack = 0;
4569 if (PL_lex_state == LEX_INTERPNORMAL) {
4570 if (PL_lex_brackets == 0) {
4571 if (PL_expect & XFAKEBRACK) {
4572 PL_expect &= XENUMMASK;
4573 PL_lex_state = LEX_INTERPEND;
4578 PL_thiswhite = newSVpvs("");
4579 sv_catpvn(PL_thiswhite,"}",1);
4582 return yylex(); /* ignore fake brackets */
4584 if (*s == '-' && s[1] == '>')
4585 PL_lex_state = LEX_INTERPENDMAYBE;
4586 else if (*s != '[' && *s != '{')
4587 PL_lex_state = LEX_INTERPEND;
4590 if (PL_expect & XFAKEBRACK) {
4591 PL_expect &= XENUMMASK;
4593 return yylex(); /* ignore fake brackets */
4595 start_force(PL_curforce);
4597 curmad('X', newSVpvn(s-1,1));
4598 CURMAD('_', PL_thiswhite);
4603 PL_thistoken = newSVpvs("");
4611 if (PL_expect == XOPERATOR) {
4612 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
4613 && isIDFIRST_lazy_if(s,UTF))
4615 CopLINE_dec(PL_curcop);
4616 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
4617 CopLINE_inc(PL_curcop);
4622 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4624 PL_expect = XOPERATOR;
4625 force_ident(PL_tokenbuf, '&');
4629 yylval.ival = (OPpENTERSUB_AMPER<<8);
4641 const char tmp = *s++;
4648 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
4649 && strchr("+-*/%.^&|<",tmp))
4650 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4651 "Reversed %c= operator",(int)tmp);
4653 if (PL_expect == XSTATE && isALPHA(tmp) &&
4654 (s == PL_linestart+1 || s[-2] == '\n') )
4656 if (PL_in_eval && !PL_rsfp) {
4661 if (strnEQ(s,"=cut",4)) {
4677 PL_thiswhite = newSVpvs("");
4678 sv_catpvn(PL_thiswhite, PL_linestart,
4679 PL_bufend - PL_linestart);
4683 PL_doextract = TRUE;
4687 if (PL_lex_brackets < PL_lex_formbrack) {
4689 #ifdef PERL_STRICT_CR
4690 while (SPACE_OR_TAB(*t))
4692 while (SPACE_OR_TAB(*t) || *t == '\r')
4695 if (*t == '\n' || *t == '#') {
4706 const char tmp = *s++;
4708 /* was this !=~ where !~ was meant?
4709 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
4711 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
4712 const char *t = s+1;
4714 while (t < PL_bufend && isSPACE(*t))
4717 if (*t == '/' || *t == '?' ||
4718 ((*t == 'm' || *t == 's' || *t == 'y')
4719 && !isALNUM(t[1])) ||
4720 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
4721 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4722 "!=~ should be !~");
4732 if (PL_expect != XOPERATOR) {
4733 if (s[1] != '<' && !strchr(s,'>'))
4736 s = scan_heredoc(s);
4738 s = scan_inputsymbol(s);
4739 TERM(sublex_start());
4745 SHop(OP_LEFT_SHIFT);
4759 const char tmp = *s++;
4761 SHop(OP_RIGHT_SHIFT);
4762 else if (tmp == '=')
4771 if (PL_expect == XOPERATOR) {
4772 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
4774 deprecate_old(commaless_variable_list);
4775 return REPORT(','); /* grandfather non-comma-format format */
4779 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
4780 PL_tokenbuf[0] = '@';
4781 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
4782 sizeof PL_tokenbuf - 1, FALSE);
4783 if (PL_expect == XOPERATOR)
4784 no_op("Array length", s);
4785 if (!PL_tokenbuf[1])
4787 PL_expect = XOPERATOR;
4788 PL_pending_ident = '#';
4792 PL_tokenbuf[0] = '$';
4793 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4794 sizeof PL_tokenbuf - 1, FALSE);
4795 if (PL_expect == XOPERATOR)
4797 if (!PL_tokenbuf[1]) {
4799 yyerror("Final $ should be \\$ or $name");
4803 /* This kludge not intended to be bulletproof. */
4804 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
4805 yylval.opval = newSVOP(OP_CONST, 0,
4806 newSViv(CopARYBASE_get(&PL_compiling)));
4807 yylval.opval->op_private = OPpCONST_ARYBASE;
4813 const char tmp = *s;
4814 if (PL_lex_state == LEX_NORMAL)
4817 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
4818 && intuit_more(s)) {
4820 PL_tokenbuf[0] = '@';
4821 if (ckWARN(WARN_SYNTAX)) {
4824 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
4827 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4828 while (t < PL_bufend && *t != ']')
4830 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4831 "Multidimensional syntax %.*s not supported",
4832 (int)((t - PL_bufptr) + 1), PL_bufptr);
4836 else if (*s == '{') {
4838 PL_tokenbuf[0] = '%';
4839 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
4840 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
4842 char tmpbuf[sizeof PL_tokenbuf];
4845 } while (isSPACE(*t));
4846 if (isIDFIRST_lazy_if(t,UTF)) {
4848 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
4852 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
4853 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4854 "You need to quote \"%s\"",
4861 PL_expect = XOPERATOR;
4862 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
4863 const bool islop = (PL_last_lop == PL_oldoldbufptr);
4864 if (!islop || PL_last_lop_op == OP_GREPSTART)
4865 PL_expect = XOPERATOR;
4866 else if (strchr("$@\"'`q", *s))
4867 PL_expect = XTERM; /* e.g. print $fh "foo" */
4868 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
4869 PL_expect = XTERM; /* e.g. print $fh &sub */
4870 else if (isIDFIRST_lazy_if(s,UTF)) {
4871 char tmpbuf[sizeof PL_tokenbuf];
4873 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
4874 if ((t2 = keyword(tmpbuf, len, 0))) {
4875 /* binary operators exclude handle interpretations */
4887 PL_expect = XTERM; /* e.g. print $fh length() */
4892 PL_expect = XTERM; /* e.g. print $fh subr() */
4895 else if (isDIGIT(*s))
4896 PL_expect = XTERM; /* e.g. print $fh 3 */
4897 else if (*s == '.' && isDIGIT(s[1]))
4898 PL_expect = XTERM; /* e.g. print $fh .3 */
4899 else if ((*s == '?' || *s == '-' || *s == '+')
4900 && !isSPACE(s[1]) && s[1] != '=')
4901 PL_expect = XTERM; /* e.g. print $fh -1 */
4902 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
4904 PL_expect = XTERM; /* e.g. print $fh /.../
4905 XXX except DORDOR operator
4907 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
4909 PL_expect = XTERM; /* print $fh <<"EOF" */
4912 PL_pending_ident = '$';
4916 if (PL_expect == XOPERATOR)
4918 PL_tokenbuf[0] = '@';
4919 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
4920 if (!PL_tokenbuf[1]) {
4923 if (PL_lex_state == LEX_NORMAL)
4925 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
4927 PL_tokenbuf[0] = '%';
4929 /* Warn about @ where they meant $. */
4930 if (*s == '[' || *s == '{') {
4931 if (ckWARN(WARN_SYNTAX)) {
4932 const char *t = s + 1;
4933 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
4935 if (*t == '}' || *t == ']') {
4937 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
4938 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
4939 "Scalar value %.*s better written as $%.*s",
4940 (int)(t-PL_bufptr), PL_bufptr,
4941 (int)(t-PL_bufptr-1), PL_bufptr+1);
4946 PL_pending_ident = '@';
4949 case '/': /* may be division, defined-or, or pattern */
4950 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
4954 case '?': /* may either be conditional or pattern */
4955 if(PL_expect == XOPERATOR) {
4963 /* A // operator. */
4973 /* Disable warning on "study /blah/" */
4974 if (PL_oldoldbufptr == PL_last_uni
4975 && (*PL_last_uni != 's' || s - PL_last_uni < 5
4976 || memNE(PL_last_uni, "study", 5)
4977 || isALNUM_lazy_if(PL_last_uni+5,UTF)
4980 s = scan_pat(s,OP_MATCH);
4981 TERM(sublex_start());
4985 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
4986 #ifdef PERL_STRICT_CR
4989 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
4991 && (s == PL_linestart || s[-1] == '\n') )
4993 PL_lex_formbrack = 0;
4997 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5003 yylval.ival = OPf_SPECIAL;
5009 if (PL_expect != XOPERATOR)
5014 case '0': case '1': case '2': case '3': case '4':
5015 case '5': case '6': case '7': case '8': case '9':
5016 s = scan_num(s, &yylval);
5017 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5018 if (PL_expect == XOPERATOR)
5023 s = scan_str(s,!!PL_madskills,FALSE);
5024 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5025 if (PL_expect == XOPERATOR) {
5026 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5028 deprecate_old(commaless_variable_list);
5029 return REPORT(','); /* grandfather non-comma-format format */
5036 yylval.ival = OP_CONST;
5037 TERM(sublex_start());
5040 s = scan_str(s,!!PL_madskills,FALSE);
5041 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5042 if (PL_expect == XOPERATOR) {
5043 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5045 deprecate_old(commaless_variable_list);
5046 return REPORT(','); /* grandfather non-comma-format format */
5053 yylval.ival = OP_CONST;
5054 /* FIXME. I think that this can be const if char *d is replaced by
5055 more localised variables. */
5056 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5057 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5058 yylval.ival = OP_STRINGIFY;
5062 TERM(sublex_start());
5065 s = scan_str(s,!!PL_madskills,FALSE);
5066 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5067 if (PL_expect == XOPERATOR)
5068 no_op("Backticks",s);
5071 readpipe_override();
5072 TERM(sublex_start());
5076 if (PL_lex_inwhat && isDIGIT(*s) && ckWARN(WARN_SYNTAX))
5077 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5079 if (PL_expect == XOPERATOR)
5080 no_op("Backslash",s);
5084 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5085 char *start = s + 2;
5086 while (isDIGIT(*start) || *start == '_')
5088 if (*start == '.' && isDIGIT(start[1])) {
5089 s = scan_num(s, &yylval);
5092 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5093 else if (!isALPHA(*start) && (PL_expect == XTERM
5094 || PL_expect == XREF || PL_expect == XSTATE
5095 || PL_expect == XTERMORDORDOR)) {
5096 /* XXX Use gv_fetchpvn rather than stomping on a const string */
5097 const char c = *start;
5100 gv = gv_fetchpv(s, 0, SVt_PVCV);
5103 s = scan_num(s, &yylval);
5110 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5152 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5154 /* Some keywords can be followed by any delimiter, including ':' */
5155 tmp = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5156 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5157 (PL_tokenbuf[0] == 'q' &&
5158 strchr("qwxr", PL_tokenbuf[1])))));
5160 /* x::* is just a word, unless x is "CORE" */
5161 if (!tmp && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5165 while (d < PL_bufend && isSPACE(*d))
5166 d++; /* no comments skipped here, or s### is misparsed */
5168 /* Is this a label? */
5169 if (!tmp && PL_expect == XSTATE
5170 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5172 yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5177 /* Check for keywords */
5178 tmp = keyword(PL_tokenbuf, len, 0);
5180 /* Is this a word before a => operator? */
5181 if (*d == '=' && d[1] == '>') {
5184 = (OP*)newSVOP(OP_CONST, 0,
5185 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5186 yylval.opval->op_private = OPpCONST_BARE;
5190 if (tmp < 0) { /* second-class keyword? */
5191 GV *ogv = NULL; /* override (winner) */
5192 GV *hgv = NULL; /* hidden (loser) */
5193 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5195 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5198 if (GvIMPORTED_CV(gv))
5200 else if (! CvMETHOD(cv))
5204 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5205 (gv = *gvp) && isGV_with_GP(gv) &&
5206 GvCVu(gv) && GvIMPORTED_CV(gv))
5213 tmp = 0; /* overridden by import or by GLOBAL */
5216 && -tmp==KEY_lock /* XXX generalizable kludge */
5219 tmp = 0; /* any sub overrides "weak" keyword */
5221 else { /* no override */
5223 if (tmp == KEY_dump && ckWARN(WARN_MISC)) {
5224 Perl_warner(aTHX_ packWARN(WARN_MISC),
5225 "dump() better written as CORE::dump()");
5229 if (hgv && tmp != KEY_x && tmp != KEY_CORE
5230 && ckWARN(WARN_AMBIGUOUS)) /* never ambiguous */
5231 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5232 "Ambiguous call resolved as CORE::%s(), %s",
5233 GvENAME(hgv), "qualify as such or use &");
5240 default: /* not a keyword */
5241 /* Trade off - by using this evil construction we can pull the
5242 variable gv into the block labelled keylookup. If not, then
5243 we have to give it function scope so that the goto from the
5244 earlier ':' case doesn't bypass the initialisation. */
5246 just_a_word_zero_gv:
5254 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5257 SV *nextPL_nextwhite = 0;
5261 /* Get the rest if it looks like a package qualifier */
5263 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5265 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5268 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5269 *s == '\'' ? "'" : "::");
5274 if (PL_expect == XOPERATOR) {
5275 if (PL_bufptr == PL_linestart) {
5276 CopLINE_dec(PL_curcop);
5277 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), PL_warn_nosemi);
5278 CopLINE_inc(PL_curcop);
5281 no_op("Bareword",s);
5284 /* Look for a subroutine with this name in current package,
5285 unless name is "Foo::", in which case Foo is a bearword
5286 (and a package name). */
5288 if (len > 2 && !PL_madskills &&
5289 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5291 if (ckWARN(WARN_BAREWORD)
5292 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5293 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5294 "Bareword \"%s\" refers to nonexistent package",
5297 PL_tokenbuf[len] = '\0';
5303 /* Mustn't actually add anything to a symbol table.
5304 But also don't want to "initialise" any placeholder
5305 constants that might already be there into full
5306 blown PVGVs with attached PVCV. */
5307 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5308 GV_NOADD_NOINIT, SVt_PVCV);
5313 /* if we saw a global override before, get the right name */
5316 sv = newSVpvs("CORE::GLOBAL::");
5317 sv_catpv(sv,PL_tokenbuf);
5320 /* If len is 0, newSVpv does strlen(), which is correct.
5321 If len is non-zero, then it will be the true length,
5322 and so the scalar will be created correctly. */
5323 sv = newSVpv(PL_tokenbuf,len);
5326 if (PL_madskills && !PL_thistoken) {
5327 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5328 PL_thistoken = newSVpv(start,s - start);
5329 PL_realtokenstart = s - SvPVX(PL_linestr);
5333 /* Presume this is going to be a bareword of some sort. */
5336 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
5337 yylval.opval->op_private = OPpCONST_BARE;
5338 /* UTF-8 package name? */
5339 if (UTF && !IN_BYTES &&
5340 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
5343 /* And if "Foo::", then that's what it certainly is. */
5348 /* Do the explicit type check so that we don't need to force
5349 the initialisation of the symbol table to have a real GV.
5350 Beware - gv may not really be a PVGV, cv may not really be
5351 a PVCV, (because of the space optimisations that gv_init
5352 understands) But they're true if for this symbol there is
5353 respectively a typeglob and a subroutine.
5355 cv = gv ? ((SvTYPE(gv) == SVt_PVGV)
5356 /* Real typeglob, so get the real subroutine: */
5358 /* A proxy for a subroutine in this package? */
5359 : SvOK(gv) ? (CV *) gv : NULL)
5362 /* See if it's the indirect object for a list operator. */
5364 if (PL_oldoldbufptr &&
5365 PL_oldoldbufptr < PL_bufptr &&
5366 (PL_oldoldbufptr == PL_last_lop
5367 || PL_oldoldbufptr == PL_last_uni) &&
5368 /* NO SKIPSPACE BEFORE HERE! */
5369 (PL_expect == XREF ||
5370 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
5372 bool immediate_paren = *s == '(';
5374 /* (Now we can afford to cross potential line boundary.) */
5375 s = SKIPSPACE2(s,nextPL_nextwhite);
5377 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
5380 /* Two barewords in a row may indicate method call. */
5382 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
5383 (tmp = intuit_method(s, gv, cv)))
5386 /* If not a declared subroutine, it's an indirect object. */
5387 /* (But it's an indir obj regardless for sort.) */
5388 /* Also, if "_" follows a filetest operator, it's a bareword */
5391 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
5393 (PL_last_lop_op != OP_MAPSTART &&
5394 PL_last_lop_op != OP_GREPSTART))))
5395 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
5396 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
5399 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
5404 PL_expect = XOPERATOR;
5407 s = SKIPSPACE2(s,nextPL_nextwhite);
5408 PL_nextwhite = nextPL_nextwhite;
5413 /* Is this a word before a => operator? */
5414 if (*s == '=' && s[1] == '>' && !pkgname) {
5416 sv_setpv(((SVOP*)yylval.opval)->op_sv, PL_tokenbuf);
5417 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
5418 SvUTF8_on(((SVOP*)yylval.opval)->op_sv);
5422 /* If followed by a paren, it's certainly a subroutine. */
5427 while (SPACE_OR_TAB(*d))
5429 if (*d == ')' && (sv = gv_const_sv(gv))) {
5433 char *par = SvPVX(PL_linestr) + PL_realtokenstart;
5434 sv_catpvn(PL_thistoken, par, s - par);
5436 sv_free(PL_nextwhite);
5447 PL_nextwhite = PL_thiswhite;
5450 start_force(PL_curforce);
5452 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5453 PL_expect = XOPERATOR;
5456 PL_nextwhite = nextPL_nextwhite;
5457 curmad('X', PL_thistoken);
5458 PL_thistoken = newSVpvs("");
5466 /* If followed by var or block, call it a method (unless sub) */
5468 if ((*s == '$' || *s == '{') && (!gv || !cv)) {
5469 PL_last_lop = PL_oldbufptr;
5470 PL_last_lop_op = OP_METHOD;
5474 /* If followed by a bareword, see if it looks like indir obj. */
5477 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
5478 && (tmp = intuit_method(s, gv, cv)))
5481 /* Not a method, so call it a subroutine (if defined) */
5484 if (lastchar == '-' && ckWARN_d(WARN_AMBIGUOUS))
5485 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5486 "Ambiguous use of -%s resolved as -&%s()",
5487 PL_tokenbuf, PL_tokenbuf);
5488 /* Check for a constant sub */
5489 if ((sv = gv_const_sv(gv)) && !PL_madskills) {
5491 SvREFCNT_dec(((SVOP*)yylval.opval)->op_sv);
5492 ((SVOP*)yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
5493 yylval.opval->op_private = 0;
5497 /* Resolve to GV now. */
5498 if (SvTYPE(gv) != SVt_PVGV) {
5499 gv = gv_fetchpv(PL_tokenbuf, 0, SVt_PVCV);
5500 assert (SvTYPE(gv) == SVt_PVGV);
5501 /* cv must have been some sort of placeholder, so
5502 now needs replacing with a real code reference. */
5506 op_free(yylval.opval);
5507 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5508 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5509 PL_last_lop = PL_oldbufptr;
5510 PL_last_lop_op = OP_ENTERSUB;
5511 /* Is there a prototype? */
5519 const char *proto = SvPV_const((SV*)cv, protolen);
5522 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
5524 while (*proto == ';')
5526 if (*proto == '&' && *s == '{') {
5527 sv_setpv(PL_subname,
5530 "__ANON__" : "__ANON__::__ANON__"));
5537 PL_nextwhite = PL_thiswhite;
5540 start_force(PL_curforce);
5541 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5544 PL_nextwhite = nextPL_nextwhite;
5545 curmad('X', PL_thistoken);
5546 PL_thistoken = newSVpvs("");
5553 /* Guess harder when madskills require "best effort". */
5554 if (PL_madskills && (!gv || !GvCVu(gv))) {
5555 int probable_sub = 0;
5556 if (strchr("\"'`$@%0123456789!*+{[<", *s))
5558 else if (isALPHA(*s)) {
5562 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
5563 if (!keyword(tmpbuf, tmplen, 0))
5566 while (d < PL_bufend && isSPACE(*d))
5568 if (*d == '=' && d[1] == '>')
5573 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
5574 op_free(yylval.opval);
5575 yylval.opval = newCVREF(0, newGVOP(OP_GV, 0, gv));
5576 yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
5577 PL_last_lop = PL_oldbufptr;
5578 PL_last_lop_op = OP_ENTERSUB;
5579 PL_nextwhite = PL_thiswhite;
5581 start_force(PL_curforce);
5582 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5584 PL_nextwhite = nextPL_nextwhite;
5585 curmad('X', PL_thistoken);
5586 PL_thistoken = newSVpvs("");
5591 NEXTVAL_NEXTTOKE.opval = yylval.opval;
5598 /* Call it a bare word */
5600 if (PL_hints & HINT_STRICT_SUBS)
5601 yylval.opval->op_private |= OPpCONST_STRICT;
5604 if (lastchar != '-') {
5605 if (ckWARN(WARN_RESERVED)) {
5609 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
5610 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
5617 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')
5618 && ckWARN_d(WARN_AMBIGUOUS)) {
5619 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5620 "Operator or semicolon missing before %c%s",
5621 lastchar, PL_tokenbuf);
5622 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5623 "Ambiguous use of %c resolved as operator %c",
5624 lastchar, lastchar);
5630 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5631 newSVpv(CopFILE(PL_curcop),0));
5635 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5636 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
5639 case KEY___PACKAGE__:
5640 yylval.opval = (OP*)newSVOP(OP_CONST, 0,
5642 ? newSVhek(HvNAME_HEK(PL_curstash))
5649 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
5650 const char *pname = "main";
5651 if (PL_tokenbuf[2] == 'D')
5652 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
5653 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
5657 GvIOp(gv) = newIO();
5658 IoIFP(GvIOp(gv)) = PL_rsfp;
5659 #if defined(HAS_FCNTL) && defined(F_SETFD)
5661 const int fd = PerlIO_fileno(PL_rsfp);
5662 fcntl(fd,F_SETFD,fd >= 3);
5665 /* Mark this internal pseudo-handle as clean */
5666 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
5668 IoTYPE(GvIOp(gv)) = IoTYPE_PIPE;
5669 else if ((PerlIO*)PL_rsfp == PerlIO_stdin())
5670 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
5672 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
5673 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
5674 /* if the script was opened in binmode, we need to revert
5675 * it to text mode for compatibility; but only iff it has CRs
5676 * XXX this is a questionable hack at best. */
5677 if (PL_bufend-PL_bufptr > 2
5678 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
5681 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
5682 loc = PerlIO_tell(PL_rsfp);
5683 (void)PerlIO_seek(PL_rsfp, 0L, 0);
5686 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
5688 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
5689 #endif /* NETWARE */
5690 #ifdef PERLIO_IS_STDIO /* really? */
5691 # if defined(__BORLANDC__)
5692 /* XXX see note in do_binmode() */
5693 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
5697 PerlIO_seek(PL_rsfp, loc, 0);
5701 #ifdef PERLIO_LAYERS
5704 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
5705 else if (PL_encoding) {
5712 XPUSHs(PL_encoding);
5714 call_method("name", G_SCALAR);
5718 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
5719 Perl_form(aTHX_ ":encoding(%"SVf")",
5728 if (PL_realtokenstart >= 0) {
5729 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
5731 PL_endwhite = newSVpvs("");
5732 sv_catsv(PL_endwhite, PL_thiswhite);
5734 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
5735 PL_realtokenstart = -1;
5737 while ((s = filter_gets(PL_endwhite, PL_rsfp,
5738 SvCUR(PL_endwhite))) != Nullch) ;
5753 if (PL_expect == XSTATE) {
5760 if (*s == ':' && s[1] == ':') {
5763 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5764 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
5765 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
5768 else if (tmp == KEY_require || tmp == KEY_do)
5769 /* that's a way to remember we saw "CORE::" */
5782 LOP(OP_ACCEPT,XTERM);
5788 LOP(OP_ATAN2,XTERM);
5794 LOP(OP_BINMODE,XTERM);
5797 LOP(OP_BLESS,XTERM);
5806 /* When 'use switch' is in effect, continue has a dual
5807 life as a control operator. */
5809 if (!FEATURE_IS_ENABLED("switch"))
5812 /* We have to disambiguate the two senses of
5813 "continue". If the next token is a '{' then
5814 treat it as the start of a continue block;
5815 otherwise treat it as a control operator.
5827 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
5844 if (!PL_cryptseen) {
5845 PL_cryptseen = TRUE;
5849 LOP(OP_CRYPT,XTERM);
5852 LOP(OP_CHMOD,XTERM);
5855 LOP(OP_CHOWN,XTERM);
5858 LOP(OP_CONNECT,XTERM);
5877 s = force_word(s,WORD,TRUE,TRUE,FALSE);
5878 if (orig_keyword == KEY_do) {
5887 PL_hints |= HINT_BLOCK_SCOPE;
5897 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
5898 LOP(OP_DBMOPEN,XTERM);
5904 s = force_word(s,WORD,TRUE,FALSE,FALSE);
5911 yylval.ival = CopLINE(PL_curcop);
5927 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
5928 UNIBRACK(OP_ENTEREVAL);
5946 case KEY_endhostent:
5952 case KEY_endservent:
5955 case KEY_endprotoent:
5966 yylval.ival = CopLINE(PL_curcop);
5968 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
5971 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
5974 if ((PL_bufend - p) >= 3 &&
5975 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
5977 else if ((PL_bufend - p) >= 4 &&
5978 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
5981 if (isIDFIRST_lazy_if(p,UTF)) {
5982 p = scan_ident(p, PL_bufend,
5983 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5987 Perl_croak(aTHX_ "Missing $ on loop variable");
5989 s = SvPVX(PL_linestr) + soff;
5995 LOP(OP_FORMLINE,XTERM);
6001 LOP(OP_FCNTL,XTERM);
6007 LOP(OP_FLOCK,XTERM);
6016 LOP(OP_GREPSTART, XREF);
6019 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6034 case KEY_getpriority:
6035 LOP(OP_GETPRIORITY,XTERM);
6037 case KEY_getprotobyname:
6040 case KEY_getprotobynumber:
6041 LOP(OP_GPBYNUMBER,XTERM);
6043 case KEY_getprotoent:
6055 case KEY_getpeername:
6056 UNI(OP_GETPEERNAME);
6058 case KEY_gethostbyname:
6061 case KEY_gethostbyaddr:
6062 LOP(OP_GHBYADDR,XTERM);
6064 case KEY_gethostent:
6067 case KEY_getnetbyname:
6070 case KEY_getnetbyaddr:
6071 LOP(OP_GNBYADDR,XTERM);
6076 case KEY_getservbyname:
6077 LOP(OP_GSBYNAME,XTERM);
6079 case KEY_getservbyport:
6080 LOP(OP_GSBYPORT,XTERM);
6082 case KEY_getservent:
6085 case KEY_getsockname:
6086 UNI(OP_GETSOCKNAME);
6088 case KEY_getsockopt:
6089 LOP(OP_GSOCKOPT,XTERM);
6104 yylval.ival = CopLINE(PL_curcop);
6115 yylval.ival = CopLINE(PL_curcop);
6119 LOP(OP_INDEX,XTERM);
6125 LOP(OP_IOCTL,XTERM);
6137 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6169 LOP(OP_LISTEN,XTERM);
6178 s = scan_pat(s,OP_MATCH);
6179 TERM(sublex_start());
6182 LOP(OP_MAPSTART, XREF);
6185 LOP(OP_MKDIR,XTERM);
6188 LOP(OP_MSGCTL,XTERM);
6191 LOP(OP_MSGGET,XTERM);
6194 LOP(OP_MSGRCV,XTERM);
6197 LOP(OP_MSGSND,XTERM);
6202 PL_in_my = (U16)tmp;
6204 if (isIDFIRST_lazy_if(s,UTF)) {
6208 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6209 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6211 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6212 if (!PL_in_my_stash) {
6215 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6219 if (PL_madskills) { /* just add type to declarator token */
6220 sv_catsv(PL_thistoken, PL_nextwhite);
6222 sv_catpvn(PL_thistoken, start, s - start);
6230 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6237 s = tokenize_use(0, s);
6241 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6248 if (isIDFIRST_lazy_if(s,UTF)) {
6250 for (d = s; isALNUM_lazy_if(d,UTF);)
6252 for (t=d; isSPACE(*t);)
6254 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6256 && !(t[0] == '=' && t[1] == '>')
6258 int parms_len = (int)(d-s);
6259 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6260 "Precedence problem: open %.*s should be open(%.*s)",
6261 parms_len, s, parms_len, s);
6267 yylval.ival = OP_OR;
6277 LOP(OP_OPEN_DIR,XTERM);
6280 checkcomma(s,PL_tokenbuf,"filehandle");
6284 checkcomma(s,PL_tokenbuf,"filehandle");
6303 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6307 LOP(OP_PIPE_OP,XTERM);
6310 s = scan_str(s,!!PL_madskills,FALSE);
6313 yylval.ival = OP_CONST;
6314 TERM(sublex_start());
6320 s = scan_str(s,!!PL_madskills,FALSE);
6323 PL_expect = XOPERATOR;
6325 if (SvCUR(PL_lex_stuff)) {
6328 d = SvPV_force(PL_lex_stuff, len);
6330 for (; isSPACE(*d) && len; --len, ++d)
6335 if (!warned && ckWARN(WARN_QW)) {
6336 for (; !isSPACE(*d) && len; --len, ++d) {
6338 Perl_warner(aTHX_ packWARN(WARN_QW),
6339 "Possible attempt to separate words with commas");
6342 else if (*d == '#') {
6343 Perl_warner(aTHX_ packWARN(WARN_QW),
6344 "Possible attempt to put comments in qw() list");
6350 for (; !isSPACE(*d) && len; --len, ++d)
6353 sv = newSVpvn(b, d-b);
6354 if (DO_UTF8(PL_lex_stuff))
6356 words = append_elem(OP_LIST, words,
6357 newSVOP(OP_CONST, 0, tokeq(sv)));
6361 start_force(PL_curforce);
6362 NEXTVAL_NEXTTOKE.opval = words;
6367 SvREFCNT_dec(PL_lex_stuff);
6368 PL_lex_stuff = NULL;
6374 s = scan_str(s,!!PL_madskills,FALSE);
6377 yylval.ival = OP_STRINGIFY;
6378 if (SvIVX(PL_lex_stuff) == '\'')
6379 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
6380 TERM(sublex_start());
6383 s = scan_pat(s,OP_QR);
6384 TERM(sublex_start());
6387 s = scan_str(s,!!PL_madskills,FALSE);
6390 readpipe_override();
6391 TERM(sublex_start());
6399 s = force_version(s, FALSE);
6401 else if (*s != 'v' || !isDIGIT(s[1])
6402 || (s = force_version(s, TRUE), *s == 'v'))
6404 *PL_tokenbuf = '\0';
6405 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6406 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
6407 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
6409 yyerror("<> should be quotes");
6411 if (orig_keyword == KEY_require) {
6419 PL_last_uni = PL_oldbufptr;
6420 PL_last_lop_op = OP_REQUIRE;
6422 return REPORT( (int)REQUIRE );
6428 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6432 LOP(OP_RENAME,XTERM);
6441 LOP(OP_RINDEX,XTERM);
6451 UNIDOR(OP_READLINE);
6455 UNIDOR(OP_BACKTICK);
6464 LOP(OP_REVERSE,XTERM);
6467 UNIDOR(OP_READLINK);
6475 TERM(sublex_start());
6477 TOKEN(1); /* force error */
6480 checkcomma(s,PL_tokenbuf,"filehandle");
6490 LOP(OP_SELECT,XTERM);
6496 LOP(OP_SEMCTL,XTERM);
6499 LOP(OP_SEMGET,XTERM);
6502 LOP(OP_SEMOP,XTERM);
6508 LOP(OP_SETPGRP,XTERM);
6510 case KEY_setpriority:
6511 LOP(OP_SETPRIORITY,XTERM);
6513 case KEY_sethostent:
6519 case KEY_setservent:
6522 case KEY_setprotoent:
6532 LOP(OP_SEEKDIR,XTERM);
6534 case KEY_setsockopt:
6535 LOP(OP_SSOCKOPT,XTERM);
6541 LOP(OP_SHMCTL,XTERM);
6544 LOP(OP_SHMGET,XTERM);
6547 LOP(OP_SHMREAD,XTERM);
6550 LOP(OP_SHMWRITE,XTERM);
6553 LOP(OP_SHUTDOWN,XTERM);
6562 LOP(OP_SOCKET,XTERM);
6564 case KEY_socketpair:
6565 LOP(OP_SOCKPAIR,XTERM);
6568 checkcomma(s,PL_tokenbuf,"subroutine name");
6570 if (*s == ';' || *s == ')') /* probably a close */
6571 Perl_croak(aTHX_ "sort is now a reserved word");
6573 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6577 LOP(OP_SPLIT,XTERM);
6580 LOP(OP_SPRINTF,XTERM);
6583 LOP(OP_SPLICE,XTERM);
6598 LOP(OP_SUBSTR,XTERM);
6604 char tmpbuf[sizeof PL_tokenbuf];
6605 SSize_t tboffset = 0;
6606 expectation attrful;
6607 bool have_name, have_proto;
6608 const int key = tmp;
6613 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6614 SV *subtoken = newSVpvn(tstart, s - tstart);
6618 s = SKIPSPACE2(s,tmpwhite);
6623 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
6624 (*s == ':' && s[1] == ':'))
6631 attrful = XATTRBLOCK;
6632 /* remember buffer pos'n for later force_word */
6633 tboffset = s - PL_oldbufptr;
6634 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6637 nametoke = newSVpvn(s, d - s);
6639 if (memchr(tmpbuf, ':', len))
6640 sv_setpvn(PL_subname, tmpbuf, len);
6642 sv_setsv(PL_subname,PL_curstname);
6643 sv_catpvs(PL_subname,"::");
6644 sv_catpvn(PL_subname,tmpbuf,len);
6651 CURMAD('X', nametoke);
6652 CURMAD('_', tmpwhite);
6653 (void) force_word(PL_oldbufptr + tboffset, WORD,
6656 s = SKIPSPACE2(d,tmpwhite);
6663 Perl_croak(aTHX_ "Missing name in \"my sub\"");
6664 PL_expect = XTERMBLOCK;
6665 attrful = XATTRTERM;
6666 sv_setpvn(PL_subname,"?",1);
6670 if (key == KEY_format) {
6672 PL_lex_formbrack = PL_lex_brackets + 1;
6674 PL_thistoken = subtoken;
6678 (void) force_word(PL_oldbufptr + tboffset, WORD,
6684 /* Look for a prototype */
6687 bool bad_proto = FALSE;
6688 const bool warnsyntax = ckWARN(WARN_SYNTAX);
6690 s = scan_str(s,!!PL_madskills,FALSE);
6692 Perl_croak(aTHX_ "Prototype not terminated");
6693 /* strip spaces and check for bad characters */
6694 d = SvPVX(PL_lex_stuff);
6696 for (p = d; *p; ++p) {
6699 if (warnsyntax && !strchr("$@%*;[]&\\_", *p))
6705 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6706 "Illegal character in prototype for %"SVf" : %s",
6707 SVfARG(PL_subname), d);
6708 SvCUR_set(PL_lex_stuff, tmp);
6713 CURMAD('q', PL_thisopen);
6714 CURMAD('_', tmpwhite);
6715 CURMAD('=', PL_thisstuff);
6716 CURMAD('Q', PL_thisclose);
6717 NEXTVAL_NEXTTOKE.opval =
6718 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6719 PL_lex_stuff = Nullsv;
6722 s = SKIPSPACE2(s,tmpwhite);
6730 if (*s == ':' && s[1] != ':')
6731 PL_expect = attrful;
6732 else if (*s != '{' && key == KEY_sub) {
6734 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
6736 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
6743 curmad('^', newSVpvs(""));
6744 CURMAD('_', tmpwhite);
6748 PL_thistoken = subtoken;
6751 NEXTVAL_NEXTTOKE.opval =
6752 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
6753 PL_lex_stuff = NULL;
6758 sv_setpv(PL_subname,
6760 (PL_curstash ? "__ANON__" : "__ANON__::__ANON__"));
6764 (void) force_word(PL_oldbufptr + tboffset, WORD,
6774 LOP(OP_SYSTEM,XREF);
6777 LOP(OP_SYMLINK,XTERM);
6780 LOP(OP_SYSCALL,XTERM);
6783 LOP(OP_SYSOPEN,XTERM);
6786 LOP(OP_SYSSEEK,XTERM);
6789 LOP(OP_SYSREAD,XTERM);
6792 LOP(OP_SYSWRITE,XTERM);
6796 TERM(sublex_start());
6817 LOP(OP_TRUNCATE,XTERM);
6829 yylval.ival = CopLINE(PL_curcop);
6833 yylval.ival = CopLINE(PL_curcop);
6837 LOP(OP_UNLINK,XTERM);
6843 LOP(OP_UNPACK,XTERM);
6846 LOP(OP_UTIME,XTERM);
6852 LOP(OP_UNSHIFT,XTERM);
6855 s = tokenize_use(1, s);
6865 yylval.ival = CopLINE(PL_curcop);
6869 yylval.ival = CopLINE(PL_curcop);
6873 PL_hints |= HINT_BLOCK_SCOPE;
6880 LOP(OP_WAITPID,XTERM);
6889 ctl_l[0] = toCTRL('L');
6891 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
6894 /* Make sure $^L is defined */
6895 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
6900 if (PL_expect == XOPERATOR)
6906 yylval.ival = OP_XOR;
6911 TERM(sublex_start());
6916 #pragma segment Main
6920 S_pending_ident(pTHX)
6925 /* pit holds the identifier we read and pending_ident is reset */
6926 char pit = PL_pending_ident;
6927 PL_pending_ident = 0;
6929 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
6930 DEBUG_T({ PerlIO_printf(Perl_debug_log,
6931 "### Pending identifier '%s'\n", PL_tokenbuf); });
6933 /* if we're in a my(), we can't allow dynamics here.
6934 $foo'bar has already been turned into $foo::bar, so
6935 just check for colons.
6937 if it's a legal name, the OP is a PADANY.
6940 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
6941 if (strchr(PL_tokenbuf,':'))
6942 yyerror(Perl_form(aTHX_ "No package name allowed for "
6943 "variable %s in \"our\"",
6945 tmp = allocmy(PL_tokenbuf);
6948 if (strchr(PL_tokenbuf,':'))
6949 yyerror(Perl_form(aTHX_ PL_no_myglob,
6950 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
6952 yylval.opval = newOP(OP_PADANY, 0);
6953 yylval.opval->op_targ = allocmy(PL_tokenbuf);
6959 build the ops for accesses to a my() variable.
6961 Deny my($a) or my($b) in a sort block, *if* $a or $b is
6962 then used in a comparison. This catches most, but not
6963 all cases. For instance, it catches
6964 sort { my($a); $a <=> $b }
6966 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
6967 (although why you'd do that is anyone's guess).
6970 if (!strchr(PL_tokenbuf,':')) {
6972 tmp = pad_findmy(PL_tokenbuf);
6973 if (tmp != NOT_IN_PAD) {
6974 /* might be an "our" variable" */
6975 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
6976 /* build ops for a bareword */
6977 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
6978 HEK * const stashname = HvNAME_HEK(stash);
6979 SV * const sym = newSVhek(stashname);
6980 sv_catpvs(sym, "::");
6981 sv_catpv(sym, PL_tokenbuf+1);
6982 yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
6983 yylval.opval->op_private = OPpCONST_ENTERED;
6986 ? (GV_ADDMULTI | GV_ADDINEVAL)
6989 ((PL_tokenbuf[0] == '$') ? SVt_PV
6990 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
6995 /* if it's a sort block and they're naming $a or $b */
6996 if (PL_last_lop_op == OP_SORT &&
6997 PL_tokenbuf[0] == '$' &&
6998 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
7001 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
7002 d < PL_bufend && *d != '\n';
7005 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7006 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7012 yylval.opval = newOP(OP_PADANY, 0);
7013 yylval.opval->op_targ = tmp;
7019 Whine if they've said @foo in a doublequoted string,
7020 and @foo isn't a variable we can find in the symbol
7023 if (pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7024 GV *gv = gv_fetchpv(PL_tokenbuf+1, 0, SVt_PVAV);
7025 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7026 && ckWARN(WARN_AMBIGUOUS)
7027 /* DO NOT warn for @- and @+ */
7028 && !( PL_tokenbuf[2] == '\0' &&
7029 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7032 /* Downgraded from fatal to warning 20000522 mjd */
7033 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7034 "Possible unintended interpolation of %s in string",
7039 /* build ops for a bareword */
7040 yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpv(PL_tokenbuf+1, 0));
7041 yylval.opval->op_private = OPpCONST_ENTERED;
7044 /* If the identifier refers to a stash, don't autovivify it.
7045 * Change 24660 had the side effect of causing symbol table
7046 * hashes to always be defined, even if they were freshly
7047 * created and the only reference in the entire program was
7048 * the single statement with the defined %foo::bar:: test.
7049 * It appears that all code in the wild doing this actually
7050 * wants to know whether sub-packages have been loaded, so
7051 * by avoiding auto-vivifying symbol tables, we ensure that
7052 * defined %foo::bar:: continues to be false, and the existing
7053 * tests still give the expected answers, even though what
7054 * they're actually testing has now changed subtly.
7056 (*PL_tokenbuf == '%' && *(d = PL_tokenbuf + strlen(PL_tokenbuf) - 1) == ':' && d[-1] == ':'
7058 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7059 ((PL_tokenbuf[0] == '$') ? SVt_PV
7060 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7066 * The following code was generated by perl_keyword.pl.
7070 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7075 case 1: /* 5 tokens of length 1 */
7107 case 2: /* 18 tokens of length 2 */
7253 case 3: /* 29 tokens of length 3 */
7257 if (name[1] == 'N' &&
7320 if (name[1] == 'i' &&
7342 return (all_keywords || FEATURE_IS_ENABLED("err") ? -KEY_err : 0);
7360 if (name[1] == 'o' &&
7369 if (name[1] == 'e' &&
7378 if (name[1] == 'n' &&
7387 if (name[1] == 'o' &&
7396 if (name[1] == 'a' &&
7405 if (name[1] == 'o' &&
7467 if (name[1] == 'e' &&
7481 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
7507 if (name[1] == 'i' &&
7516 if (name[1] == 's' &&
7525 if (name[1] == 'e' &&
7534 if (name[1] == 'o' &&
7546 case 4: /* 41 tokens of length 4 */
7550 if (name[1] == 'O' &&
7560 if (name[1] == 'N' &&
7570 if (name[1] == 'i' &&
7580 if (name[1] == 'h' &&
7590 if (name[1] == 'u' &&
7603 if (name[2] == 'c' &&
7612 if (name[2] == 's' &&
7621 if (name[2] == 'a' &&
7657 if (name[1] == 'o' &&
7670 if (name[2] == 't' &&
7679 if (name[2] == 'o' &&
7688 if (name[2] == 't' &&
7697 if (name[2] == 'e' &&
7710 if (name[1] == 'o' &&
7723 if (name[2] == 'y' &&
7732 if (name[2] == 'l' &&
7748 if (name[2] == 's' &&
7757 if (name[2] == 'n' &&
7766 if (name[2] == 'c' &&
7779 if (name[1] == 'e' &&
7789 if (name[1] == 'p' &&
7802 if (name[2] == 'c' &&
7811 if (name[2] == 'p' &&
7820 if (name[2] == 's' &&
7836 if (name[2] == 'n' &&
7906 if (name[2] == 'r' &&
7915 if (name[2] == 'r' &&
7924 if (name[2] == 'a' &&
7940 if (name[2] == 'l' &&
8002 if (name[2] == 'e' &&
8005 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8018 case 5: /* 39 tokens of length 5 */
8022 if (name[1] == 'E' &&
8033 if (name[1] == 'H' &&
8047 if (name[2] == 'a' &&
8057 if (name[2] == 'a' &&
8074 if (name[2] == 'e' &&
8084 if (name[2] == 'e' &&
8088 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8104 if (name[3] == 'i' &&
8113 if (name[3] == 'o' &&
8149 if (name[2] == 'o' &&
8159 if (name[2] == 'y' &&
8173 if (name[1] == 'l' &&
8187 if (name[2] == 'n' &&
8197 if (name[2] == 'o' &&
8211 if (name[1] == 'i' &&
8216 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8225 if (name[2] == 'd' &&
8235 if (name[2] == 'c' &&
8252 if (name[2] == 'c' &&
8262 if (name[2] == 't' &&
8276 if (name[1] == 'k' &&
8287 if (name[1] == 'r' &&
8301 if (name[2] == 's' &&
8311 if (name[2] == 'd' &&
8328 if (name[2] == 'm' &&
8338 if (name[2] == 'i' &&
8348 if (name[2] == 'e' &&
8358 if (name[2] == 'l' &&
8368 if (name[2] == 'a' &&
8381 if (name[3] == 't' &&
8384 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
8390 if (name[3] == 'd' &&
8407 if (name[1] == 'i' &&
8421 if (name[2] == 'a' &&
8434 if (name[3] == 'e' &&
8469 if (name[2] == 'i' &&
8486 if (name[2] == 'i' &&
8496 if (name[2] == 'i' &&
8513 case 6: /* 33 tokens of length 6 */
8517 if (name[1] == 'c' &&
8532 if (name[2] == 'l' &&
8543 if (name[2] == 'r' &&
8558 if (name[1] == 'e' &&
8573 if (name[2] == 's' &&
8578 if(ckWARN_d(WARN_SYNTAX))
8579 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
8585 if (name[2] == 'i' &&
8603 if (name[2] == 'l' &&
8614 if (name[2] == 'r' &&
8629 if (name[1] == 'm' &&
8644 if (name[2] == 'n' &&
8655 if (name[2] == 's' &&
8670 if (name[1] == 's' &&
8676 if (name[4] == 't' &&
8685 if (name[4] == 'e' &&
8694 if (name[4] == 'c' &&
8703 if (name[4] == 'n' &&
8719 if (name[1] == 'r' &&
8737 if (name[3] == 'a' &&
8747 if (name[3] == 'u' &&
8761 if (name[2] == 'n' &&
8779 if (name[2] == 'a' &&
8793 if (name[3] == 'e' &&
8806 if (name[4] == 't' &&
8815 if (name[4] == 'e' &&
8837 if (name[4] == 't' &&
8846 if (name[4] == 'e' &&
8862 if (name[2] == 'c' &&
8873 if (name[2] == 'l' &&
8884 if (name[2] == 'b' &&
8895 if (name[2] == 's' &&
8918 if (name[4] == 's' &&
8927 if (name[4] == 'n' &&
8940 if (name[3] == 'a' &&
8957 if (name[1] == 'a' &&
8972 case 7: /* 29 tokens of length 7 */
8976 if (name[1] == 'E' &&
8989 if (name[1] == '_' &&
9002 if (name[1] == 'i' &&
9009 return -KEY_binmode;
9015 if (name[1] == 'o' &&
9022 return -KEY_connect;
9031 if (name[2] == 'm' &&
9037 return -KEY_dbmopen;
9048 if (name[4] == 'u' &&
9052 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9058 if (name[4] == 'n' &&
9079 if (name[1] == 'o' &&
9092 if (name[1] == 'e' &&
9099 if (name[5] == 'r' &&
9102 return -KEY_getpgrp;
9108 if (name[5] == 'i' &&
9111 return -KEY_getppid;
9124 if (name[1] == 'c' &&
9131 return -KEY_lcfirst;
9137 if (name[1] == 'p' &&
9144 return -KEY_opendir;
9150 if (name[1] == 'a' &&
9168 if (name[3] == 'd' &&
9173 return -KEY_readdir;
9179 if (name[3] == 'u' &&
9190 if (name[3] == 'e' &&
9195 return -KEY_reverse;
9214 if (name[3] == 'k' &&
9219 return -KEY_seekdir;
9225 if (name[3] == 'p' &&
9230 return -KEY_setpgrp;
9240 if (name[2] == 'm' &&
9246 return -KEY_shmread;
9252 if (name[2] == 'r' &&
9258 return -KEY_sprintf;
9267 if (name[3] == 'l' &&
9272 return -KEY_symlink;
9281 if (name[4] == 'a' &&
9285 return -KEY_syscall;
9291 if (name[4] == 'p' &&
9295 return -KEY_sysopen;
9301 if (name[4] == 'e' &&
9305 return -KEY_sysread;
9311 if (name[4] == 'e' &&
9315 return -KEY_sysseek;
9333 if (name[1] == 'e' &&
9340 return -KEY_telldir;
9349 if (name[2] == 'f' &&
9355 return -KEY_ucfirst;
9361 if (name[2] == 's' &&
9367 return -KEY_unshift;
9377 if (name[1] == 'a' &&
9384 return -KEY_waitpid;
9393 case 8: /* 26 tokens of length 8 */
9397 if (name[1] == 'U' &&
9405 return KEY_AUTOLOAD;
9416 if (name[3] == 'A' &&
9422 return KEY___DATA__;
9428 if (name[3] == 'I' &&
9434 return -KEY___FILE__;
9440 if (name[3] == 'I' &&
9446 return -KEY___LINE__;
9462 if (name[2] == 'o' &&
9469 return -KEY_closedir;
9475 if (name[2] == 'n' &&
9482 return -KEY_continue;
9492 if (name[1] == 'b' &&
9500 return -KEY_dbmclose;
9506 if (name[1] == 'n' &&
9512 if (name[4] == 'r' &&
9517 return -KEY_endgrent;
9523 if (name[4] == 'w' &&
9528 return -KEY_endpwent;
9541 if (name[1] == 'o' &&
9549 return -KEY_formline;
9555 if (name[1] == 'e' &&
9566 if (name[6] == 'n' &&
9569 return -KEY_getgrent;
9575 if (name[6] == 'i' &&
9578 return -KEY_getgrgid;
9584 if (name[6] == 'a' &&
9587 return -KEY_getgrnam;
9600 if (name[4] == 'o' &&
9605 return -KEY_getlogin;
9616 if (name[6] == 'n' &&
9619 return -KEY_getpwent;
9625 if (name[6] == 'a' &&
9628 return -KEY_getpwnam;
9634 if (name[6] == 'i' &&
9637 return -KEY_getpwuid;
9657 if (name[1] == 'e' &&
9664 if (name[5] == 'i' &&
9671 return -KEY_readline;
9676 return -KEY_readlink;
9687 if (name[5] == 'i' &&
9691 return -KEY_readpipe;
9712 if (name[4] == 'r' &&
9717 return -KEY_setgrent;
9723 if (name[4] == 'w' &&
9728 return -KEY_setpwent;
9744 if (name[3] == 'w' &&
9750 return -KEY_shmwrite;
9756 if (name[3] == 't' &&
9762 return -KEY_shutdown;
9772 if (name[2] == 's' &&
9779 return -KEY_syswrite;
9789 if (name[1] == 'r' &&
9797 return -KEY_truncate;
9806 case 9: /* 9 tokens of length 9 */
9810 if (name[1] == 'N' &&
9819 return KEY_UNITCHECK;
9825 if (name[1] == 'n' &&
9834 return -KEY_endnetent;
9840 if (name[1] == 'e' &&
9849 return -KEY_getnetent;
9855 if (name[1] == 'o' &&
9864 return -KEY_localtime;
9870 if (name[1] == 'r' &&
9879 return KEY_prototype;
9885 if (name[1] == 'u' &&
9894 return -KEY_quotemeta;
9900 if (name[1] == 'e' &&
9909 return -KEY_rewinddir;
9915 if (name[1] == 'e' &&
9924 return -KEY_setnetent;
9930 if (name[1] == 'a' &&
9939 return -KEY_wantarray;
9948 case 10: /* 9 tokens of length 10 */
9952 if (name[1] == 'n' &&
9958 if (name[4] == 'o' &&
9965 return -KEY_endhostent;
9971 if (name[4] == 'e' &&
9978 return -KEY_endservent;
9991 if (name[1] == 'e' &&
9997 if (name[4] == 'o' &&
10004 return -KEY_gethostent;
10013 if (name[5] == 'r' &&
10019 return -KEY_getservent;
10025 if (name[5] == 'c' &&
10031 return -KEY_getsockopt;
10051 if (name[2] == 't')
10056 if (name[4] == 'o' &&
10063 return -KEY_sethostent;
10072 if (name[5] == 'r' &&
10078 return -KEY_setservent;
10084 if (name[5] == 'c' &&
10090 return -KEY_setsockopt;
10107 if (name[2] == 'c' &&
10116 return -KEY_socketpair;
10129 case 11: /* 8 tokens of length 11 */
10133 if (name[1] == '_' &&
10143 { /* __PACKAGE__ */
10144 return -KEY___PACKAGE__;
10150 if (name[1] == 'n' &&
10160 { /* endprotoent */
10161 return -KEY_endprotoent;
10167 if (name[1] == 'e' &&
10176 if (name[5] == 'e' &&
10182 { /* getpeername */
10183 return -KEY_getpeername;
10192 if (name[6] == 'o' &&
10197 { /* getpriority */
10198 return -KEY_getpriority;
10204 if (name[6] == 't' &&
10209 { /* getprotoent */
10210 return -KEY_getprotoent;
10224 if (name[4] == 'o' &&
10231 { /* getsockname */
10232 return -KEY_getsockname;
10245 if (name[1] == 'e' &&
10253 if (name[6] == 'o' &&
10258 { /* setpriority */
10259 return -KEY_setpriority;
10265 if (name[6] == 't' &&
10270 { /* setprotoent */
10271 return -KEY_setprotoent;
10287 case 12: /* 2 tokens of length 12 */
10288 if (name[0] == 'g' &&
10300 if (name[9] == 'd' &&
10303 { /* getnetbyaddr */
10304 return -KEY_getnetbyaddr;
10310 if (name[9] == 'a' &&
10313 { /* getnetbyname */
10314 return -KEY_getnetbyname;
10326 case 13: /* 4 tokens of length 13 */
10327 if (name[0] == 'g' &&
10334 if (name[4] == 'o' &&
10343 if (name[10] == 'd' &&
10346 { /* gethostbyaddr */
10347 return -KEY_gethostbyaddr;
10353 if (name[10] == 'a' &&
10356 { /* gethostbyname */
10357 return -KEY_gethostbyname;
10370 if (name[4] == 'e' &&
10379 if (name[10] == 'a' &&
10382 { /* getservbyname */
10383 return -KEY_getservbyname;
10389 if (name[10] == 'o' &&
10392 { /* getservbyport */
10393 return -KEY_getservbyport;
10412 case 14: /* 1 tokens of length 14 */
10413 if (name[0] == 'g' &&
10427 { /* getprotobyname */
10428 return -KEY_getprotobyname;
10433 case 16: /* 1 tokens of length 16 */
10434 if (name[0] == 'g' &&
10450 { /* getprotobynumber */
10451 return -KEY_getprotobynumber;
10465 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
10469 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
10470 if (ckWARN(WARN_SYNTAX)) {
10473 for (w = s+2; *w && level; w++) {
10476 else if (*w == ')')
10479 while (isSPACE(*w))
10481 /* the list of chars below is for end of statements or
10482 * block / parens, boolean operators (&&, ||, //) and branch
10483 * constructs (or, and, if, until, unless, while, err, for).
10484 * Not a very solid hack... */
10485 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
10486 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
10487 "%s (...) interpreted as function",name);
10490 while (s < PL_bufend && isSPACE(*s))
10494 while (s < PL_bufend && isSPACE(*s))
10496 if (isIDFIRST_lazy_if(s,UTF)) {
10497 const char * const w = s++;
10498 while (isALNUM_lazy_if(s,UTF))
10500 while (s < PL_bufend && isSPACE(*s))
10504 if (keyword(w, s - w, 0))
10507 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
10508 if (gv && GvCVu(gv))
10510 Perl_croak(aTHX_ "No comma allowed after %s", what);
10515 /* Either returns sv, or mortalizes sv and returns a new SV*.
10516 Best used as sv=new_constant(..., sv, ...).
10517 If s, pv are NULL, calls subroutine with one argument,
10518 and type is used with error messages only. */
10521 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, SV *sv, SV *pv,
10525 HV * const table = GvHV(PL_hintgv); /* ^H */
10529 const char *why1 = "", *why2 = "", *why3 = "";
10531 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
10534 why2 = (const char *)
10535 (strEQ(key,"charnames")
10536 ? "(possibly a missing \"use charnames ...\")"
10538 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
10539 (type ? type: "undef"), why2);
10541 /* This is convoluted and evil ("goto considered harmful")
10542 * but I do not understand the intricacies of all the different
10543 * failure modes of %^H in here. The goal here is to make
10544 * the most probable error message user-friendly. --jhi */
10549 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
10550 (type ? type: "undef"), why1, why2, why3);
10552 yyerror(SvPVX_const(msg));
10556 cvp = hv_fetch(table, key, strlen(key), FALSE);
10557 if (!cvp || !SvOK(*cvp)) {
10560 why3 = "} is not defined";
10563 sv_2mortal(sv); /* Parent created it permanently */
10566 pv = sv_2mortal(newSVpvn(s, len));
10568 typesv = sv_2mortal(newSVpv(type, 0));
10570 typesv = &PL_sv_undef;
10572 PUSHSTACKi(PERLSI_OVERLOAD);
10584 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
10588 /* Check the eval first */
10589 if (!PL_in_eval && SvTRUE(ERRSV)) {
10590 sv_catpvs(ERRSV, "Propagated");
10591 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
10593 res = SvREFCNT_inc_simple(sv);
10597 SvREFCNT_inc_simple_void(res);
10606 why1 = "Call to &{$^H{";
10608 why3 = "}} did not return a defined value";
10616 /* Returns a NUL terminated string, with the length of the string written to
10620 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
10623 register char *d = dest;
10624 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
10627 Perl_croak(aTHX_ ident_too_long);
10628 if (isALNUM(*s)) /* UTF handled below */
10630 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
10635 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
10639 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10640 char *t = s + UTF8SKIP(s);
10642 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10646 Perl_croak(aTHX_ ident_too_long);
10647 Copy(s, d, len, char);
10660 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
10663 char *bracket = NULL;
10665 register char *d = dest;
10666 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
10671 while (isDIGIT(*s)) {
10673 Perl_croak(aTHX_ ident_too_long);
10680 Perl_croak(aTHX_ ident_too_long);
10681 if (isALNUM(*s)) /* UTF handled below */
10683 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
10688 else if (*s == ':' && s[1] == ':') {
10692 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
10693 char *t = s + UTF8SKIP(s);
10694 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
10696 if (d + (t - s) > e)
10697 Perl_croak(aTHX_ ident_too_long);
10698 Copy(s, d, t - s, char);
10709 if (PL_lex_state != LEX_NORMAL)
10710 PL_lex_state = LEX_INTERPENDMAYBE;
10713 if (*s == '$' && s[1] &&
10714 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
10727 if (*d == '^' && *s && isCONTROLVAR(*s)) {
10732 if (isSPACE(s[-1])) {
10734 const char ch = *s++;
10735 if (!SPACE_OR_TAB(ch)) {
10741 if (isIDFIRST_lazy_if(d,UTF)) {
10745 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
10746 end += UTF8SKIP(end);
10747 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
10748 end += UTF8SKIP(end);
10750 Copy(s, d, end - s, char);
10755 while ((isALNUM(*s) || *s == ':') && d < e)
10758 Perl_croak(aTHX_ ident_too_long);
10761 while (s < send && SPACE_OR_TAB(*s))
10763 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
10764 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
10765 const char * const brack =
10767 ((*s == '[') ? "[...]" : "{...}");
10768 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10769 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
10770 funny, dest, brack, funny, dest, brack);
10773 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
10777 /* Handle extended ${^Foo} variables
10778 * 1999-02-27 mjd-perl-patch@plover.com */
10779 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
10783 while (isALNUM(*s) && d < e) {
10787 Perl_croak(aTHX_ ident_too_long);
10792 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
10793 PL_lex_state = LEX_INTERPEND;
10796 if (PL_lex_state == LEX_NORMAL) {
10797 if (ckWARN(WARN_AMBIGUOUS) &&
10798 (keyword(dest, d - dest, 0)
10799 || get_cvn_flags(dest, d - dest, 0)))
10803 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
10804 "Ambiguous use of %c{%s} resolved to %c%s",
10805 funny, dest, funny, dest);
10810 s = bracket; /* let the parser handle it */
10814 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
10815 PL_lex_state = LEX_INTERPEND;
10820 Perl_pmflag(pTHX_ U32* pmfl, int ch)
10822 PERL_UNUSED_CONTEXT;
10826 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
10827 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
10828 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
10829 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
10830 case KEEPCOPY_PAT_MOD: *pmfl |= PMf_KEEPCOPY; break;
10836 S_scan_pat(pTHX_ char *start, I32 type)
10840 char *s = scan_str(start,!!PL_madskills,FALSE);
10841 const char * const valid_flags =
10842 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
10849 const char * const delimiter = skipspace(start);
10853 ? "Search pattern not terminated or ternary operator parsed as search pattern"
10854 : "Search pattern not terminated" ));
10857 pm = (PMOP*)newPMOP(type, 0);
10858 if (PL_multi_open == '?') {
10859 /* This is the only point in the code that sets PMf_ONCE: */
10860 pm->op_pmflags |= PMf_ONCE;
10862 /* Hence it's safe to do this bit of PMOP book-keeping here, which
10863 allows us to restrict the list needed by reset to just the ??
10865 assert(type != OP_TRANS);
10867 MAGIC *mg = mg_find((SV*)PL_curstash, PERL_MAGIC_symtab);
10870 mg = sv_magicext((SV*)PL_curstash, 0, PERL_MAGIC_symtab, 0, 0,
10873 elements = mg->mg_len / sizeof(PMOP**);
10874 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
10875 ((PMOP**)mg->mg_ptr) [elements++] = pm;
10876 mg->mg_len = elements * sizeof(PMOP**);
10877 PmopSTASH_set(pm,PL_curstash);
10883 while (*s && strchr(valid_flags, *s))
10884 pmflag(&pm->op_pmflags,*s++);
10886 if (PL_madskills && modstart != s) {
10887 SV* tmptoken = newSVpvn(modstart, s - modstart);
10888 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
10891 /* issue a warning if /c is specified,but /g is not */
10892 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL)
10893 && ckWARN(WARN_REGEXP))
10895 Perl_warner(aTHX_ packWARN(WARN_REGEXP),
10896 "Use of /c modifier is meaningless without /g" );
10899 PL_lex_op = (OP*)pm;
10900 yylval.ival = OP_MATCH;
10905 S_scan_subst(pTHX_ char *start)
10916 yylval.ival = OP_NULL;
10918 s = scan_str(start,!!PL_madskills,FALSE);
10921 Perl_croak(aTHX_ "Substitution pattern not terminated");
10923 if (s[-1] == PL_multi_open)
10926 if (PL_madskills) {
10927 CURMAD('q', PL_thisopen);
10928 CURMAD('_', PL_thiswhite);
10929 CURMAD('E', PL_thisstuff);
10930 CURMAD('Q', PL_thisclose);
10931 PL_realtokenstart = s - SvPVX(PL_linestr);
10935 first_start = PL_multi_start;
10936 s = scan_str(s,!!PL_madskills,FALSE);
10938 if (PL_lex_stuff) {
10939 SvREFCNT_dec(PL_lex_stuff);
10940 PL_lex_stuff = NULL;
10942 Perl_croak(aTHX_ "Substitution replacement not terminated");
10944 PL_multi_start = first_start; /* so whole substitution is taken together */
10946 pm = (PMOP*)newPMOP(OP_SUBST, 0);
10949 if (PL_madskills) {
10950 CURMAD('z', PL_thisopen);
10951 CURMAD('R', PL_thisstuff);
10952 CURMAD('Z', PL_thisclose);
10958 if (*s == EXEC_PAT_MOD) {
10962 else if (strchr(S_PAT_MODS, *s))
10963 pmflag(&pm->op_pmflags,*s++);
10969 if (PL_madskills) {
10971 curmad('m', newSVpvn(modstart, s - modstart));
10972 append_madprops(PL_thismad, (OP*)pm, 0);
10976 if ((pm->op_pmflags & PMf_CONTINUE) && ckWARN(WARN_REGEXP)) {
10977 Perl_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
10981 SV * const repl = newSVpvs("");
10983 PL_sublex_info.super_bufptr = s;
10984 PL_sublex_info.super_bufend = PL_bufend;
10986 pm->op_pmflags |= PMf_EVAL;
10988 sv_catpv(repl, (const char *)(es ? "eval " : "do "));
10989 sv_catpvs(repl, "{");
10990 sv_catsv(repl, PL_lex_repl);
10991 if (strchr(SvPVX(PL_lex_repl), '#'))
10992 sv_catpvs(repl, "\n");
10993 sv_catpvs(repl, "}");
10995 SvREFCNT_dec(PL_lex_repl);
10996 PL_lex_repl = repl;
10999 PL_lex_op = (OP*)pm;
11000 yylval.ival = OP_SUBST;
11005 S_scan_trans(pTHX_ char *start)
11018 yylval.ival = OP_NULL;
11020 s = scan_str(start,!!PL_madskills,FALSE);
11022 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11024 if (s[-1] == PL_multi_open)
11027 if (PL_madskills) {
11028 CURMAD('q', PL_thisopen);
11029 CURMAD('_', PL_thiswhite);
11030 CURMAD('E', PL_thisstuff);
11031 CURMAD('Q', PL_thisclose);
11032 PL_realtokenstart = s - SvPVX(PL_linestr);
11036 s = scan_str(s,!!PL_madskills,FALSE);
11038 if (PL_lex_stuff) {
11039 SvREFCNT_dec(PL_lex_stuff);
11040 PL_lex_stuff = NULL;
11042 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11044 if (PL_madskills) {
11045 CURMAD('z', PL_thisopen);
11046 CURMAD('R', PL_thisstuff);
11047 CURMAD('Z', PL_thisclose);
11050 complement = del = squash = 0;
11057 complement = OPpTRANS_COMPLEMENT;
11060 del = OPpTRANS_DELETE;
11063 squash = OPpTRANS_SQUASH;
11072 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11073 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11074 o->op_private &= ~OPpTRANS_ALL;
11075 o->op_private |= del|squash|complement|
11076 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11077 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11080 yylval.ival = OP_TRANS;
11083 if (PL_madskills) {
11085 curmad('m', newSVpvn(modstart, s - modstart));
11086 append_madprops(PL_thismad, o, 0);
11095 S_scan_heredoc(pTHX_ register char *s)
11099 I32 op_type = OP_SCALAR;
11103 const char *found_newline;
11107 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11109 I32 stuffstart = s - SvPVX(PL_linestr);
11112 PL_realtokenstart = -1;
11117 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11121 while (SPACE_OR_TAB(*peek))
11123 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11126 s = delimcpy(d, e, s, PL_bufend, term, &len);
11136 if (!isALNUM_lazy_if(s,UTF))
11137 deprecate_old("bare << to mean <<\"\"");
11138 for (; isALNUM_lazy_if(s,UTF); s++) {
11143 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11144 Perl_croak(aTHX_ "Delimiter for here document is too long");
11147 len = d - PL_tokenbuf;
11150 if (PL_madskills) {
11151 tstart = PL_tokenbuf + !outer;
11152 PL_thisclose = newSVpvn(tstart, len - !outer);
11153 tstart = SvPVX(PL_linestr) + stuffstart;
11154 PL_thisopen = newSVpvn(tstart, s - tstart);
11155 stuffstart = s - SvPVX(PL_linestr);
11158 #ifndef PERL_STRICT_CR
11159 d = strchr(s, '\r');
11161 char * const olds = s;
11163 while (s < PL_bufend) {
11169 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11178 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11185 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11186 herewas = newSVpvn(s,PL_bufend-s);
11190 herewas = newSVpvn(s-1,found_newline-s+1);
11193 herewas = newSVpvn(s,found_newline-s);
11197 if (PL_madskills) {
11198 tstart = SvPVX(PL_linestr) + stuffstart;
11200 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11202 PL_thisstuff = newSVpvn(tstart, s - tstart);
11205 s += SvCUR(herewas);
11208 stuffstart = s - SvPVX(PL_linestr);
11214 tmpstr = newSV_type(SVt_PVIV);
11215 SvGROW(tmpstr, 80);
11216 if (term == '\'') {
11217 op_type = OP_CONST;
11218 SvIV_set(tmpstr, -1);
11220 else if (term == '`') {
11221 op_type = OP_BACKTICK;
11222 SvIV_set(tmpstr, '\\');
11226 PL_multi_start = CopLINE(PL_curcop);
11227 PL_multi_open = PL_multi_close = '<';
11228 term = *PL_tokenbuf;
11229 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11230 char * const bufptr = PL_sublex_info.super_bufptr;
11231 char * const bufend = PL_sublex_info.super_bufend;
11232 char * const olds = s - SvCUR(herewas);
11233 s = strchr(bufptr, '\n');
11237 while (s < bufend &&
11238 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11240 CopLINE_inc(PL_curcop);
11243 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11244 missingterm(PL_tokenbuf);
11246 sv_setpvn(herewas,bufptr,d-bufptr+1);
11247 sv_setpvn(tmpstr,d+1,s-d);
11249 sv_catpvn(herewas,s,bufend-s);
11250 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11257 while (s < PL_bufend &&
11258 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11260 CopLINE_inc(PL_curcop);
11262 if (s >= PL_bufend) {
11263 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11264 missingterm(PL_tokenbuf);
11266 sv_setpvn(tmpstr,d+1,s-d);
11268 if (PL_madskills) {
11270 sv_catpvn(PL_thisstuff, d + 1, s - d);
11272 PL_thisstuff = newSVpvn(d + 1, s - d);
11273 stuffstart = s - SvPVX(PL_linestr);
11277 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
11279 sv_catpvn(herewas,s,PL_bufend-s);
11280 sv_setsv(PL_linestr,herewas);
11281 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
11282 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11283 PL_last_lop = PL_last_uni = NULL;
11286 sv_setpvn(tmpstr,"",0); /* avoid "uninitialized" warning */
11287 while (s >= PL_bufend) { /* multiple line string? */
11289 if (PL_madskills) {
11290 tstart = SvPVX(PL_linestr) + stuffstart;
11292 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11294 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11298 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11299 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11300 missingterm(PL_tokenbuf);
11303 stuffstart = s - SvPVX(PL_linestr);
11305 CopLINE_inc(PL_curcop);
11306 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11307 PL_last_lop = PL_last_uni = NULL;
11308 #ifndef PERL_STRICT_CR
11309 if (PL_bufend - PL_linestart >= 2) {
11310 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
11311 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
11313 PL_bufend[-2] = '\n';
11315 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11317 else if (PL_bufend[-1] == '\r')
11318 PL_bufend[-1] = '\n';
11320 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
11321 PL_bufend[-1] = '\n';
11323 if (PERLDB_LINE && PL_curstash != PL_debstash)
11324 update_debugger_info(PL_linestr, NULL, 0);
11325 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
11326 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
11327 *(SvPVX(PL_linestr) + off ) = ' ';
11328 sv_catsv(PL_linestr,herewas);
11329 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11330 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
11334 sv_catsv(tmpstr,PL_linestr);
11339 PL_multi_end = CopLINE(PL_curcop);
11340 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
11341 SvPV_shrink_to_cur(tmpstr);
11343 SvREFCNT_dec(herewas);
11345 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
11347 else if (PL_encoding)
11348 sv_recode_to_utf8(tmpstr, PL_encoding);
11350 PL_lex_stuff = tmpstr;
11351 yylval.ival = op_type;
11355 /* scan_inputsymbol
11356 takes: current position in input buffer
11357 returns: new position in input buffer
11358 side-effects: yylval and lex_op are set.
11363 <FH> read from filehandle
11364 <pkg::FH> read from package qualified filehandle
11365 <pkg'FH> read from package qualified filehandle
11366 <$fh> read from filehandle in $fh
11367 <*.h> filename glob
11372 S_scan_inputsymbol(pTHX_ char *start)
11375 register char *s = start; /* current position in buffer */
11379 char *d = PL_tokenbuf; /* start of temp holding space */
11380 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
11382 end = strchr(s, '\n');
11385 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
11387 /* die if we didn't have space for the contents of the <>,
11388 or if it didn't end, or if we see a newline
11391 if (len >= (I32)sizeof PL_tokenbuf)
11392 Perl_croak(aTHX_ "Excessively long <> operator");
11394 Perl_croak(aTHX_ "Unterminated <> operator");
11399 Remember, only scalar variables are interpreted as filehandles by
11400 this code. Anything more complex (e.g., <$fh{$num}>) will be
11401 treated as a glob() call.
11402 This code makes use of the fact that except for the $ at the front,
11403 a scalar variable and a filehandle look the same.
11405 if (*d == '$' && d[1]) d++;
11407 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
11408 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
11411 /* If we've tried to read what we allow filehandles to look like, and
11412 there's still text left, then it must be a glob() and not a getline.
11413 Use scan_str to pull out the stuff between the <> and treat it
11414 as nothing more than a string.
11417 if (d - PL_tokenbuf != len) {
11418 yylval.ival = OP_GLOB;
11420 s = scan_str(start,!!PL_madskills,FALSE);
11422 Perl_croak(aTHX_ "Glob not terminated");
11426 bool readline_overriden = FALSE;
11429 /* we're in a filehandle read situation */
11432 /* turn <> into <ARGV> */
11434 Copy("ARGV",d,5,char);
11436 /* Check whether readline() is overriden */
11437 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
11439 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
11441 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
11442 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
11443 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
11444 readline_overriden = TRUE;
11446 /* if <$fh>, create the ops to turn the variable into a
11450 /* try to find it in the pad for this block, otherwise find
11451 add symbol table ops
11453 const PADOFFSET tmp = pad_findmy(d);
11454 if (tmp != NOT_IN_PAD) {
11455 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
11456 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
11457 HEK * const stashname = HvNAME_HEK(stash);
11458 SV * const sym = sv_2mortal(newSVhek(stashname));
11459 sv_catpvs(sym, "::");
11460 sv_catpv(sym, d+1);
11465 OP * const o = newOP(OP_PADSV, 0);
11467 PL_lex_op = readline_overriden
11468 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11469 append_elem(OP_LIST, o,
11470 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
11471 : (OP*)newUNOP(OP_READLINE, 0, o);
11480 ? (GV_ADDMULTI | GV_ADDINEVAL)
11483 PL_lex_op = readline_overriden
11484 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11485 append_elem(OP_LIST,
11486 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
11487 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11488 : (OP*)newUNOP(OP_READLINE, 0,
11489 newUNOP(OP_RV2SV, 0,
11490 newGVOP(OP_GV, 0, gv)));
11492 if (!readline_overriden)
11493 PL_lex_op->op_flags |= OPf_SPECIAL;
11494 /* we created the ops in PL_lex_op, so make yylval.ival a null op */
11495 yylval.ival = OP_NULL;
11498 /* If it's none of the above, it must be a literal filehandle
11499 (<Foo::BAR> or <FOO>) so build a simple readline OP */
11501 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
11502 PL_lex_op = readline_overriden
11503 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
11504 append_elem(OP_LIST,
11505 newGVOP(OP_GV, 0, gv),
11506 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
11507 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
11508 yylval.ival = OP_NULL;
11517 takes: start position in buffer
11518 keep_quoted preserve \ on the embedded delimiter(s)
11519 keep_delims preserve the delimiters around the string
11520 returns: position to continue reading from buffer
11521 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
11522 updates the read buffer.
11524 This subroutine pulls a string out of the input. It is called for:
11525 q single quotes q(literal text)
11526 ' single quotes 'literal text'
11527 qq double quotes qq(interpolate $here please)
11528 " double quotes "interpolate $here please"
11529 qx backticks qx(/bin/ls -l)
11530 ` backticks `/bin/ls -l`
11531 qw quote words @EXPORT_OK = qw( func() $spam )
11532 m// regexp match m/this/
11533 s/// regexp substitute s/this/that/
11534 tr/// string transliterate tr/this/that/
11535 y/// string transliterate y/this/that/
11536 ($*@) sub prototypes sub foo ($)
11537 (stuff) sub attr parameters sub foo : attr(stuff)
11538 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
11540 In most of these cases (all but <>, patterns and transliterate)
11541 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
11542 calls scan_str(). s/// makes yylex() call scan_subst() which calls
11543 scan_str(). tr/// and y/// make yylex() call scan_trans() which
11546 It skips whitespace before the string starts, and treats the first
11547 character as the delimiter. If the delimiter is one of ([{< then
11548 the corresponding "close" character )]}> is used as the closing
11549 delimiter. It allows quoting of delimiters, and if the string has
11550 balanced delimiters ([{<>}]) it allows nesting.
11552 On success, the SV with the resulting string is put into lex_stuff or,
11553 if that is already non-NULL, into lex_repl. The second case occurs only
11554 when parsing the RHS of the special constructs s/// and tr/// (y///).
11555 For convenience, the terminating delimiter character is stuffed into
11560 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
11563 SV *sv; /* scalar value: string */
11564 const char *tmps; /* temp string, used for delimiter matching */
11565 register char *s = start; /* current position in the buffer */
11566 register char term; /* terminating character */
11567 register char *to; /* current position in the sv's data */
11568 I32 brackets = 1; /* bracket nesting level */
11569 bool has_utf8 = FALSE; /* is there any utf8 content? */
11570 I32 termcode; /* terminating char. code */
11571 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
11572 STRLEN termlen; /* length of terminating string */
11573 int last_off = 0; /* last position for nesting bracket */
11579 /* skip space before the delimiter */
11585 if (PL_realtokenstart >= 0) {
11586 stuffstart = PL_realtokenstart;
11587 PL_realtokenstart = -1;
11590 stuffstart = start - SvPVX(PL_linestr);
11592 /* mark where we are, in case we need to report errors */
11595 /* after skipping whitespace, the next character is the terminator */
11598 termcode = termstr[0] = term;
11602 termcode = utf8_to_uvchr((U8*)s, &termlen);
11603 Copy(s, termstr, termlen, U8);
11604 if (!UTF8_IS_INVARIANT(term))
11608 /* mark where we are */
11609 PL_multi_start = CopLINE(PL_curcop);
11610 PL_multi_open = term;
11612 /* find corresponding closing delimiter */
11613 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
11614 termcode = termstr[0] = term = tmps[5];
11616 PL_multi_close = term;
11618 /* create a new SV to hold the contents. 79 is the SV's initial length.
11619 What a random number. */
11620 sv = newSV_type(SVt_PVIV);
11622 SvIV_set(sv, termcode);
11623 (void)SvPOK_only(sv); /* validate pointer */
11625 /* move past delimiter and try to read a complete string */
11627 sv_catpvn(sv, s, termlen);
11630 tstart = SvPVX(PL_linestr) + stuffstart;
11631 if (!PL_thisopen && !keep_delims) {
11632 PL_thisopen = newSVpvn(tstart, s - tstart);
11633 stuffstart = s - SvPVX(PL_linestr);
11637 if (PL_encoding && !UTF) {
11641 int offset = s - SvPVX_const(PL_linestr);
11642 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
11643 &offset, (char*)termstr, termlen);
11644 const char * const ns = SvPVX_const(PL_linestr) + offset;
11645 char * const svlast = SvEND(sv) - 1;
11647 for (; s < ns; s++) {
11648 if (*s == '\n' && !PL_rsfp)
11649 CopLINE_inc(PL_curcop);
11652 goto read_more_line;
11654 /* handle quoted delimiters */
11655 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
11657 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
11659 if ((svlast-1 - t) % 2) {
11660 if (!keep_quoted) {
11661 *(svlast-1) = term;
11663 SvCUR_set(sv, SvCUR(sv) - 1);
11668 if (PL_multi_open == PL_multi_close) {
11674 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
11675 /* At here, all closes are "was quoted" one,
11676 so we don't check PL_multi_close. */
11678 if (!keep_quoted && *(t+1) == PL_multi_open)
11683 else if (*t == PL_multi_open)
11691 SvCUR_set(sv, w - SvPVX_const(sv));
11693 last_off = w - SvPVX(sv);
11694 if (--brackets <= 0)
11699 if (!keep_delims) {
11700 SvCUR_set(sv, SvCUR(sv) - 1);
11706 /* extend sv if need be */
11707 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
11708 /* set 'to' to the next character in the sv's string */
11709 to = SvPVX(sv)+SvCUR(sv);
11711 /* if open delimiter is the close delimiter read unbridle */
11712 if (PL_multi_open == PL_multi_close) {
11713 for (; s < PL_bufend; s++,to++) {
11714 /* embedded newlines increment the current line number */
11715 if (*s == '\n' && !PL_rsfp)
11716 CopLINE_inc(PL_curcop);
11717 /* handle quoted delimiters */
11718 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
11719 if (!keep_quoted && s[1] == term)
11721 /* any other quotes are simply copied straight through */
11725 /* terminate when run out of buffer (the for() condition), or
11726 have found the terminator */
11727 else if (*s == term) {
11730 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
11733 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11739 /* if the terminator isn't the same as the start character (e.g.,
11740 matched brackets), we have to allow more in the quoting, and
11741 be prepared for nested brackets.
11744 /* read until we run out of string, or we find the terminator */
11745 for (; s < PL_bufend; s++,to++) {
11746 /* embedded newlines increment the line count */
11747 if (*s == '\n' && !PL_rsfp)
11748 CopLINE_inc(PL_curcop);
11749 /* backslashes can escape the open or closing characters */
11750 if (*s == '\\' && s+1 < PL_bufend) {
11751 if (!keep_quoted &&
11752 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
11757 /* allow nested opens and closes */
11758 else if (*s == PL_multi_close && --brackets <= 0)
11760 else if (*s == PL_multi_open)
11762 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
11767 /* terminate the copied string and update the sv's end-of-string */
11769 SvCUR_set(sv, to - SvPVX_const(sv));
11772 * this next chunk reads more into the buffer if we're not done yet
11776 break; /* handle case where we are done yet :-) */
11778 #ifndef PERL_STRICT_CR
11779 if (to - SvPVX_const(sv) >= 2) {
11780 if ((to[-2] == '\r' && to[-1] == '\n') ||
11781 (to[-2] == '\n' && to[-1] == '\r'))
11785 SvCUR_set(sv, to - SvPVX_const(sv));
11787 else if (to[-1] == '\r')
11790 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
11795 /* if we're out of file, or a read fails, bail and reset the current
11796 line marker so we can report where the unterminated string began
11799 if (PL_madskills) {
11800 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11802 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
11804 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
11808 !(PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = filter_gets(PL_linestr, PL_rsfp, 0))) {
11810 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11816 /* we read a line, so increment our line counter */
11817 CopLINE_inc(PL_curcop);
11819 /* update debugger info */
11820 if (PERLDB_LINE && PL_curstash != PL_debstash)
11821 update_debugger_info(PL_linestr, NULL, 0);
11823 /* having changed the buffer, we must update PL_bufend */
11824 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
11825 PL_last_lop = PL_last_uni = NULL;
11828 /* at this point, we have successfully read the delimited string */
11830 if (!PL_encoding || UTF) {
11832 if (PL_madskills) {
11833 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11834 const int len = s - tstart;
11836 sv_catpvn(PL_thisstuff, tstart, len);
11838 PL_thisstuff = newSVpvn(tstart, len);
11839 if (!PL_thisclose && !keep_delims)
11840 PL_thisclose = newSVpvn(s,termlen);
11845 sv_catpvn(sv, s, termlen);
11850 if (PL_madskills) {
11851 char * const tstart = SvPVX(PL_linestr) + stuffstart;
11852 const int len = s - tstart - termlen;
11854 sv_catpvn(PL_thisstuff, tstart, len);
11856 PL_thisstuff = newSVpvn(tstart, len);
11857 if (!PL_thisclose && !keep_delims)
11858 PL_thisclose = newSVpvn(s - termlen,termlen);
11862 if (has_utf8 || PL_encoding)
11865 PL_multi_end = CopLINE(PL_curcop);
11867 /* if we allocated too much space, give some back */
11868 if (SvCUR(sv) + 5 < SvLEN(sv)) {
11869 SvLEN_set(sv, SvCUR(sv) + 1);
11870 SvPV_renew(sv, SvLEN(sv));
11873 /* decide whether this is the first or second quoted string we've read
11886 takes: pointer to position in buffer
11887 returns: pointer to new position in buffer
11888 side-effects: builds ops for the constant in yylval.op
11890 Read a number in any of the formats that Perl accepts:
11892 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
11893 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
11896 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
11898 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
11901 If it reads a number without a decimal point or an exponent, it will
11902 try converting the number to an integer and see if it can do so
11903 without loss of precision.
11907 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
11910 register const char *s = start; /* current position in buffer */
11911 register char *d; /* destination in temp buffer */
11912 register char *e; /* end of temp buffer */
11913 NV nv; /* number read, as a double */
11914 SV *sv = NULL; /* place to put the converted number */
11915 bool floatit; /* boolean: int or float? */
11916 const char *lastub = NULL; /* position of last underbar */
11917 static char const number_too_long[] = "Number too long";
11919 /* We use the first character to decide what type of number this is */
11923 Perl_croak(aTHX_ "panic: scan_num");
11925 /* if it starts with a 0, it could be an octal number, a decimal in
11926 0.13 disguise, or a hexadecimal number, or a binary number. */
11930 u holds the "number so far"
11931 shift the power of 2 of the base
11932 (hex == 4, octal == 3, binary == 1)
11933 overflowed was the number more than we can hold?
11935 Shift is used when we add a digit. It also serves as an "are
11936 we in octal/hex/binary?" indicator to disallow hex characters
11937 when in octal mode.
11942 bool overflowed = FALSE;
11943 bool just_zero = TRUE; /* just plain 0 or binary number? */
11944 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
11945 static const char* const bases[5] =
11946 { "", "binary", "", "octal", "hexadecimal" };
11947 static const char* const Bases[5] =
11948 { "", "Binary", "", "Octal", "Hexadecimal" };
11949 static const char* const maxima[5] =
11951 "0b11111111111111111111111111111111",
11955 const char *base, *Base, *max;
11957 /* check for hex */
11962 } else if (s[1] == 'b') {
11967 /* check for a decimal in disguise */
11968 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
11970 /* so it must be octal */
11977 if (ckWARN(WARN_SYNTAX))
11978 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11979 "Misplaced _ in number");
11983 base = bases[shift];
11984 Base = Bases[shift];
11985 max = maxima[shift];
11987 /* read the rest of the number */
11989 /* x is used in the overflow test,
11990 b is the digit we're adding on. */
11995 /* if we don't mention it, we're done */
11999 /* _ are ignored -- but warned about if consecutive */
12001 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12002 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12003 "Misplaced _ in number");
12007 /* 8 and 9 are not octal */
12008 case '8': case '9':
12010 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12014 case '2': case '3': case '4':
12015 case '5': case '6': case '7':
12017 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12020 case '0': case '1':
12021 b = *s++ & 15; /* ASCII digit -> value of digit */
12025 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12026 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12027 /* make sure they said 0x */
12030 b = (*s++ & 7) + 9;
12032 /* Prepare to put the digit we have onto the end
12033 of the number so far. We check for overflows.
12039 x = u << shift; /* make room for the digit */
12041 if ((x >> shift) != u
12042 && !(PL_hints & HINT_NEW_BINARY)) {
12045 if (ckWARN_d(WARN_OVERFLOW))
12046 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12047 "Integer overflow in %s number",
12050 u = x | b; /* add the digit to the end */
12053 n *= nvshift[shift];
12054 /* If an NV has not enough bits in its
12055 * mantissa to represent an UV this summing of
12056 * small low-order numbers is a waste of time
12057 * (because the NV cannot preserve the
12058 * low-order bits anyway): we could just
12059 * remember when did we overflow and in the
12060 * end just multiply n by the right
12068 /* if we get here, we had success: make a scalar value from
12073 /* final misplaced underbar check */
12074 if (s[-1] == '_') {
12075 if (ckWARN(WARN_SYNTAX))
12076 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12081 if (n > 4294967295.0 && ckWARN(WARN_PORTABLE))
12082 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12083 "%s number > %s non-portable",
12089 if (u > 0xffffffff && ckWARN(WARN_PORTABLE))
12090 Perl_warner(aTHX_ packWARN(WARN_PORTABLE),
12091 "%s number > %s non-portable",
12096 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12097 sv = new_constant(start, s - start, "integer",
12099 else if (PL_hints & HINT_NEW_BINARY)
12100 sv = new_constant(start, s - start, "binary", sv, NULL, NULL);
12105 handle decimal numbers.
12106 we're also sent here when we read a 0 as the first digit
12108 case '1': case '2': case '3': case '4': case '5':
12109 case '6': case '7': case '8': case '9': case '.':
12112 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12115 /* read next group of digits and _ and copy into d */
12116 while (isDIGIT(*s) || *s == '_') {
12117 /* skip underscores, checking for misplaced ones
12121 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12122 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12123 "Misplaced _ in number");
12127 /* check for end of fixed-length buffer */
12129 Perl_croak(aTHX_ number_too_long);
12130 /* if we're ok, copy the character */
12135 /* final misplaced underbar check */
12136 if (lastub && s == lastub + 1) {
12137 if (ckWARN(WARN_SYNTAX))
12138 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12141 /* read a decimal portion if there is one. avoid
12142 3..5 being interpreted as the number 3. followed
12145 if (*s == '.' && s[1] != '.') {
12150 if (ckWARN(WARN_SYNTAX))
12151 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12152 "Misplaced _ in number");
12156 /* copy, ignoring underbars, until we run out of digits.
12158 for (; isDIGIT(*s) || *s == '_'; s++) {
12159 /* fixed length buffer check */
12161 Perl_croak(aTHX_ number_too_long);
12163 if (lastub && s == lastub + 1 && ckWARN(WARN_SYNTAX))
12164 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12165 "Misplaced _ in number");
12171 /* fractional part ending in underbar? */
12172 if (s[-1] == '_') {
12173 if (ckWARN(WARN_SYNTAX))
12174 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12175 "Misplaced _ in number");
12177 if (*s == '.' && isDIGIT(s[1])) {
12178 /* oops, it's really a v-string, but without the "v" */
12184 /* read exponent part, if present */
12185 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12189 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12190 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12192 /* stray preinitial _ */
12194 if (ckWARN(WARN_SYNTAX))
12195 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12196 "Misplaced _ in number");
12200 /* allow positive or negative exponent */
12201 if (*s == '+' || *s == '-')
12204 /* stray initial _ */
12206 if (ckWARN(WARN_SYNTAX))
12207 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12208 "Misplaced _ in number");
12212 /* read digits of exponent */
12213 while (isDIGIT(*s) || *s == '_') {
12216 Perl_croak(aTHX_ number_too_long);
12220 if (((lastub && s == lastub + 1) ||
12221 (!isDIGIT(s[1]) && s[1] != '_'))
12222 && ckWARN(WARN_SYNTAX))
12223 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
12224 "Misplaced _ in number");
12231 /* make an sv from the string */
12235 We try to do an integer conversion first if no characters
12236 indicating "float" have been found.
12241 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12243 if (flags == IS_NUMBER_IN_UV) {
12245 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12248 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12249 if (uv <= (UV) IV_MIN)
12250 sv_setiv(sv, -(IV)uv);
12257 /* terminate the string */
12259 nv = Atof(PL_tokenbuf);
12263 if ( floatit ? (PL_hints & HINT_NEW_FLOAT) :
12264 (PL_hints & HINT_NEW_INTEGER) )
12265 sv = new_constant(PL_tokenbuf,
12268 (floatit ? "float" : "integer"),
12272 /* if it starts with a v, it could be a v-string */
12275 sv = newSV(5); /* preallocate storage space */
12276 s = scan_vstring(s, PL_bufend, sv);
12280 /* make the op for the constant and return */
12283 lvalp->opval = newSVOP(OP_CONST, 0, sv);
12285 lvalp->opval = NULL;
12291 S_scan_formline(pTHX_ register char *s)
12294 register char *eol;
12296 SV * const stuff = newSVpvs("");
12297 bool needargs = FALSE;
12298 bool eofmt = FALSE;
12300 char *tokenstart = s;
12303 if (PL_madskills) {
12304 savewhite = PL_thiswhite;
12309 while (!needargs) {
12312 #ifdef PERL_STRICT_CR
12313 while (SPACE_OR_TAB(*t))
12316 while (SPACE_OR_TAB(*t) || *t == '\r')
12319 if (*t == '\n' || t == PL_bufend) {
12324 if (PL_in_eval && !PL_rsfp) {
12325 eol = (char *) memchr(s,'\n',PL_bufend-s);
12330 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12332 for (t = s; t < eol; t++) {
12333 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
12335 goto enough; /* ~~ must be first line in formline */
12337 if (*t == '@' || *t == '^')
12341 sv_catpvn(stuff, s, eol-s);
12342 #ifndef PERL_STRICT_CR
12343 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
12344 char *end = SvPVX(stuff) + SvCUR(stuff);
12347 SvCUR_set(stuff, SvCUR(stuff) - 1);
12357 if (PL_madskills) {
12359 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
12361 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
12364 s = filter_gets(PL_linestr, PL_rsfp, 0);
12366 tokenstart = PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12368 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = SvPVX(PL_linestr);
12370 PL_bufend = PL_bufptr + SvCUR(PL_linestr);
12371 PL_last_lop = PL_last_uni = NULL;
12380 if (SvCUR(stuff)) {
12383 PL_lex_state = LEX_NORMAL;
12384 start_force(PL_curforce);
12385 NEXTVAL_NEXTTOKE.ival = 0;
12389 PL_lex_state = LEX_FORMLINE;
12391 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
12393 else if (PL_encoding)
12394 sv_recode_to_utf8(stuff, PL_encoding);
12396 start_force(PL_curforce);
12397 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
12399 start_force(PL_curforce);
12400 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
12404 SvREFCNT_dec(stuff);
12406 PL_lex_formbrack = 0;
12410 if (PL_madskills) {
12412 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
12414 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
12415 PL_thiswhite = savewhite;
12427 PL_cshlen = strlen(PL_cshname);
12429 #if defined(USE_ITHREADS)
12430 PERL_UNUSED_CONTEXT;
12436 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
12439 const I32 oldsavestack_ix = PL_savestack_ix;
12440 CV* const outsidecv = PL_compcv;
12443 assert(SvTYPE(PL_compcv) == SVt_PVCV);
12445 SAVEI32(PL_subline);
12446 save_item(PL_subname);
12447 SAVESPTR(PL_compcv);
12449 PL_compcv = (CV*)newSV_type(is_format ? SVt_PVFM : SVt_PVCV);
12450 CvFLAGS(PL_compcv) |= flags;
12452 PL_subline = CopLINE(PL_curcop);
12453 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
12454 CvOUTSIDE(PL_compcv) = (CV*)SvREFCNT_inc_simple(outsidecv);
12455 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
12457 return oldsavestack_ix;
12461 #pragma segment Perl_yylex
12464 Perl_yywarn(pTHX_ const char *s)
12467 PL_in_eval |= EVAL_WARNONLY;
12469 PL_in_eval &= ~EVAL_WARNONLY;
12474 Perl_yyerror(pTHX_ const char *s)
12477 const char *where = NULL;
12478 const char *context = NULL;
12481 int yychar = PL_parser->yychar;
12483 if (!yychar || (yychar == ';' && !PL_rsfp))
12485 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
12486 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
12487 PL_oldbufptr != PL_bufptr) {
12490 The code below is removed for NetWare because it abends/crashes on NetWare
12491 when the script has error such as not having the closing quotes like:
12492 if ($var eq "value)
12493 Checking of white spaces is anyway done in NetWare code.
12496 while (isSPACE(*PL_oldoldbufptr))
12499 context = PL_oldoldbufptr;
12500 contlen = PL_bufptr - PL_oldoldbufptr;
12502 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
12503 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
12506 The code below is removed for NetWare because it abends/crashes on NetWare
12507 when the script has error such as not having the closing quotes like:
12508 if ($var eq "value)
12509 Checking of white spaces is anyway done in NetWare code.
12512 while (isSPACE(*PL_oldbufptr))
12515 context = PL_oldbufptr;
12516 contlen = PL_bufptr - PL_oldbufptr;
12518 else if (yychar > 255)
12519 where = "next token ???";
12520 else if (yychar == -2) { /* YYEMPTY */
12521 if (PL_lex_state == LEX_NORMAL ||
12522 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
12523 where = "at end of line";
12524 else if (PL_lex_inpat)
12525 where = "within pattern";
12527 where = "within string";
12530 SV * const where_sv = sv_2mortal(newSVpvs("next char "));
12532 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
12533 else if (isPRINT_LC(yychar))
12534 Perl_sv_catpvf(aTHX_ where_sv, "%c", yychar);
12536 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
12537 where = SvPVX_const(where_sv);
12539 msg = sv_2mortal(newSVpv(s, 0));
12540 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
12541 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
12543 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
12545 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
12546 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
12547 Perl_sv_catpvf(aTHX_ msg,
12548 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
12549 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
12552 if (PL_in_eval & EVAL_WARNONLY && ckWARN_d(WARN_SYNTAX))
12553 Perl_warner(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
12556 if (PL_error_count >= 10) {
12557 if (PL_in_eval && SvCUR(ERRSV))
12558 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
12559 SVfARG(ERRSV), OutCopFILE(PL_curcop));
12561 Perl_croak(aTHX_ "%s has too many errors.\n",
12562 OutCopFILE(PL_curcop));
12565 PL_in_my_stash = NULL;
12569 #pragma segment Main
12573 S_swallow_bom(pTHX_ U8 *s)
12576 const STRLEN slen = SvCUR(PL_linestr);
12579 if (s[1] == 0xFE) {
12580 /* UTF-16 little-endian? (or UTF32-LE?) */
12581 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
12582 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
12583 #ifndef PERL_NO_UTF16_FILTER
12584 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
12587 if (PL_bufend > (char*)s) {
12591 filter_add(utf16rev_textfilter, NULL);
12592 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12593 utf16_to_utf8_reversed(s, news,
12594 PL_bufend - (char*)s - 1,
12596 sv_setpvn(PL_linestr, (const char*)news, newlen);
12598 s = (U8*)SvPVX(PL_linestr);
12599 Copy(news, s, newlen, U8);
12603 SvUTF8_on(PL_linestr);
12604 s = (U8*)SvPVX(PL_linestr);
12606 /* FIXME - is this a general bug fix? */
12609 PL_bufend = SvPVX(PL_linestr) + newlen;
12612 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
12617 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
12618 #ifndef PERL_NO_UTF16_FILTER
12619 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
12622 if (PL_bufend > (char *)s) {
12626 filter_add(utf16_textfilter, NULL);
12627 Newx(news, (PL_bufend - (char*)s) * 3 / 2 + 1, U8);
12628 utf16_to_utf8(s, news,
12629 PL_bufend - (char*)s,
12631 sv_setpvn(PL_linestr, (const char*)news, newlen);
12633 SvUTF8_on(PL_linestr);
12634 s = (U8*)SvPVX(PL_linestr);
12635 PL_bufend = SvPVX(PL_linestr) + newlen;
12638 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
12643 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
12644 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12645 s += 3; /* UTF-8 */
12651 if (s[2] == 0xFE && s[3] == 0xFF) {
12652 /* UTF-32 big-endian */
12653 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
12656 else if (s[2] == 0 && s[3] != 0) {
12659 * are a good indicator of UTF-16BE. */
12660 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
12666 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
12667 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
12668 s += 4; /* UTF-8 */
12674 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
12677 * are a good indicator of UTF-16LE. */
12678 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
12686 #ifndef PERL_NO_UTF16_FILTER
12688 utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12691 const STRLEN old = SvCUR(sv);
12692 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12693 DEBUG_P(PerlIO_printf(Perl_debug_log,
12694 "utf16_textfilter(%p): %d %d (%d)\n",
12695 FPTR2DPTR(void *, utf16_textfilter),
12696 idx, maxlen, (int) count));
12700 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12701 Copy(SvPVX_const(sv), tmps, old, char);
12702 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12703 SvCUR(sv) - old, &newlen);
12704 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12706 DEBUG_P({sv_dump(sv);});
12711 utf16rev_textfilter(pTHX_ int idx, SV *sv, int maxlen)
12714 const STRLEN old = SvCUR(sv);
12715 const I32 count = FILTER_READ(idx+1, sv, maxlen);
12716 DEBUG_P(PerlIO_printf(Perl_debug_log,
12717 "utf16rev_textfilter(%p): %d %d (%d)\n",
12718 FPTR2DPTR(void *, utf16rev_textfilter),
12719 idx, maxlen, (int) count));
12723 Newx(tmps, SvCUR(sv) * 3 / 2 + 1, U8);
12724 Copy(SvPVX_const(sv), tmps, old, char);
12725 utf16_to_utf8((U8*)SvPVX_const(sv) + old, tmps + old,
12726 SvCUR(sv) - old, &newlen);
12727 sv_usepvn(sv, (char*)tmps, (STRLEN)newlen + old);
12729 DEBUG_P({ sv_dump(sv); });
12735 Returns a pointer to the next character after the parsed
12736 vstring, as well as updating the passed in sv.
12738 Function must be called like
12741 s = scan_vstring(s,e,sv);
12743 where s and e are the start and end of the string.
12744 The sv should already be large enough to store the vstring
12745 passed in, for performance reasons.
12750 Perl_scan_vstring(pTHX_ const char *s, const char *e, SV *sv)
12753 const char *pos = s;
12754 const char *start = s;
12755 if (*pos == 'v') pos++; /* get past 'v' */
12756 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12758 if ( *pos != '.') {
12759 /* this may not be a v-string if followed by => */
12760 const char *next = pos;
12761 while (next < e && isSPACE(*next))
12763 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
12764 /* return string not v-string */
12765 sv_setpvn(sv,(char *)s,pos-s);
12766 return (char *)pos;
12770 if (!isALPHA(*pos)) {
12771 U8 tmpbuf[UTF8_MAXBYTES+1];
12774 s++; /* get past 'v' */
12776 sv_setpvn(sv, "", 0);
12779 /* this is atoi() that tolerates underscores */
12782 const char *end = pos;
12784 while (--end >= s) {
12786 const UV orev = rev;
12787 rev += (*end - '0') * mult;
12789 if (orev > rev && ckWARN_d(WARN_OVERFLOW))
12790 Perl_warner(aTHX_ packWARN(WARN_OVERFLOW),
12791 "Integer overflow in decimal number");
12795 if (rev > 0x7FFFFFFF)
12796 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
12798 /* Append native character for the rev point */
12799 tmpend = uvchr_to_utf8(tmpbuf, rev);
12800 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
12801 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
12803 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
12809 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
12813 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
12821 * c-indentation-style: bsd
12822 * c-basic-offset: 4
12823 * indent-tabs-mode: t
12826 * ex: set ts=8 sts=4 sw=4 noet: