3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * 'It all comes from here, the stench and the peril.' --Frodo
14 * [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
18 * This file is the lexer for Perl. It's closely linked to the
21 * The main routine is yylex(), which returns the next token.
25 =head1 Lexer interface
27 This is the lower layer of the Perl parser, managing characters and tokens.
29 =for apidoc AmU|yy_parser *|PL_parser
31 Pointer to a structure encapsulating the state of the parsing operation
32 currently in progress. The pointer can be locally changed to perform
33 a nested parse without interfering with the state of an outer parse.
34 Individual members of C<PL_parser> have their own documentation.
40 #define PERL_IN_TOKE_C
43 #define new_constant(a,b,c,d,e,f,g) \
44 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
46 #define pl_yylval (PL_parser->yylval)
48 /* YYINITDEPTH -- initial size of the parser's stacks. */
49 #define YYINITDEPTH 200
51 /* XXX temporary backwards compatibility */
52 #define PL_lex_brackets (PL_parser->lex_brackets)
53 #define PL_lex_brackstack (PL_parser->lex_brackstack)
54 #define PL_lex_casemods (PL_parser->lex_casemods)
55 #define PL_lex_casestack (PL_parser->lex_casestack)
56 #define PL_lex_defer (PL_parser->lex_defer)
57 #define PL_lex_dojoin (PL_parser->lex_dojoin)
58 #define PL_lex_expect (PL_parser->lex_expect)
59 #define PL_lex_formbrack (PL_parser->lex_formbrack)
60 #define PL_lex_inpat (PL_parser->lex_inpat)
61 #define PL_lex_inwhat (PL_parser->lex_inwhat)
62 #define PL_lex_op (PL_parser->lex_op)
63 #define PL_lex_repl (PL_parser->lex_repl)
64 #define PL_lex_starts (PL_parser->lex_starts)
65 #define PL_lex_stuff (PL_parser->lex_stuff)
66 #define PL_multi_start (PL_parser->multi_start)
67 #define PL_multi_open (PL_parser->multi_open)
68 #define PL_multi_close (PL_parser->multi_close)
69 #define PL_pending_ident (PL_parser->pending_ident)
70 #define PL_preambled (PL_parser->preambled)
71 #define PL_sublex_info (PL_parser->sublex_info)
72 #define PL_linestr (PL_parser->linestr)
73 #define PL_expect (PL_parser->expect)
74 #define PL_copline (PL_parser->copline)
75 #define PL_bufptr (PL_parser->bufptr)
76 #define PL_oldbufptr (PL_parser->oldbufptr)
77 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
78 #define PL_linestart (PL_parser->linestart)
79 #define PL_bufend (PL_parser->bufend)
80 #define PL_last_uni (PL_parser->last_uni)
81 #define PL_last_lop (PL_parser->last_lop)
82 #define PL_last_lop_op (PL_parser->last_lop_op)
83 #define PL_lex_state (PL_parser->lex_state)
84 #define PL_rsfp (PL_parser->rsfp)
85 #define PL_rsfp_filters (PL_parser->rsfp_filters)
86 #define PL_in_my (PL_parser->in_my)
87 #define PL_in_my_stash (PL_parser->in_my_stash)
88 #define PL_tokenbuf (PL_parser->tokenbuf)
89 #define PL_multi_end (PL_parser->multi_end)
90 #define PL_error_count (PL_parser->error_count)
93 # define PL_endwhite (PL_parser->endwhite)
94 # define PL_faketokens (PL_parser->faketokens)
95 # define PL_lasttoke (PL_parser->lasttoke)
96 # define PL_nextwhite (PL_parser->nextwhite)
97 # define PL_realtokenstart (PL_parser->realtokenstart)
98 # define PL_skipwhite (PL_parser->skipwhite)
99 # define PL_thisclose (PL_parser->thisclose)
100 # define PL_thismad (PL_parser->thismad)
101 # define PL_thisopen (PL_parser->thisopen)
102 # define PL_thisstuff (PL_parser->thisstuff)
103 # define PL_thistoken (PL_parser->thistoken)
104 # define PL_thiswhite (PL_parser->thiswhite)
105 # define PL_thiswhite (PL_parser->thiswhite)
106 # define PL_nexttoke (PL_parser->nexttoke)
107 # define PL_curforce (PL_parser->curforce)
109 # define PL_nexttoke (PL_parser->nexttoke)
110 # define PL_nexttype (PL_parser->nexttype)
111 # define PL_nextval (PL_parser->nextval)
114 /* This can't be done with embed.fnc, because struct yy_parser contains a
115 member named pending_ident, which clashes with the generated #define */
117 S_pending_ident(pTHX);
119 static const char ident_too_long[] = "Identifier too long";
122 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
123 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
125 # define CURMAD(slot,sv)
126 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
129 #define XFAKEBRACK 128
130 #define XENUMMASK 127
132 #ifdef USE_UTF8_SCRIPTS
133 # define UTF (!IN_BYTES)
135 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
138 /* The maximum number of characters preceding the unrecognized one to display */
139 #define UNRECOGNIZED_PRECEDE_COUNT 10
141 /* In variables named $^X, these are the legal values for X.
142 * 1999-02-27 mjd-perl-patch@plover.com */
143 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
145 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
147 /* LEX_* are values for PL_lex_state, the state of the lexer.
148 * They are arranged oddly so that the guard on the switch statement
149 * can get by with a single comparison (if the compiler is smart enough).
152 /* #define LEX_NOTPARSING 11 is done in perl.h. */
154 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
155 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
156 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
157 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
158 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
160 /* at end of code, eg "$x" followed by: */
161 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
162 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
164 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
165 string or after \E, $foo, etc */
166 #define LEX_INTERPCONST 2 /* NOT USED */
167 #define LEX_FORMLINE 1 /* expecting a format line */
168 #define LEX_KNOWNEXT 0 /* next token known; just return it */
172 static const char* const lex_state_names[] = {
191 #include "keywords.h"
193 /* CLINE is a macro that ensures PL_copline has a sane value */
198 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
201 # define SKIPSPACE0(s) skipspace0(s)
202 # define SKIPSPACE1(s) skipspace1(s)
203 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
204 # define PEEKSPACE(s) skipspace2(s,0)
206 # define SKIPSPACE0(s) skipspace(s)
207 # define SKIPSPACE1(s) skipspace(s)
208 # define SKIPSPACE2(s,tsv) skipspace(s)
209 # define PEEKSPACE(s) skipspace(s)
213 * Convenience functions to return different tokens and prime the
214 * lexer for the next token. They all take an argument.
216 * TOKEN : generic token (used for '(', DOLSHARP, etc)
217 * OPERATOR : generic operator
218 * AOPERATOR : assignment operator
219 * PREBLOCK : beginning the block after an if, while, foreach, ...
220 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
221 * PREREF : *EXPR where EXPR is not a simple identifier
222 * TERM : expression term
223 * LOOPX : loop exiting command (goto, last, dump, etc)
224 * FTST : file test operator
225 * FUN0 : zero-argument function
226 * FUN1 : not used, except for not, which isn't a UNIOP
227 * BOop : bitwise or or xor
229 * SHop : shift operator
230 * PWop : power operator
231 * PMop : pattern-matching operator
232 * Aop : addition-level operator
233 * Mop : multiplication-level operator
234 * Eop : equality-testing operator
235 * Rop : relational operator <= != gt
237 * Also see LOP and lop() below.
240 #ifdef DEBUGGING /* Serve -DT. */
241 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
243 # define REPORT(retval) (retval)
246 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
247 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
248 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
249 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
250 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
251 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
252 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
253 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
254 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
255 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
256 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
257 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
258 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
259 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
260 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
261 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
262 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
263 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
264 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
265 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
267 /* This bit of chicanery makes a unary function followed by
268 * a parenthesis into a function with one argument, highest precedence.
269 * The UNIDOR macro is for unary functions that can be followed by the //
270 * operator (such as C<shift // 0>).
272 #define UNI2(f,x) { \
273 pl_yylval.ival = f; \
276 PL_last_uni = PL_oldbufptr; \
277 PL_last_lop_op = f; \
279 return REPORT( (int)FUNC1 ); \
281 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
283 #define UNI(f) UNI2(f,XTERM)
284 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
286 #define UNIBRACK(f) { \
287 pl_yylval.ival = f; \
289 PL_last_uni = PL_oldbufptr; \
291 return REPORT( (int)FUNC1 ); \
293 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
296 /* grandfather return to old style */
297 #define OLDLOP(f) return(pl_yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
301 /* how to interpret the pl_yylval associated with the token */
305 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
311 static struct debug_tokens {
313 enum token_type type;
315 } const debug_tokens[] =
317 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
318 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
319 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
320 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
321 { ARROW, TOKENTYPE_NONE, "ARROW" },
322 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
323 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
324 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
325 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
326 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
327 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
328 { DO, TOKENTYPE_NONE, "DO" },
329 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
330 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
331 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
332 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
333 { ELSE, TOKENTYPE_NONE, "ELSE" },
334 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
335 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
336 { FOR, TOKENTYPE_IVAL, "FOR" },
337 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
338 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
339 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
340 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
341 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
342 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
343 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
344 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
345 { IF, TOKENTYPE_IVAL, "IF" },
346 { LABEL, TOKENTYPE_PVAL, "LABEL" },
347 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
348 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
349 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
350 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
351 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
352 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
353 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
354 { MY, TOKENTYPE_IVAL, "MY" },
355 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
356 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
357 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
358 { OROP, TOKENTYPE_IVAL, "OROP" },
359 { OROR, TOKENTYPE_NONE, "OROR" },
360 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
361 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
362 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
363 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
364 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
365 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
366 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
367 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
368 { PREINC, TOKENTYPE_NONE, "PREINC" },
369 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
370 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
371 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
372 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
373 { SUB, TOKENTYPE_NONE, "SUB" },
374 { THING, TOKENTYPE_OPVAL, "THING" },
375 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
376 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
377 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
378 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
379 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
380 { USE, TOKENTYPE_IVAL, "USE" },
381 { WHEN, TOKENTYPE_IVAL, "WHEN" },
382 { WHILE, TOKENTYPE_IVAL, "WHILE" },
383 { WORD, TOKENTYPE_OPVAL, "WORD" },
384 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
385 { 0, TOKENTYPE_NONE, NULL }
388 /* dump the returned token in rv, plus any optional arg in pl_yylval */
391 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
395 PERL_ARGS_ASSERT_TOKEREPORT;
398 const char *name = NULL;
399 enum token_type type = TOKENTYPE_NONE;
400 const struct debug_tokens *p;
401 SV* const report = newSVpvs("<== ");
403 for (p = debug_tokens; p->token; p++) {
404 if (p->token == (int)rv) {
411 Perl_sv_catpv(aTHX_ report, name);
412 else if ((char)rv > ' ' && (char)rv < '~')
413 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
415 sv_catpvs(report, "EOF");
417 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
420 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
423 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
425 case TOKENTYPE_OPNUM:
426 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
427 PL_op_name[lvalp->ival]);
430 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
432 case TOKENTYPE_OPVAL:
434 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
435 PL_op_name[lvalp->opval->op_type]);
436 if (lvalp->opval->op_type == OP_CONST) {
437 Perl_sv_catpvf(aTHX_ report, " %s",
438 SvPEEK(cSVOPx_sv(lvalp->opval)));
443 sv_catpvs(report, "(opval=null)");
446 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
452 /* print the buffer with suitable escapes */
455 S_printbuf(pTHX_ const char *const fmt, const char *const s)
457 SV* const tmp = newSVpvs("");
459 PERL_ARGS_ASSERT_PRINTBUF;
461 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
468 S_deprecate_commaless_var_list(pTHX) {
470 deprecate("comma-less variable list");
471 return REPORT(','); /* grandfather non-comma-format format */
477 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
478 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
482 S_ao(pTHX_ int toketype)
485 if (*PL_bufptr == '=') {
487 if (toketype == ANDAND)
488 pl_yylval.ival = OP_ANDASSIGN;
489 else if (toketype == OROR)
490 pl_yylval.ival = OP_ORASSIGN;
491 else if (toketype == DORDOR)
492 pl_yylval.ival = OP_DORASSIGN;
500 * When Perl expects an operator and finds something else, no_op
501 * prints the warning. It always prints "<something> found where
502 * operator expected. It prints "Missing semicolon on previous line?"
503 * if the surprise occurs at the start of the line. "do you need to
504 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
505 * where the compiler doesn't know if foo is a method call or a function.
506 * It prints "Missing operator before end of line" if there's nothing
507 * after the missing operator, or "... before <...>" if there is something
508 * after the missing operator.
512 S_no_op(pTHX_ const char *const what, char *s)
515 char * const oldbp = PL_bufptr;
516 const bool is_first = (PL_oldbufptr == PL_linestart);
518 PERL_ARGS_ASSERT_NO_OP;
524 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
525 if (ckWARN_d(WARN_SYNTAX)) {
527 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
528 "\t(Missing semicolon on previous line?)\n");
529 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
531 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
533 if (t < PL_bufptr && isSPACE(*t))
534 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
535 "\t(Do you need to predeclare %.*s?)\n",
536 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
540 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
541 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
549 * Complain about missing quote/regexp/heredoc terminator.
550 * If it's called with NULL then it cauterizes the line buffer.
551 * If we're in a delimited string and the delimiter is a control
552 * character, it's reformatted into a two-char sequence like ^C.
557 S_missingterm(pTHX_ char *s)
563 char * const nl = strrchr(s,'\n');
567 else if (isCNTRL(PL_multi_close)) {
569 tmpbuf[1] = (char)toCTRL(PL_multi_close);
574 *tmpbuf = (char)PL_multi_close;
578 q = strchr(s,'"') ? '\'' : '"';
579 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
582 #define FEATURE_IS_ENABLED(name) \
583 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
584 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
585 /* The longest string we pass in. */
586 #define MAX_FEATURE_LEN (sizeof("switch")-1)
589 * S_feature_is_enabled
590 * Check whether the named feature is enabled.
593 S_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
596 HV * const hinthv = GvHV(PL_hintgv);
597 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
599 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
601 assert(namelen <= MAX_FEATURE_LEN);
602 memcpy(&he_name[8], name, namelen);
604 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
608 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
609 * utf16-to-utf8-reversed.
612 #ifdef PERL_CR_FILTER
616 register const char *s = SvPVX_const(sv);
617 register const char * const e = s + SvCUR(sv);
619 PERL_ARGS_ASSERT_STRIP_RETURN;
621 /* outer loop optimized to do nothing if there are no CR-LFs */
623 if (*s++ == '\r' && *s == '\n') {
624 /* hit a CR-LF, need to copy the rest */
625 register char *d = s - 1;
628 if (*s == '\r' && s[1] == '\n')
639 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
641 const I32 count = FILTER_READ(idx+1, sv, maxlen);
642 if (count > 0 && !maxlen)
653 * Create a parser object and initialise its parser and lexer fields
655 * rsfp is the opened file handle to read from (if any),
657 * line holds any initial content already read from the file (or in
658 * the case of no file, such as an eval, the whole contents);
660 * new_filter indicates that this is a new file and it shouldn't inherit
661 * the filters from the current parser (ie require).
665 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
668 const char *s = NULL;
670 yy_parser *parser, *oparser;
672 /* create and initialise a parser */
674 Newxz(parser, 1, yy_parser);
675 parser->old_parser = oparser = PL_parser;
678 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
679 parser->ps = parser->stack;
680 parser->stack_size = YYINITDEPTH;
682 parser->stack->state = 0;
683 parser->yyerrstatus = 0;
684 parser->yychar = YYEMPTY; /* Cause a token to be read. */
686 /* on scope exit, free this parser and restore any outer one */
688 parser->saved_curcop = PL_curcop;
690 /* initialise lexer state */
693 parser->curforce = -1;
695 parser->nexttoke = 0;
697 parser->error_count = oparser ? oparser->error_count : 0;
698 parser->copline = NOLINE;
699 parser->lex_state = LEX_NORMAL;
700 parser->expect = XSTATE;
702 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
703 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
705 Newx(parser->lex_brackstack, 120, char);
706 Newx(parser->lex_casestack, 12, char);
707 *parser->lex_casestack = '\0';
710 s = SvPV_const(line, len);
716 parser->linestr = newSVpvs("\n;");
717 } else if (SvREADONLY(line) || s[len-1] != ';') {
718 parser->linestr = newSVsv(line);
720 sv_catpvs(parser->linestr, "\n;");
723 SvREFCNT_inc_simple_void_NN(line);
724 parser->linestr = line;
726 parser->oldoldbufptr =
729 parser->linestart = SvPVX(parser->linestr);
730 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
731 parser->last_lop = parser->last_uni = NULL;
735 /* delete a parser object */
738 Perl_parser_free(pTHX_ const yy_parser *parser)
740 PERL_ARGS_ASSERT_PARSER_FREE;
742 PL_curcop = parser->saved_curcop;
743 SvREFCNT_dec(parser->linestr);
745 if (parser->rsfp == PerlIO_stdin())
746 PerlIO_clearerr(parser->rsfp);
747 else if (parser->rsfp && (!parser->old_parser ||
748 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
749 PerlIO_close(parser->rsfp);
750 SvREFCNT_dec(parser->rsfp_filters);
752 Safefree(parser->stack);
753 Safefree(parser->lex_brackstack);
754 Safefree(parser->lex_casestack);
755 PL_parser = parser->old_parser;
762 * Finalizer for lexing operations. Must be called when the parser is
763 * done with the lexer.
770 PL_doextract = FALSE;
774 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
776 Buffer scalar containing the chunk currently under consideration of the
777 text currently being lexed. This is always a plain string scalar (for
778 which C<SvPOK> is true). It is not intended to be used as a scalar by
779 normal scalar means; instead refer to the buffer directly by the pointer
780 variables described below.
782 The lexer maintains various C<char*> pointers to things in the
783 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
784 reallocated, all of these pointers must be updated. Don't attempt to
785 do this manually, but rather use L</lex_grow_linestr> if you need to
786 reallocate the buffer.
788 The content of the text chunk in the buffer is commonly exactly one
789 complete line of input, up to and including a newline terminator,
790 but there are situations where it is otherwise. The octets of the
791 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
792 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
793 flag on this scalar, which may disagree with it.
795 For direct examination of the buffer, the variable
796 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
797 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
798 of these pointers is usually preferable to examination of the scalar
799 through normal scalar means.
801 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
803 Direct pointer to the end of the chunk of text currently being lexed, the
804 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
805 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
806 always located at the end of the buffer, and does not count as part of
807 the buffer's contents.
809 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
811 Points to the current position of lexing inside the lexer buffer.
812 Characters around this point may be freely examined, within
813 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
814 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
815 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
817 Lexing code (whether in the Perl core or not) moves this pointer past
818 the characters that it consumes. It is also expected to perform some
819 bookkeeping whenever a newline character is consumed. This movement
820 can be more conveniently performed by the function L</lex_read_to>,
821 which handles newlines appropriately.
823 Interpretation of the buffer's octets can be abstracted out by
824 using the slightly higher-level functions L</lex_peek_unichar> and
825 L</lex_read_unichar>.
827 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
829 Points to the start of the current line inside the lexer buffer.
830 This is useful for indicating at which column an error occurred, and
831 not much else. This must be updated by any lexing code that consumes
832 a newline; the function L</lex_read_to> handles this detail.
838 =for apidoc Amx|bool|lex_bufutf8
840 Indicates whether the octets in the lexer buffer
841 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
842 of Unicode characters. If not, they should be interpreted as Latin-1
843 characters. This is analogous to the C<SvUTF8> flag for scalars.
845 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
846 contains valid UTF-8. Lexing code must be robust in the face of invalid
849 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
850 is significant, but not the whole story regarding the input character
851 encoding. Normally, when a file is being read, the scalar contains octets
852 and its C<SvUTF8> flag is off, but the octets should be interpreted as
853 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
854 however, the scalar may have the C<SvUTF8> flag on, and in this case its
855 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
856 is in effect. This logic may change in the future; use this function
857 instead of implementing the logic yourself.
863 Perl_lex_bufutf8(pTHX)
869 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
871 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
872 at least I<len> octets (including terminating NUL). Returns a
873 pointer to the reallocated buffer. This is necessary before making
874 any direct modification of the buffer that would increase its length.
875 L</lex_stuff_pvn> provides a more convenient way to insert text into
878 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
879 this function updates all of the lexer's variables that point directly
886 Perl_lex_grow_linestr(pTHX_ STRLEN len)
890 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
891 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
892 linestr = PL_parser->linestr;
893 buf = SvPVX(linestr);
894 if (len <= SvLEN(linestr))
896 bufend_pos = PL_parser->bufend - buf;
897 bufptr_pos = PL_parser->bufptr - buf;
898 oldbufptr_pos = PL_parser->oldbufptr - buf;
899 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
900 linestart_pos = PL_parser->linestart - buf;
901 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
902 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
903 buf = sv_grow(linestr, len);
904 PL_parser->bufend = buf + bufend_pos;
905 PL_parser->bufptr = buf + bufptr_pos;
906 PL_parser->oldbufptr = buf + oldbufptr_pos;
907 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
908 PL_parser->linestart = buf + linestart_pos;
909 if (PL_parser->last_uni)
910 PL_parser->last_uni = buf + last_uni_pos;
911 if (PL_parser->last_lop)
912 PL_parser->last_lop = buf + last_lop_pos;
917 =for apidoc Amx|void|lex_stuff_pvn|char *pv|STRLEN len|U32 flags
919 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
920 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
921 reallocating the buffer if necessary. This means that lexing code that
922 runs later will see the characters as if they had appeared in the input.
923 It is not recommended to do this as part of normal parsing, and most
924 uses of this facility run the risk of the inserted characters being
925 interpreted in an unintended manner.
927 The string to be inserted is represented by I<len> octets starting
928 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
929 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
930 The characters are recoded for the lexer buffer, according to how the
931 buffer is currently being interpreted (L</lex_bufutf8>). If a string
932 to be interpreted is available as a Perl scalar, the L</lex_stuff_sv>
933 function is more convenient.
939 Perl_lex_stuff_pvn(pTHX_ char *pv, STRLEN len, U32 flags)
942 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
943 if (flags & ~(LEX_STUFF_UTF8))
944 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
946 if (flags & LEX_STUFF_UTF8) {
950 char *p, *e = pv+len;
951 for (p = pv; p != e; p++)
952 highhalf += !!(((U8)*p) & 0x80);
955 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
956 bufptr = PL_parser->bufptr;
957 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
958 PL_parser->bufend += len+highhalf;
959 for (p = pv; p != e; p++) {
962 *bufptr++ = (char)(0xc0 | (c >> 6));
963 *bufptr++ = (char)(0x80 | (c & 0x3f));
970 if (flags & LEX_STUFF_UTF8) {
972 char *p, *e = pv+len;
973 for (p = pv; p != e; p++) {
976 Perl_croak(aTHX_ "Lexing code attempted to stuff "
977 "non-Latin-1 character into Latin-1 input");
978 } else if (c >= 0xc2 && p+1 != e &&
979 (((U8)p[1]) & 0xc0) == 0x80) {
982 } else if (c >= 0x80) {
983 /* malformed UTF-8 */
985 SAVESPTR(PL_warnhook);
986 PL_warnhook = PERL_WARNHOOK_FATAL;
987 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
993 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
994 bufptr = PL_parser->bufptr;
995 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
996 PL_parser->bufend += len-highhalf;
997 for (p = pv; p != e; p++) {
1000 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1003 *bufptr++ = (char)c;
1008 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1009 bufptr = PL_parser->bufptr;
1010 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1011 PL_parser->bufend += len;
1012 Copy(pv, bufptr, len, char);
1018 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1020 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1021 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1022 reallocating the buffer if necessary. This means that lexing code that
1023 runs later will see the characters as if they had appeared in the input.
1024 It is not recommended to do this as part of normal parsing, and most
1025 uses of this facility run the risk of the inserted characters being
1026 interpreted in an unintended manner.
1028 The string to be inserted is the string value of I<sv>. The characters
1029 are recoded for the lexer buffer, according to how the buffer is currently
1030 being interpreted (L</lex_bufutf8>). If a string to be interpreted is
1031 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1032 need to construct a scalar.
1038 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1042 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1044 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1046 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1050 =for apidoc Amx|void|lex_unstuff|char *ptr
1052 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1053 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1054 This hides the discarded text from any lexing code that runs later,
1055 as if the text had never appeared.
1057 This is not the normal way to consume lexed text. For that, use
1064 Perl_lex_unstuff(pTHX_ char *ptr)
1068 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1069 buf = PL_parser->bufptr;
1071 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1074 bufend = PL_parser->bufend;
1076 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1077 unstuff_len = ptr - buf;
1078 Move(ptr, buf, bufend+1-ptr, char);
1079 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1080 PL_parser->bufend = bufend - unstuff_len;
1084 =for apidoc Amx|void|lex_read_to|char *ptr
1086 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1087 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1088 performing the correct bookkeeping whenever a newline character is passed.
1089 This is the normal way to consume lexed text.
1091 Interpretation of the buffer's octets can be abstracted out by
1092 using the slightly higher-level functions L</lex_peek_unichar> and
1093 L</lex_read_unichar>.
1099 Perl_lex_read_to(pTHX_ char *ptr)
1102 PERL_ARGS_ASSERT_LEX_READ_TO;
1103 s = PL_parser->bufptr;
1104 if (ptr < s || ptr > PL_parser->bufend)
1105 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1106 for (; s != ptr; s++)
1108 CopLINE_inc(PL_curcop);
1109 PL_parser->linestart = s+1;
1111 PL_parser->bufptr = ptr;
1115 =for apidoc Amx|void|lex_discard_to|char *ptr
1117 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1118 up to I<ptr>. The remaining content of the buffer will be moved, and
1119 all pointers into the buffer updated appropriately. I<ptr> must not
1120 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1121 it is not permitted to discard text that has yet to be lexed.
1123 Normally it is not necessarily to do this directly, because it suffices to
1124 use the implicit discarding behaviour of L</lex_next_chunk> and things
1125 based on it. However, if a token stretches across multiple lines,
1126 and the lexing code has kept multiple lines of text in the buffer fof
1127 that purpose, then after completion of the token it would be wise to
1128 explicitly discard the now-unneeded earlier lines, to avoid future
1129 multi-line tokens growing the buffer without bound.
1135 Perl_lex_discard_to(pTHX_ char *ptr)
1139 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1140 buf = SvPVX(PL_parser->linestr);
1142 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1145 if (ptr > PL_parser->bufptr)
1146 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1147 discard_len = ptr - buf;
1148 if (PL_parser->oldbufptr < ptr)
1149 PL_parser->oldbufptr = ptr;
1150 if (PL_parser->oldoldbufptr < ptr)
1151 PL_parser->oldoldbufptr = ptr;
1152 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1153 PL_parser->last_uni = NULL;
1154 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1155 PL_parser->last_lop = NULL;
1156 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1157 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1158 PL_parser->bufend -= discard_len;
1159 PL_parser->bufptr -= discard_len;
1160 PL_parser->oldbufptr -= discard_len;
1161 PL_parser->oldoldbufptr -= discard_len;
1162 if (PL_parser->last_uni)
1163 PL_parser->last_uni -= discard_len;
1164 if (PL_parser->last_lop)
1165 PL_parser->last_lop -= discard_len;
1169 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1171 Reads in the next chunk of text to be lexed, appending it to
1172 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1173 looked to the end of the current chunk and wants to know more. It is
1174 usual, but not necessary, for lexing to have consumed the entirety of
1175 the current chunk at this time.
1177 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1178 chunk (i.e., the current chunk has been entirely consumed), normally the
1179 current chunk will be discarded at the same time that the new chunk is
1180 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1181 will not be discarded. If the current chunk has not been entirely
1182 consumed, then it will not be discarded regardless of the flag.
1184 Returns true if some new text was added to the buffer, or false if the
1185 buffer has reached the end of the input text.
1190 #define LEX_FAKE_EOF 0x80000000
1193 Perl_lex_next_chunk(pTHX_ U32 flags)
1197 STRLEN old_bufend_pos, new_bufend_pos;
1198 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1199 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1201 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1202 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1204 flags |= LEX_KEEP_PREVIOUS;
1205 #endif /* PERL_MAD */
1206 linestr = PL_parser->linestr;
1207 buf = SvPVX(linestr);
1208 if (!(flags & LEX_KEEP_PREVIOUS) &&
1209 PL_parser->bufptr == PL_parser->bufend) {
1210 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1212 if (PL_parser->last_uni != PL_parser->bufend)
1213 PL_parser->last_uni = NULL;
1214 if (PL_parser->last_lop != PL_parser->bufend)
1215 PL_parser->last_lop = NULL;
1216 last_uni_pos = last_lop_pos = 0;
1220 old_bufend_pos = PL_parser->bufend - buf;
1221 bufptr_pos = PL_parser->bufptr - buf;
1222 oldbufptr_pos = PL_parser->oldbufptr - buf;
1223 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1224 linestart_pos = PL_parser->linestart - buf;
1225 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1226 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1228 if (flags & LEX_FAKE_EOF) {
1230 } else if (!PL_parser->rsfp) {
1232 } else if (filter_gets(linestr, old_bufend_pos)) {
1236 /* End of real input. Close filehandle (unless it was STDIN),
1237 * then add implicit termination.
1239 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1240 PerlIO_clearerr(PL_parser->rsfp);
1241 else if (PL_parser->rsfp)
1242 (void)PerlIO_close(PL_parser->rsfp);
1243 PL_parser->rsfp = NULL;
1244 PL_doextract = FALSE;
1246 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1249 if (!PL_in_eval && PL_minus_p) {
1251 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1252 PL_minus_n = PL_minus_p = 0;
1253 } else if (!PL_in_eval && PL_minus_n) {
1254 sv_catpvs(linestr, /*{*/";}");
1257 sv_catpvs(linestr, ";");
1260 buf = SvPVX(linestr);
1261 new_bufend_pos = SvCUR(linestr);
1262 PL_parser->bufend = buf + new_bufend_pos;
1263 PL_parser->bufptr = buf + bufptr_pos;
1264 PL_parser->oldbufptr = buf + oldbufptr_pos;
1265 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1266 PL_parser->linestart = buf + linestart_pos;
1267 if (PL_parser->last_uni)
1268 PL_parser->last_uni = buf + last_uni_pos;
1269 if (PL_parser->last_lop)
1270 PL_parser->last_lop = buf + last_lop_pos;
1271 if (got_some && (PERLDB_LINE || PERLDB_SAVESRC) &&
1272 PL_curstash != PL_debstash) {
1273 /* debugger active and we're not compiling the debugger code,
1274 * so store the line into the debugger's array of lines
1276 update_debugger_info(NULL, buf+old_bufend_pos,
1277 new_bufend_pos-old_bufend_pos);
1283 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1285 Looks ahead one (Unicode) character in the text currently being lexed.
1286 Returns the codepoint (unsigned integer value) of the next character,
1287 or -1 if lexing has reached the end of the input text. To consume the
1288 peeked character, use L</lex_read_unichar>.
1290 If the next character is in (or extends into) the next chunk of input
1291 text, the next chunk will be read in. Normally the current chunk will be
1292 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1293 then the current chunk will not be discarded.
1295 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1296 is encountered, an exception is generated.
1302 Perl_lex_peek_unichar(pTHX_ U32 flags)
1305 if (flags & ~(LEX_KEEP_PREVIOUS))
1306 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1307 s = PL_parser->bufptr;
1308 bufend = PL_parser->bufend;
1314 if (!lex_next_chunk(flags))
1316 s = PL_parser->bufptr;
1317 bufend = PL_parser->bufend;
1323 len = PL_utf8skip[head];
1324 while ((STRLEN)(bufend-s) < len) {
1325 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1327 s = PL_parser->bufptr;
1328 bufend = PL_parser->bufend;
1331 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1332 if (retlen == (STRLEN)-1) {
1333 /* malformed UTF-8 */
1335 SAVESPTR(PL_warnhook);
1336 PL_warnhook = PERL_WARNHOOK_FATAL;
1337 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1343 if (!lex_next_chunk(flags))
1345 s = PL_parser->bufptr;
1352 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1354 Reads the next (Unicode) character in the text currently being lexed.
1355 Returns the codepoint (unsigned integer value) of the character read,
1356 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1357 if lexing has reached the end of the input text. To non-destructively
1358 examine the next character, use L</lex_peek_unichar> instead.
1360 If the next character is in (or extends into) the next chunk of input
1361 text, the next chunk will be read in. Normally the current chunk will be
1362 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1363 then the current chunk will not be discarded.
1365 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1366 is encountered, an exception is generated.
1372 Perl_lex_read_unichar(pTHX_ U32 flags)
1375 if (flags & ~(LEX_KEEP_PREVIOUS))
1376 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1377 c = lex_peek_unichar(flags);
1380 CopLINE_inc(PL_curcop);
1381 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1387 =for apidoc Amx|void|lex_read_space|U32 flags
1389 Reads optional spaces, in Perl style, in the text currently being
1390 lexed. The spaces may include ordinary whitespace characters and
1391 Perl-style comments. C<#line> directives are processed if encountered.
1392 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1393 at a non-space character (or the end of the input text).
1395 If spaces extend into the next chunk of input text, the next chunk will
1396 be read in. Normally the current chunk will be discarded at the same
1397 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1398 chunk will not be discarded.
1404 Perl_lex_read_space(pTHX_ U32 flags)
1407 bool need_incline = 0;
1408 if (flags & ~(LEX_KEEP_PREVIOUS))
1409 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1412 sv_free(PL_skipwhite);
1413 PL_skipwhite = NULL;
1416 PL_skipwhite = newSVpvs("");
1417 #endif /* PERL_MAD */
1418 s = PL_parser->bufptr;
1419 bufend = PL_parser->bufend;
1425 } while (!(c == '\n' || (c == 0 && s == bufend)));
1426 } else if (c == '\n') {
1428 PL_parser->linestart = s;
1433 } else if (isSPACE(c)) {
1435 } else if (c == 0 && s == bufend) {
1439 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1440 #endif /* PERL_MAD */
1441 PL_parser->bufptr = s;
1442 CopLINE_inc(PL_curcop);
1443 got_more = lex_next_chunk(flags);
1444 CopLINE_dec(PL_curcop);
1445 s = PL_parser->bufptr;
1446 bufend = PL_parser->bufend;
1449 if (need_incline && PL_parser->rsfp) {
1459 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1460 #endif /* PERL_MAD */
1461 PL_parser->bufptr = s;
1466 * This subroutine has nothing to do with tilting, whether at windmills
1467 * or pinball tables. Its name is short for "increment line". It
1468 * increments the current line number in CopLINE(PL_curcop) and checks
1469 * to see whether the line starts with a comment of the form
1470 * # line 500 "foo.pm"
1471 * If so, it sets the current line number and file to the values in the comment.
1475 S_incline(pTHX_ const char *s)
1482 PERL_ARGS_ASSERT_INCLINE;
1484 CopLINE_inc(PL_curcop);
1487 while (SPACE_OR_TAB(*s))
1489 if (strnEQ(s, "line", 4))
1493 if (SPACE_OR_TAB(*s))
1497 while (SPACE_OR_TAB(*s))
1505 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1507 while (SPACE_OR_TAB(*s))
1509 if (*s == '"' && (t = strchr(s+1, '"'))) {
1515 while (!isSPACE(*t))
1519 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1521 if (*e != '\n' && *e != '\0')
1522 return; /* false alarm */
1525 const STRLEN len = t - s;
1526 #ifndef USE_ITHREADS
1527 SV *const temp_sv = CopFILESV(PL_curcop);
1532 cf = SvPVX(temp_sv);
1533 tmplen = SvCUR(temp_sv);
1539 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
1540 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1541 * to *{"::_<newfilename"} */
1542 /* However, the long form of evals is only turned on by the
1543 debugger - usually they're "(eval %lu)" */
1547 STRLEN tmplen2 = len;
1548 if (tmplen + 2 <= sizeof smallbuf)
1551 Newx(tmpbuf, tmplen + 2, char);
1554 memcpy(tmpbuf + 2, cf, tmplen);
1556 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1561 if (tmplen2 + 2 <= sizeof smallbuf)
1564 Newx(tmpbuf2, tmplen2 + 2, char);
1566 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1567 /* Either they malloc'd it, or we malloc'd it,
1568 so no prefix is present in ours. */
1573 memcpy(tmpbuf2 + 2, s, tmplen2);
1576 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1578 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1579 /* adjust ${"::_<newfilename"} to store the new file name */
1580 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1581 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1582 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1585 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1587 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1590 CopFILE_free(PL_curcop);
1591 CopFILE_setn(PL_curcop, s, len);
1593 CopLINE_set(PL_curcop, atoi(n)-1);
1597 /* skip space before PL_thistoken */
1600 S_skipspace0(pTHX_ register char *s)
1602 PERL_ARGS_ASSERT_SKIPSPACE0;
1609 PL_thiswhite = newSVpvs("");
1610 sv_catsv(PL_thiswhite, PL_skipwhite);
1611 sv_free(PL_skipwhite);
1614 PL_realtokenstart = s - SvPVX(PL_linestr);
1618 /* skip space after PL_thistoken */
1621 S_skipspace1(pTHX_ register char *s)
1623 const char *start = s;
1624 I32 startoff = start - SvPVX(PL_linestr);
1626 PERL_ARGS_ASSERT_SKIPSPACE1;
1631 start = SvPVX(PL_linestr) + startoff;
1632 if (!PL_thistoken && PL_realtokenstart >= 0) {
1633 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1634 PL_thistoken = newSVpvn(tstart, start - tstart);
1636 PL_realtokenstart = -1;
1639 PL_nextwhite = newSVpvs("");
1640 sv_catsv(PL_nextwhite, PL_skipwhite);
1641 sv_free(PL_skipwhite);
1648 S_skipspace2(pTHX_ register char *s, SV **svp)
1651 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1652 const I32 startoff = s - SvPVX(PL_linestr);
1654 PERL_ARGS_ASSERT_SKIPSPACE2;
1657 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1658 if (!PL_madskills || !svp)
1660 start = SvPVX(PL_linestr) + startoff;
1661 if (!PL_thistoken && PL_realtokenstart >= 0) {
1662 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1663 PL_thistoken = newSVpvn(tstart, start - tstart);
1664 PL_realtokenstart = -1;
1668 *svp = newSVpvs("");
1669 sv_setsv(*svp, PL_skipwhite);
1670 sv_free(PL_skipwhite);
1679 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1681 AV *av = CopFILEAVx(PL_curcop);
1683 SV * const sv = newSV_type(SVt_PVMG);
1685 sv_setsv(sv, orig_sv);
1687 sv_setpvn(sv, buf, len);
1690 av_store(av, (I32)CopLINE(PL_curcop), sv);
1696 * Called to gobble the appropriate amount and type of whitespace.
1697 * Skips comments as well.
1701 S_skipspace(pTHX_ register char *s)
1705 #endif /* PERL_MAD */
1706 PERL_ARGS_ASSERT_SKIPSPACE;
1709 sv_free(PL_skipwhite);
1710 PL_skipwhite = NULL;
1712 #endif /* PERL_MAD */
1713 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1714 while (s < PL_bufend && SPACE_OR_TAB(*s))
1716 } else if (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE) {
1717 while (isSPACE(*s) && *s != '\n')
1722 } while (s != PL_bufend && *s != '\n');
1727 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1729 lex_read_space(LEX_KEEP_PREVIOUS);
1731 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1732 if (PL_linestart > PL_bufptr)
1733 PL_bufptr = PL_linestart;
1738 PL_skipwhite = newSVpvn(start, s-start);
1739 #endif /* PERL_MAD */
1745 * Check the unary operators to ensure there's no ambiguity in how they're
1746 * used. An ambiguous piece of code would be:
1748 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1749 * the +5 is its argument.
1759 if (PL_oldoldbufptr != PL_last_uni)
1761 while (isSPACE(*PL_last_uni))
1764 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1766 if ((t = strchr(s, '(')) && t < PL_bufptr)
1769 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1770 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1771 (int)(s - PL_last_uni), PL_last_uni);
1775 * LOP : macro to build a list operator. Its behaviour has been replaced
1776 * with a subroutine, S_lop() for which LOP is just another name.
1779 #define LOP(f,x) return lop(f,x,s)
1783 * Build a list operator (or something that might be one). The rules:
1784 * - if we have a next token, then it's a list operator [why?]
1785 * - if the next thing is an opening paren, then it's a function
1786 * - else it's a list operator
1790 S_lop(pTHX_ I32 f, int x, char *s)
1794 PERL_ARGS_ASSERT_LOP;
1800 PL_last_lop = PL_oldbufptr;
1801 PL_last_lop_op = (OPCODE)f;
1804 return REPORT(LSTOP);
1807 return REPORT(LSTOP);
1810 return REPORT(FUNC);
1813 return REPORT(FUNC);
1815 return REPORT(LSTOP);
1821 * Sets up for an eventual force_next(). start_force(0) basically does
1822 * an unshift, while start_force(-1) does a push. yylex removes items
1827 S_start_force(pTHX_ int where)
1831 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1832 where = PL_lasttoke;
1833 assert(PL_curforce < 0 || PL_curforce == where);
1834 if (PL_curforce != where) {
1835 for (i = PL_lasttoke; i > where; --i) {
1836 PL_nexttoke[i] = PL_nexttoke[i-1];
1840 if (PL_curforce < 0) /* in case of duplicate start_force() */
1841 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1842 PL_curforce = where;
1845 curmad('^', newSVpvs(""));
1846 CURMAD('_', PL_nextwhite);
1851 S_curmad(pTHX_ char slot, SV *sv)
1857 if (PL_curforce < 0)
1858 where = &PL_thismad;
1860 where = &PL_nexttoke[PL_curforce].next_mad;
1866 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1868 else if (PL_encoding) {
1869 sv_recode_to_utf8(sv, PL_encoding);
1874 /* keep a slot open for the head of the list? */
1875 if (slot != '_' && *where && (*where)->mad_key == '^') {
1876 (*where)->mad_key = slot;
1877 sv_free(MUTABLE_SV(((*where)->mad_val)));
1878 (*where)->mad_val = (void*)sv;
1881 addmad(newMADsv(slot, sv), where, 0);
1884 # define start_force(where) NOOP
1885 # define curmad(slot, sv) NOOP
1890 * When the lexer realizes it knows the next token (for instance,
1891 * it is reordering tokens for the parser) then it can call S_force_next
1892 * to know what token to return the next time the lexer is called. Caller
1893 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1894 * and possibly PL_expect to ensure the lexer handles the token correctly.
1898 S_force_next(pTHX_ I32 type)
1903 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1904 tokereport(type, &NEXTVAL_NEXTTOKE);
1908 if (PL_curforce < 0)
1909 start_force(PL_lasttoke);
1910 PL_nexttoke[PL_curforce].next_type = type;
1911 if (PL_lex_state != LEX_KNOWNEXT)
1912 PL_lex_defer = PL_lex_state;
1913 PL_lex_state = LEX_KNOWNEXT;
1914 PL_lex_expect = PL_expect;
1917 PL_nexttype[PL_nexttoke] = type;
1919 if (PL_lex_state != LEX_KNOWNEXT) {
1920 PL_lex_defer = PL_lex_state;
1921 PL_lex_expect = PL_expect;
1922 PL_lex_state = LEX_KNOWNEXT;
1928 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1931 SV * const sv = newSVpvn_utf8(start, len,
1934 && !is_ascii_string((const U8*)start, len)
1935 && is_utf8_string((const U8*)start, len));
1941 * When the lexer knows the next thing is a word (for instance, it has
1942 * just seen -> and it knows that the next char is a word char, then
1943 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1947 * char *start : buffer position (must be within PL_linestr)
1948 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1949 * int check_keyword : if true, Perl checks to make sure the word isn't
1950 * a keyword (do this if the word is a label, e.g. goto FOO)
1951 * int allow_pack : if true, : characters will also be allowed (require,
1952 * use, etc. do this)
1953 * int allow_initial_tick : used by the "sub" lexer only.
1957 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1963 PERL_ARGS_ASSERT_FORCE_WORD;
1965 start = SKIPSPACE1(start);
1967 if (isIDFIRST_lazy_if(s,UTF) ||
1968 (allow_pack && *s == ':') ||
1969 (allow_initial_tick && *s == '\'') )
1971 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1972 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1974 start_force(PL_curforce);
1976 curmad('X', newSVpvn(start,s-start));
1977 if (token == METHOD) {
1982 PL_expect = XOPERATOR;
1986 curmad('g', newSVpvs( "forced" ));
1987 NEXTVAL_NEXTTOKE.opval
1988 = (OP*)newSVOP(OP_CONST,0,
1989 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1990 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1998 * Called when the lexer wants $foo *foo &foo etc, but the program
1999 * text only contains the "foo" portion. The first argument is a pointer
2000 * to the "foo", and the second argument is the type symbol to prefix.
2001 * Forces the next token to be a "WORD".
2002 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2006 S_force_ident(pTHX_ register const char *s, int kind)
2010 PERL_ARGS_ASSERT_FORCE_IDENT;
2013 const STRLEN len = strlen(s);
2014 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
2015 start_force(PL_curforce);
2016 NEXTVAL_NEXTTOKE.opval = o;
2019 o->op_private = OPpCONST_ENTERED;
2020 /* XXX see note in pp_entereval() for why we forgo typo
2021 warnings if the symbol must be introduced in an eval.
2023 gv_fetchpvn_flags(s, len,
2024 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2026 kind == '$' ? SVt_PV :
2027 kind == '@' ? SVt_PVAV :
2028 kind == '%' ? SVt_PVHV :
2036 Perl_str_to_version(pTHX_ SV *sv)
2041 const char *start = SvPV_const(sv,len);
2042 const char * const end = start + len;
2043 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2045 PERL_ARGS_ASSERT_STR_TO_VERSION;
2047 while (start < end) {
2051 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2056 retval += ((NV)n)/nshift;
2065 * Forces the next token to be a version number.
2066 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2067 * and if "guessing" is TRUE, then no new token is created (and the caller
2068 * must use an alternative parsing method).
2072 S_force_version(pTHX_ char *s, int guessing)
2078 I32 startoff = s - SvPVX(PL_linestr);
2081 PERL_ARGS_ASSERT_FORCE_VERSION;
2089 while (isDIGIT(*d) || *d == '_' || *d == '.')
2093 start_force(PL_curforce);
2094 curmad('X', newSVpvn(s,d-s));
2097 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
2099 s = scan_num(s, &pl_yylval);
2100 version = pl_yylval.opval;
2101 ver = cSVOPx(version)->op_sv;
2102 if (SvPOK(ver) && !SvNIOK(ver)) {
2103 SvUPGRADE(ver, SVt_PVNV);
2104 SvNV_set(ver, str_to_version(ver));
2105 SvNOK_on(ver); /* hint that it is a version */
2108 else if (guessing) {
2111 sv_free(PL_nextwhite); /* let next token collect whitespace */
2113 s = SvPVX(PL_linestr) + startoff;
2121 if (PL_madskills && !version) {
2122 sv_free(PL_nextwhite); /* let next token collect whitespace */
2124 s = SvPVX(PL_linestr) + startoff;
2127 /* NOTE: The parser sees the package name and the VERSION swapped */
2128 start_force(PL_curforce);
2129 NEXTVAL_NEXTTOKE.opval = version;
2137 * Tokenize a quoted string passed in as an SV. It finds the next
2138 * chunk, up to end of string or a backslash. It may make a new
2139 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2144 S_tokeq(pTHX_ SV *sv)
2148 register char *send;
2153 PERL_ARGS_ASSERT_TOKEQ;
2158 s = SvPV_force(sv, len);
2159 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2162 while (s < send && *s != '\\')
2167 if ( PL_hints & HINT_NEW_STRING ) {
2168 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2172 if (s + 1 < send && (s[1] == '\\'))
2173 s++; /* all that, just for this */
2178 SvCUR_set(sv, d - SvPVX_const(sv));
2180 if ( PL_hints & HINT_NEW_STRING )
2181 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2186 * Now come three functions related to double-quote context,
2187 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2188 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2189 * interact with PL_lex_state, and create fake ( ... ) argument lists
2190 * to handle functions and concatenation.
2191 * They assume that whoever calls them will be setting up a fake
2192 * join call, because each subthing puts a ',' after it. This lets
2195 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2197 * (I'm not sure whether the spurious commas at the end of lcfirst's
2198 * arguments and join's arguments are created or not).
2203 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2205 * Pattern matching will set PL_lex_op to the pattern-matching op to
2206 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2208 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2210 * Everything else becomes a FUNC.
2212 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2213 * had an OP_CONST or OP_READLINE). This just sets us up for a
2214 * call to S_sublex_push().
2218 S_sublex_start(pTHX)
2221 register const I32 op_type = pl_yylval.ival;
2223 if (op_type == OP_NULL) {
2224 pl_yylval.opval = PL_lex_op;
2228 if (op_type == OP_CONST || op_type == OP_READLINE) {
2229 SV *sv = tokeq(PL_lex_stuff);
2231 if (SvTYPE(sv) == SVt_PVIV) {
2232 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2234 const char * const p = SvPV_const(sv, len);
2235 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2239 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2240 PL_lex_stuff = NULL;
2241 /* Allow <FH> // "foo" */
2242 if (op_type == OP_READLINE)
2243 PL_expect = XTERMORDORDOR;
2246 else if (op_type == OP_BACKTICK && PL_lex_op) {
2247 /* readpipe() vas overriden */
2248 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2249 pl_yylval.opval = PL_lex_op;
2251 PL_lex_stuff = NULL;
2255 PL_sublex_info.super_state = PL_lex_state;
2256 PL_sublex_info.sub_inwhat = (U16)op_type;
2257 PL_sublex_info.sub_op = PL_lex_op;
2258 PL_lex_state = LEX_INTERPPUSH;
2262 pl_yylval.opval = PL_lex_op;
2272 * Create a new scope to save the lexing state. The scope will be
2273 * ended in S_sublex_done. Returns a '(', starting the function arguments
2274 * to the uc, lc, etc. found before.
2275 * Sets PL_lex_state to LEX_INTERPCONCAT.
2284 PL_lex_state = PL_sublex_info.super_state;
2285 SAVEBOOL(PL_lex_dojoin);
2286 SAVEI32(PL_lex_brackets);
2287 SAVEI32(PL_lex_casemods);
2288 SAVEI32(PL_lex_starts);
2289 SAVEI8(PL_lex_state);
2290 SAVEVPTR(PL_lex_inpat);
2291 SAVEI16(PL_lex_inwhat);
2292 SAVECOPLINE(PL_curcop);
2293 SAVEPPTR(PL_bufptr);
2294 SAVEPPTR(PL_bufend);
2295 SAVEPPTR(PL_oldbufptr);
2296 SAVEPPTR(PL_oldoldbufptr);
2297 SAVEPPTR(PL_last_lop);
2298 SAVEPPTR(PL_last_uni);
2299 SAVEPPTR(PL_linestart);
2300 SAVESPTR(PL_linestr);
2301 SAVEGENERICPV(PL_lex_brackstack);
2302 SAVEGENERICPV(PL_lex_casestack);
2304 PL_linestr = PL_lex_stuff;
2305 PL_lex_stuff = NULL;
2307 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2308 = SvPVX(PL_linestr);
2309 PL_bufend += SvCUR(PL_linestr);
2310 PL_last_lop = PL_last_uni = NULL;
2311 SAVEFREESV(PL_linestr);
2313 PL_lex_dojoin = FALSE;
2314 PL_lex_brackets = 0;
2315 Newx(PL_lex_brackstack, 120, char);
2316 Newx(PL_lex_casestack, 12, char);
2317 PL_lex_casemods = 0;
2318 *PL_lex_casestack = '\0';
2320 PL_lex_state = LEX_INTERPCONCAT;
2321 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2323 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2324 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2325 PL_lex_inpat = PL_sublex_info.sub_op;
2327 PL_lex_inpat = NULL;
2334 * Restores lexer state after a S_sublex_push.
2341 if (!PL_lex_starts++) {
2342 SV * const sv = newSVpvs("");
2343 if (SvUTF8(PL_linestr))
2345 PL_expect = XOPERATOR;
2346 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2350 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2351 PL_lex_state = LEX_INTERPCASEMOD;
2355 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2356 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2357 PL_linestr = PL_lex_repl;
2359 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2360 PL_bufend += SvCUR(PL_linestr);
2361 PL_last_lop = PL_last_uni = NULL;
2362 SAVEFREESV(PL_linestr);
2363 PL_lex_dojoin = FALSE;
2364 PL_lex_brackets = 0;
2365 PL_lex_casemods = 0;
2366 *PL_lex_casestack = '\0';
2368 if (SvEVALED(PL_lex_repl)) {
2369 PL_lex_state = LEX_INTERPNORMAL;
2371 /* we don't clear PL_lex_repl here, so that we can check later
2372 whether this is an evalled subst; that means we rely on the
2373 logic to ensure sublex_done() is called again only via the
2374 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2377 PL_lex_state = LEX_INTERPCONCAT;
2387 PL_endwhite = newSVpvs("");
2388 sv_catsv(PL_endwhite, PL_thiswhite);
2392 sv_setpvs(PL_thistoken,"");
2394 PL_realtokenstart = -1;
2398 PL_bufend = SvPVX(PL_linestr);
2399 PL_bufend += SvCUR(PL_linestr);
2400 PL_expect = XOPERATOR;
2401 PL_sublex_info.sub_inwhat = 0;
2409 Extracts a pattern, double-quoted string, or transliteration. This
2412 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2413 processing a pattern (PL_lex_inpat is true), a transliteration
2414 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2416 Returns a pointer to the character scanned up to. If this is
2417 advanced from the start pointer supplied (i.e. if anything was
2418 successfully parsed), will leave an OP for the substring scanned
2419 in pl_yylval. Caller must intuit reason for not parsing further
2420 by looking at the next characters herself.
2424 double-quoted style: \r and \n
2425 regexp special ones: \D \s
2428 case and quoting: \U \Q \E
2429 stops on @ and $, but not for $ as tail anchor
2431 In transliterations:
2432 characters are VERY literal, except for - not at the start or end
2433 of the string, which indicates a range. If the range is in bytes,
2434 scan_const expands the range to the full set of intermediate
2435 characters. If the range is in utf8, the hyphen is replaced with
2436 a certain range mark which will be handled by pmtrans() in op.c.
2438 In double-quoted strings:
2440 double-quoted style: \r and \n
2442 deprecated backrefs: \1 (in substitution replacements)
2443 case and quoting: \U \Q \E
2446 scan_const does *not* construct ops to handle interpolated strings.
2447 It stops processing as soon as it finds an embedded $ or @ variable
2448 and leaves it to the caller to work out what's going on.
2450 embedded arrays (whether in pattern or not) could be:
2451 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2453 $ in double-quoted strings must be the symbol of an embedded scalar.
2455 $ in pattern could be $foo or could be tail anchor. Assumption:
2456 it's a tail anchor if $ is the last thing in the string, or if it's
2457 followed by one of "()| \r\n\t"
2459 \1 (backreferences) are turned into $1
2461 The structure of the code is
2462 while (there's a character to process) {
2463 handle transliteration ranges
2464 skip regexp comments /(?#comment)/ and codes /(?{code})/
2465 skip #-initiated comments in //x patterns
2466 check for embedded arrays
2467 check for embedded scalars
2469 leave intact backslashes from leaveit (below)
2470 deprecate \1 in substitution replacements
2471 handle string-changing backslashes \l \U \Q \E, etc.
2472 switch (what was escaped) {
2473 handle \- in a transliteration (becomes a literal -)
2474 handle \132 (octal characters)
2475 handle \x15 and \x{1234} (hex characters)
2476 handle \N{name} (named characters)
2477 handle \cV (control characters)
2478 handle printf-style backslashes (\f, \r, \n, etc)
2481 } (end if backslash)
2482 handle regular character
2483 } (end while character to read)
2488 S_scan_const(pTHX_ char *start)
2491 register char *send = PL_bufend; /* end of the constant */
2492 SV *sv = newSV(send - start); /* sv for the constant. See
2493 note below on sizing. */
2494 register char *s = start; /* start of the constant */
2495 register char *d = SvPVX(sv); /* destination for copies */
2496 bool dorange = FALSE; /* are we in a translit range? */
2497 bool didrange = FALSE; /* did we just finish a range? */
2498 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
2499 I32 this_utf8 = UTF; /* Is the source string assumed
2500 to be UTF8? But, this can
2501 show as true when the source
2502 isn't utf8, as for example
2503 when it is entirely composed
2506 /* Note on sizing: The scanned constant is placed into sv, which is
2507 * initialized by newSV() assuming one byte of output for every byte of
2508 * input. This routine expects newSV() to allocate an extra byte for a
2509 * trailing NUL, which this routine will append if it gets to the end of
2510 * the input. There may be more bytes of input than output (eg., \N{LATIN
2511 * CAPITAL LETTER A}), or more output than input if the constant ends up
2512 * recoded to utf8, but each time a construct is found that might increase
2513 * the needed size, SvGROW() is called. Its size parameter each time is
2514 * based on the best guess estimate at the time, namely the length used so
2515 * far, plus the length the current construct will occupy, plus room for
2516 * the trailing NUL, plus one byte for every input byte still unscanned */
2520 UV literal_endpoint = 0;
2521 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2524 PERL_ARGS_ASSERT_SCAN_CONST;
2526 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2527 /* If we are doing a trans and we know we want UTF8 set expectation */
2528 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2529 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2533 while (s < send || dorange) {
2534 /* get transliterations out of the way (they're most literal) */
2535 if (PL_lex_inwhat == OP_TRANS) {
2536 /* expand a range A-Z to the full set of characters. AIE! */
2538 I32 i; /* current expanded character */
2539 I32 min; /* first character in range */
2540 I32 max; /* last character in range */
2551 char * const c = (char*)utf8_hop((U8*)d, -1);
2555 *c = (char)UTF_TO_NATIVE(0xff);
2556 /* mark the range as done, and continue */
2562 i = d - SvPVX_const(sv); /* remember current offset */
2565 SvLEN(sv) + (has_utf8 ?
2566 (512 - UTF_CONTINUATION_MARK +
2569 /* How many two-byte within 0..255: 128 in UTF-8,
2570 * 96 in UTF-8-mod. */
2572 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2574 d = SvPVX(sv) + i; /* refresh d after realloc */
2578 for (j = 0; j <= 1; j++) {
2579 char * const c = (char*)utf8_hop((U8*)d, -1);
2580 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2586 max = (U8)0xff; /* only to \xff */
2587 uvmax = uv; /* \x{100} to uvmax */
2589 d = c; /* eat endpoint chars */
2594 d -= 2; /* eat the first char and the - */
2595 min = (U8)*d; /* first char in range */
2596 max = (U8)d[1]; /* last char in range */
2603 "Invalid range \"%c-%c\" in transliteration operator",
2604 (char)min, (char)max);
2608 if (literal_endpoint == 2 &&
2609 ((isLOWER(min) && isLOWER(max)) ||
2610 (isUPPER(min) && isUPPER(max)))) {
2612 for (i = min; i <= max; i++)
2614 *d++ = NATIVE_TO_NEED(has_utf8,i);
2616 for (i = min; i <= max; i++)
2618 *d++ = NATIVE_TO_NEED(has_utf8,i);
2623 for (i = min; i <= max; i++)
2626 const U8 ch = (U8)NATIVE_TO_UTF(i);
2627 if (UNI_IS_INVARIANT(ch))
2630 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2631 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2640 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2642 *d++ = (char)UTF_TO_NATIVE(0xff);
2644 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2648 /* mark the range as done, and continue */
2652 literal_endpoint = 0;
2657 /* range begins (ignore - as first or last char) */
2658 else if (*s == '-' && s+1 < send && s != start) {
2660 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2667 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2677 literal_endpoint = 0;
2678 native_range = TRUE;
2683 /* if we get here, we're not doing a transliteration */
2685 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2686 except for the last char, which will be done separately. */
2687 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2689 while (s+1 < send && *s != ')')
2690 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2692 else if (s[2] == '{' /* This should match regcomp.c */
2693 || (s[2] == '?' && s[3] == '{'))
2696 char *regparse = s + (s[2] == '{' ? 3 : 4);
2699 while (count && (c = *regparse)) {
2700 if (c == '\\' && regparse[1])
2708 if (*regparse != ')')
2709 regparse--; /* Leave one char for continuation. */
2710 while (s < regparse)
2711 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2715 /* likewise skip #-initiated comments in //x patterns */
2716 else if (*s == '#' && PL_lex_inpat &&
2717 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2718 while (s+1 < send && *s != '\n')
2719 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2722 /* check for embedded arrays
2723 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2725 else if (*s == '@' && s[1]) {
2726 if (isALNUM_lazy_if(s+1,UTF))
2728 if (strchr(":'{$", s[1]))
2730 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2731 break; /* in regexp, neither @+ nor @- are interpolated */
2734 /* check for embedded scalars. only stop if we're sure it's a
2737 else if (*s == '$') {
2738 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2740 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2742 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2743 "Possible unintended interpolation of $\\ in regex");
2745 break; /* in regexp, $ might be tail anchor */
2749 /* End of else if chain - OP_TRANS rejoin rest */
2752 if (*s == '\\' && s+1 < send) {
2755 /* deprecate \1 in strings and substitution replacements */
2756 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2757 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2759 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2764 /* string-change backslash escapes */
2765 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2769 /* skip any other backslash escapes in a pattern */
2770 else if (PL_lex_inpat) {
2771 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2772 goto default_action;
2775 /* if we get here, it's either a quoted -, or a digit */
2778 /* quoted - in transliterations */
2780 if (PL_lex_inwhat == OP_TRANS) {
2787 if ((isALPHA(*s) || isDIGIT(*s)))
2788 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2789 "Unrecognized escape \\%c passed through",
2791 /* default action is to copy the quoted character */
2792 goto default_action;
2795 /* eg. \132 indicates the octal constant 0x132 */
2796 case '0': case '1': case '2': case '3':
2797 case '4': case '5': case '6': case '7':
2801 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2804 goto NUM_ESCAPE_INSERT;
2806 /* eg. \x24 indicates the hex constant 0x24 */
2810 char* const e = strchr(s, '}');
2811 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2812 PERL_SCAN_DISALLOW_PREFIX;
2817 yyerror("Missing right brace on \\x{}");
2821 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2827 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2828 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2834 /* Insert oct, hex, or \N{U+...} escaped character. There will
2835 * always be enough room in sv since such escapes will be
2836 * longer than any UTF-8 sequence they can end up as, except if
2837 * they force us to recode the rest of the string into utf8 */
2839 /* Here uv is the ordinal of the next character being added in
2840 * unicode (converted from native). (It has to be done before
2841 * here because \N is interpreted as unicode, and oct and hex
2843 if (!UNI_IS_INVARIANT(uv)) {
2844 if (!has_utf8 && uv > 255) {
2845 /* Might need to recode whatever we have accumulated so
2846 * far if it contains any chars variant in utf8 or
2849 SvCUR_set(sv, d - SvPVX_const(sv));
2852 /* See Note on sizing above. */
2853 sv_utf8_upgrade_flags_grow(sv,
2854 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2855 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2856 d = SvPVX(sv) + SvCUR(sv);
2861 d = (char*)uvuni_to_utf8((U8*)d, uv);
2862 if (PL_lex_inwhat == OP_TRANS &&
2863 PL_sublex_info.sub_op) {
2864 PL_sublex_info.sub_op->op_private |=
2865 (PL_lex_repl ? OPpTRANS_FROM_UTF
2869 if (uv > 255 && !dorange)
2870 native_range = FALSE;
2882 /* \N{LATIN SMALL LETTER A} is a named character, and so is
2887 char* e = strchr(s, '}');
2893 yyerror("Missing right brace on \\N{}");
2897 if (e > s + 2 && s[1] == 'U' && s[2] == '+') {
2898 /* \N{U+...} The ... is a unicode value even on EBCDIC
2900 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2901 PERL_SCAN_DISALLOW_PREFIX;
2904 uv = grok_hex(s, &len, &flags, NULL);
2905 if ( e > s && len != (STRLEN)(e - s) ) {
2909 goto NUM_ESCAPE_INSERT;
2911 res = newSVpvn(s + 1, e - s - 1);
2912 res = new_constant( NULL, 0, "charnames",
2913 res, NULL, s - 2, e - s + 3 );
2915 sv_utf8_upgrade(res);
2916 str = SvPV_const(res,len);
2917 #ifdef EBCDIC_NEVER_MIND
2918 /* charnames uses pack U and that has been
2919 * recently changed to do the below uni->native
2920 * mapping, so this would be redundant (and wrong,
2921 * the code point would be doubly converted).
2922 * But leave this in just in case the pack U change
2923 * gets revoked, but the semantics is still
2924 * desireable for charnames. --jhi */
2926 UV uv = utf8_to_uvchr((const U8*)str, 0);
2929 U8 tmpbuf[UTF8_MAXBYTES+1], *d;
2931 d = uvchr_to_utf8(tmpbuf, UNI_TO_NATIVE(uv));
2932 sv_setpvn(res, (char *)tmpbuf, d - tmpbuf);
2933 str = SvPV_const(res, len);
2937 /* If destination is not in utf8 but this new character is,
2938 * recode the dest to utf8 */
2939 if (!has_utf8 && SvUTF8(res)) {
2940 SvCUR_set(sv, d - SvPVX_const(sv));
2943 /* See Note on sizing above. */
2944 sv_utf8_upgrade_flags_grow(sv,
2945 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2946 len + (STRLEN)(send - s) + 1);
2947 d = SvPVX(sv) + SvCUR(sv);
2949 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
2951 /* See Note on sizing above. (NOTE: SvCUR() is not set
2952 * correctly here). */
2953 const STRLEN off = d - SvPVX_const(sv);
2954 d = SvGROW(sv, off + len + (STRLEN)(send - s) + 1) + off;
2958 native_range = FALSE; /* \N{} is guessed to be Unicode */
2960 Copy(str, d, len, char);
2967 yyerror("Missing braces on \\N{}");
2970 /* \c is a control character */
2979 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
2982 yyerror("Missing control char name in \\c");
2986 /* printf-style backslashes, formfeeds, newlines, etc */
2988 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
2991 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
2994 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
2997 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3000 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3003 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3006 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3012 } /* end if (backslash) */
3019 /* If we started with encoded form, or already know we want it,
3020 then encode the next character */
3021 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3025 /* One might think that it is wasted effort in the case of the
3026 * source being utf8 (this_utf8 == TRUE) to take the next character
3027 * in the source, convert it to an unsigned value, and then convert
3028 * it back again. But the source has not been validated here. The
3029 * routine that does the conversion checks for errors like
3032 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3033 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3035 SvCUR_set(sv, d - SvPVX_const(sv));
3038 /* See Note on sizing above. */
3039 sv_utf8_upgrade_flags_grow(sv,
3040 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3041 need + (STRLEN)(send - s) + 1);
3042 d = SvPVX(sv) + SvCUR(sv);
3044 } else if (need > len) {
3045 /* encoded value larger than old, may need extra space (NOTE:
3046 * SvCUR() is not set correctly here). See Note on sizing
3048 const STRLEN off = d - SvPVX_const(sv);
3049 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3053 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3055 if (uv > 255 && !dorange)
3056 native_range = FALSE;
3060 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3062 } /* while loop to process each character */
3064 /* terminate the string and set up the sv */
3066 SvCUR_set(sv, d - SvPVX_const(sv));
3067 if (SvCUR(sv) >= SvLEN(sv))
3068 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3071 if (PL_encoding && !has_utf8) {
3072 sv_recode_to_utf8(sv, PL_encoding);
3078 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3079 PL_sublex_info.sub_op->op_private |=
3080 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3084 /* shrink the sv if we allocated more than we used */
3085 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3086 SvPV_shrink_to_cur(sv);
3089 /* return the substring (via pl_yylval) only if we parsed anything */
3090 if (s > PL_bufptr) {
3091 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3092 const char *const key = PL_lex_inpat ? "qr" : "q";
3093 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3097 if (PL_lex_inwhat == OP_TRANS) {
3100 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3108 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3111 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3118 * Returns TRUE if there's more to the expression (e.g., a subscript),
3121 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3123 * ->[ and ->{ return TRUE
3124 * { and [ outside a pattern are always subscripts, so return TRUE
3125 * if we're outside a pattern and it's not { or [, then return FALSE
3126 * if we're in a pattern and the first char is a {
3127 * {4,5} (any digits around the comma) returns FALSE
3128 * if we're in a pattern and the first char is a [
3130 * [SOMETHING] has a funky algorithm to decide whether it's a
3131 * character class or not. It has to deal with things like
3132 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3133 * anything else returns TRUE
3136 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3139 S_intuit_more(pTHX_ register char *s)
3143 PERL_ARGS_ASSERT_INTUIT_MORE;
3145 if (PL_lex_brackets)
3147 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3149 if (*s != '{' && *s != '[')
3154 /* In a pattern, so maybe we have {n,m}. */
3171 /* On the other hand, maybe we have a character class */
3174 if (*s == ']' || *s == '^')
3177 /* this is terrifying, and it works */
3178 int weight = 2; /* let's weigh the evidence */
3180 unsigned char un_char = 255, last_un_char;
3181 const char * const send = strchr(s,']');
3182 char tmpbuf[sizeof PL_tokenbuf * 4];
3184 if (!send) /* has to be an expression */
3187 Zero(seen,256,char);
3190 else if (isDIGIT(*s)) {
3192 if (isDIGIT(s[1]) && s[2] == ']')
3198 for (; s < send; s++) {
3199 last_un_char = un_char;
3200 un_char = (unsigned char)*s;
3205 weight -= seen[un_char] * 10;
3206 if (isALNUM_lazy_if(s+1,UTF)) {
3208 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3209 len = (int)strlen(tmpbuf);
3210 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
3215 else if (*s == '$' && s[1] &&
3216 strchr("[#!%*<>()-=",s[1])) {
3217 if (/*{*/ strchr("])} =",s[2]))
3226 if (strchr("wds]",s[1]))
3228 else if (seen[(U8)'\''] || seen[(U8)'"'])
3230 else if (strchr("rnftbxcav",s[1]))
3232 else if (isDIGIT(s[1])) {
3234 while (s[1] && isDIGIT(s[1]))
3244 if (strchr("aA01! ",last_un_char))
3246 if (strchr("zZ79~",s[1]))
3248 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3249 weight -= 5; /* cope with negative subscript */
3252 if (!isALNUM(last_un_char)
3253 && !(last_un_char == '$' || last_un_char == '@'
3254 || last_un_char == '&')
3255 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3260 if (keyword(tmpbuf, d - tmpbuf, 0))
3263 if (un_char == last_un_char + 1)
3265 weight -= seen[un_char];
3270 if (weight >= 0) /* probably a character class */
3280 * Does all the checking to disambiguate
3282 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3283 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3285 * First argument is the stuff after the first token, e.g. "bar".
3287 * Not a method if bar is a filehandle.
3288 * Not a method if foo is a subroutine prototyped to take a filehandle.
3289 * Not a method if it's really "Foo $bar"
3290 * Method if it's "foo $bar"
3291 * Not a method if it's really "print foo $bar"
3292 * Method if it's really "foo package::" (interpreted as package->foo)
3293 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3294 * Not a method if bar is a filehandle or package, but is quoted with
3299 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3302 char *s = start + (*start == '$');
3303 char tmpbuf[sizeof PL_tokenbuf];
3310 PERL_ARGS_ASSERT_INTUIT_METHOD;
3313 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3317 const char *proto = SvPVX_const(cv);
3328 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3329 /* start is the beginning of the possible filehandle/object,
3330 * and s is the end of it
3331 * tmpbuf is a copy of it
3334 if (*start == '$') {
3335 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3336 isUPPER(*PL_tokenbuf))
3339 len = start - SvPVX(PL_linestr);
3343 start = SvPVX(PL_linestr) + len;
3347 return *s == '(' ? FUNCMETH : METHOD;
3349 if (!keyword(tmpbuf, len, 0)) {
3350 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3354 soff = s - SvPVX(PL_linestr);
3358 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
3359 if (indirgv && GvCVu(indirgv))
3361 /* filehandle or package name makes it a method */
3362 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
3364 soff = s - SvPVX(PL_linestr);
3367 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3368 return 0; /* no assumptions -- "=>" quotes bearword */
3370 start_force(PL_curforce);
3371 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3372 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3373 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3375 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3380 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3382 return *s == '(' ? FUNCMETH : METHOD;
3388 /* Encoded script support. filter_add() effectively inserts a
3389 * 'pre-processing' function into the current source input stream.
3390 * Note that the filter function only applies to the current source file
3391 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3393 * The datasv parameter (which may be NULL) can be used to pass
3394 * private data to this instance of the filter. The filter function
3395 * can recover the SV using the FILTER_DATA macro and use it to
3396 * store private buffers and state information.
3398 * The supplied datasv parameter is upgraded to a PVIO type
3399 * and the IoDIRP/IoANY field is used to store the function pointer,
3400 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3401 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3402 * private use must be set using malloc'd pointers.
3406 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3415 if (!PL_rsfp_filters)
3416 PL_rsfp_filters = newAV();
3419 SvUPGRADE(datasv, SVt_PVIO);
3420 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3421 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3422 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3423 FPTR2DPTR(void *, IoANY(datasv)),
3424 SvPV_nolen(datasv)));
3425 av_unshift(PL_rsfp_filters, 1);
3426 av_store(PL_rsfp_filters, 0, datasv) ;
3431 /* Delete most recently added instance of this filter function. */
3433 Perl_filter_del(pTHX_ filter_t funcp)
3438 PERL_ARGS_ASSERT_FILTER_DEL;
3441 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3442 FPTR2DPTR(void*, funcp)));
3444 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3446 /* if filter is on top of stack (usual case) just pop it off */
3447 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3448 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3449 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
3450 IoANY(datasv) = (void *)NULL;
3451 sv_free(av_pop(PL_rsfp_filters));
3455 /* we need to search for the correct entry and clear it */
3456 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3460 /* Invoke the idxth filter function for the current rsfp. */
3461 /* maxlen 0 = read one text line */
3463 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3468 /* This API is bad. It should have been using unsigned int for maxlen.
3469 Not sure if we want to change the API, but if not we should sanity
3470 check the value here. */
3471 const unsigned int correct_length
3480 PERL_ARGS_ASSERT_FILTER_READ;
3482 if (!PL_parser || !PL_rsfp_filters)
3484 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3485 /* Provide a default input filter to make life easy. */
3486 /* Note that we append to the line. This is handy. */
3487 DEBUG_P(PerlIO_printf(Perl_debug_log,
3488 "filter_read %d: from rsfp\n", idx));
3489 if (correct_length) {
3492 const int old_len = SvCUR(buf_sv);
3494 /* ensure buf_sv is large enough */
3495 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3496 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3497 correct_length)) <= 0) {
3498 if (PerlIO_error(PL_rsfp))
3499 return -1; /* error */
3501 return 0 ; /* end of file */
3503 SvCUR_set(buf_sv, old_len + len) ;
3504 SvPVX(buf_sv)[old_len + len] = '\0';
3507 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3508 if (PerlIO_error(PL_rsfp))
3509 return -1; /* error */
3511 return 0 ; /* end of file */
3514 return SvCUR(buf_sv);
3516 /* Skip this filter slot if filter has been deleted */
3517 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3518 DEBUG_P(PerlIO_printf(Perl_debug_log,
3519 "filter_read %d: skipped (filter deleted)\n",
3521 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
3523 /* Get function pointer hidden within datasv */
3524 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
3525 DEBUG_P(PerlIO_printf(Perl_debug_log,
3526 "filter_read %d: via function %p (%s)\n",
3527 idx, (void*)datasv, SvPV_nolen_const(datasv)));
3528 /* Call function. The function is expected to */
3529 /* call "FILTER_READ(idx+1, buf_sv)" first. */
3530 /* Return: <0:error, =0:eof, >0:not eof */
3531 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
3535 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
3539 PERL_ARGS_ASSERT_FILTER_GETS;
3541 #ifdef PERL_CR_FILTER
3542 if (!PL_rsfp_filters) {
3543 filter_add(S_cr_textfilter,NULL);
3546 if (PL_rsfp_filters) {
3548 SvCUR_set(sv, 0); /* start with empty line */
3549 if (FILTER_READ(0, sv, 0) > 0)
3550 return ( SvPVX(sv) ) ;
3555 return (sv_gets(sv, PL_rsfp, append));
3559 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3564 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3566 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3570 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3571 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3573 return GvHV(gv); /* Foo:: */
3576 /* use constant CLASS => 'MyClass' */
3577 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3578 if (gv && GvCV(gv)) {
3579 SV * const sv = cv_const_sv(GvCV(gv));
3581 pkgname = SvPV_const(sv, len);
3584 return gv_stashpvn(pkgname, len, 0);
3588 * S_readpipe_override
3589 * Check whether readpipe() is overriden, and generates the appropriate
3590 * optree, provided sublex_start() is called afterwards.
3593 S_readpipe_override(pTHX)
3596 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3597 pl_yylval.ival = OP_BACKTICK;
3599 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3601 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3602 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3603 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3605 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3606 append_elem(OP_LIST,
3607 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3608 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3615 * The intent of this yylex wrapper is to minimize the changes to the
3616 * tokener when we aren't interested in collecting madprops. It remains
3617 * to be seen how successful this strategy will be...
3624 char *s = PL_bufptr;
3626 /* make sure PL_thiswhite is initialized */
3630 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3631 if (PL_pending_ident)
3632 return S_pending_ident(aTHX);
3634 /* previous token ate up our whitespace? */
3635 if (!PL_lasttoke && PL_nextwhite) {
3636 PL_thiswhite = PL_nextwhite;
3640 /* isolate the token, and figure out where it is without whitespace */
3641 PL_realtokenstart = -1;
3645 assert(PL_curforce < 0);
3647 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3648 if (!PL_thistoken) {
3649 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3650 PL_thistoken = newSVpvs("");
3652 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3653 PL_thistoken = newSVpvn(tstart, s - tstart);
3656 if (PL_thismad) /* install head */
3657 CURMAD('X', PL_thistoken);
3660 /* last whitespace of a sublex? */
3661 if (optype == ')' && PL_endwhite) {
3662 CURMAD('X', PL_endwhite);
3667 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3668 if (!PL_thiswhite && !PL_endwhite && !optype) {
3669 sv_free(PL_thistoken);
3674 /* put off final whitespace till peg */
3675 if (optype == ';' && !PL_rsfp) {
3676 PL_nextwhite = PL_thiswhite;
3679 else if (PL_thisopen) {
3680 CURMAD('q', PL_thisopen);
3682 sv_free(PL_thistoken);
3686 /* Store actual token text as madprop X */
3687 CURMAD('X', PL_thistoken);
3691 /* add preceding whitespace as madprop _ */
3692 CURMAD('_', PL_thiswhite);
3696 /* add quoted material as madprop = */
3697 CURMAD('=', PL_thisstuff);
3701 /* add terminating quote as madprop Q */
3702 CURMAD('Q', PL_thisclose);
3706 /* special processing based on optype */
3710 /* opval doesn't need a TOKEN since it can already store mp */
3720 if (pl_yylval.opval)
3721 append_madprops(PL_thismad, pl_yylval.opval, 0);
3729 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3738 /* remember any fake bracket that lexer is about to discard */
3739 if (PL_lex_brackets == 1 &&
3740 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
3743 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3746 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
3747 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3750 break; /* don't bother looking for trailing comment */
3759 /* attach a trailing comment to its statement instead of next token */
3763 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
3765 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
3767 if (*s == '\n' || *s == '#') {
3768 while (s < PL_bufend && *s != '\n')
3772 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
3773 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
3790 /* Create new token struct. Note: opvals return early above. */
3791 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
3798 S_tokenize_use(pTHX_ int is_use, char *s) {
3801 PERL_ARGS_ASSERT_TOKENIZE_USE;
3803 if (PL_expect != XSTATE)
3804 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
3805 is_use ? "use" : "no"));
3807 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
3808 s = force_version(s, TRUE);
3809 if (*s == ';' || (s = SKIPSPACE1(s), *s == ';')) {
3810 start_force(PL_curforce);
3811 NEXTVAL_NEXTTOKE.opval = NULL;
3814 else if (*s == 'v') {
3815 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3816 s = force_version(s, FALSE);
3820 s = force_word(s,WORD,FALSE,TRUE,FALSE);
3821 s = force_version(s, FALSE);
3823 pl_yylval.ival = is_use;
3827 static const char* const exp_name[] =
3828 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
3829 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
3836 Works out what to call the token just pulled out of the input
3837 stream. The yacc parser takes care of taking the ops we return and
3838 stitching them into a tree.
3844 if read an identifier
3845 if we're in a my declaration
3846 croak if they tried to say my($foo::bar)
3847 build the ops for a my() declaration
3848 if it's an access to a my() variable
3849 are we in a sort block?
3850 croak if my($a); $a <=> $b
3851 build ops for access to a my() variable
3852 if in a dq string, and they've said @foo and we can't find @foo
3854 build ops for a bareword
3855 if we already built the token before, use it.
3860 #pragma segment Perl_yylex
3866 register char *s = PL_bufptr;
3871 /* orig_keyword, gvp, and gv are initialized here because
3872 * jump to the label just_a_word_zero can bypass their
3873 * initialization later. */
3874 I32 orig_keyword = 0;
3879 SV* tmp = newSVpvs("");
3880 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
3881 (IV)CopLINE(PL_curcop),
3882 lex_state_names[PL_lex_state],
3883 exp_name[PL_expect],
3884 pv_display(tmp, s, strlen(s), 0, 60));
3887 /* check if there's an identifier for us to look at */
3888 if (PL_pending_ident)
3889 return REPORT(S_pending_ident(aTHX));
3891 /* no identifier pending identification */
3893 switch (PL_lex_state) {
3895 case LEX_NORMAL: /* Some compilers will produce faster */
3896 case LEX_INTERPNORMAL: /* code if we comment these out. */
3900 /* when we've already built the next token, just pull it out of the queue */
3904 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
3906 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
3907 PL_nexttoke[PL_lasttoke].next_mad = 0;
3908 if (PL_thismad && PL_thismad->mad_key == '_') {
3909 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
3910 PL_thismad->mad_val = 0;
3911 mad_free(PL_thismad);
3916 PL_lex_state = PL_lex_defer;
3917 PL_expect = PL_lex_expect;
3918 PL_lex_defer = LEX_NORMAL;
3919 if (!PL_nexttoke[PL_lasttoke].next_type)
3924 pl_yylval = PL_nextval[PL_nexttoke];
3926 PL_lex_state = PL_lex_defer;
3927 PL_expect = PL_lex_expect;
3928 PL_lex_defer = LEX_NORMAL;
3932 /* FIXME - can these be merged? */
3933 return(PL_nexttoke[PL_lasttoke].next_type);
3935 return REPORT(PL_nexttype[PL_nexttoke]);
3938 /* interpolated case modifiers like \L \U, including \Q and \E.
3939 when we get here, PL_bufptr is at the \
3941 case LEX_INTERPCASEMOD:
3943 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
3944 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
3946 /* handle \E or end of string */
3947 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
3949 if (PL_lex_casemods) {
3950 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
3951 PL_lex_casestack[PL_lex_casemods] = '\0';
3953 if (PL_bufptr != PL_bufend
3954 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
3956 PL_lex_state = LEX_INTERPCONCAT;
3959 PL_thistoken = newSVpvs("\\E");
3965 while (PL_bufptr != PL_bufend &&
3966 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
3968 PL_thiswhite = newSVpvs("");
3969 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
3973 if (PL_bufptr != PL_bufend)
3976 PL_lex_state = LEX_INTERPCONCAT;
3980 DEBUG_T({ PerlIO_printf(Perl_debug_log,
3981 "### Saw case modifier\n"); });
3983 if (s[1] == '\\' && s[2] == 'E') {
3986 PL_thiswhite = newSVpvs("");
3987 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
3990 PL_lex_state = LEX_INTERPCONCAT;
3995 if (!PL_madskills) /* when just compiling don't need correct */
3996 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
3997 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
3998 if ((*s == 'L' || *s == 'U') &&
3999 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4000 PL_lex_casestack[--PL_lex_casemods] = '\0';
4003 if (PL_lex_casemods > 10)
4004 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4005 PL_lex_casestack[PL_lex_casemods++] = *s;
4006 PL_lex_casestack[PL_lex_casemods] = '\0';
4007 PL_lex_state = LEX_INTERPCONCAT;
4008 start_force(PL_curforce);
4009 NEXTVAL_NEXTTOKE.ival = 0;
4011 start_force(PL_curforce);
4013 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4015 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4017 NEXTVAL_NEXTTOKE.ival = OP_LC;
4019 NEXTVAL_NEXTTOKE.ival = OP_UC;
4021 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4023 Perl_croak(aTHX_ "panic: yylex");
4025 SV* const tmpsv = newSVpvs("\\ ");
4026 /* replace the space with the character we want to escape
4028 SvPVX(tmpsv)[1] = *s;
4034 if (PL_lex_starts) {
4040 sv_free(PL_thistoken);
4041 PL_thistoken = newSVpvs("");
4044 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4045 if (PL_lex_casemods == 1 && PL_lex_inpat)
4054 case LEX_INTERPPUSH:
4055 return REPORT(sublex_push());
4057 case LEX_INTERPSTART:
4058 if (PL_bufptr == PL_bufend)
4059 return REPORT(sublex_done());
4060 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4061 "### Interpolated variable\n"); });
4063 PL_lex_dojoin = (*PL_bufptr == '@');
4064 PL_lex_state = LEX_INTERPNORMAL;
4065 if (PL_lex_dojoin) {
4066 start_force(PL_curforce);
4067 NEXTVAL_NEXTTOKE.ival = 0;
4069 start_force(PL_curforce);
4070 force_ident("\"", '$');
4071 start_force(PL_curforce);
4072 NEXTVAL_NEXTTOKE.ival = 0;
4074 start_force(PL_curforce);
4075 NEXTVAL_NEXTTOKE.ival = 0;
4077 start_force(PL_curforce);
4078 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4081 if (PL_lex_starts++) {
4086 sv_free(PL_thistoken);
4087 PL_thistoken = newSVpvs("");
4090 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4091 if (!PL_lex_casemods && PL_lex_inpat)
4098 case LEX_INTERPENDMAYBE:
4099 if (intuit_more(PL_bufptr)) {
4100 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4106 if (PL_lex_dojoin) {
4107 PL_lex_dojoin = FALSE;
4108 PL_lex_state = LEX_INTERPCONCAT;
4112 sv_free(PL_thistoken);
4113 PL_thistoken = newSVpvs("");
4118 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4119 && SvEVALED(PL_lex_repl))
4121 if (PL_bufptr != PL_bufend)
4122 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4126 case LEX_INTERPCONCAT:
4128 if (PL_lex_brackets)
4129 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4131 if (PL_bufptr == PL_bufend)
4132 return REPORT(sublex_done());
4134 if (SvIVX(PL_linestr) == '\'') {
4135 SV *sv = newSVsv(PL_linestr);
4138 else if ( PL_hints & HINT_NEW_RE )
4139 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4140 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4144 s = scan_const(PL_bufptr);
4146 PL_lex_state = LEX_INTERPCASEMOD;
4148 PL_lex_state = LEX_INTERPSTART;
4151 if (s != PL_bufptr) {
4152 start_force(PL_curforce);
4154 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4156 NEXTVAL_NEXTTOKE = pl_yylval;
4159 if (PL_lex_starts++) {
4163 sv_free(PL_thistoken);
4164 PL_thistoken = newSVpvs("");
4167 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4168 if (!PL_lex_casemods && PL_lex_inpat)
4181 PL_lex_state = LEX_NORMAL;
4182 s = scan_formline(PL_bufptr);
4183 if (!PL_lex_formbrack)
4189 PL_oldoldbufptr = PL_oldbufptr;
4195 sv_free(PL_thistoken);
4198 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4202 if (isIDFIRST_lazy_if(s,UTF))
4205 unsigned char c = *s;
4206 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4207 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4208 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4213 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4217 goto fake_eof; /* emulate EOF on ^D or ^Z */
4226 if (PL_lex_brackets) {
4227 yyerror((const char *)
4229 ? "Format not terminated"
4230 : "Missing right curly or square bracket"));
4232 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4233 "### Tokener got EOF\n");
4237 if (s++ < PL_bufend)
4238 goto retry; /* ignore stray nulls */
4241 if (!PL_in_eval && !PL_preambled) {
4242 PL_preambled = TRUE;
4248 /* Generate a string of Perl code to load the debugger.
4249 * If PERL5DB is set, it will return the contents of that,
4250 * otherwise a compile-time require of perl5db.pl. */
4252 const char * const pdb = PerlEnv_getenv("PERL5DB");
4255 sv_setpv(PL_linestr, pdb);
4256 sv_catpvs(PL_linestr,";");
4258 SETERRNO(0,SS_NORMAL);
4259 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4262 sv_setpvs(PL_linestr,"");
4263 if (PL_preambleav) {
4264 SV **svp = AvARRAY(PL_preambleav);
4265 SV **const end = svp + AvFILLp(PL_preambleav);
4267 sv_catsv(PL_linestr, *svp);
4269 sv_catpvs(PL_linestr, ";");
4271 sv_free(MUTABLE_SV(PL_preambleav));
4272 PL_preambleav = NULL;
4275 sv_catpvs(PL_linestr,
4276 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4277 if (PL_minus_n || PL_minus_p) {
4278 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4280 sv_catpvs(PL_linestr,"chomp;");
4283 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4284 || *PL_splitstr == '"')
4285 && strchr(PL_splitstr + 1, *PL_splitstr))
4286 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4288 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4289 bytes can be used as quoting characters. :-) */
4290 const char *splits = PL_splitstr;
4291 sv_catpvs(PL_linestr, "our @F=split(q\0");
4294 if (*splits == '\\')
4295 sv_catpvn(PL_linestr, splits, 1);
4296 sv_catpvn(PL_linestr, splits, 1);
4297 } while (*splits++);
4298 /* This loop will embed the trailing NUL of
4299 PL_linestr as the last thing it does before
4301 sv_catpvs(PL_linestr, ");");
4305 sv_catpvs(PL_linestr,"our @F=split(' ');");
4308 sv_catpvs(PL_linestr, "\n");
4309 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4310 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4311 PL_last_lop = PL_last_uni = NULL;
4312 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4313 update_debugger_info(PL_linestr, NULL, 0);
4318 bof = PL_rsfp ? TRUE : FALSE;
4321 fake_eof = LEX_FAKE_EOF;
4323 PL_bufptr = PL_bufend;
4324 if (!lex_next_chunk(fake_eof)) {
4326 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4330 PL_realtokenstart = -1;
4333 /* If it looks like the start of a BOM or raw UTF-16,
4334 * check if it in fact is. */
4340 #ifdef PERLIO_IS_STDIO
4341 # ifdef __GNU_LIBRARY__
4342 # if __GNU_LIBRARY__ == 1 /* Linux glibc5 */
4343 # define FTELL_FOR_PIPE_IS_BROKEN
4347 # if __GLIBC__ == 1 /* maybe some glibc5 release had it like this? */
4348 # define FTELL_FOR_PIPE_IS_BROKEN
4353 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
4355 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4356 s = swallow_bom((U8*)s);
4360 /* Incest with pod. */
4363 sv_catsv(PL_thiswhite, PL_linestr);
4365 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4366 sv_setpvs(PL_linestr, "");
4367 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4368 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4369 PL_last_lop = PL_last_uni = NULL;
4370 PL_doextract = FALSE;
4374 } while (PL_doextract);
4375 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4376 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4377 update_debugger_info(PL_linestr, NULL, 0);
4378 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4379 PL_last_lop = PL_last_uni = NULL;
4380 if (CopLINE(PL_curcop) == 1) {
4381 while (s < PL_bufend && isSPACE(*s))
4383 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4387 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4391 if (*s == '#' && *(s+1) == '!')
4393 #ifdef ALTERNATE_SHEBANG
4395 static char const as[] = ALTERNATE_SHEBANG;
4396 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4397 d = s + (sizeof(as) - 1);
4399 #endif /* ALTERNATE_SHEBANG */
4408 while (*d && !isSPACE(*d))
4412 #ifdef ARG_ZERO_IS_SCRIPT
4413 if (ipathend > ipath) {
4415 * HP-UX (at least) sets argv[0] to the script name,
4416 * which makes $^X incorrect. And Digital UNIX and Linux,
4417 * at least, set argv[0] to the basename of the Perl
4418 * interpreter. So, having found "#!", we'll set it right.
4420 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4422 assert(SvPOK(x) || SvGMAGICAL(x));
4423 if (sv_eq(x, CopFILESV(PL_curcop))) {
4424 sv_setpvn(x, ipath, ipathend - ipath);
4430 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4431 const char * const lstart = SvPV_const(x,llen);
4433 bstart += blen - llen;
4434 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4435 sv_setpvn(x, ipath, ipathend - ipath);
4440 TAINT_NOT; /* $^X is always tainted, but that's OK */
4442 #endif /* ARG_ZERO_IS_SCRIPT */
4447 d = instr(s,"perl -");
4449 d = instr(s,"perl");
4451 /* avoid getting into infinite loops when shebang
4452 * line contains "Perl" rather than "perl" */
4454 for (d = ipathend-4; d >= ipath; --d) {
4455 if ((*d == 'p' || *d == 'P')
4456 && !ibcmp(d, "perl", 4))
4466 #ifdef ALTERNATE_SHEBANG
4468 * If the ALTERNATE_SHEBANG on this system starts with a
4469 * character that can be part of a Perl expression, then if
4470 * we see it but not "perl", we're probably looking at the
4471 * start of Perl code, not a request to hand off to some
4472 * other interpreter. Similarly, if "perl" is there, but
4473 * not in the first 'word' of the line, we assume the line
4474 * contains the start of the Perl program.
4476 if (d && *s != '#') {
4477 const char *c = ipath;
4478 while (*c && !strchr("; \t\r\n\f\v#", *c))
4481 d = NULL; /* "perl" not in first word; ignore */
4483 *s = '#'; /* Don't try to parse shebang line */
4485 #endif /* ALTERNATE_SHEBANG */
4490 !instr(s,"indir") &&
4491 instr(PL_origargv[0],"perl"))
4498 while (s < PL_bufend && isSPACE(*s))
4500 if (s < PL_bufend) {
4501 Newx(newargv,PL_origargc+3,char*);
4503 while (s < PL_bufend && !isSPACE(*s))
4506 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4509 newargv = PL_origargv;
4512 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4514 Perl_croak(aTHX_ "Can't exec %s", ipath);
4517 while (*d && !isSPACE(*d))
4519 while (SPACE_OR_TAB(*d))
4523 const bool switches_done = PL_doswitches;
4524 const U32 oldpdb = PL_perldb;
4525 const bool oldn = PL_minus_n;
4526 const bool oldp = PL_minus_p;
4530 bool baduni = FALSE;
4532 const char *d2 = d1 + 1;
4533 if (parse_unicode_opts((const char **)&d2)
4537 if (baduni || *d1 == 'M' || *d1 == 'm') {
4538 const char * const m = d1;
4539 while (*d1 && !isSPACE(*d1))
4541 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4544 d1 = moreswitches(d1);
4546 if (PL_doswitches && !switches_done) {
4547 int argc = PL_origargc;
4548 char **argv = PL_origargv;
4551 } while (argc && argv[0][0] == '-' && argv[0][1]);
4552 init_argv_symbols(argc,argv);
4554 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4555 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4556 /* if we have already added "LINE: while (<>) {",
4557 we must not do it again */
4559 sv_setpvs(PL_linestr, "");
4560 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4561 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4562 PL_last_lop = PL_last_uni = NULL;
4563 PL_preambled = FALSE;
4564 if (PERLDB_LINE || PERLDB_SAVESRC)
4565 (void)gv_fetchfile(PL_origfilename);
4572 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4574 PL_lex_state = LEX_FORMLINE;
4579 #ifdef PERL_STRICT_CR
4580 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4582 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4584 case ' ': case '\t': case '\f': case 013:
4586 PL_realtokenstart = -1;
4588 PL_thiswhite = newSVpvs("");
4589 sv_catpvn(PL_thiswhite, s, 1);
4596 PL_realtokenstart = -1;
4600 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4601 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4602 /* handle eval qq[#line 1 "foo"\n ...] */
4603 CopLINE_dec(PL_curcop);
4606 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4608 if (!PL_in_eval || PL_rsfp)
4613 while (d < PL_bufend && *d != '\n')
4617 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4618 Perl_croak(aTHX_ "panic: input overflow");
4621 PL_thiswhite = newSVpvn(s, d - s);
4626 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4628 PL_lex_state = LEX_FORMLINE;
4634 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4635 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4638 TOKEN(PEG); /* make sure any #! line is accessible */
4643 /* if (PL_madskills && PL_lex_formbrack) { */
4645 while (d < PL_bufend && *d != '\n')
4649 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4650 Perl_croak(aTHX_ "panic: input overflow");
4651 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4653 PL_thiswhite = newSVpvs("");
4654 if (CopLINE(PL_curcop) == 1) {
4655 sv_setpvs(PL_thiswhite, "");
4658 sv_catpvn(PL_thiswhite, s, d - s);
4672 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4680 while (s < PL_bufend && SPACE_OR_TAB(*s))
4683 if (strnEQ(s,"=>",2)) {
4684 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4685 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4686 OPERATOR('-'); /* unary minus */
4688 PL_last_uni = PL_oldbufptr;
4690 case 'r': ftst = OP_FTEREAD; break;
4691 case 'w': ftst = OP_FTEWRITE; break;
4692 case 'x': ftst = OP_FTEEXEC; break;
4693 case 'o': ftst = OP_FTEOWNED; break;
4694 case 'R': ftst = OP_FTRREAD; break;
4695 case 'W': ftst = OP_FTRWRITE; break;
4696 case 'X': ftst = OP_FTREXEC; break;
4697 case 'O': ftst = OP_FTROWNED; break;
4698 case 'e': ftst = OP_FTIS; break;
4699 case 'z': ftst = OP_FTZERO; break;
4700 case 's': ftst = OP_FTSIZE; break;
4701 case 'f': ftst = OP_FTFILE; break;
4702 case 'd': ftst = OP_FTDIR; break;
4703 case 'l': ftst = OP_FTLINK; break;
4704 case 'p': ftst = OP_FTPIPE; break;
4705 case 'S': ftst = OP_FTSOCK; break;
4706 case 'u': ftst = OP_FTSUID; break;
4707 case 'g': ftst = OP_FTSGID; break;
4708 case 'k': ftst = OP_FTSVTX; break;
4709 case 'b': ftst = OP_FTBLK; break;
4710 case 'c': ftst = OP_FTCHR; break;
4711 case 't': ftst = OP_FTTTY; break;
4712 case 'T': ftst = OP_FTTEXT; break;
4713 case 'B': ftst = OP_FTBINARY; break;
4714 case 'M': case 'A': case 'C':
4715 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4717 case 'M': ftst = OP_FTMTIME; break;
4718 case 'A': ftst = OP_FTATIME; break;
4719 case 'C': ftst = OP_FTCTIME; break;
4727 PL_last_lop_op = (OPCODE)ftst;
4728 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4729 "### Saw file test %c\n", (int)tmp);
4734 /* Assume it was a minus followed by a one-letter named
4735 * subroutine call (or a -bareword), then. */
4736 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4737 "### '-%c' looked like a file test but was not\n",
4744 const char tmp = *s++;
4747 if (PL_expect == XOPERATOR)
4752 else if (*s == '>') {
4755 if (isIDFIRST_lazy_if(s,UTF)) {
4756 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
4764 if (PL_expect == XOPERATOR)
4767 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4769 OPERATOR('-'); /* unary minus */
4775 const char tmp = *s++;
4778 if (PL_expect == XOPERATOR)
4783 if (PL_expect == XOPERATOR)
4786 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
4793 if (PL_expect != XOPERATOR) {
4794 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
4795 PL_expect = XOPERATOR;
4796 force_ident(PL_tokenbuf, '*');
4809 if (PL_expect == XOPERATOR) {
4813 PL_tokenbuf[0] = '%';
4814 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
4815 sizeof PL_tokenbuf - 1, FALSE);
4816 if (!PL_tokenbuf[1]) {
4819 PL_pending_ident = '%';
4828 const char tmp = *s++;
4833 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
4840 const char tmp = *s++;
4846 goto just_a_word_zero_gv;
4849 switch (PL_expect) {
4855 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
4857 PL_bufptr = s; /* update in case we back off */
4859 deprecate(":= for an empty attribute list");
4866 PL_expect = XTERMBLOCK;
4869 stuffstart = s - SvPVX(PL_linestr) - 1;
4873 while (isIDFIRST_lazy_if(s,UTF)) {
4876 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
4877 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
4878 if (tmp < 0) tmp = -tmp;
4893 sv = newSVpvn(s, len);
4895 d = scan_str(d,TRUE,TRUE);
4897 /* MUST advance bufptr here to avoid bogus
4898 "at end of line" context messages from yyerror().
4900 PL_bufptr = s + len;
4901 yyerror("Unterminated attribute parameter in attribute list");
4905 return REPORT(0); /* EOF indicator */
4909 sv_catsv(sv, PL_lex_stuff);
4910 attrs = append_elem(OP_LIST, attrs,
4911 newSVOP(OP_CONST, 0, sv));
4912 SvREFCNT_dec(PL_lex_stuff);
4913 PL_lex_stuff = NULL;
4916 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
4918 if (PL_in_my == KEY_our) {
4919 deprecate(":unique");
4922 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
4925 /* NOTE: any CV attrs applied here need to be part of
4926 the CVf_BUILTIN_ATTRS define in cv.h! */
4927 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
4929 CvLVALUE_on(PL_compcv);
4931 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
4933 deprecate(":locked");
4935 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
4937 CvMETHOD_on(PL_compcv);
4939 /* After we've set the flags, it could be argued that
4940 we don't need to do the attributes.pm-based setting
4941 process, and shouldn't bother appending recognized
4942 flags. To experiment with that, uncomment the
4943 following "else". (Note that's already been
4944 uncommented. That keeps the above-applied built-in
4945 attributes from being intercepted (and possibly
4946 rejected) by a package's attribute routines, but is
4947 justified by the performance win for the common case
4948 of applying only built-in attributes.) */
4950 attrs = append_elem(OP_LIST, attrs,
4951 newSVOP(OP_CONST, 0,
4955 if (*s == ':' && s[1] != ':')
4958 break; /* require real whitespace or :'s */
4959 /* XXX losing whitespace on sequential attributes here */
4963 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
4964 if (*s != ';' && *s != '}' && *s != tmp
4965 && (tmp != '=' || *s != ')')) {
4966 const char q = ((*s == '\'') ? '"' : '\'');
4967 /* If here for an expression, and parsed no attrs, back
4969 if (tmp == '=' && !attrs) {
4973 /* MUST advance bufptr here to avoid bogus "at end of line"
4974 context messages from yyerror().
4977 yyerror( (const char *)
4979 ? Perl_form(aTHX_ "Invalid separator character "
4980 "%c%c%c in attribute list", q, *s, q)
4981 : "Unterminated attribute list" ) );
4989 start_force(PL_curforce);
4990 NEXTVAL_NEXTTOKE.opval = attrs;
4991 CURMAD('_', PL_nextwhite);
4996 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
4997 (s - SvPVX(PL_linestr)) - stuffstart);
5005 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5006 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5014 const char tmp = *s++;
5019 const char tmp = *s++;
5027 if (PL_lex_brackets <= 0)
5028 yyerror("Unmatched right square bracket");
5031 if (PL_lex_state == LEX_INTERPNORMAL) {
5032 if (PL_lex_brackets == 0) {
5033 if (*s == '-' && s[1] == '>')
5034 PL_lex_state = LEX_INTERPENDMAYBE;
5035 else if (*s != '[' && *s != '{')
5036 PL_lex_state = LEX_INTERPEND;
5043 if (PL_lex_brackets > 100) {
5044 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5046 switch (PL_expect) {
5048 if (PL_lex_formbrack) {
5052 if (PL_oldoldbufptr == PL_last_lop)
5053 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5055 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5056 OPERATOR(HASHBRACK);
5058 while (s < PL_bufend && SPACE_OR_TAB(*s))
5061 PL_tokenbuf[0] = '\0';
5062 if (d < PL_bufend && *d == '-') {
5063 PL_tokenbuf[0] = '-';
5065 while (d < PL_bufend && SPACE_OR_TAB(*d))
5068 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5069 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5071 while (d < PL_bufend && SPACE_OR_TAB(*d))
5074 const char minus = (PL_tokenbuf[0] == '-');
5075 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5083 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5088 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5093 if (PL_oldoldbufptr == PL_last_lop)
5094 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5096 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5099 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5101 /* This hack is to get the ${} in the message. */
5103 yyerror("syntax error");
5106 OPERATOR(HASHBRACK);
5108 /* This hack serves to disambiguate a pair of curlies
5109 * as being a block or an anon hash. Normally, expectation
5110 * determines that, but in cases where we're not in a
5111 * position to expect anything in particular (like inside
5112 * eval"") we have to resolve the ambiguity. This code
5113 * covers the case where the first term in the curlies is a
5114 * quoted string. Most other cases need to be explicitly
5115 * disambiguated by prepending a "+" before the opening
5116 * curly in order to force resolution as an anon hash.
5118 * XXX should probably propagate the outer expectation
5119 * into eval"" to rely less on this hack, but that could
5120 * potentially break current behavior of eval"".
5124 if (*s == '\'' || *s == '"' || *s == '`') {
5125 /* common case: get past first string, handling escapes */
5126 for (t++; t < PL_bufend && *t != *s;)
5127 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5131 else if (*s == 'q') {
5134 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5137 /* skip q//-like construct */
5139 char open, close, term;
5142 while (t < PL_bufend && isSPACE(*t))
5144 /* check for q => */
5145 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5146 OPERATOR(HASHBRACK);
5150 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5154 for (t++; t < PL_bufend; t++) {
5155 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5157 else if (*t == open)
5161 for (t++; t < PL_bufend; t++) {
5162 if (*t == '\\' && t+1 < PL_bufend)
5164 else if (*t == close && --brackets <= 0)
5166 else if (*t == open)
5173 /* skip plain q word */
5174 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5177 else if (isALNUM_lazy_if(t,UTF)) {
5179 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5182 while (t < PL_bufend && isSPACE(*t))
5184 /* if comma follows first term, call it an anon hash */
5185 /* XXX it could be a comma expression with loop modifiers */
5186 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5187 || (*t == '=' && t[1] == '>')))
5188 OPERATOR(HASHBRACK);
5189 if (PL_expect == XREF)
5192 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5198 pl_yylval.ival = CopLINE(PL_curcop);
5199 if (isSPACE(*s) || *s == '#')
5200 PL_copline = NOLINE; /* invalidate current command line number */
5205 if (PL_lex_brackets <= 0)
5206 yyerror("Unmatched right curly bracket");
5208 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5209 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5210 PL_lex_formbrack = 0;
5211 if (PL_lex_state == LEX_INTERPNORMAL) {
5212 if (PL_lex_brackets == 0) {
5213 if (PL_expect & XFAKEBRACK) {
5214 PL_expect &= XENUMMASK;
5215 PL_lex_state = LEX_INTERPEND;
5220 PL_thiswhite = newSVpvs("");
5221 sv_catpvs(PL_thiswhite,"}");
5224 return yylex(); /* ignore fake brackets */
5226 if (*s == '-' && s[1] == '>')
5227 PL_lex_state = LEX_INTERPENDMAYBE;
5228 else if (*s != '[' && *s != '{')
5229 PL_lex_state = LEX_INTERPEND;
5232 if (PL_expect & XFAKEBRACK) {
5233 PL_expect &= XENUMMASK;
5235 return yylex(); /* ignore fake brackets */
5237 start_force(PL_curforce);
5239 curmad('X', newSVpvn(s-1,1));
5240 CURMAD('_', PL_thiswhite);
5245 PL_thistoken = newSVpvs("");
5253 if (PL_expect == XOPERATOR) {
5254 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5255 && isIDFIRST_lazy_if(s,UTF))
5257 CopLINE_dec(PL_curcop);
5258 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5259 CopLINE_inc(PL_curcop);
5264 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5266 PL_expect = XOPERATOR;
5267 force_ident(PL_tokenbuf, '&');
5271 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5283 const char tmp = *s++;
5290 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5291 && strchr("+-*/%.^&|<",tmp))
5292 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5293 "Reversed %c= operator",(int)tmp);
5295 if (PL_expect == XSTATE && isALPHA(tmp) &&
5296 (s == PL_linestart+1 || s[-2] == '\n') )
5298 if (PL_in_eval && !PL_rsfp) {
5303 if (strnEQ(s,"=cut",4)) {
5319 PL_thiswhite = newSVpvs("");
5320 sv_catpvn(PL_thiswhite, PL_linestart,
5321 PL_bufend - PL_linestart);
5325 PL_doextract = TRUE;
5329 if (PL_lex_brackets < PL_lex_formbrack) {
5331 #ifdef PERL_STRICT_CR
5332 while (SPACE_OR_TAB(*t))
5334 while (SPACE_OR_TAB(*t) || *t == '\r')
5337 if (*t == '\n' || *t == '#') {
5348 const char tmp = *s++;
5350 /* was this !=~ where !~ was meant?
5351 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5353 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5354 const char *t = s+1;
5356 while (t < PL_bufend && isSPACE(*t))
5359 if (*t == '/' || *t == '?' ||
5360 ((*t == 'm' || *t == 's' || *t == 'y')
5361 && !isALNUM(t[1])) ||
5362 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5363 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5364 "!=~ should be !~");
5374 if (PL_expect != XOPERATOR) {
5375 if (s[1] != '<' && !strchr(s,'>'))
5378 s = scan_heredoc(s);
5380 s = scan_inputsymbol(s);
5381 TERM(sublex_start());
5387 SHop(OP_LEFT_SHIFT);
5401 const char tmp = *s++;
5403 SHop(OP_RIGHT_SHIFT);
5404 else if (tmp == '=')
5413 if (PL_expect == XOPERATOR) {
5414 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5415 return deprecate_commaless_var_list();
5419 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
5420 PL_tokenbuf[0] = '@';
5421 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
5422 sizeof PL_tokenbuf - 1, FALSE);
5423 if (PL_expect == XOPERATOR)
5424 no_op("Array length", s);
5425 if (!PL_tokenbuf[1])
5427 PL_expect = XOPERATOR;
5428 PL_pending_ident = '#';
5432 PL_tokenbuf[0] = '$';
5433 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5434 sizeof PL_tokenbuf - 1, FALSE);
5435 if (PL_expect == XOPERATOR)
5437 if (!PL_tokenbuf[1]) {
5439 yyerror("Final $ should be \\$ or $name");
5443 /* This kludge not intended to be bulletproof. */
5444 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
5445 pl_yylval.opval = newSVOP(OP_CONST, 0,
5446 newSViv(CopARYBASE_get(&PL_compiling)));
5447 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
5453 const char tmp = *s;
5454 if (PL_lex_state == LEX_NORMAL)
5457 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
5458 && intuit_more(s)) {
5460 PL_tokenbuf[0] = '@';
5461 if (ckWARN(WARN_SYNTAX)) {
5464 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
5467 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5468 while (t < PL_bufend && *t != ']')
5470 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5471 "Multidimensional syntax %.*s not supported",
5472 (int)((t - PL_bufptr) + 1), PL_bufptr);
5476 else if (*s == '{') {
5478 PL_tokenbuf[0] = '%';
5479 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
5480 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
5482 char tmpbuf[sizeof PL_tokenbuf];
5485 } while (isSPACE(*t));
5486 if (isIDFIRST_lazy_if(t,UTF)) {
5488 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
5492 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
5493 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5494 "You need to quote \"%s\"",
5501 PL_expect = XOPERATOR;
5502 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
5503 const bool islop = (PL_last_lop == PL_oldoldbufptr);
5504 if (!islop || PL_last_lop_op == OP_GREPSTART)
5505 PL_expect = XOPERATOR;
5506 else if (strchr("$@\"'`q", *s))
5507 PL_expect = XTERM; /* e.g. print $fh "foo" */
5508 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
5509 PL_expect = XTERM; /* e.g. print $fh &sub */
5510 else if (isIDFIRST_lazy_if(s,UTF)) {
5511 char tmpbuf[sizeof PL_tokenbuf];
5513 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5514 if ((t2 = keyword(tmpbuf, len, 0))) {
5515 /* binary operators exclude handle interpretations */
5527 PL_expect = XTERM; /* e.g. print $fh length() */
5532 PL_expect = XTERM; /* e.g. print $fh subr() */
5535 else if (isDIGIT(*s))
5536 PL_expect = XTERM; /* e.g. print $fh 3 */
5537 else if (*s == '.' && isDIGIT(s[1]))
5538 PL_expect = XTERM; /* e.g. print $fh .3 */
5539 else if ((*s == '?' || *s == '-' || *s == '+')
5540 && !isSPACE(s[1]) && s[1] != '=')
5541 PL_expect = XTERM; /* e.g. print $fh -1 */
5542 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5544 PL_expect = XTERM; /* e.g. print $fh /.../
5545 XXX except DORDOR operator
5547 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5549 PL_expect = XTERM; /* print $fh <<"EOF" */
5552 PL_pending_ident = '$';
5556 if (PL_expect == XOPERATOR)
5558 PL_tokenbuf[0] = '@';
5559 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5560 if (!PL_tokenbuf[1]) {
5563 if (PL_lex_state == LEX_NORMAL)
5565 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5567 PL_tokenbuf[0] = '%';
5569 /* Warn about @ where they meant $. */
5570 if (*s == '[' || *s == '{') {
5571 if (ckWARN(WARN_SYNTAX)) {
5572 const char *t = s + 1;
5573 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5575 if (*t == '}' || *t == ']') {
5577 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5578 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5579 "Scalar value %.*s better written as $%.*s",
5580 (int)(t-PL_bufptr), PL_bufptr,
5581 (int)(t-PL_bufptr-1), PL_bufptr+1);
5586 PL_pending_ident = '@';
5589 case '/': /* may be division, defined-or, or pattern */
5590 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5594 case '?': /* may either be conditional or pattern */
5595 if (PL_expect == XOPERATOR) {
5603 /* A // operator. */
5613 /* Disable warning on "study /blah/" */
5614 if (PL_oldoldbufptr == PL_last_uni
5615 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5616 || memNE(PL_last_uni, "study", 5)
5617 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5620 s = scan_pat(s,OP_MATCH);
5621 TERM(sublex_start());
5625 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5626 #ifdef PERL_STRICT_CR
5629 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5631 && (s == PL_linestart || s[-1] == '\n') )
5633 PL_lex_formbrack = 0;
5637 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5641 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5647 pl_yylval.ival = OPf_SPECIAL;
5653 if (PL_expect != XOPERATOR)
5658 case '0': case '1': case '2': case '3': case '4':
5659 case '5': case '6': case '7': case '8': case '9':
5660 s = scan_num(s, &pl_yylval);
5661 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5662 if (PL_expect == XOPERATOR)
5667 s = scan_str(s,!!PL_madskills,FALSE);
5668 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5669 if (PL_expect == XOPERATOR) {
5670 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5671 return deprecate_commaless_var_list();
5678 pl_yylval.ival = OP_CONST;
5679 TERM(sublex_start());
5682 s = scan_str(s,!!PL_madskills,FALSE);
5683 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5684 if (PL_expect == XOPERATOR) {
5685 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5686 return deprecate_commaless_var_list();
5693 pl_yylval.ival = OP_CONST;
5694 /* FIXME. I think that this can be const if char *d is replaced by
5695 more localised variables. */
5696 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5697 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5698 pl_yylval.ival = OP_STRINGIFY;
5702 TERM(sublex_start());
5705 s = scan_str(s,!!PL_madskills,FALSE);
5706 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5707 if (PL_expect == XOPERATOR)
5708 no_op("Backticks",s);
5711 readpipe_override();
5712 TERM(sublex_start());
5716 if (PL_lex_inwhat && isDIGIT(*s))
5717 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5719 if (PL_expect == XOPERATOR)
5720 no_op("Backslash",s);
5724 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5725 char *start = s + 2;
5726 while (isDIGIT(*start) || *start == '_')
5728 if (*start == '.' && isDIGIT(start[1])) {
5729 s = scan_num(s, &pl_yylval);
5732 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5733 else if (!isALPHA(*start) && (PL_expect == XTERM
5734 || PL_expect == XREF || PL_expect == XSTATE
5735 || PL_expect == XTERMORDORDOR)) {
5736 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5738 s = scan_num(s, &pl_yylval);
5745 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
5788 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5790 /* Some keywords can be followed by any delimiter, including ':' */
5791 anydelim = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
5792 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
5793 (PL_tokenbuf[0] == 'q' &&
5794 strchr("qwxr", PL_tokenbuf[1])))));
5796 /* x::* is just a word, unless x is "CORE" */
5797 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
5801 while (d < PL_bufend && isSPACE(*d))
5802 d++; /* no comments skipped here, or s### is misparsed */
5804 /* Is this a word before a => operator? */
5805 if (*d == '=' && d[1] == '>') {
5808 = (OP*)newSVOP(OP_CONST, 0,
5809 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
5810 pl_yylval.opval->op_private = OPpCONST_BARE;
5814 /* Check for plugged-in keyword */
5818 char *saved_bufptr = PL_bufptr;
5820 result = CALL_FPTR(PL_keyword_plugin)(aTHX_ PL_tokenbuf, len, &o);
5822 if (result == KEYWORD_PLUGIN_DECLINE) {
5823 /* not a plugged-in keyword */
5824 PL_bufptr = saved_bufptr;
5825 } else if (result == KEYWORD_PLUGIN_STMT) {
5826 pl_yylval.opval = o;
5829 return REPORT(PLUGSTMT);
5830 } else if (result == KEYWORD_PLUGIN_EXPR) {
5831 pl_yylval.opval = o;
5833 PL_expect = XOPERATOR;
5834 return REPORT(PLUGEXPR);
5836 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
5841 /* Check for built-in keyword */
5842 tmp = keyword(PL_tokenbuf, len, 0);
5844 /* Is this a label? */
5845 if (!anydelim && PL_expect == XSTATE
5846 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
5848 Perl_croak(aTHX_ "Can't use keyword '%s' as a label", PL_tokenbuf);
5850 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
5855 if (tmp < 0) { /* second-class keyword? */
5856 GV *ogv = NULL; /* override (winner) */
5857 GV *hgv = NULL; /* hidden (loser) */
5858 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
5860 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
5863 if (GvIMPORTED_CV(gv))
5865 else if (! CvMETHOD(cv))
5869 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
5870 (gv = *gvp) && isGV_with_GP(gv) &&
5871 GvCVu(gv) && GvIMPORTED_CV(gv))
5878 tmp = 0; /* overridden by import or by GLOBAL */
5881 && -tmp==KEY_lock /* XXX generalizable kludge */
5884 tmp = 0; /* any sub overrides "weak" keyword */
5886 else { /* no override */
5888 if (tmp == KEY_dump) {
5889 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
5890 "dump() better written as CORE::dump()");
5894 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
5895 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
5896 "Ambiguous call resolved as CORE::%s(), %s",
5897 GvENAME(hgv), "qualify as such or use &");
5904 default: /* not a keyword */
5905 /* Trade off - by using this evil construction we can pull the
5906 variable gv into the block labelled keylookup. If not, then
5907 we have to give it function scope so that the goto from the
5908 earlier ':' case doesn't bypass the initialisation. */
5910 just_a_word_zero_gv:
5918 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
5922 SV *nextPL_nextwhite = 0;
5926 /* Get the rest if it looks like a package qualifier */
5928 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
5930 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
5933 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
5934 *s == '\'' ? "'" : "::");
5939 if (PL_expect == XOPERATOR) {
5940 if (PL_bufptr == PL_linestart) {
5941 CopLINE_dec(PL_curcop);
5942 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5943 CopLINE_inc(PL_curcop);
5946 no_op("Bareword",s);
5949 /* Look for a subroutine with this name in current package,
5950 unless name is "Foo::", in which case Foo is a bearword
5951 (and a package name). */
5953 if (len > 2 && !PL_madskills &&
5954 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
5956 if (ckWARN(WARN_BAREWORD)
5957 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
5958 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
5959 "Bareword \"%s\" refers to nonexistent package",
5962 PL_tokenbuf[len] = '\0';
5968 /* Mustn't actually add anything to a symbol table.
5969 But also don't want to "initialise" any placeholder
5970 constants that might already be there into full
5971 blown PVGVs with attached PVCV. */
5972 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
5973 GV_NOADD_NOINIT, SVt_PVCV);
5978 /* if we saw a global override before, get the right name */
5981 sv = newSVpvs("CORE::GLOBAL::");
5982 sv_catpv(sv,PL_tokenbuf);
5985 /* If len is 0, newSVpv does strlen(), which is correct.
5986 If len is non-zero, then it will be the true length,
5987 and so the scalar will be created correctly. */
5988 sv = newSVpv(PL_tokenbuf,len);
5991 if (PL_madskills && !PL_thistoken) {
5992 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
5993 PL_thistoken = newSVpvn(start,s - start);
5994 PL_realtokenstart = s - SvPVX(PL_linestr);
5998 /* Presume this is going to be a bareword of some sort. */
6001 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6002 pl_yylval.opval->op_private = OPpCONST_BARE;
6003 /* UTF-8 package name? */
6004 if (UTF && !IN_BYTES &&
6005 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
6008 /* And if "Foo::", then that's what it certainly is. */
6015 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc(sv));
6016 const_op->op_private = OPpCONST_BARE;
6017 rv2cv_op = newCVREF(0, const_op);
6019 if (rv2cv_op->op_type == OP_RV2CV &&
6020 (rv2cv_op->op_flags & OPf_KIDS)) {
6021 OP *rv_op = cUNOPx(rv2cv_op)->op_first;
6022 switch (rv_op->op_type) {
6024 SV *sv = cSVOPx_sv(rv_op);
6025 if (SvROK(sv) && SvTYPE(SvRV(sv)) == SVt_PVCV)
6029 GV *gv = cGVOPx_gv(rv_op);
6030 CV *maybe_cv = GvCVu(gv);
6031 if (maybe_cv && SvTYPE((SV*)maybe_cv) == SVt_PVCV)
6037 /* See if it's the indirect object for a list operator. */
6039 if (PL_oldoldbufptr &&
6040 PL_oldoldbufptr < PL_bufptr &&
6041 (PL_oldoldbufptr == PL_last_lop
6042 || PL_oldoldbufptr == PL_last_uni) &&
6043 /* NO SKIPSPACE BEFORE HERE! */
6044 (PL_expect == XREF ||
6045 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6047 bool immediate_paren = *s == '(';
6049 /* (Now we can afford to cross potential line boundary.) */
6050 s = SKIPSPACE2(s,nextPL_nextwhite);
6052 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6055 /* Two barewords in a row may indicate method call. */
6057 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6058 (tmp = intuit_method(s, gv, cv))) {
6063 /* If not a declared subroutine, it's an indirect object. */
6064 /* (But it's an indir obj regardless for sort.) */
6065 /* Also, if "_" follows a filetest operator, it's a bareword */
6068 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6070 (PL_last_lop_op != OP_MAPSTART &&
6071 PL_last_lop_op != OP_GREPSTART))))
6072 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6073 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6076 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6081 PL_expect = XOPERATOR;
6084 s = SKIPSPACE2(s,nextPL_nextwhite);
6085 PL_nextwhite = nextPL_nextwhite;
6090 /* Is this a word before a => operator? */
6091 if (*s == '=' && s[1] == '>' && !pkgname) {
6094 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6095 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6096 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6100 /* If followed by a paren, it's certainly a subroutine. */
6105 while (SPACE_OR_TAB(*d))
6107 if (*d == ')' && (sv = cv_const_sv(cv))) {
6114 PL_nextwhite = PL_thiswhite;
6117 start_force(PL_curforce);
6119 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6120 PL_expect = XOPERATOR;
6123 PL_nextwhite = nextPL_nextwhite;
6124 curmad('X', PL_thistoken);
6125 PL_thistoken = newSVpvs("");
6134 /* If followed by var or block, call it a method (unless sub) */
6136 if ((*s == '$' || *s == '{') && !cv) {
6138 PL_last_lop = PL_oldbufptr;
6139 PL_last_lop_op = OP_METHOD;
6143 /* If followed by a bareword, see if it looks like indir obj. */
6146 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6147 && (tmp = intuit_method(s, gv, cv))) {
6152 /* Not a method, so call it a subroutine (if defined) */
6155 if (lastchar == '-')
6156 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6157 "Ambiguous use of -%s resolved as -&%s()",
6158 PL_tokenbuf, PL_tokenbuf);
6159 /* Check for a constant sub */
6160 if ((sv = cv_const_sv(cv))) {
6163 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6164 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6165 pl_yylval.opval->op_private = 0;
6169 op_free(pl_yylval.opval);
6170 pl_yylval.opval = rv2cv_op;
6171 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6172 PL_last_lop = PL_oldbufptr;
6173 PL_last_lop_op = OP_ENTERSUB;
6174 /* Is there a prototype? */
6182 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
6185 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
6187 while (*proto == ';')
6189 if (*proto == '&' && *s == '{') {
6191 sv_setpvs(PL_subname, "__ANON__");
6193 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6200 PL_nextwhite = PL_thiswhite;
6203 start_force(PL_curforce);
6204 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6207 PL_nextwhite = nextPL_nextwhite;
6208 curmad('X', PL_thistoken);
6209 PL_thistoken = newSVpvs("");
6216 /* Guess harder when madskills require "best effort". */
6217 if (PL_madskills && (!gv || !GvCVu(gv))) {
6218 int probable_sub = 0;
6219 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6221 else if (isALPHA(*s)) {
6225 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6226 if (!keyword(tmpbuf, tmplen, 0))
6229 while (d < PL_bufend && isSPACE(*d))
6231 if (*d == '=' && d[1] == '>')
6236 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
6237 op_free(pl_yylval.opval);
6238 pl_yylval.opval = rv2cv_op;
6239 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6240 PL_last_lop = PL_oldbufptr;
6241 PL_last_lop_op = OP_ENTERSUB;
6242 PL_nextwhite = PL_thiswhite;
6244 start_force(PL_curforce);
6245 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6247 PL_nextwhite = nextPL_nextwhite;
6248 curmad('X', PL_thistoken);
6249 PL_thistoken = newSVpvs("");
6254 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6261 /* Call it a bare word */
6263 if (PL_hints & HINT_STRICT_SUBS)
6264 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6267 /* after "print" and similar functions (corresponding to
6268 * "F? L" in opcode.pl), whatever wasn't already parsed as
6269 * a filehandle should be subject to "strict subs".
6270 * Likewise for the optional indirect-object argument to system
6271 * or exec, which can't be a bareword */
6272 if ((PL_last_lop_op == OP_PRINT
6273 || PL_last_lop_op == OP_PRTF
6274 || PL_last_lop_op == OP_SAY
6275 || PL_last_lop_op == OP_SYSTEM
6276 || PL_last_lop_op == OP_EXEC)
6277 && (PL_hints & HINT_STRICT_SUBS))
6278 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6279 if (lastchar != '-') {
6280 if (ckWARN(WARN_RESERVED)) {
6284 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
6285 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6293 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
6294 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6295 "Operator or semicolon missing before %c%s",
6296 lastchar, PL_tokenbuf);
6297 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6298 "Ambiguous use of %c resolved as operator %c",
6299 lastchar, lastchar);
6305 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6306 newSVpv(CopFILE(PL_curcop),0));
6310 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6311 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
6314 case KEY___PACKAGE__:
6315 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6317 ? newSVhek(HvNAME_HEK(PL_curstash))
6324 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6325 const char *pname = "main";
6326 if (PL_tokenbuf[2] == 'D')
6327 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
6328 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
6332 GvIOp(gv) = newIO();
6333 IoIFP(GvIOp(gv)) = PL_rsfp;
6334 #if defined(HAS_FCNTL) && defined(F_SETFD)
6336 const int fd = PerlIO_fileno(PL_rsfp);
6337 fcntl(fd,F_SETFD,fd >= 3);
6340 /* Mark this internal pseudo-handle as clean */
6341 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6342 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6343 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6345 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6346 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6347 /* if the script was opened in binmode, we need to revert
6348 * it to text mode for compatibility; but only iff it has CRs
6349 * XXX this is a questionable hack at best. */
6350 if (PL_bufend-PL_bufptr > 2
6351 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6354 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6355 loc = PerlIO_tell(PL_rsfp);
6356 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6359 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6361 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6362 #endif /* NETWARE */
6363 #ifdef PERLIO_IS_STDIO /* really? */
6364 # if defined(__BORLANDC__)
6365 /* XXX see note in do_binmode() */
6366 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
6370 PerlIO_seek(PL_rsfp, loc, 0);
6374 #ifdef PERLIO_LAYERS
6377 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6378 else if (PL_encoding) {
6385 XPUSHs(PL_encoding);
6387 call_method("name", G_SCALAR);
6391 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
6392 Perl_form(aTHX_ ":encoding(%"SVf")",
6401 if (PL_realtokenstart >= 0) {
6402 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6404 PL_endwhite = newSVpvs("");
6405 sv_catsv(PL_endwhite, PL_thiswhite);
6407 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
6408 PL_realtokenstart = -1;
6410 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
6426 if (PL_expect == XSTATE) {
6433 if (*s == ':' && s[1] == ':') {
6436 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6437 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
6438 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
6441 else if (tmp == KEY_require || tmp == KEY_do)
6442 /* that's a way to remember we saw "CORE::" */
6455 LOP(OP_ACCEPT,XTERM);
6461 LOP(OP_ATAN2,XTERM);
6467 LOP(OP_BINMODE,XTERM);
6470 LOP(OP_BLESS,XTERM);
6479 /* When 'use switch' is in effect, continue has a dual
6480 life as a control operator. */
6482 if (!FEATURE_IS_ENABLED("switch"))
6485 /* We have to disambiguate the two senses of
6486 "continue". If the next token is a '{' then
6487 treat it as the start of a continue block;
6488 otherwise treat it as a control operator.
6500 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
6517 if (!PL_cryptseen) {
6518 PL_cryptseen = TRUE;
6522 LOP(OP_CRYPT,XTERM);
6525 LOP(OP_CHMOD,XTERM);
6528 LOP(OP_CHOWN,XTERM);
6531 LOP(OP_CONNECT,XTERM);
6550 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6551 if (orig_keyword == KEY_do) {
6560 PL_hints |= HINT_BLOCK_SCOPE;
6570 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
6571 LOP(OP_DBMOPEN,XTERM);
6577 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6584 pl_yylval.ival = CopLINE(PL_curcop);
6600 PL_expect = (*s == '{') ? XTERMBLOCK : XTERM;
6601 UNIBRACK(OP_ENTEREVAL);
6615 case KEY_endhostent:
6621 case KEY_endservent:
6624 case KEY_endprotoent:
6635 pl_yylval.ival = CopLINE(PL_curcop);
6637 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6640 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6643 if ((PL_bufend - p) >= 3 &&
6644 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6646 else if ((PL_bufend - p) >= 4 &&
6647 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6650 if (isIDFIRST_lazy_if(p,UTF)) {
6651 p = scan_ident(p, PL_bufend,
6652 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6656 Perl_croak(aTHX_ "Missing $ on loop variable");
6658 s = SvPVX(PL_linestr) + soff;
6664 LOP(OP_FORMLINE,XTERM);
6670 LOP(OP_FCNTL,XTERM);
6676 LOP(OP_FLOCK,XTERM);
6685 LOP(OP_GREPSTART, XREF);
6688 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6703 case KEY_getpriority:
6704 LOP(OP_GETPRIORITY,XTERM);
6706 case KEY_getprotobyname:
6709 case KEY_getprotobynumber:
6710 LOP(OP_GPBYNUMBER,XTERM);
6712 case KEY_getprotoent:
6724 case KEY_getpeername:
6725 UNI(OP_GETPEERNAME);
6727 case KEY_gethostbyname:
6730 case KEY_gethostbyaddr:
6731 LOP(OP_GHBYADDR,XTERM);
6733 case KEY_gethostent:
6736 case KEY_getnetbyname:
6739 case KEY_getnetbyaddr:
6740 LOP(OP_GNBYADDR,XTERM);
6745 case KEY_getservbyname:
6746 LOP(OP_GSBYNAME,XTERM);
6748 case KEY_getservbyport:
6749 LOP(OP_GSBYPORT,XTERM);
6751 case KEY_getservent:
6754 case KEY_getsockname:
6755 UNI(OP_GETSOCKNAME);
6757 case KEY_getsockopt:
6758 LOP(OP_GSOCKOPT,XTERM);
6773 pl_yylval.ival = CopLINE(PL_curcop);
6783 pl_yylval.ival = CopLINE(PL_curcop);
6787 LOP(OP_INDEX,XTERM);
6793 LOP(OP_IOCTL,XTERM);
6805 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6837 LOP(OP_LISTEN,XTERM);
6846 s = scan_pat(s,OP_MATCH);
6847 TERM(sublex_start());
6850 LOP(OP_MAPSTART, XREF);
6853 LOP(OP_MKDIR,XTERM);
6856 LOP(OP_MSGCTL,XTERM);
6859 LOP(OP_MSGGET,XTERM);
6862 LOP(OP_MSGRCV,XTERM);
6865 LOP(OP_MSGSND,XTERM);
6870 PL_in_my = (U16)tmp;
6872 if (isIDFIRST_lazy_if(s,UTF)) {
6876 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
6877 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
6879 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
6880 if (!PL_in_my_stash) {
6883 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
6887 if (PL_madskills) { /* just add type to declarator token */
6888 sv_catsv(PL_thistoken, PL_nextwhite);
6890 sv_catpvn(PL_thistoken, start, s - start);
6898 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6905 s = tokenize_use(0, s);
6909 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
6916 if (isIDFIRST_lazy_if(s,UTF)) {
6918 for (d = s; isALNUM_lazy_if(d,UTF);)
6920 for (t=d; isSPACE(*t);)
6922 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
6924 && !(t[0] == '=' && t[1] == '>')
6926 int parms_len = (int)(d-s);
6927 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
6928 "Precedence problem: open %.*s should be open(%.*s)",
6929 parms_len, s, parms_len, s);
6935 pl_yylval.ival = OP_OR;
6945 LOP(OP_OPEN_DIR,XTERM);
6948 checkcomma(s,PL_tokenbuf,"filehandle");
6952 checkcomma(s,PL_tokenbuf,"filehandle");
6971 s = force_word(s,WORD,FALSE,TRUE,FALSE);
6972 s = force_version(s, FALSE);
6976 LOP(OP_PIPE_OP,XTERM);
6979 s = scan_str(s,!!PL_madskills,FALSE);
6982 pl_yylval.ival = OP_CONST;
6983 TERM(sublex_start());
6989 s = scan_str(s,!!PL_madskills,FALSE);
6992 PL_expect = XOPERATOR;
6994 if (SvCUR(PL_lex_stuff)) {
6997 d = SvPV_force(PL_lex_stuff, len);
6999 for (; isSPACE(*d) && len; --len, ++d)
7004 if (!warned && ckWARN(WARN_QW)) {
7005 for (; !isSPACE(*d) && len; --len, ++d) {
7007 Perl_warner(aTHX_ packWARN(WARN_QW),
7008 "Possible attempt to separate words with commas");
7011 else if (*d == '#') {
7012 Perl_warner(aTHX_ packWARN(WARN_QW),
7013 "Possible attempt to put comments in qw() list");
7019 for (; !isSPACE(*d) && len; --len, ++d)
7022 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7023 words = append_elem(OP_LIST, words,
7024 newSVOP(OP_CONST, 0, tokeq(sv)));
7028 start_force(PL_curforce);
7029 NEXTVAL_NEXTTOKE.opval = words;
7034 SvREFCNT_dec(PL_lex_stuff);
7035 PL_lex_stuff = NULL;
7041 s = scan_str(s,!!PL_madskills,FALSE);
7044 pl_yylval.ival = OP_STRINGIFY;
7045 if (SvIVX(PL_lex_stuff) == '\'')
7046 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
7047 TERM(sublex_start());
7050 s = scan_pat(s,OP_QR);
7051 TERM(sublex_start());
7054 s = scan_str(s,!!PL_madskills,FALSE);
7057 readpipe_override();
7058 TERM(sublex_start());
7066 s = force_version(s, FALSE);
7068 else if (*s != 'v' || !isDIGIT(s[1])
7069 || (s = force_version(s, TRUE), *s == 'v'))
7071 *PL_tokenbuf = '\0';
7072 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7073 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7074 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7076 yyerror("<> should be quotes");
7078 if (orig_keyword == KEY_require) {
7086 PL_last_uni = PL_oldbufptr;
7087 PL_last_lop_op = OP_REQUIRE;
7089 return REPORT( (int)REQUIRE );
7095 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7099 LOP(OP_RENAME,XTERM);
7108 LOP(OP_RINDEX,XTERM);
7117 UNIDOR(OP_READLINE);
7120 UNIDOR(OP_BACKTICK);
7129 LOP(OP_REVERSE,XTERM);
7132 UNIDOR(OP_READLINK);
7139 if (pl_yylval.opval)
7140 TERM(sublex_start());
7142 TOKEN(1); /* force error */
7145 checkcomma(s,PL_tokenbuf,"filehandle");
7155 LOP(OP_SELECT,XTERM);
7161 LOP(OP_SEMCTL,XTERM);
7164 LOP(OP_SEMGET,XTERM);
7167 LOP(OP_SEMOP,XTERM);
7173 LOP(OP_SETPGRP,XTERM);
7175 case KEY_setpriority:
7176 LOP(OP_SETPRIORITY,XTERM);
7178 case KEY_sethostent:
7184 case KEY_setservent:
7187 case KEY_setprotoent:
7197 LOP(OP_SEEKDIR,XTERM);
7199 case KEY_setsockopt:
7200 LOP(OP_SSOCKOPT,XTERM);
7206 LOP(OP_SHMCTL,XTERM);
7209 LOP(OP_SHMGET,XTERM);
7212 LOP(OP_SHMREAD,XTERM);
7215 LOP(OP_SHMWRITE,XTERM);
7218 LOP(OP_SHUTDOWN,XTERM);
7227 LOP(OP_SOCKET,XTERM);
7229 case KEY_socketpair:
7230 LOP(OP_SOCKPAIR,XTERM);
7233 checkcomma(s,PL_tokenbuf,"subroutine name");
7235 if (*s == ';' || *s == ')') /* probably a close */
7236 Perl_croak(aTHX_ "sort is now a reserved word");
7238 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7242 LOP(OP_SPLIT,XTERM);
7245 LOP(OP_SPRINTF,XTERM);
7248 LOP(OP_SPLICE,XTERM);
7263 LOP(OP_SUBSTR,XTERM);
7269 char tmpbuf[sizeof PL_tokenbuf];
7270 SSize_t tboffset = 0;
7271 expectation attrful;
7272 bool have_name, have_proto;
7273 const int key = tmp;
7278 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7279 SV *subtoken = newSVpvn(tstart, s - tstart);
7283 s = SKIPSPACE2(s,tmpwhite);
7288 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7289 (*s == ':' && s[1] == ':'))
7292 SV *nametoke = NULL;
7296 attrful = XATTRBLOCK;
7297 /* remember buffer pos'n for later force_word */
7298 tboffset = s - PL_oldbufptr;
7299 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7302 nametoke = newSVpvn(s, d - s);
7304 if (memchr(tmpbuf, ':', len))
7305 sv_setpvn(PL_subname, tmpbuf, len);
7307 sv_setsv(PL_subname,PL_curstname);
7308 sv_catpvs(PL_subname,"::");
7309 sv_catpvn(PL_subname,tmpbuf,len);
7316 CURMAD('X', nametoke);
7317 CURMAD('_', tmpwhite);
7318 (void) force_word(PL_oldbufptr + tboffset, WORD,
7321 s = SKIPSPACE2(d,tmpwhite);
7328 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7329 PL_expect = XTERMBLOCK;
7330 attrful = XATTRTERM;
7331 sv_setpvs(PL_subname,"?");
7335 if (key == KEY_format) {
7337 PL_lex_formbrack = PL_lex_brackets + 1;
7339 PL_thistoken = subtoken;
7343 (void) force_word(PL_oldbufptr + tboffset, WORD,
7349 /* Look for a prototype */
7352 bool bad_proto = FALSE;
7353 bool in_brackets = FALSE;
7354 char greedy_proto = ' ';
7355 bool proto_after_greedy_proto = FALSE;
7356 bool must_be_last = FALSE;
7357 bool underscore = FALSE;
7358 bool seen_underscore = FALSE;
7359 const bool warnsyntax = ckWARN(WARN_SYNTAX);
7361 s = scan_str(s,!!PL_madskills,FALSE);
7363 Perl_croak(aTHX_ "Prototype not terminated");
7364 /* strip spaces and check for bad characters */
7365 d = SvPVX(PL_lex_stuff);
7367 for (p = d; *p; ++p) {
7373 proto_after_greedy_proto = TRUE;
7374 if (!strchr("$@%*;[]&\\_", *p)) {
7386 else if ( *p == ']' ) {
7387 in_brackets = FALSE;
7389 else if ( (*p == '@' || *p == '%') &&
7390 ( tmp < 2 || d[tmp-2] != '\\' ) &&
7392 must_be_last = TRUE;
7395 else if ( *p == '_' ) {
7396 underscore = seen_underscore = TRUE;
7403 if (proto_after_greedy_proto)
7404 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
7405 "Prototype after '%c' for %"SVf" : %s",
7406 greedy_proto, SVfARG(PL_subname), d);
7408 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
7409 "Illegal character %sin prototype for %"SVf" : %s",
7410 seen_underscore ? "after '_' " : "",
7411 SVfARG(PL_subname), d);
7412 SvCUR_set(PL_lex_stuff, tmp);
7417 CURMAD('q', PL_thisopen);
7418 CURMAD('_', tmpwhite);
7419 CURMAD('=', PL_thisstuff);
7420 CURMAD('Q', PL_thisclose);
7421 NEXTVAL_NEXTTOKE.opval =
7422 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7423 PL_lex_stuff = NULL;
7426 s = SKIPSPACE2(s,tmpwhite);
7434 if (*s == ':' && s[1] != ':')
7435 PL_expect = attrful;
7436 else if (*s != '{' && key == KEY_sub) {
7438 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
7440 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
7447 curmad('^', newSVpvs(""));
7448 CURMAD('_', tmpwhite);
7452 PL_thistoken = subtoken;
7455 NEXTVAL_NEXTTOKE.opval =
7456 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7457 PL_lex_stuff = NULL;
7463 sv_setpvs(PL_subname, "__ANON__");
7465 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7469 (void) force_word(PL_oldbufptr + tboffset, WORD,
7478 LOP(OP_SYSTEM,XREF);
7481 LOP(OP_SYMLINK,XTERM);
7484 LOP(OP_SYSCALL,XTERM);
7487 LOP(OP_SYSOPEN,XTERM);
7490 LOP(OP_SYSSEEK,XTERM);
7493 LOP(OP_SYSREAD,XTERM);
7496 LOP(OP_SYSWRITE,XTERM);
7500 TERM(sublex_start());
7521 LOP(OP_TRUNCATE,XTERM);
7533 pl_yylval.ival = CopLINE(PL_curcop);
7537 pl_yylval.ival = CopLINE(PL_curcop);
7541 LOP(OP_UNLINK,XTERM);
7547 LOP(OP_UNPACK,XTERM);
7550 LOP(OP_UTIME,XTERM);
7556 LOP(OP_UNSHIFT,XTERM);
7559 s = tokenize_use(1, s);
7569 pl_yylval.ival = CopLINE(PL_curcop);
7573 pl_yylval.ival = CopLINE(PL_curcop);
7577 PL_hints |= HINT_BLOCK_SCOPE;
7584 LOP(OP_WAITPID,XTERM);
7593 ctl_l[0] = toCTRL('L');
7595 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7598 /* Make sure $^L is defined */
7599 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7604 if (PL_expect == XOPERATOR)
7610 pl_yylval.ival = OP_XOR;
7615 TERM(sublex_start());
7620 #pragma segment Main
7624 S_pending_ident(pTHX)
7629 /* pit holds the identifier we read and pending_ident is reset */
7630 char pit = PL_pending_ident;
7631 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7632 /* All routes through this function want to know if there is a colon. */
7633 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7634 PL_pending_ident = 0;
7636 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7637 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7638 "### Pending identifier '%s'\n", PL_tokenbuf); });
7640 /* if we're in a my(), we can't allow dynamics here.
7641 $foo'bar has already been turned into $foo::bar, so
7642 just check for colons.
7644 if it's a legal name, the OP is a PADANY.
7647 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7649 yyerror(Perl_form(aTHX_ "No package name allowed for "
7650 "variable %s in \"our\"",
7652 tmp = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7656 yyerror(Perl_form(aTHX_ PL_no_myglob,
7657 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7659 pl_yylval.opval = newOP(OP_PADANY, 0);
7660 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7666 build the ops for accesses to a my() variable.
7668 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7669 then used in a comparison. This catches most, but not
7670 all cases. For instance, it catches
7671 sort { my($a); $a <=> $b }
7673 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7674 (although why you'd do that is anyone's guess).
7679 tmp = pad_findmy(PL_tokenbuf, tokenbuf_len, 0);
7680 if (tmp != NOT_IN_PAD) {
7681 /* might be an "our" variable" */
7682 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7683 /* build ops for a bareword */
7684 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7685 HEK * const stashname = HvNAME_HEK(stash);
7686 SV * const sym = newSVhek(stashname);
7687 sv_catpvs(sym, "::");
7688 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7689 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7690 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7693 ? (GV_ADDMULTI | GV_ADDINEVAL)
7696 ((PL_tokenbuf[0] == '$') ? SVt_PV
7697 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7702 /* if it's a sort block and they're naming $a or $b */
7703 if (PL_last_lop_op == OP_SORT &&
7704 PL_tokenbuf[0] == '$' &&
7705 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
7708 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
7709 d < PL_bufend && *d != '\n';
7712 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7713 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7719 pl_yylval.opval = newOP(OP_PADANY, 0);
7720 pl_yylval.opval->op_targ = tmp;
7726 Whine if they've said @foo in a doublequoted string,
7727 and @foo isn't a variable we can find in the symbol
7730 if (ckWARN(WARN_AMBIGUOUS) &&
7731 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7732 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7734 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7735 /* DO NOT warn for @- and @+ */
7736 && !( PL_tokenbuf[2] == '\0' &&
7737 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7740 /* Downgraded from fatal to warning 20000522 mjd */
7741 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7742 "Possible unintended interpolation of %s in string",
7747 /* build ops for a bareword */
7748 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
7750 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7752 PL_tokenbuf + 1, tokenbuf_len - 1,
7753 /* If the identifier refers to a stash, don't autovivify it.
7754 * Change 24660 had the side effect of causing symbol table
7755 * hashes to always be defined, even if they were freshly
7756 * created and the only reference in the entire program was
7757 * the single statement with the defined %foo::bar:: test.
7758 * It appears that all code in the wild doing this actually
7759 * wants to know whether sub-packages have been loaded, so
7760 * by avoiding auto-vivifying symbol tables, we ensure that
7761 * defined %foo::bar:: continues to be false, and the existing
7762 * tests still give the expected answers, even though what
7763 * they're actually testing has now changed subtly.
7765 (*PL_tokenbuf == '%'
7766 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
7769 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
7770 ((PL_tokenbuf[0] == '$') ? SVt_PV
7771 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7777 * The following code was generated by perl_keyword.pl.
7781 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
7785 PERL_ARGS_ASSERT_KEYWORD;
7789 case 1: /* 5 tokens of length 1 */
7821 case 2: /* 18 tokens of length 2 */
7967 case 3: /* 29 tokens of length 3 */
7971 if (name[1] == 'N' &&
8034 if (name[1] == 'i' &&
8066 if (name[1] == 'o' &&
8075 if (name[1] == 'e' &&
8084 if (name[1] == 'n' &&
8093 if (name[1] == 'o' &&
8102 if (name[1] == 'a' &&
8111 if (name[1] == 'o' &&
8173 if (name[1] == 'e' &&
8187 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
8213 if (name[1] == 'i' &&
8222 if (name[1] == 's' &&
8231 if (name[1] == 'e' &&
8240 if (name[1] == 'o' &&
8252 case 4: /* 41 tokens of length 4 */
8256 if (name[1] == 'O' &&
8266 if (name[1] == 'N' &&
8276 if (name[1] == 'i' &&
8286 if (name[1] == 'h' &&
8296 if (name[1] == 'u' &&
8309 if (name[2] == 'c' &&
8318 if (name[2] == 's' &&
8327 if (name[2] == 'a' &&
8363 if (name[1] == 'o' &&
8376 if (name[2] == 't' &&
8385 if (name[2] == 'o' &&
8394 if (name[2] == 't' &&
8403 if (name[2] == 'e' &&
8416 if (name[1] == 'o' &&
8429 if (name[2] == 'y' &&
8438 if (name[2] == 'l' &&
8454 if (name[2] == 's' &&
8463 if (name[2] == 'n' &&
8472 if (name[2] == 'c' &&
8485 if (name[1] == 'e' &&
8495 if (name[1] == 'p' &&
8508 if (name[2] == 'c' &&
8517 if (name[2] == 'p' &&
8526 if (name[2] == 's' &&
8542 if (name[2] == 'n' &&
8612 if (name[2] == 'r' &&
8621 if (name[2] == 'r' &&
8630 if (name[2] == 'a' &&
8646 if (name[2] == 'l' &&
8708 if (name[2] == 'e' &&
8711 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8724 case 5: /* 39 tokens of length 5 */
8728 if (name[1] == 'E' &&
8739 if (name[1] == 'H' &&
8753 if (name[2] == 'a' &&
8763 if (name[2] == 'a' &&
8780 if (name[2] == 'e' &&
8790 if (name[2] == 'e' &&
8794 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
8810 if (name[3] == 'i' &&
8819 if (name[3] == 'o' &&
8855 if (name[2] == 'o' &&
8865 if (name[2] == 'y' &&
8879 if (name[1] == 'l' &&
8893 if (name[2] == 'n' &&
8903 if (name[2] == 'o' &&
8917 if (name[1] == 'i' &&
8922 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
8931 if (name[2] == 'd' &&
8941 if (name[2] == 'c' &&
8958 if (name[2] == 'c' &&
8968 if (name[2] == 't' &&
8982 if (name[1] == 'k' &&
8993 if (name[1] == 'r' &&
9007 if (name[2] == 's' &&
9017 if (name[2] == 'd' &&
9034 if (name[2] == 'm' &&
9044 if (name[2] == 'i' &&
9054 if (name[2] == 'e' &&
9064 if (name[2] == 'l' &&
9074 if (name[2] == 'a' &&
9087 if (name[3] == 't' &&
9090 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
9096 if (name[3] == 'd' &&
9113 if (name[1] == 'i' &&
9127 if (name[2] == 'a' &&
9140 if (name[3] == 'e' &&
9175 if (name[2] == 'i' &&
9192 if (name[2] == 'i' &&
9202 if (name[2] == 'i' &&
9219 case 6: /* 33 tokens of length 6 */
9223 if (name[1] == 'c' &&
9238 if (name[2] == 'l' &&
9249 if (name[2] == 'r' &&
9264 if (name[1] == 'e' &&
9279 if (name[2] == 's' &&
9284 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
9290 if (name[2] == 'i' &&
9308 if (name[2] == 'l' &&
9319 if (name[2] == 'r' &&
9334 if (name[1] == 'm' &&
9349 if (name[2] == 'n' &&
9360 if (name[2] == 's' &&
9375 if (name[1] == 's' &&
9381 if (name[4] == 't' &&
9390 if (name[4] == 'e' &&
9399 if (name[4] == 'c' &&
9408 if (name[4] == 'n' &&
9424 if (name[1] == 'r' &&
9442 if (name[3] == 'a' &&
9452 if (name[3] == 'u' &&
9466 if (name[2] == 'n' &&
9484 if (name[2] == 'a' &&
9498 if (name[3] == 'e' &&
9511 if (name[4] == 't' &&
9520 if (name[4] == 'e' &&
9542 if (name[4] == 't' &&
9551 if (name[4] == 'e' &&
9567 if (name[2] == 'c' &&
9578 if (name[2] == 'l' &&
9589 if (name[2] == 'b' &&
9600 if (name[2] == 's' &&
9623 if (name[4] == 's' &&
9632 if (name[4] == 'n' &&
9645 if (name[3] == 'a' &&
9662 if (name[1] == 'a' &&
9677 case 7: /* 29 tokens of length 7 */
9681 if (name[1] == 'E' &&
9694 if (name[1] == '_' &&
9707 if (name[1] == 'i' &&
9714 return -KEY_binmode;
9720 if (name[1] == 'o' &&
9727 return -KEY_connect;
9736 if (name[2] == 'm' &&
9742 return -KEY_dbmopen;
9753 if (name[4] == 'u' &&
9757 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
9763 if (name[4] == 'n' &&
9784 if (name[1] == 'o' &&
9797 if (name[1] == 'e' &&
9804 if (name[5] == 'r' &&
9807 return -KEY_getpgrp;
9813 if (name[5] == 'i' &&
9816 return -KEY_getppid;
9829 if (name[1] == 'c' &&
9836 return -KEY_lcfirst;
9842 if (name[1] == 'p' &&
9849 return -KEY_opendir;
9855 if (name[1] == 'a' &&
9873 if (name[3] == 'd' &&
9878 return -KEY_readdir;
9884 if (name[3] == 'u' &&
9895 if (name[3] == 'e' &&
9900 return -KEY_reverse;
9919 if (name[3] == 'k' &&
9924 return -KEY_seekdir;
9930 if (name[3] == 'p' &&
9935 return -KEY_setpgrp;
9945 if (name[2] == 'm' &&
9951 return -KEY_shmread;
9957 if (name[2] == 'r' &&
9963 return -KEY_sprintf;
9972 if (name[3] == 'l' &&
9977 return -KEY_symlink;
9986 if (name[4] == 'a' &&
9990 return -KEY_syscall;
9996 if (name[4] == 'p' &&
10000 return -KEY_sysopen;
10006 if (name[4] == 'e' &&
10010 return -KEY_sysread;
10016 if (name[4] == 'e' &&
10020 return -KEY_sysseek;
10038 if (name[1] == 'e' &&
10045 return -KEY_telldir;
10054 if (name[2] == 'f' &&
10060 return -KEY_ucfirst;
10066 if (name[2] == 's' &&
10072 return -KEY_unshift;
10082 if (name[1] == 'a' &&
10089 return -KEY_waitpid;
10098 case 8: /* 26 tokens of length 8 */
10102 if (name[1] == 'U' &&
10110 return KEY_AUTOLOAD;
10116 if (name[1] == '_')
10121 if (name[3] == 'A' &&
10127 return KEY___DATA__;
10133 if (name[3] == 'I' &&
10139 return -KEY___FILE__;
10145 if (name[3] == 'I' &&
10151 return -KEY___LINE__;
10167 if (name[2] == 'o' &&
10174 return -KEY_closedir;
10180 if (name[2] == 'n' &&
10187 return -KEY_continue;
10197 if (name[1] == 'b' &&
10205 return -KEY_dbmclose;
10211 if (name[1] == 'n' &&
10217 if (name[4] == 'r' &&
10222 return -KEY_endgrent;
10228 if (name[4] == 'w' &&
10233 return -KEY_endpwent;
10246 if (name[1] == 'o' &&
10254 return -KEY_formline;
10260 if (name[1] == 'e' &&
10266 if (name[4] == 'r')
10271 if (name[6] == 'n' &&
10274 return -KEY_getgrent;
10280 if (name[6] == 'i' &&
10283 return -KEY_getgrgid;
10289 if (name[6] == 'a' &&
10292 return -KEY_getgrnam;
10305 if (name[4] == 'o' &&
10310 return -KEY_getlogin;
10316 if (name[4] == 'w')
10321 if (name[6] == 'n' &&
10324 return -KEY_getpwent;
10330 if (name[6] == 'a' &&
10333 return -KEY_getpwnam;
10339 if (name[6] == 'i' &&
10342 return -KEY_getpwuid;
10362 if (name[1] == 'e' &&
10369 if (name[5] == 'i' &&
10376 return -KEY_readline;
10381 return -KEY_readlink;
10392 if (name[5] == 'i' &&
10396 return -KEY_readpipe;
10412 if (name[2] == 't')
10417 if (name[4] == 'r' &&
10422 return -KEY_setgrent;
10428 if (name[4] == 'w' &&
10433 return -KEY_setpwent;
10449 if (name[3] == 'w' &&
10455 return -KEY_shmwrite;
10461 if (name[3] == 't' &&
10467 return -KEY_shutdown;
10477 if (name[2] == 's' &&
10484 return -KEY_syswrite;
10494 if (name[1] == 'r' &&
10502 return -KEY_truncate;
10511 case 9: /* 9 tokens of length 9 */
10515 if (name[1] == 'N' &&
10524 return KEY_UNITCHECK;
10530 if (name[1] == 'n' &&
10539 return -KEY_endnetent;
10545 if (name[1] == 'e' &&
10554 return -KEY_getnetent;
10560 if (name[1] == 'o' &&
10569 return -KEY_localtime;
10575 if (name[1] == 'r' &&
10584 return KEY_prototype;
10590 if (name[1] == 'u' &&
10599 return -KEY_quotemeta;
10605 if (name[1] == 'e' &&
10614 return -KEY_rewinddir;
10620 if (name[1] == 'e' &&
10629 return -KEY_setnetent;
10635 if (name[1] == 'a' &&
10644 return -KEY_wantarray;
10653 case 10: /* 9 tokens of length 10 */
10657 if (name[1] == 'n' &&
10663 if (name[4] == 'o' &&
10670 return -KEY_endhostent;
10676 if (name[4] == 'e' &&
10683 return -KEY_endservent;
10696 if (name[1] == 'e' &&
10702 if (name[4] == 'o' &&
10709 return -KEY_gethostent;
10718 if (name[5] == 'r' &&
10724 return -KEY_getservent;
10730 if (name[5] == 'c' &&
10736 return -KEY_getsockopt;
10756 if (name[2] == 't')
10761 if (name[4] == 'o' &&
10768 return -KEY_sethostent;
10777 if (name[5] == 'r' &&
10783 return -KEY_setservent;
10789 if (name[5] == 'c' &&
10795 return -KEY_setsockopt;
10812 if (name[2] == 'c' &&
10821 return -KEY_socketpair;
10834 case 11: /* 8 tokens of length 11 */
10838 if (name[1] == '_' &&
10848 { /* __PACKAGE__ */
10849 return -KEY___PACKAGE__;
10855 if (name[1] == 'n' &&
10865 { /* endprotoent */
10866 return -KEY_endprotoent;
10872 if (name[1] == 'e' &&
10881 if (name[5] == 'e' &&
10887 { /* getpeername */
10888 return -KEY_getpeername;
10897 if (name[6] == 'o' &&
10902 { /* getpriority */
10903 return -KEY_getpriority;
10909 if (name[6] == 't' &&
10914 { /* getprotoent */
10915 return -KEY_getprotoent;
10929 if (name[4] == 'o' &&
10936 { /* getsockname */
10937 return -KEY_getsockname;
10950 if (name[1] == 'e' &&
10958 if (name[6] == 'o' &&
10963 { /* setpriority */
10964 return -KEY_setpriority;
10970 if (name[6] == 't' &&
10975 { /* setprotoent */
10976 return -KEY_setprotoent;
10992 case 12: /* 2 tokens of length 12 */
10993 if (name[0] == 'g' &&
11005 if (name[9] == 'd' &&
11008 { /* getnetbyaddr */
11009 return -KEY_getnetbyaddr;
11015 if (name[9] == 'a' &&
11018 { /* getnetbyname */
11019 return -KEY_getnetbyname;
11031 case 13: /* 4 tokens of length 13 */
11032 if (name[0] == 'g' &&
11039 if (name[4] == 'o' &&
11048 if (name[10] == 'd' &&
11051 { /* gethostbyaddr */
11052 return -KEY_gethostbyaddr;
11058 if (name[10] == 'a' &&
11061 { /* gethostbyname */
11062 return -KEY_gethostbyname;
11075 if (name[4] == 'e' &&
11084 if (name[10] == 'a' &&
11087 { /* getservbyname */
11088 return -KEY_getservbyname;
11094 if (name[10] == 'o' &&
11097 { /* getservbyport */
11098 return -KEY_getservbyport;
11117 case 14: /* 1 tokens of length 14 */
11118 if (name[0] == 'g' &&
11132 { /* getprotobyname */
11133 return -KEY_getprotobyname;
11138 case 16: /* 1 tokens of length 16 */
11139 if (name[0] == 'g' &&
11155 { /* getprotobynumber */
11156 return -KEY_getprotobynumber;
11170 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
11174 PERL_ARGS_ASSERT_CHECKCOMMA;
11176 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
11177 if (ckWARN(WARN_SYNTAX)) {
11180 for (w = s+2; *w && level; w++) {
11183 else if (*w == ')')
11186 while (isSPACE(*w))
11188 /* the list of chars below is for end of statements or
11189 * block / parens, boolean operators (&&, ||, //) and branch
11190 * constructs (or, and, if, until, unless, while, err, for).
11191 * Not a very solid hack... */
11192 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
11193 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11194 "%s (...) interpreted as function",name);
11197 while (s < PL_bufend && isSPACE(*s))
11201 while (s < PL_bufend && isSPACE(*s))
11203 if (isIDFIRST_lazy_if(s,UTF)) {
11204 const char * const w = s++;
11205 while (isALNUM_lazy_if(s,UTF))
11207 while (s < PL_bufend && isSPACE(*s))
11211 if (keyword(w, s - w, 0))
11214 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
11215 if (gv && GvCVu(gv))
11217 Perl_croak(aTHX_ "No comma allowed after %s", what);
11222 /* Either returns sv, or mortalizes sv and returns a new SV*.
11223 Best used as sv=new_constant(..., sv, ...).
11224 If s, pv are NULL, calls subroutine with one argument,
11225 and type is used with error messages only. */
11228 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
11229 SV *sv, SV *pv, const char *type, STRLEN typelen)
11232 HV * const table = GvHV(PL_hintgv); /* ^H */
11236 const char *why1 = "", *why2 = "", *why3 = "";
11238 PERL_ARGS_ASSERT_NEW_CONSTANT;
11240 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
11243 why2 = (const char *)
11244 (strEQ(key,"charnames")
11245 ? "(possibly a missing \"use charnames ...\")"
11247 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
11248 (type ? type: "undef"), why2);
11250 /* This is convoluted and evil ("goto considered harmful")
11251 * but I do not understand the intricacies of all the different
11252 * failure modes of %^H in here. The goal here is to make
11253 * the most probable error message user-friendly. --jhi */
11258 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
11259 (type ? type: "undef"), why1, why2, why3);
11261 yyerror(SvPVX_const(msg));
11265 cvp = hv_fetch(table, key, keylen, FALSE);
11266 if (!cvp || !SvOK(*cvp)) {
11269 why3 = "} is not defined";
11272 sv_2mortal(sv); /* Parent created it permanently */
11275 pv = newSVpvn_flags(s, len, SVs_TEMP);
11277 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
11279 typesv = &PL_sv_undef;
11281 PUSHSTACKi(PERLSI_OVERLOAD);
11293 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
11297 /* Check the eval first */
11298 if (!PL_in_eval && SvTRUE(ERRSV)) {
11299 sv_catpvs(ERRSV, "Propagated");
11300 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
11302 res = SvREFCNT_inc_simple(sv);
11306 SvREFCNT_inc_simple_void(res);
11315 why1 = "Call to &{$^H{";
11317 why3 = "}} did not return a defined value";
11325 /* Returns a NUL terminated string, with the length of the string written to
11329 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
11332 register char *d = dest;
11333 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11335 PERL_ARGS_ASSERT_SCAN_WORD;
11339 Perl_croak(aTHX_ ident_too_long);
11340 if (isALNUM(*s)) /* UTF handled below */
11342 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
11347 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
11351 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11352 char *t = s + UTF8SKIP(s);
11354 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11358 Perl_croak(aTHX_ ident_too_long);
11359 Copy(s, d, len, char);
11372 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
11375 char *bracket = NULL;
11377 register char *d = dest;
11378 register char * const e = d + destlen + 3; /* two-character token, ending NUL */
11380 PERL_ARGS_ASSERT_SCAN_IDENT;
11385 while (isDIGIT(*s)) {
11387 Perl_croak(aTHX_ ident_too_long);
11394 Perl_croak(aTHX_ ident_too_long);
11395 if (isALNUM(*s)) /* UTF handled below */
11397 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
11402 else if (*s == ':' && s[1] == ':') {
11406 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11407 char *t = s + UTF8SKIP(s);
11408 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11410 if (d + (t - s) > e)
11411 Perl_croak(aTHX_ ident_too_long);
11412 Copy(s, d, t - s, char);
11423 if (PL_lex_state != LEX_NORMAL)
11424 PL_lex_state = LEX_INTERPENDMAYBE;
11427 if (*s == '$' && s[1] &&
11428 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
11441 if (*d == '^' && *s && isCONTROLVAR(*s)) {
11446 if (isSPACE(s[-1])) {
11448 const char ch = *s++;
11449 if (!SPACE_OR_TAB(ch)) {
11455 if (isIDFIRST_lazy_if(d,UTF)) {
11459 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
11460 end += UTF8SKIP(end);
11461 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
11462 end += UTF8SKIP(end);
11464 Copy(s, d, end - s, char);
11469 while ((isALNUM(*s) || *s == ':') && d < e)
11472 Perl_croak(aTHX_ ident_too_long);
11475 while (s < send && SPACE_OR_TAB(*s))
11477 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
11478 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
11479 const char * const brack =
11481 ((*s == '[') ? "[...]" : "{...}");
11482 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11483 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
11484 funny, dest, brack, funny, dest, brack);
11487 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
11491 /* Handle extended ${^Foo} variables
11492 * 1999-02-27 mjd-perl-patch@plover.com */
11493 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
11497 while (isALNUM(*s) && d < e) {
11501 Perl_croak(aTHX_ ident_too_long);
11506 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
11507 PL_lex_state = LEX_INTERPEND;
11510 if (PL_lex_state == LEX_NORMAL) {
11511 if (ckWARN(WARN_AMBIGUOUS) &&
11512 (keyword(dest, d - dest, 0)
11513 || get_cvn_flags(dest, d - dest, 0)))
11517 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11518 "Ambiguous use of %c{%s} resolved to %c%s",
11519 funny, dest, funny, dest);
11524 s = bracket; /* let the parser handle it */
11528 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
11529 PL_lex_state = LEX_INTERPEND;
11534 S_pmflag(U32 pmfl, const char ch) {
11536 CASE_STD_PMMOD_FLAGS_PARSE_SET(&pmfl);
11537 case GLOBAL_PAT_MOD: pmfl |= PMf_GLOBAL; break;
11538 case CONTINUE_PAT_MOD: pmfl |= PMf_CONTINUE; break;
11539 case ONCE_PAT_MOD: pmfl |= PMf_KEEP; break;
11540 case KEEPCOPY_PAT_MOD: pmfl |= PMf_KEEPCOPY; break;
11546 Perl_pmflag(pTHX_ U32* pmfl, int ch)
11548 PERL_ARGS_ASSERT_PMFLAG;
11550 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
11551 "Perl_pmflag() is deprecated, and will be removed from the XS API");
11554 *pmfl = S_pmflag(*pmfl, (char)ch);
11559 S_scan_pat(pTHX_ char *start, I32 type)
11563 char *s = scan_str(start,!!PL_madskills,FALSE);
11564 const char * const valid_flags =
11565 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
11570 PERL_ARGS_ASSERT_SCAN_PAT;
11573 const char * const delimiter = skipspace(start);
11577 ? "Search pattern not terminated or ternary operator parsed as search pattern"
11578 : "Search pattern not terminated" ));
11581 pm = (PMOP*)newPMOP(type, 0);
11582 if (PL_multi_open == '?') {
11583 /* This is the only point in the code that sets PMf_ONCE: */
11584 pm->op_pmflags |= PMf_ONCE;
11586 /* Hence it's safe to do this bit of PMOP book-keeping here, which
11587 allows us to restrict the list needed by reset to just the ??
11589 assert(type != OP_TRANS);
11591 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11594 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11597 elements = mg->mg_len / sizeof(PMOP**);
11598 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11599 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11600 mg->mg_len = elements * sizeof(PMOP**);
11601 PmopSTASH_set(pm,PL_curstash);
11607 while (*s && strchr(valid_flags, *s))
11608 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11610 if (PL_madskills && modstart != s) {
11611 SV* tmptoken = newSVpvn(modstart, s - modstart);
11612 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11615 /* issue a warning if /c is specified,but /g is not */
11616 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11618 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11619 "Use of /c modifier is meaningless without /g" );
11622 PL_lex_op = (OP*)pm;
11623 pl_yylval.ival = OP_MATCH;
11628 S_scan_subst(pTHX_ char *start)
11639 PERL_ARGS_ASSERT_SCAN_SUBST;
11641 pl_yylval.ival = OP_NULL;
11643 s = scan_str(start,!!PL_madskills,FALSE);
11646 Perl_croak(aTHX_ "Substitution pattern not terminated");
11648 if (s[-1] == PL_multi_open)
11651 if (PL_madskills) {
11652 CURMAD('q', PL_thisopen);
11653 CURMAD('_', PL_thiswhite);
11654 CURMAD('E', PL_thisstuff);
11655 CURMAD('Q', PL_thisclose);
11656 PL_realtokenstart = s - SvPVX(PL_linestr);
11660 first_start = PL_multi_start;
11661 s = scan_str(s,!!PL_madskills,FALSE);
11663 if (PL_lex_stuff) {
11664 SvREFCNT_dec(PL_lex_stuff);
11665 PL_lex_stuff = NULL;
11667 Perl_croak(aTHX_ "Substitution replacement not terminated");
11669 PL_multi_start = first_start; /* so whole substitution is taken together */
11671 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11674 if (PL_madskills) {
11675 CURMAD('z', PL_thisopen);
11676 CURMAD('R', PL_thisstuff);
11677 CURMAD('Z', PL_thisclose);
11683 if (*s == EXEC_PAT_MOD) {
11687 else if (strchr(S_PAT_MODS, *s))
11688 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11694 if (PL_madskills) {
11696 curmad('m', newSVpvn(modstart, s - modstart));
11697 append_madprops(PL_thismad, (OP*)pm, 0);
11701 if ((pm->op_pmflags & PMf_CONTINUE)) {
11702 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
11706 SV * const repl = newSVpvs("");
11708 PL_sublex_info.super_bufptr = s;
11709 PL_sublex_info.super_bufend = PL_bufend;
11711 pm->op_pmflags |= PMf_EVAL;
11714 sv_catpvs(repl, "eval ");
11716 sv_catpvs(repl, "do ");
11718 sv_catpvs(repl, "{");
11719 sv_catsv(repl, PL_lex_repl);
11720 if (strchr(SvPVX(PL_lex_repl), '#'))
11721 sv_catpvs(repl, "\n");
11722 sv_catpvs(repl, "}");
11724 SvREFCNT_dec(PL_lex_repl);
11725 PL_lex_repl = repl;
11728 PL_lex_op = (OP*)pm;
11729 pl_yylval.ival = OP_SUBST;
11734 S_scan_trans(pTHX_ char *start)
11747 PERL_ARGS_ASSERT_SCAN_TRANS;
11749 pl_yylval.ival = OP_NULL;
11751 s = scan_str(start,!!PL_madskills,FALSE);
11753 Perl_croak(aTHX_ "Transliteration pattern not terminated");
11755 if (s[-1] == PL_multi_open)
11758 if (PL_madskills) {
11759 CURMAD('q', PL_thisopen);
11760 CURMAD('_', PL_thiswhite);
11761 CURMAD('E', PL_thisstuff);
11762 CURMAD('Q', PL_thisclose);
11763 PL_realtokenstart = s - SvPVX(PL_linestr);
11767 s = scan_str(s,!!PL_madskills,FALSE);
11769 if (PL_lex_stuff) {
11770 SvREFCNT_dec(PL_lex_stuff);
11771 PL_lex_stuff = NULL;
11773 Perl_croak(aTHX_ "Transliteration replacement not terminated");
11775 if (PL_madskills) {
11776 CURMAD('z', PL_thisopen);
11777 CURMAD('R', PL_thisstuff);
11778 CURMAD('Z', PL_thisclose);
11781 complement = del = squash = 0;
11788 complement = OPpTRANS_COMPLEMENT;
11791 del = OPpTRANS_DELETE;
11794 squash = OPpTRANS_SQUASH;
11803 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
11804 o = newPVOP(OP_TRANS, 0, (char*)tbl);
11805 o->op_private &= ~OPpTRANS_ALL;
11806 o->op_private |= del|squash|complement|
11807 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
11808 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
11811 pl_yylval.ival = OP_TRANS;
11814 if (PL_madskills) {
11816 curmad('m', newSVpvn(modstart, s - modstart));
11817 append_madprops(PL_thismad, o, 0);
11826 S_scan_heredoc(pTHX_ register char *s)
11830 I32 op_type = OP_SCALAR;
11834 const char *found_newline;
11838 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
11840 I32 stuffstart = s - SvPVX(PL_linestr);
11843 PL_realtokenstart = -1;
11846 PERL_ARGS_ASSERT_SCAN_HEREDOC;
11850 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
11854 while (SPACE_OR_TAB(*peek))
11856 if (*peek == '`' || *peek == '\'' || *peek =='"') {
11859 s = delimcpy(d, e, s, PL_bufend, term, &len);
11869 if (!isALNUM_lazy_if(s,UTF))
11870 deprecate("bare << to mean <<\"\"");
11871 for (; isALNUM_lazy_if(s,UTF); s++) {
11876 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
11877 Perl_croak(aTHX_ "Delimiter for here document is too long");
11880 len = d - PL_tokenbuf;
11883 if (PL_madskills) {
11884 tstart = PL_tokenbuf + !outer;
11885 PL_thisclose = newSVpvn(tstart, len - !outer);
11886 tstart = SvPVX(PL_linestr) + stuffstart;
11887 PL_thisopen = newSVpvn(tstart, s - tstart);
11888 stuffstart = s - SvPVX(PL_linestr);
11891 #ifndef PERL_STRICT_CR
11892 d = strchr(s, '\r');
11894 char * const olds = s;
11896 while (s < PL_bufend) {
11902 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
11911 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
11918 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
11919 herewas = newSVpvn(s,PL_bufend-s);
11923 herewas = newSVpvn(s-1,found_newline-s+1);
11926 herewas = newSVpvn(s,found_newline-s);
11930 if (PL_madskills) {
11931 tstart = SvPVX(PL_linestr) + stuffstart;
11933 sv_catpvn(PL_thisstuff, tstart, s - tstart);
11935 PL_thisstuff = newSVpvn(tstart, s - tstart);
11938 s += SvCUR(herewas);
11941 stuffstart = s - SvPVX(PL_linestr);
11947 tmpstr = newSV_type(SVt_PVIV);
11948 SvGROW(tmpstr, 80);
11949 if (term == '\'') {
11950 op_type = OP_CONST;
11951 SvIV_set(tmpstr, -1);
11953 else if (term == '`') {
11954 op_type = OP_BACKTICK;
11955 SvIV_set(tmpstr, '\\');
11959 PL_multi_start = CopLINE(PL_curcop);
11960 PL_multi_open = PL_multi_close = '<';
11961 term = *PL_tokenbuf;
11962 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
11963 char * const bufptr = PL_sublex_info.super_bufptr;
11964 char * const bufend = PL_sublex_info.super_bufend;
11965 char * const olds = s - SvCUR(herewas);
11966 s = strchr(bufptr, '\n');
11970 while (s < bufend &&
11971 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11973 CopLINE_inc(PL_curcop);
11976 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11977 missingterm(PL_tokenbuf);
11979 sv_setpvn(herewas,bufptr,d-bufptr+1);
11980 sv_setpvn(tmpstr,d+1,s-d);
11982 sv_catpvn(herewas,s,bufend-s);
11983 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
11990 while (s < PL_bufend &&
11991 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
11993 CopLINE_inc(PL_curcop);
11995 if (s >= PL_bufend) {
11996 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
11997 missingterm(PL_tokenbuf);
11999 sv_setpvn(tmpstr,d+1,s-d);
12001 if (PL_madskills) {
12003 sv_catpvn(PL_thisstuff, d + 1, s - d);
12005 PL_thisstuff = newSVpvn(d + 1, s - d);
12006 stuffstart = s - SvPVX(PL_linestr);
12010 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
12012 sv_catpvn(herewas,s,PL_bufend-s);
12013 sv_setsv(PL_linestr,herewas);
12014 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
12015 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12016 PL_last_lop = PL_last_uni = NULL;
12019 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
12020 while (s >= PL_bufend) { /* multiple line string? */
12022 if (PL_madskills) {
12023 tstart = SvPVX(PL_linestr) + stuffstart;
12025 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12027 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12031 if (!outer || !lex_next_chunk(0)) {
12032 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12033 missingterm(PL_tokenbuf);
12037 stuffstart = s - SvPVX(PL_linestr);
12039 CopLINE_inc(PL_curcop);
12040 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12041 PL_last_lop = PL_last_uni = NULL;
12042 #ifndef PERL_STRICT_CR
12043 if (PL_bufend - PL_linestart >= 2) {
12044 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
12045 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
12047 PL_bufend[-2] = '\n';
12049 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12051 else if (PL_bufend[-1] == '\r')
12052 PL_bufend[-1] = '\n';
12054 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
12055 PL_bufend[-1] = '\n';
12057 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
12058 update_debugger_info(PL_linestr, NULL, 0);
12059 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
12060 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
12061 *(SvPVX(PL_linestr) + off ) = ' ';
12062 sv_catsv(PL_linestr,herewas);
12063 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12064 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
12068 sv_catsv(tmpstr,PL_linestr);
12073 PL_multi_end = CopLINE(PL_curcop);
12074 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
12075 SvPV_shrink_to_cur(tmpstr);
12077 SvREFCNT_dec(herewas);
12079 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
12081 else if (PL_encoding)
12082 sv_recode_to_utf8(tmpstr, PL_encoding);
12084 PL_lex_stuff = tmpstr;
12085 pl_yylval.ival = op_type;
12089 /* scan_inputsymbol
12090 takes: current position in input buffer
12091 returns: new position in input buffer
12092 side-effects: pl_yylval and lex_op are set.
12097 <FH> read from filehandle
12098 <pkg::FH> read from package qualified filehandle
12099 <pkg'FH> read from package qualified filehandle
12100 <$fh> read from filehandle in $fh
12101 <*.h> filename glob
12106 S_scan_inputsymbol(pTHX_ char *start)
12109 register char *s = start; /* current position in buffer */
12112 char *d = PL_tokenbuf; /* start of temp holding space */
12113 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
12115 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
12117 end = strchr(s, '\n');
12120 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
12122 /* die if we didn't have space for the contents of the <>,
12123 or if it didn't end, or if we see a newline
12126 if (len >= (I32)sizeof PL_tokenbuf)
12127 Perl_croak(aTHX_ "Excessively long <> operator");
12129 Perl_croak(aTHX_ "Unterminated <> operator");
12134 Remember, only scalar variables are interpreted as filehandles by
12135 this code. Anything more complex (e.g., <$fh{$num}>) will be
12136 treated as a glob() call.
12137 This code makes use of the fact that except for the $ at the front,
12138 a scalar variable and a filehandle look the same.
12140 if (*d == '$' && d[1]) d++;
12142 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
12143 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
12146 /* If we've tried to read what we allow filehandles to look like, and
12147 there's still text left, then it must be a glob() and not a getline.
12148 Use scan_str to pull out the stuff between the <> and treat it
12149 as nothing more than a string.
12152 if (d - PL_tokenbuf != len) {
12153 pl_yylval.ival = OP_GLOB;
12154 s = scan_str(start,!!PL_madskills,FALSE);
12156 Perl_croak(aTHX_ "Glob not terminated");
12160 bool readline_overriden = FALSE;
12163 /* we're in a filehandle read situation */
12166 /* turn <> into <ARGV> */
12168 Copy("ARGV",d,5,char);
12170 /* Check whether readline() is overriden */
12171 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
12173 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
12175 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
12176 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
12177 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
12178 readline_overriden = TRUE;
12180 /* if <$fh>, create the ops to turn the variable into a
12184 /* try to find it in the pad for this block, otherwise find
12185 add symbol table ops
12187 const PADOFFSET tmp = pad_findmy(d, len, 0);
12188 if (tmp != NOT_IN_PAD) {
12189 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
12190 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
12191 HEK * const stashname = HvNAME_HEK(stash);
12192 SV * const sym = sv_2mortal(newSVhek(stashname));
12193 sv_catpvs(sym, "::");
12194 sv_catpv(sym, d+1);
12199 OP * const o = newOP(OP_PADSV, 0);
12201 PL_lex_op = readline_overriden
12202 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12203 append_elem(OP_LIST, o,
12204 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
12205 : (OP*)newUNOP(OP_READLINE, 0, o);
12214 ? (GV_ADDMULTI | GV_ADDINEVAL)
12217 PL_lex_op = readline_overriden
12218 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12219 append_elem(OP_LIST,
12220 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
12221 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12222 : (OP*)newUNOP(OP_READLINE, 0,
12223 newUNOP(OP_RV2SV, 0,
12224 newGVOP(OP_GV, 0, gv)));
12226 if (!readline_overriden)
12227 PL_lex_op->op_flags |= OPf_SPECIAL;
12228 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
12229 pl_yylval.ival = OP_NULL;
12232 /* If it's none of the above, it must be a literal filehandle
12233 (<Foo::BAR> or <FOO>) so build a simple readline OP */
12235 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
12236 PL_lex_op = readline_overriden
12237 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12238 append_elem(OP_LIST,
12239 newGVOP(OP_GV, 0, gv),
12240 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12241 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
12242 pl_yylval.ival = OP_NULL;
12251 takes: start position in buffer
12252 keep_quoted preserve \ on the embedded delimiter(s)
12253 keep_delims preserve the delimiters around the string
12254 returns: position to continue reading from buffer
12255 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
12256 updates the read buffer.
12258 This subroutine pulls a string out of the input. It is called for:
12259 q single quotes q(literal text)
12260 ' single quotes 'literal text'
12261 qq double quotes qq(interpolate $here please)
12262 " double quotes "interpolate $here please"
12263 qx backticks qx(/bin/ls -l)
12264 ` backticks `/bin/ls -l`
12265 qw quote words @EXPORT_OK = qw( func() $spam )
12266 m// regexp match m/this/
12267 s/// regexp substitute s/this/that/
12268 tr/// string transliterate tr/this/that/
12269 y/// string transliterate y/this/that/
12270 ($*@) sub prototypes sub foo ($)
12271 (stuff) sub attr parameters sub foo : attr(stuff)
12272 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
12274 In most of these cases (all but <>, patterns and transliterate)
12275 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
12276 calls scan_str(). s/// makes yylex() call scan_subst() which calls
12277 scan_str(). tr/// and y/// make yylex() call scan_trans() which
12280 It skips whitespace before the string starts, and treats the first
12281 character as the delimiter. If the delimiter is one of ([{< then
12282 the corresponding "close" character )]}> is used as the closing
12283 delimiter. It allows quoting of delimiters, and if the string has
12284 balanced delimiters ([{<>}]) it allows nesting.
12286 On success, the SV with the resulting string is put into lex_stuff or,
12287 if that is already non-NULL, into lex_repl. The second case occurs only
12288 when parsing the RHS of the special constructs s/// and tr/// (y///).
12289 For convenience, the terminating delimiter character is stuffed into
12294 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
12297 SV *sv; /* scalar value: string */
12298 const char *tmps; /* temp string, used for delimiter matching */
12299 register char *s = start; /* current position in the buffer */
12300 register char term; /* terminating character */
12301 register char *to; /* current position in the sv's data */
12302 I32 brackets = 1; /* bracket nesting level */
12303 bool has_utf8 = FALSE; /* is there any utf8 content? */
12304 I32 termcode; /* terminating char. code */
12305 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
12306 STRLEN termlen; /* length of terminating string */
12307 int last_off = 0; /* last position for nesting bracket */
12313 PERL_ARGS_ASSERT_SCAN_STR;
12315 /* skip space before the delimiter */
12321 if (PL_realtokenstart >= 0) {
12322 stuffstart = PL_realtokenstart;
12323 PL_realtokenstart = -1;
12326 stuffstart = start - SvPVX(PL_linestr);
12328 /* mark where we are, in case we need to report errors */
12331 /* after skipping whitespace, the next character is the terminator */
12334 termcode = termstr[0] = term;
12338 termcode = utf8_to_uvchr((U8*)s, &termlen);
12339 Copy(s, termstr, termlen, U8);
12340 if (!UTF8_IS_INVARIANT(term))
12344 /* mark where we are */
12345 PL_multi_start = CopLINE(PL_curcop);
12346 PL_multi_open = term;
12348 /* find corresponding closing delimiter */
12349 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
12350 termcode = termstr[0] = term = tmps[5];
12352 PL_multi_close = term;
12354 /* create a new SV to hold the contents. 79 is the SV's initial length.
12355 What a random number. */
12356 sv = newSV_type(SVt_PVIV);
12358 SvIV_set(sv, termcode);
12359 (void)SvPOK_only(sv); /* validate pointer */
12361 /* move past delimiter and try to read a complete string */
12363 sv_catpvn(sv, s, termlen);
12366 tstart = SvPVX(PL_linestr) + stuffstart;
12367 if (!PL_thisopen && !keep_delims) {
12368 PL_thisopen = newSVpvn(tstart, s - tstart);
12369 stuffstart = s - SvPVX(PL_linestr);
12373 if (PL_encoding && !UTF) {
12377 int offset = s - SvPVX_const(PL_linestr);
12378 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
12379 &offset, (char*)termstr, termlen);
12380 const char * const ns = SvPVX_const(PL_linestr) + offset;
12381 char * const svlast = SvEND(sv) - 1;
12383 for (; s < ns; s++) {
12384 if (*s == '\n' && !PL_rsfp)
12385 CopLINE_inc(PL_curcop);
12388 goto read_more_line;
12390 /* handle quoted delimiters */
12391 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
12393 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
12395 if ((svlast-1 - t) % 2) {
12396 if (!keep_quoted) {
12397 *(svlast-1) = term;
12399 SvCUR_set(sv, SvCUR(sv) - 1);
12404 if (PL_multi_open == PL_multi_close) {
12410 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
12411 /* At here, all closes are "was quoted" one,
12412 so we don't check PL_multi_close. */
12414 if (!keep_quoted && *(t+1) == PL_multi_open)
12419 else if (*t == PL_multi_open)
12427 SvCUR_set(sv, w - SvPVX_const(sv));
12429 last_off = w - SvPVX(sv);
12430 if (--brackets <= 0)
12435 if (!keep_delims) {
12436 SvCUR_set(sv, SvCUR(sv) - 1);
12442 /* extend sv if need be */
12443 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
12444 /* set 'to' to the next character in the sv's string */
12445 to = SvPVX(sv)+SvCUR(sv);
12447 /* if open delimiter is the close delimiter read unbridle */
12448 if (PL_multi_open == PL_multi_close) {
12449 for (; s < PL_bufend; s++,to++) {
12450 /* embedded newlines increment the current line number */
12451 if (*s == '\n' && !PL_rsfp)
12452 CopLINE_inc(PL_curcop);
12453 /* handle quoted delimiters */
12454 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
12455 if (!keep_quoted && s[1] == term)
12457 /* any other quotes are simply copied straight through */
12461 /* terminate when run out of buffer (the for() condition), or
12462 have found the terminator */
12463 else if (*s == term) {
12466 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
12469 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12475 /* if the terminator isn't the same as the start character (e.g.,
12476 matched brackets), we have to allow more in the quoting, and
12477 be prepared for nested brackets.
12480 /* read until we run out of string, or we find the terminator */
12481 for (; s < PL_bufend; s++,to++) {
12482 /* embedded newlines increment the line count */
12483 if (*s == '\n' && !PL_rsfp)
12484 CopLINE_inc(PL_curcop);
12485 /* backslashes can escape the open or closing characters */
12486 if (*s == '\\' && s+1 < PL_bufend) {
12487 if (!keep_quoted &&
12488 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
12493 /* allow nested opens and closes */
12494 else if (*s == PL_multi_close && --brackets <= 0)
12496 else if (*s == PL_multi_open)
12498 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12503 /* terminate the copied string and update the sv's end-of-string */
12505 SvCUR_set(sv, to - SvPVX_const(sv));
12508 * this next chunk reads more into the buffer if we're not done yet
12512 break; /* handle case where we are done yet :-) */
12514 #ifndef PERL_STRICT_CR
12515 if (to - SvPVX_const(sv) >= 2) {
12516 if ((to[-2] == '\r' && to[-1] == '\n') ||
12517 (to[-2] == '\n' && to[-1] == '\r'))
12521 SvCUR_set(sv, to - SvPVX_const(sv));
12523 else if (to[-1] == '\r')
12526 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
12531 /* if we're out of file, or a read fails, bail and reset the current
12532 line marker so we can report where the unterminated string began
12535 if (PL_madskills) {
12536 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12538 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12540 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12543 CopLINE_inc(PL_curcop);
12544 PL_bufptr = PL_bufend;
12545 if (!lex_next_chunk(0)) {
12547 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12556 /* at this point, we have successfully read the delimited string */
12558 if (!PL_encoding || UTF) {
12560 if (PL_madskills) {
12561 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12562 const int len = s - tstart;
12564 sv_catpvn(PL_thisstuff, tstart, len);
12566 PL_thisstuff = newSVpvn(tstart, len);
12567 if (!PL_thisclose && !keep_delims)
12568 PL_thisclose = newSVpvn(s,termlen);
12573 sv_catpvn(sv, s, termlen);
12578 if (PL_madskills) {
12579 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12580 const int len = s - tstart - termlen;
12582 sv_catpvn(PL_thisstuff, tstart, len);
12584 PL_thisstuff = newSVpvn(tstart, len);
12585 if (!PL_thisclose && !keep_delims)
12586 PL_thisclose = newSVpvn(s - termlen,termlen);
12590 if (has_utf8 || PL_encoding)
12593 PL_multi_end = CopLINE(PL_curcop);
12595 /* if we allocated too much space, give some back */
12596 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12597 SvLEN_set(sv, SvCUR(sv) + 1);
12598 SvPV_renew(sv, SvLEN(sv));
12601 /* decide whether this is the first or second quoted string we've read
12614 takes: pointer to position in buffer
12615 returns: pointer to new position in buffer
12616 side-effects: builds ops for the constant in pl_yylval.op
12618 Read a number in any of the formats that Perl accepts:
12620 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12621 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12624 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12626 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12629 If it reads a number without a decimal point or an exponent, it will
12630 try converting the number to an integer and see if it can do so
12631 without loss of precision.
12635 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12638 register const char *s = start; /* current position in buffer */
12639 register char *d; /* destination in temp buffer */
12640 register char *e; /* end of temp buffer */
12641 NV nv; /* number read, as a double */
12642 SV *sv = NULL; /* place to put the converted number */
12643 bool floatit; /* boolean: int or float? */
12644 const char *lastub = NULL; /* position of last underbar */
12645 static char const number_too_long[] = "Number too long";
12647 PERL_ARGS_ASSERT_SCAN_NUM;
12649 /* We use the first character to decide what type of number this is */
12653 Perl_croak(aTHX_ "panic: scan_num");
12655 /* if it starts with a 0, it could be an octal number, a decimal in
12656 0.13 disguise, or a hexadecimal number, or a binary number. */
12660 u holds the "number so far"
12661 shift the power of 2 of the base
12662 (hex == 4, octal == 3, binary == 1)
12663 overflowed was the number more than we can hold?
12665 Shift is used when we add a digit. It also serves as an "are
12666 we in octal/hex/binary?" indicator to disallow hex characters
12667 when in octal mode.
12672 bool overflowed = FALSE;
12673 bool just_zero = TRUE; /* just plain 0 or binary number? */
12674 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12675 static const char* const bases[5] =
12676 { "", "binary", "", "octal", "hexadecimal" };
12677 static const char* const Bases[5] =
12678 { "", "Binary", "", "Octal", "Hexadecimal" };
12679 static const char* const maxima[5] =
12681 "0b11111111111111111111111111111111",
12685 const char *base, *Base, *max;
12687 /* check for hex */
12692 } else if (s[1] == 'b') {
12697 /* check for a decimal in disguise */
12698 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
12700 /* so it must be octal */
12707 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12708 "Misplaced _ in number");
12712 base = bases[shift];
12713 Base = Bases[shift];
12714 max = maxima[shift];
12716 /* read the rest of the number */
12718 /* x is used in the overflow test,
12719 b is the digit we're adding on. */
12724 /* if we don't mention it, we're done */
12728 /* _ are ignored -- but warned about if consecutive */
12730 if (lastub && s == lastub + 1)
12731 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12732 "Misplaced _ in number");
12736 /* 8 and 9 are not octal */
12737 case '8': case '9':
12739 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
12743 case '2': case '3': case '4':
12744 case '5': case '6': case '7':
12746 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
12749 case '0': case '1':
12750 b = *s++ & 15; /* ASCII digit -> value of digit */
12754 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
12755 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
12756 /* make sure they said 0x */
12759 b = (*s++ & 7) + 9;
12761 /* Prepare to put the digit we have onto the end
12762 of the number so far. We check for overflows.
12768 x = u << shift; /* make room for the digit */
12770 if ((x >> shift) != u
12771 && !(PL_hints & HINT_NEW_BINARY)) {
12774 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
12775 "Integer overflow in %s number",
12778 u = x | b; /* add the digit to the end */
12781 n *= nvshift[shift];
12782 /* If an NV has not enough bits in its
12783 * mantissa to represent an UV this summing of
12784 * small low-order numbers is a waste of time
12785 * (because the NV cannot preserve the
12786 * low-order bits anyway): we could just
12787 * remember when did we overflow and in the
12788 * end just multiply n by the right
12796 /* if we get here, we had success: make a scalar value from
12801 /* final misplaced underbar check */
12802 if (s[-1] == '_') {
12803 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12808 if (n > 4294967295.0)
12809 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12810 "%s number > %s non-portable",
12816 if (u > 0xffffffff)
12817 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
12818 "%s number > %s non-portable",
12823 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
12824 sv = new_constant(start, s - start, "integer",
12825 sv, NULL, NULL, 0);
12826 else if (PL_hints & HINT_NEW_BINARY)
12827 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
12832 handle decimal numbers.
12833 we're also sent here when we read a 0 as the first digit
12835 case '1': case '2': case '3': case '4': case '5':
12836 case '6': case '7': case '8': case '9': case '.':
12839 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
12842 /* read next group of digits and _ and copy into d */
12843 while (isDIGIT(*s) || *s == '_') {
12844 /* skip underscores, checking for misplaced ones
12848 if (lastub && s == lastub + 1)
12849 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12850 "Misplaced _ in number");
12854 /* check for end of fixed-length buffer */
12856 Perl_croak(aTHX_ number_too_long);
12857 /* if we're ok, copy the character */
12862 /* final misplaced underbar check */
12863 if (lastub && s == lastub + 1) {
12864 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
12867 /* read a decimal portion if there is one. avoid
12868 3..5 being interpreted as the number 3. followed
12871 if (*s == '.' && s[1] != '.') {
12876 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12877 "Misplaced _ in number");
12881 /* copy, ignoring underbars, until we run out of digits.
12883 for (; isDIGIT(*s) || *s == '_'; s++) {
12884 /* fixed length buffer check */
12886 Perl_croak(aTHX_ number_too_long);
12888 if (lastub && s == lastub + 1)
12889 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12890 "Misplaced _ in number");
12896 /* fractional part ending in underbar? */
12897 if (s[-1] == '_') {
12898 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12899 "Misplaced _ in number");
12901 if (*s == '.' && isDIGIT(s[1])) {
12902 /* oops, it's really a v-string, but without the "v" */
12908 /* read exponent part, if present */
12909 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
12913 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
12914 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
12916 /* stray preinitial _ */
12918 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12919 "Misplaced _ in number");
12923 /* allow positive or negative exponent */
12924 if (*s == '+' || *s == '-')
12927 /* stray initial _ */
12929 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12930 "Misplaced _ in number");
12934 /* read digits of exponent */
12935 while (isDIGIT(*s) || *s == '_') {
12938 Perl_croak(aTHX_ number_too_long);
12942 if (((lastub && s == lastub + 1) ||
12943 (!isDIGIT(s[1]) && s[1] != '_')))
12944 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12945 "Misplaced _ in number");
12952 /* make an sv from the string */
12956 We try to do an integer conversion first if no characters
12957 indicating "float" have been found.
12962 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
12964 if (flags == IS_NUMBER_IN_UV) {
12966 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
12969 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
12970 if (uv <= (UV) IV_MIN)
12971 sv_setiv(sv, -(IV)uv);
12978 /* terminate the string */
12980 nv = Atof(PL_tokenbuf);
12985 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
12986 const char *const key = floatit ? "float" : "integer";
12987 const STRLEN keylen = floatit ? 5 : 7;
12988 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
12989 key, keylen, sv, NULL, NULL, 0);
12993 /* if it starts with a v, it could be a v-string */
12996 sv = newSV(5); /* preallocate storage space */
12997 s = scan_vstring(s, PL_bufend, sv);
13001 /* make the op for the constant and return */
13004 lvalp->opval = newSVOP(OP_CONST, 0, sv);
13006 lvalp->opval = NULL;
13012 S_scan_formline(pTHX_ register char *s)
13015 register char *eol;
13017 SV * const stuff = newSVpvs("");
13018 bool needargs = FALSE;
13019 bool eofmt = FALSE;
13021 char *tokenstart = s;
13022 SV* savewhite = NULL;
13024 if (PL_madskills) {
13025 savewhite = PL_thiswhite;
13030 PERL_ARGS_ASSERT_SCAN_FORMLINE;
13032 while (!needargs) {
13035 #ifdef PERL_STRICT_CR
13036 while (SPACE_OR_TAB(*t))
13039 while (SPACE_OR_TAB(*t) || *t == '\r')
13042 if (*t == '\n' || t == PL_bufend) {
13047 if (PL_in_eval && !PL_rsfp) {
13048 eol = (char *) memchr(s,'\n',PL_bufend-s);
13053 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
13055 for (t = s; t < eol; t++) {
13056 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
13058 goto enough; /* ~~ must be first line in formline */
13060 if (*t == '@' || *t == '^')
13064 sv_catpvn(stuff, s, eol-s);
13065 #ifndef PERL_STRICT_CR
13066 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
13067 char *end = SvPVX(stuff) + SvCUR(stuff);
13070 SvCUR_set(stuff, SvCUR(stuff) - 1);
13081 if (PL_madskills) {
13083 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
13085 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
13088 PL_bufptr = PL_bufend;
13089 CopLINE_inc(PL_curcop);
13090 got_some = lex_next_chunk(0);
13091 CopLINE_dec(PL_curcop);
13094 tokenstart = PL_bufptr;
13102 if (SvCUR(stuff)) {
13105 PL_lex_state = LEX_NORMAL;
13106 start_force(PL_curforce);
13107 NEXTVAL_NEXTTOKE.ival = 0;
13111 PL_lex_state = LEX_FORMLINE;
13113 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
13115 else if (PL_encoding)
13116 sv_recode_to_utf8(stuff, PL_encoding);
13118 start_force(PL_curforce);
13119 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
13121 start_force(PL_curforce);
13122 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
13126 SvREFCNT_dec(stuff);
13128 PL_lex_formbrack = 0;
13132 if (PL_madskills) {
13134 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
13136 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
13137 PL_thiswhite = savewhite;
13144 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
13147 const I32 oldsavestack_ix = PL_savestack_ix;
13148 CV* const outsidecv = PL_compcv;
13151 assert(SvTYPE(PL_compcv) == SVt_PVCV);
13153 SAVEI32(PL_subline);
13154 save_item(PL_subname);
13155 SAVESPTR(PL_compcv);
13157 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
13158 CvFLAGS(PL_compcv) |= flags;
13160 PL_subline = CopLINE(PL_curcop);
13161 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
13162 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
13163 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
13165 return oldsavestack_ix;
13169 #pragma segment Perl_yylex
13172 S_yywarn(pTHX_ const char *const s)
13176 PERL_ARGS_ASSERT_YYWARN;
13178 PL_in_eval |= EVAL_WARNONLY;
13180 PL_in_eval &= ~EVAL_WARNONLY;
13185 Perl_yyerror(pTHX_ const char *const s)
13188 const char *where = NULL;
13189 const char *context = NULL;
13192 int yychar = PL_parser->yychar;
13194 PERL_ARGS_ASSERT_YYERROR;
13196 if (!yychar || (yychar == ';' && !PL_rsfp))
13198 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
13199 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
13200 PL_oldbufptr != PL_bufptr) {
13203 The code below is removed for NetWare because it abends/crashes on NetWare
13204 when the script has error such as not having the closing quotes like:
13205 if ($var eq "value)
13206 Checking of white spaces is anyway done in NetWare code.
13209 while (isSPACE(*PL_oldoldbufptr))
13212 context = PL_oldoldbufptr;
13213 contlen = PL_bufptr - PL_oldoldbufptr;
13215 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
13216 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
13219 The code below is removed for NetWare because it abends/crashes on NetWare
13220 when the script has error such as not having the closing quotes like:
13221 if ($var eq "value)
13222 Checking of white spaces is anyway done in NetWare code.
13225 while (isSPACE(*PL_oldbufptr))
13228 context = PL_oldbufptr;
13229 contlen = PL_bufptr - PL_oldbufptr;
13231 else if (yychar > 255)
13232 where = "next token ???";
13233 else if (yychar == -2) { /* YYEMPTY */
13234 if (PL_lex_state == LEX_NORMAL ||
13235 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
13236 where = "at end of line";
13237 else if (PL_lex_inpat)
13238 where = "within pattern";
13240 where = "within string";
13243 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
13245 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
13246 else if (isPRINT_LC(yychar)) {
13247 const char string = yychar;
13248 sv_catpvn(where_sv, &string, 1);
13251 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
13252 where = SvPVX_const(where_sv);
13254 msg = sv_2mortal(newSVpv(s, 0));
13255 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
13256 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
13258 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
13260 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
13261 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
13262 Perl_sv_catpvf(aTHX_ msg,
13263 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
13264 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
13267 if (PL_in_eval & EVAL_WARNONLY) {
13268 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
13272 if (PL_error_count >= 10) {
13273 if (PL_in_eval && SvCUR(ERRSV))
13274 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
13275 SVfARG(ERRSV), OutCopFILE(PL_curcop));
13277 Perl_croak(aTHX_ "%s has too many errors.\n",
13278 OutCopFILE(PL_curcop));
13281 PL_in_my_stash = NULL;
13285 #pragma segment Main
13289 S_swallow_bom(pTHX_ U8 *s)
13292 const STRLEN slen = SvCUR(PL_linestr);
13294 PERL_ARGS_ASSERT_SWALLOW_BOM;
13298 if (s[1] == 0xFE) {
13299 /* UTF-16 little-endian? (or UTF32-LE?) */
13300 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
13301 Perl_croak(aTHX_ "Unsupported script encoding UTF32-LE");
13302 #ifndef PERL_NO_UTF16_FILTER
13303 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF16-LE script encoding (BOM)\n");
13305 if (PL_bufend > (char*)s) {
13306 s = add_utf16_textfilter(s, TRUE);
13309 Perl_croak(aTHX_ "Unsupported script encoding UTF16-LE");
13314 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
13315 #ifndef PERL_NO_UTF16_FILTER
13316 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
13318 if (PL_bufend > (char *)s) {
13319 s = add_utf16_textfilter(s, FALSE);
13322 Perl_croak(aTHX_ "Unsupported script encoding UTF16-BE");
13327 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
13328 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13329 s += 3; /* UTF-8 */
13335 if (s[2] == 0xFE && s[3] == 0xFF) {
13336 /* UTF-32 big-endian */
13337 Perl_croak(aTHX_ "Unsupported script encoding UTF32-BE");
13340 else if (s[2] == 0 && s[3] != 0) {
13343 * are a good indicator of UTF-16BE. */
13344 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
13345 s = add_utf16_textfilter(s, FALSE);
13350 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
13351 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13352 s += 4; /* UTF-8 */
13358 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
13361 * are a good indicator of UTF-16LE. */
13362 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
13363 s = add_utf16_textfilter(s, TRUE);
13370 #ifndef PERL_NO_UTF16_FILTER
13372 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
13375 SV *const filter = FILTER_DATA(idx);
13376 /* We re-use this each time round, throwing the contents away before we
13378 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
13379 SV *const utf8_buffer = filter;
13380 IV status = IoPAGE(filter);
13381 const bool reverse = (bool) IoLINES(filter);
13384 /* As we're automatically added, at the lowest level, and hence only called
13385 from this file, we can be sure that we're not called in block mode. Hence
13386 don't bother writing code to deal with block mode. */
13388 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
13391 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
13393 DEBUG_P(PerlIO_printf(Perl_debug_log,
13394 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13395 FPTR2DPTR(void *, S_utf16_textfilter),
13396 reverse ? 'l' : 'b', idx, maxlen, status,
13397 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13404 /* First, look in our buffer of existing UTF-8 data: */
13405 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
13409 } else if (status == 0) {
13411 IoPAGE(filter) = 0;
13412 nl = SvEND(utf8_buffer);
13415 STRLEN got = nl - SvPVX(utf8_buffer);
13416 /* Did we have anything to append? */
13418 sv_catpvn(sv, SvPVX(utf8_buffer), got);
13419 /* Everything else in this code works just fine if SVp_POK isn't
13420 set. This, however, needs it, and we need it to work, else
13421 we loop infinitely because the buffer is never consumed. */
13422 sv_chop(utf8_buffer, nl);
13426 /* OK, not a complete line there, so need to read some more UTF-16.
13427 Read an extra octect if the buffer currently has an odd number. */
13431 if (SvCUR(utf16_buffer) >= 2) {
13432 /* Location of the high octet of the last complete code point.
13433 Gosh, UTF-16 is a pain. All the benefits of variable length,
13434 *coupled* with all the benefits of partial reads and
13436 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
13437 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
13439 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
13443 /* We have the first half of a surrogate. Read more. */
13444 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
13447 status = FILTER_READ(idx + 1, utf16_buffer,
13448 160 + (SvCUR(utf16_buffer) & 1));
13449 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
13450 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
13453 IoPAGE(filter) = status;
13458 chars = SvCUR(utf16_buffer) >> 1;
13459 have = SvCUR(utf8_buffer);
13460 SvGROW(utf8_buffer, have + chars * 3 + 1);
13463 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
13464 (U8*)SvPVX_const(utf8_buffer) + have,
13465 chars * 2, &newlen);
13467 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
13468 (U8*)SvPVX_const(utf8_buffer) + have,
13469 chars * 2, &newlen);
13471 SvCUR_set(utf8_buffer, have + newlen);
13474 /* No need to keep this SV "well-formed" with a '\0' after the end, as
13475 it's private to us, and utf16_to_utf8{,reversed} take a
13476 (pointer,length) pair, rather than a NUL-terminated string. */
13477 if(SvCUR(utf16_buffer) & 1) {
13478 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
13479 SvCUR_set(utf16_buffer, 1);
13481 SvCUR_set(utf16_buffer, 0);
13484 DEBUG_P(PerlIO_printf(Perl_debug_log,
13485 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13487 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13488 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
13493 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
13495 SV *filter = filter_add(S_utf16_textfilter, NULL);
13497 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
13498 sv_setpvs(filter, "");
13499 IoLINES(filter) = reversed;
13500 IoPAGE(filter) = 1; /* Not EOF */
13502 /* Sadly, we have to return a valid pointer, come what may, so we have to
13503 ignore any error return from this. */
13504 SvCUR_set(PL_linestr, 0);
13505 if (FILTER_READ(0, PL_linestr, 0)) {
13506 SvUTF8_on(PL_linestr);
13508 SvUTF8_on(PL_linestr);
13510 PL_bufend = SvEND(PL_linestr);
13511 return (U8*)SvPVX(PL_linestr);
13516 Returns a pointer to the next character after the parsed
13517 vstring, as well as updating the passed in sv.
13519 Function must be called like
13522 s = scan_vstring(s,e,sv);
13524 where s and e are the start and end of the string.
13525 The sv should already be large enough to store the vstring
13526 passed in, for performance reasons.
13531 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
13534 const char *pos = s;
13535 const char *start = s;
13537 PERL_ARGS_ASSERT_SCAN_VSTRING;
13539 if (*pos == 'v') pos++; /* get past 'v' */
13540 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13542 if ( *pos != '.') {
13543 /* this may not be a v-string if followed by => */
13544 const char *next = pos;
13545 while (next < e && isSPACE(*next))
13547 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
13548 /* return string not v-string */
13549 sv_setpvn(sv,(char *)s,pos-s);
13550 return (char *)pos;
13554 if (!isALPHA(*pos)) {
13555 U8 tmpbuf[UTF8_MAXBYTES+1];
13558 s++; /* get past 'v' */
13563 /* this is atoi() that tolerates underscores */
13566 const char *end = pos;
13568 while (--end >= s) {
13570 const UV orev = rev;
13571 rev += (*end - '0') * mult;
13574 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13575 "Integer overflow in decimal number");
13579 if (rev > 0x7FFFFFFF)
13580 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
13582 /* Append native character for the rev point */
13583 tmpend = uvchr_to_utf8(tmpbuf, rev);
13584 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
13585 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
13587 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
13593 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13597 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
13604 Perl_keyword_plugin_standard(pTHX_
13605 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
13607 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
13608 PERL_UNUSED_CONTEXT;
13609 PERL_UNUSED_ARG(keyword_ptr);
13610 PERL_UNUSED_ARG(keyword_len);
13611 PERL_UNUSED_ARG(op_ptr);
13612 return KEYWORD_PLUGIN_DECLINE;
13617 * c-indentation-style: bsd
13618 * c-basic-offset: 4
13619 * indent-tabs-mode: t
13622 * ex: set ts=8 sts=4 sw=4 noet: