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("unicode_strings")-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|const 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_ const char *pv, STRLEN len, U32 flags)
943 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
944 if (flags & ~(LEX_STUFF_UTF8))
945 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
947 if (flags & LEX_STUFF_UTF8) {
951 const char *p, *e = pv+len;
952 for (p = pv; p != e; p++)
953 highhalf += !!(((U8)*p) & 0x80);
956 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
957 bufptr = PL_parser->bufptr;
958 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
959 SvCUR_set(PL_parser->linestr,
960 SvCUR(PL_parser->linestr) + len+highhalf);
961 PL_parser->bufend += len+highhalf;
962 for (p = pv; p != e; p++) {
965 *bufptr++ = (char)(0xc0 | (c >> 6));
966 *bufptr++ = (char)(0x80 | (c & 0x3f));
973 if (flags & LEX_STUFF_UTF8) {
975 const char *p, *e = pv+len;
976 for (p = pv; p != e; p++) {
979 Perl_croak(aTHX_ "Lexing code attempted to stuff "
980 "non-Latin-1 character into Latin-1 input");
981 } else if (c >= 0xc2 && p+1 != e &&
982 (((U8)p[1]) & 0xc0) == 0x80) {
985 } else if (c >= 0x80) {
986 /* malformed UTF-8 */
988 SAVESPTR(PL_warnhook);
989 PL_warnhook = PERL_WARNHOOK_FATAL;
990 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
996 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
997 bufptr = PL_parser->bufptr;
998 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
999 SvCUR_set(PL_parser->linestr,
1000 SvCUR(PL_parser->linestr) + len-highhalf);
1001 PL_parser->bufend += len-highhalf;
1002 for (p = pv; p != e; p++) {
1005 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1008 *bufptr++ = (char)c;
1013 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1014 bufptr = PL_parser->bufptr;
1015 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1016 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1017 PL_parser->bufend += len;
1018 Copy(pv, bufptr, len, char);
1024 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1026 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1027 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1028 reallocating the buffer if necessary. This means that lexing code that
1029 runs later will see the characters as if they had appeared in the input.
1030 It is not recommended to do this as part of normal parsing, and most
1031 uses of this facility run the risk of the inserted characters being
1032 interpreted in an unintended manner.
1034 The string to be inserted is the string value of I<sv>. The characters
1035 are recoded for the lexer buffer, according to how the buffer is currently
1036 being interpreted (L</lex_bufutf8>). If a string to be interpreted is
1037 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1038 need to construct a scalar.
1044 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1048 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1050 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1052 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1056 =for apidoc Amx|void|lex_unstuff|char *ptr
1058 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1059 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1060 This hides the discarded text from any lexing code that runs later,
1061 as if the text had never appeared.
1063 This is not the normal way to consume lexed text. For that, use
1070 Perl_lex_unstuff(pTHX_ char *ptr)
1074 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1075 buf = PL_parser->bufptr;
1077 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1080 bufend = PL_parser->bufend;
1082 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1083 unstuff_len = ptr - buf;
1084 Move(ptr, buf, bufend+1-ptr, char);
1085 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1086 PL_parser->bufend = bufend - unstuff_len;
1090 =for apidoc Amx|void|lex_read_to|char *ptr
1092 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1093 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1094 performing the correct bookkeeping whenever a newline character is passed.
1095 This is the normal way to consume lexed text.
1097 Interpretation of the buffer's octets can be abstracted out by
1098 using the slightly higher-level functions L</lex_peek_unichar> and
1099 L</lex_read_unichar>.
1105 Perl_lex_read_to(pTHX_ char *ptr)
1108 PERL_ARGS_ASSERT_LEX_READ_TO;
1109 s = PL_parser->bufptr;
1110 if (ptr < s || ptr > PL_parser->bufend)
1111 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1112 for (; s != ptr; s++)
1114 CopLINE_inc(PL_curcop);
1115 PL_parser->linestart = s+1;
1117 PL_parser->bufptr = ptr;
1121 =for apidoc Amx|void|lex_discard_to|char *ptr
1123 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1124 up to I<ptr>. The remaining content of the buffer will be moved, and
1125 all pointers into the buffer updated appropriately. I<ptr> must not
1126 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1127 it is not permitted to discard text that has yet to be lexed.
1129 Normally it is not necessarily to do this directly, because it suffices to
1130 use the implicit discarding behaviour of L</lex_next_chunk> and things
1131 based on it. However, if a token stretches across multiple lines,
1132 and the lexing code has kept multiple lines of text in the buffer for
1133 that purpose, then after completion of the token it would be wise to
1134 explicitly discard the now-unneeded earlier lines, to avoid future
1135 multi-line tokens growing the buffer without bound.
1141 Perl_lex_discard_to(pTHX_ char *ptr)
1145 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1146 buf = SvPVX(PL_parser->linestr);
1148 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1151 if (ptr > PL_parser->bufptr)
1152 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1153 discard_len = ptr - buf;
1154 if (PL_parser->oldbufptr < ptr)
1155 PL_parser->oldbufptr = ptr;
1156 if (PL_parser->oldoldbufptr < ptr)
1157 PL_parser->oldoldbufptr = ptr;
1158 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1159 PL_parser->last_uni = NULL;
1160 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1161 PL_parser->last_lop = NULL;
1162 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1163 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1164 PL_parser->bufend -= discard_len;
1165 PL_parser->bufptr -= discard_len;
1166 PL_parser->oldbufptr -= discard_len;
1167 PL_parser->oldoldbufptr -= discard_len;
1168 if (PL_parser->last_uni)
1169 PL_parser->last_uni -= discard_len;
1170 if (PL_parser->last_lop)
1171 PL_parser->last_lop -= discard_len;
1175 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1177 Reads in the next chunk of text to be lexed, appending it to
1178 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1179 looked to the end of the current chunk and wants to know more. It is
1180 usual, but not necessary, for lexing to have consumed the entirety of
1181 the current chunk at this time.
1183 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1184 chunk (i.e., the current chunk has been entirely consumed), normally the
1185 current chunk will be discarded at the same time that the new chunk is
1186 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1187 will not be discarded. If the current chunk has not been entirely
1188 consumed, then it will not be discarded regardless of the flag.
1190 Returns true if some new text was added to the buffer, or false if the
1191 buffer has reached the end of the input text.
1196 #define LEX_FAKE_EOF 0x80000000
1199 Perl_lex_next_chunk(pTHX_ U32 flags)
1203 STRLEN old_bufend_pos, new_bufend_pos;
1204 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1205 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1206 bool got_some_for_debugger = 0;
1208 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1209 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1210 linestr = PL_parser->linestr;
1211 buf = SvPVX(linestr);
1212 if (!(flags & LEX_KEEP_PREVIOUS) &&
1213 PL_parser->bufptr == PL_parser->bufend) {
1214 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1216 if (PL_parser->last_uni != PL_parser->bufend)
1217 PL_parser->last_uni = NULL;
1218 if (PL_parser->last_lop != PL_parser->bufend)
1219 PL_parser->last_lop = NULL;
1220 last_uni_pos = last_lop_pos = 0;
1224 old_bufend_pos = PL_parser->bufend - buf;
1225 bufptr_pos = PL_parser->bufptr - buf;
1226 oldbufptr_pos = PL_parser->oldbufptr - buf;
1227 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1228 linestart_pos = PL_parser->linestart - buf;
1229 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1230 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1232 if (flags & LEX_FAKE_EOF) {
1234 } else if (!PL_parser->rsfp) {
1236 } else if (filter_gets(linestr, old_bufend_pos)) {
1238 got_some_for_debugger = 1;
1240 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1241 sv_setpvs(linestr, "");
1243 /* End of real input. Close filehandle (unless it was STDIN),
1244 * then add implicit termination.
1246 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1247 PerlIO_clearerr(PL_parser->rsfp);
1248 else if (PL_parser->rsfp)
1249 (void)PerlIO_close(PL_parser->rsfp);
1250 PL_parser->rsfp = NULL;
1251 PL_doextract = FALSE;
1253 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1256 if (!PL_in_eval && PL_minus_p) {
1258 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1259 PL_minus_n = PL_minus_p = 0;
1260 } else if (!PL_in_eval && PL_minus_n) {
1261 sv_catpvs(linestr, /*{*/";}");
1264 sv_catpvs(linestr, ";");
1267 buf = SvPVX(linestr);
1268 new_bufend_pos = SvCUR(linestr);
1269 PL_parser->bufend = buf + new_bufend_pos;
1270 PL_parser->bufptr = buf + bufptr_pos;
1271 PL_parser->oldbufptr = buf + oldbufptr_pos;
1272 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1273 PL_parser->linestart = buf + linestart_pos;
1274 if (PL_parser->last_uni)
1275 PL_parser->last_uni = buf + last_uni_pos;
1276 if (PL_parser->last_lop)
1277 PL_parser->last_lop = buf + last_lop_pos;
1278 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1279 PL_curstash != PL_debstash) {
1280 /* debugger active and we're not compiling the debugger code,
1281 * so store the line into the debugger's array of lines
1283 update_debugger_info(NULL, buf+old_bufend_pos,
1284 new_bufend_pos-old_bufend_pos);
1290 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1292 Looks ahead one (Unicode) character in the text currently being lexed.
1293 Returns the codepoint (unsigned integer value) of the next character,
1294 or -1 if lexing has reached the end of the input text. To consume the
1295 peeked character, use L</lex_read_unichar>.
1297 If the next character is in (or extends into) the next chunk of input
1298 text, the next chunk will be read in. Normally the current chunk will be
1299 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1300 then the current chunk will not be discarded.
1302 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1303 is encountered, an exception is generated.
1309 Perl_lex_peek_unichar(pTHX_ U32 flags)
1313 if (flags & ~(LEX_KEEP_PREVIOUS))
1314 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1315 s = PL_parser->bufptr;
1316 bufend = PL_parser->bufend;
1322 if (!lex_next_chunk(flags))
1324 s = PL_parser->bufptr;
1325 bufend = PL_parser->bufend;
1331 len = PL_utf8skip[head];
1332 while ((STRLEN)(bufend-s) < len) {
1333 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1335 s = PL_parser->bufptr;
1336 bufend = PL_parser->bufend;
1339 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1340 if (retlen == (STRLEN)-1) {
1341 /* malformed UTF-8 */
1343 SAVESPTR(PL_warnhook);
1344 PL_warnhook = PERL_WARNHOOK_FATAL;
1345 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1351 if (!lex_next_chunk(flags))
1353 s = PL_parser->bufptr;
1360 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1362 Reads the next (Unicode) character in the text currently being lexed.
1363 Returns the codepoint (unsigned integer value) of the character read,
1364 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1365 if lexing has reached the end of the input text. To non-destructively
1366 examine the next character, use L</lex_peek_unichar> instead.
1368 If the next character is in (or extends into) the next chunk of input
1369 text, the next chunk will be read in. Normally the current chunk will be
1370 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1371 then the current chunk will not be discarded.
1373 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1374 is encountered, an exception is generated.
1380 Perl_lex_read_unichar(pTHX_ U32 flags)
1383 if (flags & ~(LEX_KEEP_PREVIOUS))
1384 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1385 c = lex_peek_unichar(flags);
1388 CopLINE_inc(PL_curcop);
1389 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1395 =for apidoc Amx|void|lex_read_space|U32 flags
1397 Reads optional spaces, in Perl style, in the text currently being
1398 lexed. The spaces may include ordinary whitespace characters and
1399 Perl-style comments. C<#line> directives are processed if encountered.
1400 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1401 at a non-space character (or the end of the input text).
1403 If spaces extend into the next chunk of input text, the next chunk will
1404 be read in. Normally the current chunk will be discarded at the same
1405 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1406 chunk will not be discarded.
1411 #define LEX_NO_NEXT_CHUNK 0x80000000
1414 Perl_lex_read_space(pTHX_ U32 flags)
1417 bool need_incline = 0;
1418 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1419 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1422 sv_free(PL_skipwhite);
1423 PL_skipwhite = NULL;
1426 PL_skipwhite = newSVpvs("");
1427 #endif /* PERL_MAD */
1428 s = PL_parser->bufptr;
1429 bufend = PL_parser->bufend;
1435 } while (!(c == '\n' || (c == 0 && s == bufend)));
1436 } else if (c == '\n') {
1438 PL_parser->linestart = s;
1443 } else if (isSPACE(c)) {
1445 } else if (c == 0 && s == bufend) {
1449 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1450 #endif /* PERL_MAD */
1451 if (flags & LEX_NO_NEXT_CHUNK)
1453 PL_parser->bufptr = s;
1454 CopLINE_inc(PL_curcop);
1455 got_more = lex_next_chunk(flags);
1456 CopLINE_dec(PL_curcop);
1457 s = PL_parser->bufptr;
1458 bufend = PL_parser->bufend;
1461 if (need_incline && PL_parser->rsfp) {
1471 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1472 #endif /* PERL_MAD */
1473 PL_parser->bufptr = s;
1478 * This subroutine has nothing to do with tilting, whether at windmills
1479 * or pinball tables. Its name is short for "increment line". It
1480 * increments the current line number in CopLINE(PL_curcop) and checks
1481 * to see whether the line starts with a comment of the form
1482 * # line 500 "foo.pm"
1483 * If so, it sets the current line number and file to the values in the comment.
1487 S_incline(pTHX_ const char *s)
1494 PERL_ARGS_ASSERT_INCLINE;
1496 CopLINE_inc(PL_curcop);
1499 while (SPACE_OR_TAB(*s))
1501 if (strnEQ(s, "line", 4))
1505 if (SPACE_OR_TAB(*s))
1509 while (SPACE_OR_TAB(*s))
1517 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1519 while (SPACE_OR_TAB(*s))
1521 if (*s == '"' && (t = strchr(s+1, '"'))) {
1527 while (!isSPACE(*t))
1531 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1533 if (*e != '\n' && *e != '\0')
1534 return; /* false alarm */
1537 const STRLEN len = t - s;
1538 #ifndef USE_ITHREADS
1539 SV *const temp_sv = CopFILESV(PL_curcop);
1544 cf = SvPVX(temp_sv);
1545 tmplen = SvCUR(temp_sv);
1551 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
1552 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1553 * to *{"::_<newfilename"} */
1554 /* However, the long form of evals is only turned on by the
1555 debugger - usually they're "(eval %lu)" */
1559 STRLEN tmplen2 = len;
1560 if (tmplen + 2 <= sizeof smallbuf)
1563 Newx(tmpbuf, tmplen + 2, char);
1566 memcpy(tmpbuf + 2, cf, tmplen);
1568 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1573 if (tmplen2 + 2 <= sizeof smallbuf)
1576 Newx(tmpbuf2, tmplen2 + 2, char);
1578 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1579 /* Either they malloc'd it, or we malloc'd it,
1580 so no prefix is present in ours. */
1585 memcpy(tmpbuf2 + 2, s, tmplen2);
1588 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1590 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1591 /* adjust ${"::_<newfilename"} to store the new file name */
1592 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1593 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1594 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1597 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1599 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1602 CopFILE_free(PL_curcop);
1603 CopFILE_setn(PL_curcop, s, len);
1605 CopLINE_set(PL_curcop, atoi(n)-1);
1609 /* skip space before PL_thistoken */
1612 S_skipspace0(pTHX_ register char *s)
1614 PERL_ARGS_ASSERT_SKIPSPACE0;
1621 PL_thiswhite = newSVpvs("");
1622 sv_catsv(PL_thiswhite, PL_skipwhite);
1623 sv_free(PL_skipwhite);
1626 PL_realtokenstart = s - SvPVX(PL_linestr);
1630 /* skip space after PL_thistoken */
1633 S_skipspace1(pTHX_ register char *s)
1635 const char *start = s;
1636 I32 startoff = start - SvPVX(PL_linestr);
1638 PERL_ARGS_ASSERT_SKIPSPACE1;
1643 start = SvPVX(PL_linestr) + startoff;
1644 if (!PL_thistoken && PL_realtokenstart >= 0) {
1645 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1646 PL_thistoken = newSVpvn(tstart, start - tstart);
1648 PL_realtokenstart = -1;
1651 PL_nextwhite = newSVpvs("");
1652 sv_catsv(PL_nextwhite, PL_skipwhite);
1653 sv_free(PL_skipwhite);
1660 S_skipspace2(pTHX_ register char *s, SV **svp)
1663 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1664 const I32 startoff = s - SvPVX(PL_linestr);
1666 PERL_ARGS_ASSERT_SKIPSPACE2;
1669 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1670 if (!PL_madskills || !svp)
1672 start = SvPVX(PL_linestr) + startoff;
1673 if (!PL_thistoken && PL_realtokenstart >= 0) {
1674 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1675 PL_thistoken = newSVpvn(tstart, start - tstart);
1676 PL_realtokenstart = -1;
1680 *svp = newSVpvs("");
1681 sv_setsv(*svp, PL_skipwhite);
1682 sv_free(PL_skipwhite);
1691 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1693 AV *av = CopFILEAVx(PL_curcop);
1695 SV * const sv = newSV_type(SVt_PVMG);
1697 sv_setsv(sv, orig_sv);
1699 sv_setpvn(sv, buf, len);
1702 av_store(av, (I32)CopLINE(PL_curcop), sv);
1708 * Called to gobble the appropriate amount and type of whitespace.
1709 * Skips comments as well.
1713 S_skipspace(pTHX_ register char *s)
1717 #endif /* PERL_MAD */
1718 PERL_ARGS_ASSERT_SKIPSPACE;
1721 sv_free(PL_skipwhite);
1722 PL_skipwhite = NULL;
1724 #endif /* PERL_MAD */
1725 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1726 while (s < PL_bufend && SPACE_OR_TAB(*s))
1729 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1731 lex_read_space(LEX_KEEP_PREVIOUS |
1732 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1733 LEX_NO_NEXT_CHUNK : 0));
1735 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1736 if (PL_linestart > PL_bufptr)
1737 PL_bufptr = PL_linestart;
1742 PL_skipwhite = newSVpvn(start, s-start);
1743 #endif /* PERL_MAD */
1749 * Check the unary operators to ensure there's no ambiguity in how they're
1750 * used. An ambiguous piece of code would be:
1752 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1753 * the +5 is its argument.
1763 if (PL_oldoldbufptr != PL_last_uni)
1765 while (isSPACE(*PL_last_uni))
1768 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1770 if ((t = strchr(s, '(')) && t < PL_bufptr)
1773 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1774 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1775 (int)(s - PL_last_uni), PL_last_uni);
1779 * LOP : macro to build a list operator. Its behaviour has been replaced
1780 * with a subroutine, S_lop() for which LOP is just another name.
1783 #define LOP(f,x) return lop(f,x,s)
1787 * Build a list operator (or something that might be one). The rules:
1788 * - if we have a next token, then it's a list operator [why?]
1789 * - if the next thing is an opening paren, then it's a function
1790 * - else it's a list operator
1794 S_lop(pTHX_ I32 f, int x, char *s)
1798 PERL_ARGS_ASSERT_LOP;
1804 PL_last_lop = PL_oldbufptr;
1805 PL_last_lop_op = (OPCODE)f;
1808 return REPORT(LSTOP);
1811 return REPORT(LSTOP);
1814 return REPORT(FUNC);
1817 return REPORT(FUNC);
1819 return REPORT(LSTOP);
1825 * Sets up for an eventual force_next(). start_force(0) basically does
1826 * an unshift, while start_force(-1) does a push. yylex removes items
1831 S_start_force(pTHX_ int where)
1835 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1836 where = PL_lasttoke;
1837 assert(PL_curforce < 0 || PL_curforce == where);
1838 if (PL_curforce != where) {
1839 for (i = PL_lasttoke; i > where; --i) {
1840 PL_nexttoke[i] = PL_nexttoke[i-1];
1844 if (PL_curforce < 0) /* in case of duplicate start_force() */
1845 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1846 PL_curforce = where;
1849 curmad('^', newSVpvs(""));
1850 CURMAD('_', PL_nextwhite);
1855 S_curmad(pTHX_ char slot, SV *sv)
1861 if (PL_curforce < 0)
1862 where = &PL_thismad;
1864 where = &PL_nexttoke[PL_curforce].next_mad;
1870 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1872 else if (PL_encoding) {
1873 sv_recode_to_utf8(sv, PL_encoding);
1878 /* keep a slot open for the head of the list? */
1879 if (slot != '_' && *where && (*where)->mad_key == '^') {
1880 (*where)->mad_key = slot;
1881 sv_free(MUTABLE_SV(((*where)->mad_val)));
1882 (*where)->mad_val = (void*)sv;
1885 addmad(newMADsv(slot, sv), where, 0);
1888 # define start_force(where) NOOP
1889 # define curmad(slot, sv) NOOP
1894 * When the lexer realizes it knows the next token (for instance,
1895 * it is reordering tokens for the parser) then it can call S_force_next
1896 * to know what token to return the next time the lexer is called. Caller
1897 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1898 * and possibly PL_expect to ensure the lexer handles the token correctly.
1902 S_force_next(pTHX_ I32 type)
1907 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1908 tokereport(type, &NEXTVAL_NEXTTOKE);
1912 if (PL_curforce < 0)
1913 start_force(PL_lasttoke);
1914 PL_nexttoke[PL_curforce].next_type = type;
1915 if (PL_lex_state != LEX_KNOWNEXT)
1916 PL_lex_defer = PL_lex_state;
1917 PL_lex_state = LEX_KNOWNEXT;
1918 PL_lex_expect = PL_expect;
1921 PL_nexttype[PL_nexttoke] = type;
1923 if (PL_lex_state != LEX_KNOWNEXT) {
1924 PL_lex_defer = PL_lex_state;
1925 PL_lex_expect = PL_expect;
1926 PL_lex_state = LEX_KNOWNEXT;
1932 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1935 SV * const sv = newSVpvn_utf8(start, len,
1938 && !is_ascii_string((const U8*)start, len)
1939 && is_utf8_string((const U8*)start, len));
1945 * When the lexer knows the next thing is a word (for instance, it has
1946 * just seen -> and it knows that the next char is a word char, then
1947 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1951 * char *start : buffer position (must be within PL_linestr)
1952 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1953 * int check_keyword : if true, Perl checks to make sure the word isn't
1954 * a keyword (do this if the word is a label, e.g. goto FOO)
1955 * int allow_pack : if true, : characters will also be allowed (require,
1956 * use, etc. do this)
1957 * int allow_initial_tick : used by the "sub" lexer only.
1961 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1967 PERL_ARGS_ASSERT_FORCE_WORD;
1969 start = SKIPSPACE1(start);
1971 if (isIDFIRST_lazy_if(s,UTF) ||
1972 (allow_pack && *s == ':') ||
1973 (allow_initial_tick && *s == '\'') )
1975 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1976 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1978 start_force(PL_curforce);
1980 curmad('X', newSVpvn(start,s-start));
1981 if (token == METHOD) {
1986 PL_expect = XOPERATOR;
1990 curmad('g', newSVpvs( "forced" ));
1991 NEXTVAL_NEXTTOKE.opval
1992 = (OP*)newSVOP(OP_CONST,0,
1993 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1994 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2002 * Called when the lexer wants $foo *foo &foo etc, but the program
2003 * text only contains the "foo" portion. The first argument is a pointer
2004 * to the "foo", and the second argument is the type symbol to prefix.
2005 * Forces the next token to be a "WORD".
2006 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2010 S_force_ident(pTHX_ register const char *s, int kind)
2014 PERL_ARGS_ASSERT_FORCE_IDENT;
2017 const STRLEN len = strlen(s);
2018 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
2019 start_force(PL_curforce);
2020 NEXTVAL_NEXTTOKE.opval = o;
2023 o->op_private = OPpCONST_ENTERED;
2024 /* XXX see note in pp_entereval() for why we forgo typo
2025 warnings if the symbol must be introduced in an eval.
2027 gv_fetchpvn_flags(s, len,
2028 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2030 kind == '$' ? SVt_PV :
2031 kind == '@' ? SVt_PVAV :
2032 kind == '%' ? SVt_PVHV :
2040 Perl_str_to_version(pTHX_ SV *sv)
2045 const char *start = SvPV_const(sv,len);
2046 const char * const end = start + len;
2047 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2049 PERL_ARGS_ASSERT_STR_TO_VERSION;
2051 while (start < end) {
2055 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2060 retval += ((NV)n)/nshift;
2069 * Forces the next token to be a version number.
2070 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2071 * and if "guessing" is TRUE, then no new token is created (and the caller
2072 * must use an alternative parsing method).
2076 S_force_version(pTHX_ char *s, int guessing)
2082 I32 startoff = s - SvPVX(PL_linestr);
2085 PERL_ARGS_ASSERT_FORCE_VERSION;
2093 while (isDIGIT(*d) || *d == '_' || *d == '.')
2097 start_force(PL_curforce);
2098 curmad('X', newSVpvn(s,d-s));
2101 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2103 #ifdef USE_LOCALE_NUMERIC
2104 char *loc = setlocale(LC_NUMERIC, "C");
2106 s = scan_num(s, &pl_yylval);
2107 #ifdef USE_LOCALE_NUMERIC
2108 setlocale(LC_NUMERIC, loc);
2110 version = pl_yylval.opval;
2111 ver = cSVOPx(version)->op_sv;
2112 if (SvPOK(ver) && !SvNIOK(ver)) {
2113 SvUPGRADE(ver, SVt_PVNV);
2114 SvNV_set(ver, str_to_version(ver));
2115 SvNOK_on(ver); /* hint that it is a version */
2118 else if (guessing) {
2121 sv_free(PL_nextwhite); /* let next token collect whitespace */
2123 s = SvPVX(PL_linestr) + startoff;
2131 if (PL_madskills && !version) {
2132 sv_free(PL_nextwhite); /* let next token collect whitespace */
2134 s = SvPVX(PL_linestr) + startoff;
2137 /* NOTE: The parser sees the package name and the VERSION swapped */
2138 start_force(PL_curforce);
2139 NEXTVAL_NEXTTOKE.opval = version;
2146 * S_force_strict_version
2147 * Forces the next token to be a version number using strict syntax rules.
2151 S_force_strict_version(pTHX_ char *s)
2156 I32 startoff = s - SvPVX(PL_linestr);
2158 const char *errstr = NULL;
2160 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2162 while (isSPACE(*s)) /* leading whitespace */
2165 if (is_STRICT_VERSION(s,&errstr)) {
2167 s = (char *)scan_version(s, ver, 0);
2168 version = newSVOP(OP_CONST, 0, ver);
2170 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2171 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2175 yyerror(errstr); /* version required */
2180 if (PL_madskills && !version) {
2181 sv_free(PL_nextwhite); /* let next token collect whitespace */
2183 s = SvPVX(PL_linestr) + startoff;
2186 /* NOTE: The parser sees the package name and the VERSION swapped */
2187 start_force(PL_curforce);
2188 NEXTVAL_NEXTTOKE.opval = version;
2196 * Tokenize a quoted string passed in as an SV. It finds the next
2197 * chunk, up to end of string or a backslash. It may make a new
2198 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2203 S_tokeq(pTHX_ SV *sv)
2207 register char *send;
2212 PERL_ARGS_ASSERT_TOKEQ;
2217 s = SvPV_force(sv, len);
2218 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2221 while (s < send && *s != '\\')
2226 if ( PL_hints & HINT_NEW_STRING ) {
2227 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2231 if (s + 1 < send && (s[1] == '\\'))
2232 s++; /* all that, just for this */
2237 SvCUR_set(sv, d - SvPVX_const(sv));
2239 if ( PL_hints & HINT_NEW_STRING )
2240 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2245 * Now come three functions related to double-quote context,
2246 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2247 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2248 * interact with PL_lex_state, and create fake ( ... ) argument lists
2249 * to handle functions and concatenation.
2250 * They assume that whoever calls them will be setting up a fake
2251 * join call, because each subthing puts a ',' after it. This lets
2254 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2256 * (I'm not sure whether the spurious commas at the end of lcfirst's
2257 * arguments and join's arguments are created or not).
2262 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2264 * Pattern matching will set PL_lex_op to the pattern-matching op to
2265 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2267 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2269 * Everything else becomes a FUNC.
2271 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2272 * had an OP_CONST or OP_READLINE). This just sets us up for a
2273 * call to S_sublex_push().
2277 S_sublex_start(pTHX)
2280 register const I32 op_type = pl_yylval.ival;
2282 if (op_type == OP_NULL) {
2283 pl_yylval.opval = PL_lex_op;
2287 if (op_type == OP_CONST || op_type == OP_READLINE) {
2288 SV *sv = tokeq(PL_lex_stuff);
2290 if (SvTYPE(sv) == SVt_PVIV) {
2291 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2293 const char * const p = SvPV_const(sv, len);
2294 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2298 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2299 PL_lex_stuff = NULL;
2300 /* Allow <FH> // "foo" */
2301 if (op_type == OP_READLINE)
2302 PL_expect = XTERMORDORDOR;
2305 else if (op_type == OP_BACKTICK && PL_lex_op) {
2306 /* readpipe() vas overriden */
2307 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2308 pl_yylval.opval = PL_lex_op;
2310 PL_lex_stuff = NULL;
2314 PL_sublex_info.super_state = PL_lex_state;
2315 PL_sublex_info.sub_inwhat = (U16)op_type;
2316 PL_sublex_info.sub_op = PL_lex_op;
2317 PL_lex_state = LEX_INTERPPUSH;
2321 pl_yylval.opval = PL_lex_op;
2331 * Create a new scope to save the lexing state. The scope will be
2332 * ended in S_sublex_done. Returns a '(', starting the function arguments
2333 * to the uc, lc, etc. found before.
2334 * Sets PL_lex_state to LEX_INTERPCONCAT.
2343 PL_lex_state = PL_sublex_info.super_state;
2344 SAVEBOOL(PL_lex_dojoin);
2345 SAVEI32(PL_lex_brackets);
2346 SAVEI32(PL_lex_casemods);
2347 SAVEI32(PL_lex_starts);
2348 SAVEI8(PL_lex_state);
2349 SAVEVPTR(PL_lex_inpat);
2350 SAVEI16(PL_lex_inwhat);
2351 SAVECOPLINE(PL_curcop);
2352 SAVEPPTR(PL_bufptr);
2353 SAVEPPTR(PL_bufend);
2354 SAVEPPTR(PL_oldbufptr);
2355 SAVEPPTR(PL_oldoldbufptr);
2356 SAVEPPTR(PL_last_lop);
2357 SAVEPPTR(PL_last_uni);
2358 SAVEPPTR(PL_linestart);
2359 SAVESPTR(PL_linestr);
2360 SAVEGENERICPV(PL_lex_brackstack);
2361 SAVEGENERICPV(PL_lex_casestack);
2363 PL_linestr = PL_lex_stuff;
2364 PL_lex_stuff = NULL;
2366 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2367 = SvPVX(PL_linestr);
2368 PL_bufend += SvCUR(PL_linestr);
2369 PL_last_lop = PL_last_uni = NULL;
2370 SAVEFREESV(PL_linestr);
2372 PL_lex_dojoin = FALSE;
2373 PL_lex_brackets = 0;
2374 Newx(PL_lex_brackstack, 120, char);
2375 Newx(PL_lex_casestack, 12, char);
2376 PL_lex_casemods = 0;
2377 *PL_lex_casestack = '\0';
2379 PL_lex_state = LEX_INTERPCONCAT;
2380 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2382 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2383 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2384 PL_lex_inpat = PL_sublex_info.sub_op;
2386 PL_lex_inpat = NULL;
2393 * Restores lexer state after a S_sublex_push.
2400 if (!PL_lex_starts++) {
2401 SV * const sv = newSVpvs("");
2402 if (SvUTF8(PL_linestr))
2404 PL_expect = XOPERATOR;
2405 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2409 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2410 PL_lex_state = LEX_INTERPCASEMOD;
2414 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2415 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2416 PL_linestr = PL_lex_repl;
2418 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2419 PL_bufend += SvCUR(PL_linestr);
2420 PL_last_lop = PL_last_uni = NULL;
2421 SAVEFREESV(PL_linestr);
2422 PL_lex_dojoin = FALSE;
2423 PL_lex_brackets = 0;
2424 PL_lex_casemods = 0;
2425 *PL_lex_casestack = '\0';
2427 if (SvEVALED(PL_lex_repl)) {
2428 PL_lex_state = LEX_INTERPNORMAL;
2430 /* we don't clear PL_lex_repl here, so that we can check later
2431 whether this is an evalled subst; that means we rely on the
2432 logic to ensure sublex_done() is called again only via the
2433 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2436 PL_lex_state = LEX_INTERPCONCAT;
2446 PL_endwhite = newSVpvs("");
2447 sv_catsv(PL_endwhite, PL_thiswhite);
2451 sv_setpvs(PL_thistoken,"");
2453 PL_realtokenstart = -1;
2457 PL_bufend = SvPVX(PL_linestr);
2458 PL_bufend += SvCUR(PL_linestr);
2459 PL_expect = XOPERATOR;
2460 PL_sublex_info.sub_inwhat = 0;
2468 Extracts a pattern, double-quoted string, or transliteration. This
2471 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2472 processing a pattern (PL_lex_inpat is true), a transliteration
2473 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2475 Returns a pointer to the character scanned up to. If this is
2476 advanced from the start pointer supplied (i.e. if anything was
2477 successfully parsed), will leave an OP for the substring scanned
2478 in pl_yylval. Caller must intuit reason for not parsing further
2479 by looking at the next characters herself.
2483 constants: \N{NAME} only
2484 case and quoting: \U \Q \E
2485 stops on @ and $, but not for $ as tail anchor
2487 In transliterations:
2488 characters are VERY literal, except for - not at the start or end
2489 of the string, which indicates a range. If the range is in bytes,
2490 scan_const expands the range to the full set of intermediate
2491 characters. If the range is in utf8, the hyphen is replaced with
2492 a certain range mark which will be handled by pmtrans() in op.c.
2494 In double-quoted strings:
2496 double-quoted style: \r and \n
2497 constants: \x31, etc.
2498 deprecated backrefs: \1 (in substitution replacements)
2499 case and quoting: \U \Q \E
2502 scan_const does *not* construct ops to handle interpolated strings.
2503 It stops processing as soon as it finds an embedded $ or @ variable
2504 and leaves it to the caller to work out what's going on.
2506 embedded arrays (whether in pattern or not) could be:
2507 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2509 $ in double-quoted strings must be the symbol of an embedded scalar.
2511 $ in pattern could be $foo or could be tail anchor. Assumption:
2512 it's a tail anchor if $ is the last thing in the string, or if it's
2513 followed by one of "()| \r\n\t"
2515 \1 (backreferences) are turned into $1
2517 The structure of the code is
2518 while (there's a character to process) {
2519 handle transliteration ranges
2520 skip regexp comments /(?#comment)/ and codes /(?{code})/
2521 skip #-initiated comments in //x patterns
2522 check for embedded arrays
2523 check for embedded scalars
2525 deprecate \1 in substitution replacements
2526 handle string-changing backslashes \l \U \Q \E, etc.
2527 switch (what was escaped) {
2528 handle \- in a transliteration (becomes a literal -)
2529 if a pattern and not \N{, go treat as regular character
2530 handle \132 (octal characters)
2531 handle \x15 and \x{1234} (hex characters)
2532 handle \N{name} (named characters, also \N{3,5} in a pattern)
2533 handle \cV (control characters)
2534 handle printf-style backslashes (\f, \r, \n, etc)
2537 } (end if backslash)
2538 handle regular character
2539 } (end while character to read)
2544 S_scan_const(pTHX_ char *start)
2547 register char *send = PL_bufend; /* end of the constant */
2548 SV *sv = newSV(send - start); /* sv for the constant. See
2549 note below on sizing. */
2550 register char *s = start; /* start of the constant */
2551 register char *d = SvPVX(sv); /* destination for copies */
2552 bool dorange = FALSE; /* are we in a translit range? */
2553 bool didrange = FALSE; /* did we just finish a range? */
2554 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
2555 I32 this_utf8 = UTF; /* Is the source string assumed
2556 to be UTF8? But, this can
2557 show as true when the source
2558 isn't utf8, as for example
2559 when it is entirely composed
2562 /* Note on sizing: The scanned constant is placed into sv, which is
2563 * initialized by newSV() assuming one byte of output for every byte of
2564 * input. This routine expects newSV() to allocate an extra byte for a
2565 * trailing NUL, which this routine will append if it gets to the end of
2566 * the input. There may be more bytes of input than output (eg., \N{LATIN
2567 * CAPITAL LETTER A}), or more output than input if the constant ends up
2568 * recoded to utf8, but each time a construct is found that might increase
2569 * the needed size, SvGROW() is called. Its size parameter each time is
2570 * based on the best guess estimate at the time, namely the length used so
2571 * far, plus the length the current construct will occupy, plus room for
2572 * the trailing NUL, plus one byte for every input byte still unscanned */
2576 UV literal_endpoint = 0;
2577 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2580 PERL_ARGS_ASSERT_SCAN_CONST;
2582 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2583 /* If we are doing a trans and we know we want UTF8 set expectation */
2584 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2585 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2589 while (s < send || dorange) {
2591 /* get transliterations out of the way (they're most literal) */
2592 if (PL_lex_inwhat == OP_TRANS) {
2593 /* expand a range A-Z to the full set of characters. AIE! */
2595 I32 i; /* current expanded character */
2596 I32 min; /* first character in range */
2597 I32 max; /* last character in range */
2608 char * const c = (char*)utf8_hop((U8*)d, -1);
2612 *c = (char)UTF_TO_NATIVE(0xff);
2613 /* mark the range as done, and continue */
2619 i = d - SvPVX_const(sv); /* remember current offset */
2622 SvLEN(sv) + (has_utf8 ?
2623 (512 - UTF_CONTINUATION_MARK +
2626 /* How many two-byte within 0..255: 128 in UTF-8,
2627 * 96 in UTF-8-mod. */
2629 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2631 d = SvPVX(sv) + i; /* refresh d after realloc */
2635 for (j = 0; j <= 1; j++) {
2636 char * const c = (char*)utf8_hop((U8*)d, -1);
2637 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2643 max = (U8)0xff; /* only to \xff */
2644 uvmax = uv; /* \x{100} to uvmax */
2646 d = c; /* eat endpoint chars */
2651 d -= 2; /* eat the first char and the - */
2652 min = (U8)*d; /* first char in range */
2653 max = (U8)d[1]; /* last char in range */
2660 "Invalid range \"%c-%c\" in transliteration operator",
2661 (char)min, (char)max);
2665 if (literal_endpoint == 2 &&
2666 ((isLOWER(min) && isLOWER(max)) ||
2667 (isUPPER(min) && isUPPER(max)))) {
2669 for (i = min; i <= max; i++)
2671 *d++ = NATIVE_TO_NEED(has_utf8,i);
2673 for (i = min; i <= max; i++)
2675 *d++ = NATIVE_TO_NEED(has_utf8,i);
2680 for (i = min; i <= max; i++)
2683 const U8 ch = (U8)NATIVE_TO_UTF(i);
2684 if (UNI_IS_INVARIANT(ch))
2687 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2688 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2697 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2699 *d++ = (char)UTF_TO_NATIVE(0xff);
2701 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2705 /* mark the range as done, and continue */
2709 literal_endpoint = 0;
2714 /* range begins (ignore - as first or last char) */
2715 else if (*s == '-' && s+1 < send && s != start) {
2717 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2724 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2734 literal_endpoint = 0;
2735 native_range = TRUE;
2740 /* if we get here, we're not doing a transliteration */
2742 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2743 except for the last char, which will be done separately. */
2744 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2746 while (s+1 < send && *s != ')')
2747 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2749 else if (s[2] == '{' /* This should match regcomp.c */
2750 || (s[2] == '?' && s[3] == '{'))
2753 char *regparse = s + (s[2] == '{' ? 3 : 4);
2756 while (count && (c = *regparse)) {
2757 if (c == '\\' && regparse[1])
2765 if (*regparse != ')')
2766 regparse--; /* Leave one char for continuation. */
2767 while (s < regparse)
2768 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2772 /* likewise skip #-initiated comments in //x patterns */
2773 else if (*s == '#' && PL_lex_inpat &&
2774 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2775 while (s+1 < send && *s != '\n')
2776 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2779 /* check for embedded arrays
2780 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2782 else if (*s == '@' && s[1]) {
2783 if (isALNUM_lazy_if(s+1,UTF))
2785 if (strchr(":'{$", s[1]))
2787 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2788 break; /* in regexp, neither @+ nor @- are interpolated */
2791 /* check for embedded scalars. only stop if we're sure it's a
2794 else if (*s == '$') {
2795 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2797 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2799 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2800 "Possible unintended interpolation of $\\ in regex");
2802 break; /* in regexp, $ might be tail anchor */
2806 /* End of else if chain - OP_TRANS rejoin rest */
2809 if (*s == '\\' && s+1 < send) {
2810 char* e; /* Can be used for ending '}', etc. */
2814 /* deprecate \1 in strings and substitution replacements */
2815 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2816 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2818 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2823 /* string-change backslash escapes */
2824 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2828 /* In a pattern, process \N, but skip any other backslash escapes.
2829 * This is because we don't want to translate an escape sequence
2830 * into a meta symbol and have the regex compiler use the meta
2831 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2832 * in spite of this, we do have to process \N here while the proper
2833 * charnames handler is in scope. See bugs #56444 and #62056.
2834 * There is a complication because \N in a pattern may also stand
2835 * for 'match a non-nl', and not mean a charname, in which case its
2836 * processing should be deferred to the regex compiler. To be a
2837 * charname it must be followed immediately by a '{', and not look
2838 * like \N followed by a curly quantifier, i.e., not something like
2839 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2841 else if (PL_lex_inpat
2844 || regcurly(s + 1)))
2846 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2847 goto default_action;
2852 /* quoted - in transliterations */
2854 if (PL_lex_inwhat == OP_TRANS) {
2861 if ((isALPHA(*s) || isDIGIT(*s)))
2862 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2863 "Unrecognized escape \\%c passed through",
2865 /* default action is to copy the quoted character */
2866 goto default_action;
2869 /* eg. \132 indicates the octal constant 0x132 */
2870 case '0': case '1': case '2': case '3':
2871 case '4': case '5': case '6': case '7':
2875 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2878 goto NUM_ESCAPE_INSERT;
2880 /* eg. \x24 indicates the hex constant 0x24 */
2884 char* const e = strchr(s, '}');
2885 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2886 PERL_SCAN_DISALLOW_PREFIX;
2891 yyerror("Missing right brace on \\x{}");
2895 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2901 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2902 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2908 /* Insert oct or hex escaped character. There will always be
2909 * enough room in sv since such escapes will be longer than any
2910 * UTF-8 sequence they can end up as, except if they force us
2911 * to recode the rest of the string into utf8 */
2913 /* Here uv is the ordinal of the next character being added in
2914 * unicode (converted from native). */
2915 if (!UNI_IS_INVARIANT(uv)) {
2916 if (!has_utf8 && uv > 255) {
2917 /* Might need to recode whatever we have accumulated so
2918 * far if it contains any chars variant in utf8 or
2921 SvCUR_set(sv, d - SvPVX_const(sv));
2924 /* See Note on sizing above. */
2925 sv_utf8_upgrade_flags_grow(sv,
2926 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2927 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2928 d = SvPVX(sv) + SvCUR(sv);
2933 d = (char*)uvuni_to_utf8((U8*)d, uv);
2934 if (PL_lex_inwhat == OP_TRANS &&
2935 PL_sublex_info.sub_op) {
2936 PL_sublex_info.sub_op->op_private |=
2937 (PL_lex_repl ? OPpTRANS_FROM_UTF
2941 if (uv > 255 && !dorange)
2942 native_range = FALSE;
2955 /* In a non-pattern \N must be a named character, like \N{LATIN
2956 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
2957 * mean to match a non-newline. For non-patterns, named
2958 * characters are converted to their string equivalents. In
2959 * patterns, named characters are not converted to their
2960 * ultimate forms for the same reasons that other escapes
2961 * aren't. Instead, they are converted to the \N{U+...} form
2962 * to get the value from the charnames that is in effect right
2963 * now, while preserving the fact that it was a named character
2964 * so that the regex compiler knows this */
2966 /* This section of code doesn't generally use the
2967 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
2968 * a close examination of this macro and determined it is a
2969 * no-op except on utfebcdic variant characters. Every
2970 * character generated by this that would normally need to be
2971 * enclosed by this macro is invariant, so the macro is not
2972 * needed, and would complicate use of copy(). There are other
2973 * parts of this file where the macro is used inconsistently,
2974 * but are saved by it being a no-op */
2976 /* The structure of this section of code (besides checking for
2977 * errors and upgrading to utf8) is:
2978 * Further disambiguate between the two meanings of \N, and if
2979 * not a charname, go process it elsewhere
2980 * If of form \N{U+...}, pass it through if a pattern;
2981 * otherwise convert to utf8
2982 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
2983 * pattern; otherwise convert to utf8 */
2985 /* Here, s points to the 'N'; the test below is guaranteed to
2986 * succeed if we are being called on a pattern as we already
2987 * know from a test above that the next character is a '{'.
2988 * On a non-pattern \N must mean 'named sequence, which
2989 * requires braces */
2992 yyerror("Missing braces on \\N{}");
2997 /* If there is no matching '}', it is an error. */
2998 if (! (e = strchr(s, '}'))) {
2999 if (! PL_lex_inpat) {
3000 yyerror("Missing right brace on \\N{}");
3002 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3007 /* Here it looks like a named character */
3011 /* XXX This block is temporary code. \N{} implies that the
3012 * pattern is to have Unicode semantics, and therefore
3013 * currently has to be encoded in utf8. By putting it in
3014 * utf8 now, we save a whole pass in the regular expression
3015 * compiler. Once that code is changed so Unicode
3016 * semantics doesn't necessarily have to be in utf8, this
3017 * block should be removed */
3019 SvCUR_set(sv, d - SvPVX_const(sv));
3022 /* See Note on sizing above. */
3023 sv_utf8_upgrade_flags_grow(sv,
3024 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3025 /* 5 = '\N{' + cur char + NUL */
3026 (STRLEN)(send - s) + 5);
3027 d = SvPVX(sv) + SvCUR(sv);
3032 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3033 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3034 | PERL_SCAN_DISALLOW_PREFIX;
3037 /* For \N{U+...}, the '...' is a unicode value even on
3038 * EBCDIC machines */
3039 s += 2; /* Skip to next char after the 'U+' */
3041 uv = grok_hex(s, &len, &flags, NULL);
3042 if (len == 0 || len != (STRLEN)(e - s)) {
3043 yyerror("Invalid hexadecimal number in \\N{U+...}");
3050 /* Pass through to the regex compiler unchanged. The
3051 * reason we evaluated the number above is to make sure
3052 * there wasn't a syntax error. */
3053 s -= 5; /* Include the '\N{U+' */
3054 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3057 else { /* Not a pattern: convert the hex to string */
3059 /* If destination is not in utf8, unconditionally
3060 * recode it to be so. This is because \N{} implies
3061 * Unicode semantics, and scalars have to be in utf8
3062 * to guarantee those semantics */
3064 SvCUR_set(sv, d - SvPVX_const(sv));
3067 /* See Note on sizing above. */
3068 sv_utf8_upgrade_flags_grow(
3070 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3071 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3072 d = SvPVX(sv) + SvCUR(sv);
3076 /* Add the string to the output */
3077 if (UNI_IS_INVARIANT(uv)) {
3080 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3083 else { /* Here is \N{NAME} but not \N{U+...}. */
3085 SV *res; /* result from charnames */
3086 const char *str; /* the string in 'res' */
3087 STRLEN len; /* its length */
3089 /* Get the value for NAME */
3090 res = newSVpvn(s, e - s);
3091 res = new_constant( NULL, 0, "charnames",
3092 /* includes all of: \N{...} */
3093 res, NULL, s - 3, e - s + 4 );
3095 /* Most likely res will be in utf8 already since the
3096 * standard charnames uses pack U, but a custom translator
3097 * can leave it otherwise, so make sure. XXX This can be
3098 * revisited to not have charnames use utf8 for characters
3099 * that don't need it when regexes don't have to be in utf8
3100 * for Unicode semantics. If doing so, remember EBCDIC */
3101 sv_utf8_upgrade(res);
3102 str = SvPV_const(res, len);
3104 /* Don't accept malformed input */
3105 if (! is_utf8_string((U8 *) str, len)) {
3106 yyerror("Malformed UTF-8 returned by \\N");
3108 else if (PL_lex_inpat) {
3110 if (! len) { /* The name resolved to an empty string */
3111 Copy("\\N{}", d, 4, char);
3115 /* In order to not lose information for the regex
3116 * compiler, pass the result in the specially made
3117 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3118 * the code points in hex of each character
3119 * returned by charnames */
3121 const char *str_end = str + len;
3122 STRLEN char_length; /* cur char's byte length */
3123 STRLEN output_length; /* and the number of bytes
3124 after this is translated
3126 const STRLEN off = d - SvPVX_const(sv);
3128 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3129 * max('U+', '.'); and 1 for NUL */
3130 char hex_string[2 * UTF8_MAXBYTES + 5];
3132 /* Get the first character of the result. */
3133 U32 uv = utf8n_to_uvuni((U8 *) str,
3138 /* The call to is_utf8_string() above hopefully
3139 * guarantees that there won't be an error. But
3140 * it's easy here to make sure. The function just
3141 * above warns and returns 0 if invalid utf8, but
3142 * it can also return 0 if the input is validly a
3143 * NUL. Disambiguate */
3144 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3145 uv = UNICODE_REPLACEMENT;
3148 /* Convert first code point to hex, including the
3149 * boiler plate before it */
3150 sprintf(hex_string, "\\N{U+%X", (unsigned int) uv);
3151 output_length = strlen(hex_string);
3153 /* Make sure there is enough space to hold it */
3154 d = off + SvGROW(sv, off
3156 + (STRLEN)(send - e)
3157 + 2); /* '}' + NUL */
3159 Copy(hex_string, d, output_length, char);
3162 /* For each subsequent character, append dot and
3163 * its ordinal in hex */
3164 while ((str += char_length) < str_end) {
3165 const STRLEN off = d - SvPVX_const(sv);
3166 U32 uv = utf8n_to_uvuni((U8 *) str,
3170 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3171 uv = UNICODE_REPLACEMENT;
3174 sprintf(hex_string, ".%X", (unsigned int) uv);
3175 output_length = strlen(hex_string);
3177 d = off + SvGROW(sv, off
3179 + (STRLEN)(send - e)
3180 + 2); /* '}' + NUL */
3181 Copy(hex_string, d, output_length, char);
3185 *d++ = '}'; /* Done. Add the trailing brace */
3188 else { /* Here, not in a pattern. Convert the name to a
3191 /* If destination is not in utf8, unconditionally
3192 * recode it to be so. This is because \N{} implies
3193 * Unicode semantics, and scalars have to be in utf8
3194 * to guarantee those semantics */
3196 SvCUR_set(sv, d - SvPVX_const(sv));
3199 /* See Note on sizing above. */
3200 sv_utf8_upgrade_flags_grow(sv,
3201 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3202 len + (STRLEN)(send - s) + 1);
3203 d = SvPVX(sv) + SvCUR(sv);
3205 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3207 /* See Note on sizing above. (NOTE: SvCUR() is not
3208 * set correctly here). */
3209 const STRLEN off = d - SvPVX_const(sv);
3210 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3212 Copy(str, d, len, char);
3217 /* Deprecate non-approved name syntax */
3218 if (ckWARN_d(WARN_DEPRECATED)) {
3219 bool problematic = FALSE;
3222 /* For non-ut8 input, look to see that the first
3223 * character is an alpha, then loop through the rest
3224 * checking that each is a continuation */
3226 if (! isALPHAU(*i)) problematic = TRUE;
3227 else for (i = s + 1; i < e; i++) {
3228 if (isCHARNAME_CONT(*i)) continue;
3234 /* Similarly for utf8. For invariants can check
3235 * directly. We accept anything above the latin1
3236 * range because it is immaterial to Perl if it is
3237 * correct or not, and is expensive to check. But
3238 * it is fairly easy in the latin1 range to convert
3239 * the variants into a single character and check
3241 if (UTF8_IS_INVARIANT(*i)) {
3242 if (! isALPHAU(*i)) problematic = TRUE;
3243 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3244 if (! isALPHAU(UNI_TO_NATIVE(UTF8_ACCUMULATE(*i,
3250 if (! problematic) for (i = s + UTF8SKIP(s);
3254 if (UTF8_IS_INVARIANT(*i)) {
3255 if (isCHARNAME_CONT(*i)) continue;
3256 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3258 } else if (isCHARNAME_CONT(
3260 UTF8_ACCUMULATE(*i, *(i+1)))))
3269 /* The e-i passed to the final %.*s makes sure that
3270 * should the trailing NUL be missing that this
3271 * print won't run off the end of the string */
3272 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3273 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s", i - s + 1, s, e - i, i + 1);
3276 } /* End \N{NAME} */
3279 native_range = FALSE; /* \N{} is defined to be Unicode */
3281 s = e + 1; /* Point to just after the '}' */
3284 /* \c is a control character */
3288 *d++ = grok_bslash_c(*s++, 1);
3291 yyerror("Missing control char name in \\c");
3295 /* printf-style backslashes, formfeeds, newlines, etc */
3297 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3300 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3303 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3306 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3309 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3312 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3315 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3321 } /* end if (backslash) */
3328 /* If we started with encoded form, or already know we want it,
3329 then encode the next character */
3330 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3334 /* One might think that it is wasted effort in the case of the
3335 * source being utf8 (this_utf8 == TRUE) to take the next character
3336 * in the source, convert it to an unsigned value, and then convert
3337 * it back again. But the source has not been validated here. The
3338 * routine that does the conversion checks for errors like
3341 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3342 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3344 SvCUR_set(sv, d - SvPVX_const(sv));
3347 /* See Note on sizing above. */
3348 sv_utf8_upgrade_flags_grow(sv,
3349 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3350 need + (STRLEN)(send - s) + 1);
3351 d = SvPVX(sv) + SvCUR(sv);
3353 } else if (need > len) {
3354 /* encoded value larger than old, may need extra space (NOTE:
3355 * SvCUR() is not set correctly here). See Note on sizing
3357 const STRLEN off = d - SvPVX_const(sv);
3358 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3362 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3364 if (uv > 255 && !dorange)
3365 native_range = FALSE;
3369 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3371 } /* while loop to process each character */
3373 /* terminate the string and set up the sv */
3375 SvCUR_set(sv, d - SvPVX_const(sv));
3376 if (SvCUR(sv) >= SvLEN(sv))
3377 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3380 if (PL_encoding && !has_utf8) {
3381 sv_recode_to_utf8(sv, PL_encoding);
3387 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3388 PL_sublex_info.sub_op->op_private |=
3389 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3393 /* shrink the sv if we allocated more than we used */
3394 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3395 SvPV_shrink_to_cur(sv);
3398 /* return the substring (via pl_yylval) only if we parsed anything */
3399 if (s > PL_bufptr) {
3400 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3401 const char *const key = PL_lex_inpat ? "qr" : "q";
3402 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3406 if (PL_lex_inwhat == OP_TRANS) {
3409 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3417 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3420 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3427 * Returns TRUE if there's more to the expression (e.g., a subscript),
3430 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3432 * ->[ and ->{ return TRUE
3433 * { and [ outside a pattern are always subscripts, so return TRUE
3434 * if we're outside a pattern and it's not { or [, then return FALSE
3435 * if we're in a pattern and the first char is a {
3436 * {4,5} (any digits around the comma) returns FALSE
3437 * if we're in a pattern and the first char is a [
3439 * [SOMETHING] has a funky algorithm to decide whether it's a
3440 * character class or not. It has to deal with things like
3441 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3442 * anything else returns TRUE
3445 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3448 S_intuit_more(pTHX_ register char *s)
3452 PERL_ARGS_ASSERT_INTUIT_MORE;
3454 if (PL_lex_brackets)
3456 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3458 if (*s != '{' && *s != '[')
3463 /* In a pattern, so maybe we have {n,m}. */
3480 /* On the other hand, maybe we have a character class */
3483 if (*s == ']' || *s == '^')
3486 /* this is terrifying, and it works */
3487 int weight = 2; /* let's weigh the evidence */
3489 unsigned char un_char = 255, last_un_char;
3490 const char * const send = strchr(s,']');
3491 char tmpbuf[sizeof PL_tokenbuf * 4];
3493 if (!send) /* has to be an expression */
3496 Zero(seen,256,char);
3499 else if (isDIGIT(*s)) {
3501 if (isDIGIT(s[1]) && s[2] == ']')
3507 for (; s < send; s++) {
3508 last_un_char = un_char;
3509 un_char = (unsigned char)*s;
3514 weight -= seen[un_char] * 10;
3515 if (isALNUM_lazy_if(s+1,UTF)) {
3517 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3518 len = (int)strlen(tmpbuf);
3519 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
3524 else if (*s == '$' && s[1] &&
3525 strchr("[#!%*<>()-=",s[1])) {
3526 if (/*{*/ strchr("])} =",s[2]))
3535 if (strchr("wds]",s[1]))
3537 else if (seen[(U8)'\''] || seen[(U8)'"'])
3539 else if (strchr("rnftbxcav",s[1]))
3541 else if (isDIGIT(s[1])) {
3543 while (s[1] && isDIGIT(s[1]))
3553 if (strchr("aA01! ",last_un_char))
3555 if (strchr("zZ79~",s[1]))
3557 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3558 weight -= 5; /* cope with negative subscript */
3561 if (!isALNUM(last_un_char)
3562 && !(last_un_char == '$' || last_un_char == '@'
3563 || last_un_char == '&')
3564 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3569 if (keyword(tmpbuf, d - tmpbuf, 0))
3572 if (un_char == last_un_char + 1)
3574 weight -= seen[un_char];
3579 if (weight >= 0) /* probably a character class */
3589 * Does all the checking to disambiguate
3591 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3592 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3594 * First argument is the stuff after the first token, e.g. "bar".
3596 * Not a method if bar is a filehandle.
3597 * Not a method if foo is a subroutine prototyped to take a filehandle.
3598 * Not a method if it's really "Foo $bar"
3599 * Method if it's "foo $bar"
3600 * Not a method if it's really "print foo $bar"
3601 * Method if it's really "foo package::" (interpreted as package->foo)
3602 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3603 * Not a method if bar is a filehandle or package, but is quoted with
3608 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3611 char *s = start + (*start == '$');
3612 char tmpbuf[sizeof PL_tokenbuf];
3619 PERL_ARGS_ASSERT_INTUIT_METHOD;
3622 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3626 const char *proto = SvPVX_const(cv);
3637 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3638 /* start is the beginning of the possible filehandle/object,
3639 * and s is the end of it
3640 * tmpbuf is a copy of it
3643 if (*start == '$') {
3644 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3645 isUPPER(*PL_tokenbuf))
3648 len = start - SvPVX(PL_linestr);
3652 start = SvPVX(PL_linestr) + len;
3656 return *s == '(' ? FUNCMETH : METHOD;
3658 if (!keyword(tmpbuf, len, 0)) {
3659 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3663 soff = s - SvPVX(PL_linestr);
3667 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
3668 if (indirgv && GvCVu(indirgv))
3670 /* filehandle or package name makes it a method */
3671 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
3673 soff = s - SvPVX(PL_linestr);
3676 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3677 return 0; /* no assumptions -- "=>" quotes bearword */
3679 start_force(PL_curforce);
3680 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3681 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3682 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3684 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3689 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3691 return *s == '(' ? FUNCMETH : METHOD;
3697 /* Encoded script support. filter_add() effectively inserts a
3698 * 'pre-processing' function into the current source input stream.
3699 * Note that the filter function only applies to the current source file
3700 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3702 * The datasv parameter (which may be NULL) can be used to pass
3703 * private data to this instance of the filter. The filter function
3704 * can recover the SV using the FILTER_DATA macro and use it to
3705 * store private buffers and state information.
3707 * The supplied datasv parameter is upgraded to a PVIO type
3708 * and the IoDIRP/IoANY field is used to store the function pointer,
3709 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3710 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3711 * private use must be set using malloc'd pointers.
3715 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3724 if (!PL_rsfp_filters)
3725 PL_rsfp_filters = newAV();
3728 SvUPGRADE(datasv, SVt_PVIO);
3729 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3730 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3731 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3732 FPTR2DPTR(void *, IoANY(datasv)),
3733 SvPV_nolen(datasv)));
3734 av_unshift(PL_rsfp_filters, 1);
3735 av_store(PL_rsfp_filters, 0, datasv) ;
3740 /* Delete most recently added instance of this filter function. */
3742 Perl_filter_del(pTHX_ filter_t funcp)
3747 PERL_ARGS_ASSERT_FILTER_DEL;
3750 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3751 FPTR2DPTR(void*, funcp)));
3753 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3755 /* if filter is on top of stack (usual case) just pop it off */
3756 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3757 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3758 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
3759 IoANY(datasv) = (void *)NULL;
3760 sv_free(av_pop(PL_rsfp_filters));
3764 /* we need to search for the correct entry and clear it */
3765 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3769 /* Invoke the idxth filter function for the current rsfp. */
3770 /* maxlen 0 = read one text line */
3772 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3777 /* This API is bad. It should have been using unsigned int for maxlen.
3778 Not sure if we want to change the API, but if not we should sanity
3779 check the value here. */
3780 const unsigned int correct_length
3789 PERL_ARGS_ASSERT_FILTER_READ;
3791 if (!PL_parser || !PL_rsfp_filters)
3793 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3794 /* Provide a default input filter to make life easy. */
3795 /* Note that we append to the line. This is handy. */
3796 DEBUG_P(PerlIO_printf(Perl_debug_log,
3797 "filter_read %d: from rsfp\n", idx));
3798 if (correct_length) {
3801 const int old_len = SvCUR(buf_sv);
3803 /* ensure buf_sv is large enough */
3804 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3805 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3806 correct_length)) <= 0) {
3807 if (PerlIO_error(PL_rsfp))
3808 return -1; /* error */
3810 return 0 ; /* end of file */
3812 SvCUR_set(buf_sv, old_len + len) ;
3813 SvPVX(buf_sv)[old_len + len] = '\0';
3816 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3817 if (PerlIO_error(PL_rsfp))
3818 return -1; /* error */
3820 return 0 ; /* end of file */
3823 return SvCUR(buf_sv);
3825 /* Skip this filter slot if filter has been deleted */
3826 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3827 DEBUG_P(PerlIO_printf(Perl_debug_log,
3828 "filter_read %d: skipped (filter deleted)\n",
3830 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
3832 /* Get function pointer hidden within datasv */
3833 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
3834 DEBUG_P(PerlIO_printf(Perl_debug_log,
3835 "filter_read %d: via function %p (%s)\n",
3836 idx, (void*)datasv, SvPV_nolen_const(datasv)));
3837 /* Call function. The function is expected to */
3838 /* call "FILTER_READ(idx+1, buf_sv)" first. */
3839 /* Return: <0:error, =0:eof, >0:not eof */
3840 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
3844 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
3848 PERL_ARGS_ASSERT_FILTER_GETS;
3850 #ifdef PERL_CR_FILTER
3851 if (!PL_rsfp_filters) {
3852 filter_add(S_cr_textfilter,NULL);
3855 if (PL_rsfp_filters) {
3857 SvCUR_set(sv, 0); /* start with empty line */
3858 if (FILTER_READ(0, sv, 0) > 0)
3859 return ( SvPVX(sv) ) ;
3864 return (sv_gets(sv, PL_rsfp, append));
3868 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3873 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3875 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3879 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3880 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3882 return GvHV(gv); /* Foo:: */
3885 /* use constant CLASS => 'MyClass' */
3886 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3887 if (gv && GvCV(gv)) {
3888 SV * const sv = cv_const_sv(GvCV(gv));
3890 pkgname = SvPV_const(sv, len);
3893 return gv_stashpvn(pkgname, len, 0);
3897 * S_readpipe_override
3898 * Check whether readpipe() is overriden, and generates the appropriate
3899 * optree, provided sublex_start() is called afterwards.
3902 S_readpipe_override(pTHX)
3905 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3906 pl_yylval.ival = OP_BACKTICK;
3908 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3910 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3911 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3912 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3914 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3915 append_elem(OP_LIST,
3916 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3917 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3924 * The intent of this yylex wrapper is to minimize the changes to the
3925 * tokener when we aren't interested in collecting madprops. It remains
3926 * to be seen how successful this strategy will be...
3933 char *s = PL_bufptr;
3935 /* make sure PL_thiswhite is initialized */
3939 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3940 if (PL_pending_ident)
3941 return S_pending_ident(aTHX);
3943 /* previous token ate up our whitespace? */
3944 if (!PL_lasttoke && PL_nextwhite) {
3945 PL_thiswhite = PL_nextwhite;
3949 /* isolate the token, and figure out where it is without whitespace */
3950 PL_realtokenstart = -1;
3954 assert(PL_curforce < 0);
3956 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3957 if (!PL_thistoken) {
3958 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3959 PL_thistoken = newSVpvs("");
3961 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3962 PL_thistoken = newSVpvn(tstart, s - tstart);
3965 if (PL_thismad) /* install head */
3966 CURMAD('X', PL_thistoken);
3969 /* last whitespace of a sublex? */
3970 if (optype == ')' && PL_endwhite) {
3971 CURMAD('X', PL_endwhite);
3976 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3977 if (!PL_thiswhite && !PL_endwhite && !optype) {
3978 sv_free(PL_thistoken);
3983 /* put off final whitespace till peg */
3984 if (optype == ';' && !PL_rsfp) {
3985 PL_nextwhite = PL_thiswhite;
3988 else if (PL_thisopen) {
3989 CURMAD('q', PL_thisopen);
3991 sv_free(PL_thistoken);
3995 /* Store actual token text as madprop X */
3996 CURMAD('X', PL_thistoken);
4000 /* add preceding whitespace as madprop _ */
4001 CURMAD('_', PL_thiswhite);
4005 /* add quoted material as madprop = */
4006 CURMAD('=', PL_thisstuff);
4010 /* add terminating quote as madprop Q */
4011 CURMAD('Q', PL_thisclose);
4015 /* special processing based on optype */
4019 /* opval doesn't need a TOKEN since it can already store mp */
4029 if (pl_yylval.opval)
4030 append_madprops(PL_thismad, pl_yylval.opval, 0);
4038 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4047 /* remember any fake bracket that lexer is about to discard */
4048 if (PL_lex_brackets == 1 &&
4049 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4052 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4055 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4056 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4059 break; /* don't bother looking for trailing comment */
4068 /* attach a trailing comment to its statement instead of next token */
4072 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4074 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4076 if (*s == '\n' || *s == '#') {
4077 while (s < PL_bufend && *s != '\n')
4081 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4082 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4099 /* Create new token struct. Note: opvals return early above. */
4100 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4107 S_tokenize_use(pTHX_ int is_use, char *s) {
4110 PERL_ARGS_ASSERT_TOKENIZE_USE;
4112 if (PL_expect != XSTATE)
4113 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4114 is_use ? "use" : "no"));
4116 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4117 s = force_version(s, TRUE);
4118 if (*s == ';' || *s == '}'
4119 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4120 start_force(PL_curforce);
4121 NEXTVAL_NEXTTOKE.opval = NULL;
4124 else if (*s == 'v') {
4125 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4126 s = force_version(s, FALSE);
4130 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4131 s = force_version(s, FALSE);
4133 pl_yylval.ival = is_use;
4137 static const char* const exp_name[] =
4138 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4139 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4146 Works out what to call the token just pulled out of the input
4147 stream. The yacc parser takes care of taking the ops we return and
4148 stitching them into a tree.
4154 if read an identifier
4155 if we're in a my declaration
4156 croak if they tried to say my($foo::bar)
4157 build the ops for a my() declaration
4158 if it's an access to a my() variable
4159 are we in a sort block?
4160 croak if my($a); $a <=> $b
4161 build ops for access to a my() variable
4162 if in a dq string, and they've said @foo and we can't find @foo
4164 build ops for a bareword
4165 if we already built the token before, use it.
4170 #pragma segment Perl_yylex
4176 register char *s = PL_bufptr;
4182 /* orig_keyword, gvp, and gv are initialized here because
4183 * jump to the label just_a_word_zero can bypass their
4184 * initialization later. */
4185 I32 orig_keyword = 0;
4190 SV* tmp = newSVpvs("");
4191 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4192 (IV)CopLINE(PL_curcop),
4193 lex_state_names[PL_lex_state],
4194 exp_name[PL_expect],
4195 pv_display(tmp, s, strlen(s), 0, 60));
4198 /* check if there's an identifier for us to look at */
4199 if (PL_pending_ident)
4200 return REPORT(S_pending_ident(aTHX));
4202 /* no identifier pending identification */
4204 switch (PL_lex_state) {
4206 case LEX_NORMAL: /* Some compilers will produce faster */
4207 case LEX_INTERPNORMAL: /* code if we comment these out. */
4211 /* when we've already built the next token, just pull it out of the queue */
4215 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4217 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4218 PL_nexttoke[PL_lasttoke].next_mad = 0;
4219 if (PL_thismad && PL_thismad->mad_key == '_') {
4220 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4221 PL_thismad->mad_val = 0;
4222 mad_free(PL_thismad);
4227 PL_lex_state = PL_lex_defer;
4228 PL_expect = PL_lex_expect;
4229 PL_lex_defer = LEX_NORMAL;
4230 if (!PL_nexttoke[PL_lasttoke].next_type)
4235 pl_yylval = PL_nextval[PL_nexttoke];
4237 PL_lex_state = PL_lex_defer;
4238 PL_expect = PL_lex_expect;
4239 PL_lex_defer = LEX_NORMAL;
4243 /* FIXME - can these be merged? */
4244 return(PL_nexttoke[PL_lasttoke].next_type);
4246 return REPORT(PL_nexttype[PL_nexttoke]);
4249 /* interpolated case modifiers like \L \U, including \Q and \E.
4250 when we get here, PL_bufptr is at the \
4252 case LEX_INTERPCASEMOD:
4254 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4255 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
4257 /* handle \E or end of string */
4258 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4260 if (PL_lex_casemods) {
4261 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4262 PL_lex_casestack[PL_lex_casemods] = '\0';
4264 if (PL_bufptr != PL_bufend
4265 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
4267 PL_lex_state = LEX_INTERPCONCAT;
4270 PL_thistoken = newSVpvs("\\E");
4276 while (PL_bufptr != PL_bufend &&
4277 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4279 PL_thiswhite = newSVpvs("");
4280 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4284 if (PL_bufptr != PL_bufend)
4287 PL_lex_state = LEX_INTERPCONCAT;
4291 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4292 "### Saw case modifier\n"); });
4294 if (s[1] == '\\' && s[2] == 'E') {
4297 PL_thiswhite = newSVpvs("");
4298 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4301 PL_lex_state = LEX_INTERPCONCAT;
4306 if (!PL_madskills) /* when just compiling don't need correct */
4307 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4308 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4309 if ((*s == 'L' || *s == 'U') &&
4310 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4311 PL_lex_casestack[--PL_lex_casemods] = '\0';
4314 if (PL_lex_casemods > 10)
4315 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4316 PL_lex_casestack[PL_lex_casemods++] = *s;
4317 PL_lex_casestack[PL_lex_casemods] = '\0';
4318 PL_lex_state = LEX_INTERPCONCAT;
4319 start_force(PL_curforce);
4320 NEXTVAL_NEXTTOKE.ival = 0;
4322 start_force(PL_curforce);
4324 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4326 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4328 NEXTVAL_NEXTTOKE.ival = OP_LC;
4330 NEXTVAL_NEXTTOKE.ival = OP_UC;
4332 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4334 Perl_croak(aTHX_ "panic: yylex");
4336 SV* const tmpsv = newSVpvs("\\ ");
4337 /* replace the space with the character we want to escape
4339 SvPVX(tmpsv)[1] = *s;
4345 if (PL_lex_starts) {
4351 sv_free(PL_thistoken);
4352 PL_thistoken = newSVpvs("");
4355 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4356 if (PL_lex_casemods == 1 && PL_lex_inpat)
4365 case LEX_INTERPPUSH:
4366 return REPORT(sublex_push());
4368 case LEX_INTERPSTART:
4369 if (PL_bufptr == PL_bufend)
4370 return REPORT(sublex_done());
4371 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4372 "### Interpolated variable\n"); });
4374 PL_lex_dojoin = (*PL_bufptr == '@');
4375 PL_lex_state = LEX_INTERPNORMAL;
4376 if (PL_lex_dojoin) {
4377 start_force(PL_curforce);
4378 NEXTVAL_NEXTTOKE.ival = 0;
4380 start_force(PL_curforce);
4381 force_ident("\"", '$');
4382 start_force(PL_curforce);
4383 NEXTVAL_NEXTTOKE.ival = 0;
4385 start_force(PL_curforce);
4386 NEXTVAL_NEXTTOKE.ival = 0;
4388 start_force(PL_curforce);
4389 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4392 if (PL_lex_starts++) {
4397 sv_free(PL_thistoken);
4398 PL_thistoken = newSVpvs("");
4401 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4402 if (!PL_lex_casemods && PL_lex_inpat)
4409 case LEX_INTERPENDMAYBE:
4410 if (intuit_more(PL_bufptr)) {
4411 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4417 if (PL_lex_dojoin) {
4418 PL_lex_dojoin = FALSE;
4419 PL_lex_state = LEX_INTERPCONCAT;
4423 sv_free(PL_thistoken);
4424 PL_thistoken = newSVpvs("");
4429 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4430 && SvEVALED(PL_lex_repl))
4432 if (PL_bufptr != PL_bufend)
4433 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4437 case LEX_INTERPCONCAT:
4439 if (PL_lex_brackets)
4440 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4442 if (PL_bufptr == PL_bufend)
4443 return REPORT(sublex_done());
4445 if (SvIVX(PL_linestr) == '\'') {
4446 SV *sv = newSVsv(PL_linestr);
4449 else if ( PL_hints & HINT_NEW_RE )
4450 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4451 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4455 s = scan_const(PL_bufptr);
4457 PL_lex_state = LEX_INTERPCASEMOD;
4459 PL_lex_state = LEX_INTERPSTART;
4462 if (s != PL_bufptr) {
4463 start_force(PL_curforce);
4465 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4467 NEXTVAL_NEXTTOKE = pl_yylval;
4470 if (PL_lex_starts++) {
4474 sv_free(PL_thistoken);
4475 PL_thistoken = newSVpvs("");
4478 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4479 if (!PL_lex_casemods && PL_lex_inpat)
4492 PL_lex_state = LEX_NORMAL;
4493 s = scan_formline(PL_bufptr);
4494 if (!PL_lex_formbrack)
4500 PL_oldoldbufptr = PL_oldbufptr;
4506 sv_free(PL_thistoken);
4509 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4513 if (isIDFIRST_lazy_if(s,UTF))
4516 unsigned char c = *s;
4517 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4518 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4519 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4524 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4528 goto fake_eof; /* emulate EOF on ^D or ^Z */
4537 if (PL_lex_brackets) {
4538 yyerror((const char *)
4540 ? "Format not terminated"
4541 : "Missing right curly or square bracket"));
4543 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4544 "### Tokener got EOF\n");
4548 if (s++ < PL_bufend)
4549 goto retry; /* ignore stray nulls */
4552 if (!PL_in_eval && !PL_preambled) {
4553 PL_preambled = TRUE;
4559 /* Generate a string of Perl code to load the debugger.
4560 * If PERL5DB is set, it will return the contents of that,
4561 * otherwise a compile-time require of perl5db.pl. */
4563 const char * const pdb = PerlEnv_getenv("PERL5DB");
4566 sv_setpv(PL_linestr, pdb);
4567 sv_catpvs(PL_linestr,";");
4569 SETERRNO(0,SS_NORMAL);
4570 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4573 sv_setpvs(PL_linestr,"");
4574 if (PL_preambleav) {
4575 SV **svp = AvARRAY(PL_preambleav);
4576 SV **const end = svp + AvFILLp(PL_preambleav);
4578 sv_catsv(PL_linestr, *svp);
4580 sv_catpvs(PL_linestr, ";");
4582 sv_free(MUTABLE_SV(PL_preambleav));
4583 PL_preambleav = NULL;
4586 sv_catpvs(PL_linestr,
4587 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4588 if (PL_minus_n || PL_minus_p) {
4589 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4591 sv_catpvs(PL_linestr,"chomp;");
4594 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4595 || *PL_splitstr == '"')
4596 && strchr(PL_splitstr + 1, *PL_splitstr))
4597 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4599 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4600 bytes can be used as quoting characters. :-) */
4601 const char *splits = PL_splitstr;
4602 sv_catpvs(PL_linestr, "our @F=split(q\0");
4605 if (*splits == '\\')
4606 sv_catpvn(PL_linestr, splits, 1);
4607 sv_catpvn(PL_linestr, splits, 1);
4608 } while (*splits++);
4609 /* This loop will embed the trailing NUL of
4610 PL_linestr as the last thing it does before
4612 sv_catpvs(PL_linestr, ");");
4616 sv_catpvs(PL_linestr,"our @F=split(' ');");
4619 sv_catpvs(PL_linestr, "\n");
4620 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4621 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4622 PL_last_lop = PL_last_uni = NULL;
4623 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4624 update_debugger_info(PL_linestr, NULL, 0);
4629 bof = PL_rsfp ? TRUE : FALSE;
4632 fake_eof = LEX_FAKE_EOF;
4634 PL_bufptr = PL_bufend;
4635 CopLINE_inc(PL_curcop);
4636 if (!lex_next_chunk(fake_eof)) {
4637 CopLINE_dec(PL_curcop);
4639 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4641 CopLINE_dec(PL_curcop);
4644 PL_realtokenstart = -1;
4647 /* If it looks like the start of a BOM or raw UTF-16,
4648 * check if it in fact is. */
4649 if (bof && PL_rsfp &&
4654 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
4656 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4657 s = swallow_bom((U8*)s);
4661 /* Incest with pod. */
4664 sv_catsv(PL_thiswhite, PL_linestr);
4666 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4667 sv_setpvs(PL_linestr, "");
4668 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4669 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4670 PL_last_lop = PL_last_uni = NULL;
4671 PL_doextract = FALSE;
4676 } while (PL_doextract);
4677 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4678 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4679 PL_last_lop = PL_last_uni = NULL;
4680 if (CopLINE(PL_curcop) == 1) {
4681 while (s < PL_bufend && isSPACE(*s))
4683 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4687 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4691 if (*s == '#' && *(s+1) == '!')
4693 #ifdef ALTERNATE_SHEBANG
4695 static char const as[] = ALTERNATE_SHEBANG;
4696 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4697 d = s + (sizeof(as) - 1);
4699 #endif /* ALTERNATE_SHEBANG */
4708 while (*d && !isSPACE(*d))
4712 #ifdef ARG_ZERO_IS_SCRIPT
4713 if (ipathend > ipath) {
4715 * HP-UX (at least) sets argv[0] to the script name,
4716 * which makes $^X incorrect. And Digital UNIX and Linux,
4717 * at least, set argv[0] to the basename of the Perl
4718 * interpreter. So, having found "#!", we'll set it right.
4720 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4722 assert(SvPOK(x) || SvGMAGICAL(x));
4723 if (sv_eq(x, CopFILESV(PL_curcop))) {
4724 sv_setpvn(x, ipath, ipathend - ipath);
4730 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4731 const char * const lstart = SvPV_const(x,llen);
4733 bstart += blen - llen;
4734 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4735 sv_setpvn(x, ipath, ipathend - ipath);
4740 TAINT_NOT; /* $^X is always tainted, but that's OK */
4742 #endif /* ARG_ZERO_IS_SCRIPT */
4747 d = instr(s,"perl -");
4749 d = instr(s,"perl");
4751 /* avoid getting into infinite loops when shebang
4752 * line contains "Perl" rather than "perl" */
4754 for (d = ipathend-4; d >= ipath; --d) {
4755 if ((*d == 'p' || *d == 'P')
4756 && !ibcmp(d, "perl", 4))
4766 #ifdef ALTERNATE_SHEBANG
4768 * If the ALTERNATE_SHEBANG on this system starts with a
4769 * character that can be part of a Perl expression, then if
4770 * we see it but not "perl", we're probably looking at the
4771 * start of Perl code, not a request to hand off to some
4772 * other interpreter. Similarly, if "perl" is there, but
4773 * not in the first 'word' of the line, we assume the line
4774 * contains the start of the Perl program.
4776 if (d && *s != '#') {
4777 const char *c = ipath;
4778 while (*c && !strchr("; \t\r\n\f\v#", *c))
4781 d = NULL; /* "perl" not in first word; ignore */
4783 *s = '#'; /* Don't try to parse shebang line */
4785 #endif /* ALTERNATE_SHEBANG */
4790 !instr(s,"indir") &&
4791 instr(PL_origargv[0],"perl"))
4798 while (s < PL_bufend && isSPACE(*s))
4800 if (s < PL_bufend) {
4801 Newx(newargv,PL_origargc+3,char*);
4803 while (s < PL_bufend && !isSPACE(*s))
4806 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4809 newargv = PL_origargv;
4812 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4814 Perl_croak(aTHX_ "Can't exec %s", ipath);
4817 while (*d && !isSPACE(*d))
4819 while (SPACE_OR_TAB(*d))
4823 const bool switches_done = PL_doswitches;
4824 const U32 oldpdb = PL_perldb;
4825 const bool oldn = PL_minus_n;
4826 const bool oldp = PL_minus_p;
4830 bool baduni = FALSE;
4832 const char *d2 = d1 + 1;
4833 if (parse_unicode_opts((const char **)&d2)
4837 if (baduni || *d1 == 'M' || *d1 == 'm') {
4838 const char * const m = d1;
4839 while (*d1 && !isSPACE(*d1))
4841 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4844 d1 = moreswitches(d1);
4846 if (PL_doswitches && !switches_done) {
4847 int argc = PL_origargc;
4848 char **argv = PL_origargv;
4851 } while (argc && argv[0][0] == '-' && argv[0][1]);
4852 init_argv_symbols(argc,argv);
4854 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4855 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4856 /* if we have already added "LINE: while (<>) {",
4857 we must not do it again */
4859 sv_setpvs(PL_linestr, "");
4860 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4861 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4862 PL_last_lop = PL_last_uni = NULL;
4863 PL_preambled = FALSE;
4864 if (PERLDB_LINE || PERLDB_SAVESRC)
4865 (void)gv_fetchfile(PL_origfilename);
4872 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4874 PL_lex_state = LEX_FORMLINE;
4879 #ifdef PERL_STRICT_CR
4880 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4882 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4884 case ' ': case '\t': case '\f': case 013:
4886 PL_realtokenstart = -1;
4888 PL_thiswhite = newSVpvs("");
4889 sv_catpvn(PL_thiswhite, s, 1);
4896 PL_realtokenstart = -1;
4900 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4901 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4902 /* handle eval qq[#line 1 "foo"\n ...] */
4903 CopLINE_dec(PL_curcop);
4906 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4908 if (!PL_in_eval || PL_rsfp)
4913 while (d < PL_bufend && *d != '\n')
4917 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4918 Perl_croak(aTHX_ "panic: input overflow");
4921 PL_thiswhite = newSVpvn(s, d - s);
4926 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4928 PL_lex_state = LEX_FORMLINE;
4934 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4935 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4938 TOKEN(PEG); /* make sure any #! line is accessible */
4943 /* if (PL_madskills && PL_lex_formbrack) { */
4945 while (d < PL_bufend && *d != '\n')
4949 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4950 Perl_croak(aTHX_ "panic: input overflow");
4951 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4953 PL_thiswhite = newSVpvs("");
4954 if (CopLINE(PL_curcop) == 1) {
4955 sv_setpvs(PL_thiswhite, "");
4958 sv_catpvn(PL_thiswhite, s, d - s);
4972 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4980 while (s < PL_bufend && SPACE_OR_TAB(*s))
4983 if (strnEQ(s,"=>",2)) {
4984 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4985 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4986 OPERATOR('-'); /* unary minus */
4988 PL_last_uni = PL_oldbufptr;
4990 case 'r': ftst = OP_FTEREAD; break;
4991 case 'w': ftst = OP_FTEWRITE; break;
4992 case 'x': ftst = OP_FTEEXEC; break;
4993 case 'o': ftst = OP_FTEOWNED; break;
4994 case 'R': ftst = OP_FTRREAD; break;
4995 case 'W': ftst = OP_FTRWRITE; break;
4996 case 'X': ftst = OP_FTREXEC; break;
4997 case 'O': ftst = OP_FTROWNED; break;
4998 case 'e': ftst = OP_FTIS; break;
4999 case 'z': ftst = OP_FTZERO; break;
5000 case 's': ftst = OP_FTSIZE; break;
5001 case 'f': ftst = OP_FTFILE; break;
5002 case 'd': ftst = OP_FTDIR; break;
5003 case 'l': ftst = OP_FTLINK; break;
5004 case 'p': ftst = OP_FTPIPE; break;
5005 case 'S': ftst = OP_FTSOCK; break;
5006 case 'u': ftst = OP_FTSUID; break;
5007 case 'g': ftst = OP_FTSGID; break;
5008 case 'k': ftst = OP_FTSVTX; break;
5009 case 'b': ftst = OP_FTBLK; break;
5010 case 'c': ftst = OP_FTCHR; break;
5011 case 't': ftst = OP_FTTTY; break;
5012 case 'T': ftst = OP_FTTEXT; break;
5013 case 'B': ftst = OP_FTBINARY; break;
5014 case 'M': case 'A': case 'C':
5015 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5017 case 'M': ftst = OP_FTMTIME; break;
5018 case 'A': ftst = OP_FTATIME; break;
5019 case 'C': ftst = OP_FTCTIME; break;
5027 PL_last_lop_op = (OPCODE)ftst;
5028 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5029 "### Saw file test %c\n", (int)tmp);
5034 /* Assume it was a minus followed by a one-letter named
5035 * subroutine call (or a -bareword), then. */
5036 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5037 "### '-%c' looked like a file test but was not\n",
5044 const char tmp = *s++;
5047 if (PL_expect == XOPERATOR)
5052 else if (*s == '>') {
5055 if (isIDFIRST_lazy_if(s,UTF)) {
5056 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5064 if (PL_expect == XOPERATOR)
5067 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5069 OPERATOR('-'); /* unary minus */
5075 const char tmp = *s++;
5078 if (PL_expect == XOPERATOR)
5083 if (PL_expect == XOPERATOR)
5086 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5093 if (PL_expect != XOPERATOR) {
5094 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5095 PL_expect = XOPERATOR;
5096 force_ident(PL_tokenbuf, '*');
5109 if (PL_expect == XOPERATOR) {
5113 PL_tokenbuf[0] = '%';
5114 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5115 sizeof PL_tokenbuf - 1, FALSE);
5116 if (!PL_tokenbuf[1]) {
5119 PL_pending_ident = '%';
5128 const char tmp = *s++;
5133 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5140 const char tmp = *s++;
5146 goto just_a_word_zero_gv;
5149 switch (PL_expect) {
5155 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5157 PL_bufptr = s; /* update in case we back off */
5159 deprecate(":= for an empty attribute list");
5166 PL_expect = XTERMBLOCK;
5169 stuffstart = s - SvPVX(PL_linestr) - 1;
5173 while (isIDFIRST_lazy_if(s,UTF)) {
5176 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5177 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5178 if (tmp < 0) tmp = -tmp;
5193 sv = newSVpvn(s, len);
5195 d = scan_str(d,TRUE,TRUE);
5197 /* MUST advance bufptr here to avoid bogus
5198 "at end of line" context messages from yyerror().
5200 PL_bufptr = s + len;
5201 yyerror("Unterminated attribute parameter in attribute list");
5205 return REPORT(0); /* EOF indicator */
5209 sv_catsv(sv, PL_lex_stuff);
5210 attrs = append_elem(OP_LIST, attrs,
5211 newSVOP(OP_CONST, 0, sv));
5212 SvREFCNT_dec(PL_lex_stuff);
5213 PL_lex_stuff = NULL;
5216 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5218 if (PL_in_my == KEY_our) {
5219 deprecate(":unique");
5222 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5225 /* NOTE: any CV attrs applied here need to be part of
5226 the CVf_BUILTIN_ATTRS define in cv.h! */
5227 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5229 CvLVALUE_on(PL_compcv);
5231 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5233 deprecate(":locked");
5235 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5237 CvMETHOD_on(PL_compcv);
5239 /* After we've set the flags, it could be argued that
5240 we don't need to do the attributes.pm-based setting
5241 process, and shouldn't bother appending recognized
5242 flags. To experiment with that, uncomment the
5243 following "else". (Note that's already been
5244 uncommented. That keeps the above-applied built-in
5245 attributes from being intercepted (and possibly
5246 rejected) by a package's attribute routines, but is
5247 justified by the performance win for the common case
5248 of applying only built-in attributes.) */
5250 attrs = append_elem(OP_LIST, attrs,
5251 newSVOP(OP_CONST, 0,
5255 if (*s == ':' && s[1] != ':')
5258 break; /* require real whitespace or :'s */
5259 /* XXX losing whitespace on sequential attributes here */
5263 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5264 if (*s != ';' && *s != '}' && *s != tmp
5265 && (tmp != '=' || *s != ')')) {
5266 const char q = ((*s == '\'') ? '"' : '\'');
5267 /* If here for an expression, and parsed no attrs, back
5269 if (tmp == '=' && !attrs) {
5273 /* MUST advance bufptr here to avoid bogus "at end of line"
5274 context messages from yyerror().
5277 yyerror( (const char *)
5279 ? Perl_form(aTHX_ "Invalid separator character "
5280 "%c%c%c in attribute list", q, *s, q)
5281 : "Unterminated attribute list" ) );
5289 start_force(PL_curforce);
5290 NEXTVAL_NEXTTOKE.opval = attrs;
5291 CURMAD('_', PL_nextwhite);
5296 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5297 (s - SvPVX(PL_linestr)) - stuffstart);
5305 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5306 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5314 const char tmp = *s++;
5319 const char tmp = *s++;
5327 if (PL_lex_brackets <= 0)
5328 yyerror("Unmatched right square bracket");
5331 if (PL_lex_state == LEX_INTERPNORMAL) {
5332 if (PL_lex_brackets == 0) {
5333 if (*s == '-' && s[1] == '>')
5334 PL_lex_state = LEX_INTERPENDMAYBE;
5335 else if (*s != '[' && *s != '{')
5336 PL_lex_state = LEX_INTERPEND;
5343 if (PL_lex_brackets > 100) {
5344 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5346 switch (PL_expect) {
5348 if (PL_lex_formbrack) {
5352 if (PL_oldoldbufptr == PL_last_lop)
5353 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5355 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5356 OPERATOR(HASHBRACK);
5358 while (s < PL_bufend && SPACE_OR_TAB(*s))
5361 PL_tokenbuf[0] = '\0';
5362 if (d < PL_bufend && *d == '-') {
5363 PL_tokenbuf[0] = '-';
5365 while (d < PL_bufend && SPACE_OR_TAB(*d))
5368 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5369 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5371 while (d < PL_bufend && SPACE_OR_TAB(*d))
5374 const char minus = (PL_tokenbuf[0] == '-');
5375 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5383 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5388 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5393 if (PL_oldoldbufptr == PL_last_lop)
5394 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5396 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5399 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5401 /* This hack is to get the ${} in the message. */
5403 yyerror("syntax error");
5406 OPERATOR(HASHBRACK);
5408 /* This hack serves to disambiguate a pair of curlies
5409 * as being a block or an anon hash. Normally, expectation
5410 * determines that, but in cases where we're not in a
5411 * position to expect anything in particular (like inside
5412 * eval"") we have to resolve the ambiguity. This code
5413 * covers the case where the first term in the curlies is a
5414 * quoted string. Most other cases need to be explicitly
5415 * disambiguated by prepending a "+" before the opening
5416 * curly in order to force resolution as an anon hash.
5418 * XXX should probably propagate the outer expectation
5419 * into eval"" to rely less on this hack, but that could
5420 * potentially break current behavior of eval"".
5424 if (*s == '\'' || *s == '"' || *s == '`') {
5425 /* common case: get past first string, handling escapes */
5426 for (t++; t < PL_bufend && *t != *s;)
5427 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5431 else if (*s == 'q') {
5434 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5437 /* skip q//-like construct */
5439 char open, close, term;
5442 while (t < PL_bufend && isSPACE(*t))
5444 /* check for q => */
5445 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5446 OPERATOR(HASHBRACK);
5450 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5454 for (t++; t < PL_bufend; t++) {
5455 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5457 else if (*t == open)
5461 for (t++; t < PL_bufend; t++) {
5462 if (*t == '\\' && t+1 < PL_bufend)
5464 else if (*t == close && --brackets <= 0)
5466 else if (*t == open)
5473 /* skip plain q word */
5474 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5477 else if (isALNUM_lazy_if(t,UTF)) {
5479 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5482 while (t < PL_bufend && isSPACE(*t))
5484 /* if comma follows first term, call it an anon hash */
5485 /* XXX it could be a comma expression with loop modifiers */
5486 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5487 || (*t == '=' && t[1] == '>')))
5488 OPERATOR(HASHBRACK);
5489 if (PL_expect == XREF)
5492 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5498 pl_yylval.ival = CopLINE(PL_curcop);
5499 if (isSPACE(*s) || *s == '#')
5500 PL_copline = NOLINE; /* invalidate current command line number */
5505 if (PL_lex_brackets <= 0)
5506 yyerror("Unmatched right curly bracket");
5508 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5509 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5510 PL_lex_formbrack = 0;
5511 if (PL_lex_state == LEX_INTERPNORMAL) {
5512 if (PL_lex_brackets == 0) {
5513 if (PL_expect & XFAKEBRACK) {
5514 PL_expect &= XENUMMASK;
5515 PL_lex_state = LEX_INTERPEND;
5520 PL_thiswhite = newSVpvs("");
5521 sv_catpvs(PL_thiswhite,"}");
5524 return yylex(); /* ignore fake brackets */
5526 if (*s == '-' && s[1] == '>')
5527 PL_lex_state = LEX_INTERPENDMAYBE;
5528 else if (*s != '[' && *s != '{')
5529 PL_lex_state = LEX_INTERPEND;
5532 if (PL_expect & XFAKEBRACK) {
5533 PL_expect &= XENUMMASK;
5535 return yylex(); /* ignore fake brackets */
5537 start_force(PL_curforce);
5539 curmad('X', newSVpvn(s-1,1));
5540 CURMAD('_', PL_thiswhite);
5545 PL_thistoken = newSVpvs("");
5553 if (PL_expect == XOPERATOR) {
5554 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5555 && isIDFIRST_lazy_if(s,UTF))
5557 CopLINE_dec(PL_curcop);
5558 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5559 CopLINE_inc(PL_curcop);
5564 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5566 PL_expect = XOPERATOR;
5567 force_ident(PL_tokenbuf, '&');
5571 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5583 const char tmp = *s++;
5590 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5591 && strchr("+-*/%.^&|<",tmp))
5592 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5593 "Reversed %c= operator",(int)tmp);
5595 if (PL_expect == XSTATE && isALPHA(tmp) &&
5596 (s == PL_linestart+1 || s[-2] == '\n') )
5598 if (PL_in_eval && !PL_rsfp) {
5603 if (strnEQ(s,"=cut",4)) {
5619 PL_thiswhite = newSVpvs("");
5620 sv_catpvn(PL_thiswhite, PL_linestart,
5621 PL_bufend - PL_linestart);
5625 PL_doextract = TRUE;
5629 if (PL_lex_brackets < PL_lex_formbrack) {
5631 #ifdef PERL_STRICT_CR
5632 while (SPACE_OR_TAB(*t))
5634 while (SPACE_OR_TAB(*t) || *t == '\r')
5637 if (*t == '\n' || *t == '#') {
5648 const char tmp = *s++;
5650 /* was this !=~ where !~ was meant?
5651 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5653 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5654 const char *t = s+1;
5656 while (t < PL_bufend && isSPACE(*t))
5659 if (*t == '/' || *t == '?' ||
5660 ((*t == 'm' || *t == 's' || *t == 'y')
5661 && !isALNUM(t[1])) ||
5662 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5663 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5664 "!=~ should be !~");
5674 if (PL_expect != XOPERATOR) {
5675 if (s[1] != '<' && !strchr(s,'>'))
5678 s = scan_heredoc(s);
5680 s = scan_inputsymbol(s);
5681 TERM(sublex_start());
5687 SHop(OP_LEFT_SHIFT);
5701 const char tmp = *s++;
5703 SHop(OP_RIGHT_SHIFT);
5704 else if (tmp == '=')
5713 if (PL_expect == XOPERATOR) {
5714 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5715 return deprecate_commaless_var_list();
5719 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
5720 PL_tokenbuf[0] = '@';
5721 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
5722 sizeof PL_tokenbuf - 1, FALSE);
5723 if (PL_expect == XOPERATOR)
5724 no_op("Array length", s);
5725 if (!PL_tokenbuf[1])
5727 PL_expect = XOPERATOR;
5728 PL_pending_ident = '#';
5732 PL_tokenbuf[0] = '$';
5733 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5734 sizeof PL_tokenbuf - 1, FALSE);
5735 if (PL_expect == XOPERATOR)
5737 if (!PL_tokenbuf[1]) {
5739 yyerror("Final $ should be \\$ or $name");
5743 /* This kludge not intended to be bulletproof. */
5744 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
5745 pl_yylval.opval = newSVOP(OP_CONST, 0,
5746 newSViv(CopARYBASE_get(&PL_compiling)));
5747 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
5753 const char tmp = *s;
5754 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
5757 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
5758 && intuit_more(s)) {
5760 PL_tokenbuf[0] = '@';
5761 if (ckWARN(WARN_SYNTAX)) {
5764 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
5767 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5768 while (t < PL_bufend && *t != ']')
5770 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5771 "Multidimensional syntax %.*s not supported",
5772 (int)((t - PL_bufptr) + 1), PL_bufptr);
5776 else if (*s == '{') {
5778 PL_tokenbuf[0] = '%';
5779 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
5780 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
5782 char tmpbuf[sizeof PL_tokenbuf];
5785 } while (isSPACE(*t));
5786 if (isIDFIRST_lazy_if(t,UTF)) {
5788 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
5792 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
5793 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5794 "You need to quote \"%s\"",
5801 PL_expect = XOPERATOR;
5802 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
5803 const bool islop = (PL_last_lop == PL_oldoldbufptr);
5804 if (!islop || PL_last_lop_op == OP_GREPSTART)
5805 PL_expect = XOPERATOR;
5806 else if (strchr("$@\"'`q", *s))
5807 PL_expect = XTERM; /* e.g. print $fh "foo" */
5808 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
5809 PL_expect = XTERM; /* e.g. print $fh &sub */
5810 else if (isIDFIRST_lazy_if(s,UTF)) {
5811 char tmpbuf[sizeof PL_tokenbuf];
5813 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5814 if ((t2 = keyword(tmpbuf, len, 0))) {
5815 /* binary operators exclude handle interpretations */
5827 PL_expect = XTERM; /* e.g. print $fh length() */
5832 PL_expect = XTERM; /* e.g. print $fh subr() */
5835 else if (isDIGIT(*s))
5836 PL_expect = XTERM; /* e.g. print $fh 3 */
5837 else if (*s == '.' && isDIGIT(s[1]))
5838 PL_expect = XTERM; /* e.g. print $fh .3 */
5839 else if ((*s == '?' || *s == '-' || *s == '+')
5840 && !isSPACE(s[1]) && s[1] != '=')
5841 PL_expect = XTERM; /* e.g. print $fh -1 */
5842 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5844 PL_expect = XTERM; /* e.g. print $fh /.../
5845 XXX except DORDOR operator
5847 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5849 PL_expect = XTERM; /* print $fh <<"EOF" */
5852 PL_pending_ident = '$';
5856 if (PL_expect == XOPERATOR)
5858 PL_tokenbuf[0] = '@';
5859 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5860 if (!PL_tokenbuf[1]) {
5863 if (PL_lex_state == LEX_NORMAL)
5865 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5867 PL_tokenbuf[0] = '%';
5869 /* Warn about @ where they meant $. */
5870 if (*s == '[' || *s == '{') {
5871 if (ckWARN(WARN_SYNTAX)) {
5872 const char *t = s + 1;
5873 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5875 if (*t == '}' || *t == ']') {
5877 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5878 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5879 "Scalar value %.*s better written as $%.*s",
5880 (int)(t-PL_bufptr), PL_bufptr,
5881 (int)(t-PL_bufptr-1), PL_bufptr+1);
5886 PL_pending_ident = '@';
5889 case '/': /* may be division, defined-or, or pattern */
5890 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5894 case '?': /* may either be conditional or pattern */
5895 if (PL_expect == XOPERATOR) {
5903 /* A // operator. */
5913 /* Disable warning on "study /blah/" */
5914 if (PL_oldoldbufptr == PL_last_uni
5915 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5916 || memNE(PL_last_uni, "study", 5)
5917 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5920 s = scan_pat(s,OP_MATCH);
5921 TERM(sublex_start());
5925 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5926 #ifdef PERL_STRICT_CR
5929 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5931 && (s == PL_linestart || s[-1] == '\n') )
5933 PL_lex_formbrack = 0;
5937 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5941 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5947 pl_yylval.ival = OPf_SPECIAL;
5956 case '0': case '1': case '2': case '3': case '4':
5957 case '5': case '6': case '7': case '8': case '9':
5958 s = scan_num(s, &pl_yylval);
5959 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5960 if (PL_expect == XOPERATOR)
5965 s = scan_str(s,!!PL_madskills,FALSE);
5966 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5967 if (PL_expect == XOPERATOR) {
5968 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5969 return deprecate_commaless_var_list();
5976 pl_yylval.ival = OP_CONST;
5977 TERM(sublex_start());
5980 s = scan_str(s,!!PL_madskills,FALSE);
5981 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5982 if (PL_expect == XOPERATOR) {
5983 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5984 return deprecate_commaless_var_list();
5991 pl_yylval.ival = OP_CONST;
5992 /* FIXME. I think that this can be const if char *d is replaced by
5993 more localised variables. */
5994 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5995 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5996 pl_yylval.ival = OP_STRINGIFY;
6000 TERM(sublex_start());
6003 s = scan_str(s,!!PL_madskills,FALSE);
6004 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6005 if (PL_expect == XOPERATOR)
6006 no_op("Backticks",s);
6009 readpipe_override();
6010 TERM(sublex_start());
6014 if (PL_lex_inwhat && isDIGIT(*s))
6015 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6017 if (PL_expect == XOPERATOR)
6018 no_op("Backslash",s);
6022 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6023 char *start = s + 2;
6024 while (isDIGIT(*start) || *start == '_')
6026 if (*start == '.' && isDIGIT(start[1])) {
6027 s = scan_num(s, &pl_yylval);
6030 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6031 else if (!isALPHA(*start) && (PL_expect == XTERM
6032 || PL_expect == XREF || PL_expect == XSTATE
6033 || PL_expect == XTERMORDORDOR)) {
6034 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
6036 s = scan_num(s, &pl_yylval);
6043 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6086 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6088 /* Some keywords can be followed by any delimiter, including ':' */
6089 anydelim = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
6090 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
6091 (PL_tokenbuf[0] == 'q' &&
6092 strchr("qwxr", PL_tokenbuf[1])))));
6094 /* x::* is just a word, unless x is "CORE" */
6095 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6099 while (d < PL_bufend && isSPACE(*d))
6100 d++; /* no comments skipped here, or s### is misparsed */
6102 /* Is this a word before a => operator? */
6103 if (*d == '=' && d[1] == '>') {
6106 = (OP*)newSVOP(OP_CONST, 0,
6107 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6108 pl_yylval.opval->op_private = OPpCONST_BARE;
6112 /* Check for plugged-in keyword */
6116 char *saved_bufptr = PL_bufptr;
6118 result = CALL_FPTR(PL_keyword_plugin)(aTHX_ PL_tokenbuf, len, &o);
6120 if (result == KEYWORD_PLUGIN_DECLINE) {
6121 /* not a plugged-in keyword */
6122 PL_bufptr = saved_bufptr;
6123 } else if (result == KEYWORD_PLUGIN_STMT) {
6124 pl_yylval.opval = o;
6127 return REPORT(PLUGSTMT);
6128 } else if (result == KEYWORD_PLUGIN_EXPR) {
6129 pl_yylval.opval = o;
6131 PL_expect = XOPERATOR;
6132 return REPORT(PLUGEXPR);
6134 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6139 /* Check for built-in keyword */
6140 tmp = keyword(PL_tokenbuf, len, 0);
6142 /* Is this a label? */
6143 if (!anydelim && PL_expect == XSTATE
6144 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6146 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
6151 if (tmp < 0) { /* second-class keyword? */
6152 GV *ogv = NULL; /* override (winner) */
6153 GV *hgv = NULL; /* hidden (loser) */
6154 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6156 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
6159 if (GvIMPORTED_CV(gv))
6161 else if (! CvMETHOD(cv))
6165 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
6166 (gv = *gvp) && isGV_with_GP(gv) &&
6167 GvCVu(gv) && GvIMPORTED_CV(gv))
6174 tmp = 0; /* overridden by import or by GLOBAL */
6177 && -tmp==KEY_lock /* XXX generalizable kludge */
6180 tmp = 0; /* any sub overrides "weak" keyword */
6182 else { /* no override */
6184 if (tmp == KEY_dump) {
6185 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6186 "dump() better written as CORE::dump()");
6190 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6191 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6192 "Ambiguous call resolved as CORE::%s(), %s",
6193 GvENAME(hgv), "qualify as such or use &");
6200 default: /* not a keyword */
6201 /* Trade off - by using this evil construction we can pull the
6202 variable gv into the block labelled keylookup. If not, then
6203 we have to give it function scope so that the goto from the
6204 earlier ':' case doesn't bypass the initialisation. */
6206 just_a_word_zero_gv:
6214 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6218 SV *nextPL_nextwhite = 0;
6222 /* Get the rest if it looks like a package qualifier */
6224 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6226 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6229 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
6230 *s == '\'' ? "'" : "::");
6235 if (PL_expect == XOPERATOR) {
6236 if (PL_bufptr == PL_linestart) {
6237 CopLINE_dec(PL_curcop);
6238 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6239 CopLINE_inc(PL_curcop);
6242 no_op("Bareword",s);
6245 /* Look for a subroutine with this name in current package,
6246 unless name is "Foo::", in which case Foo is a bearword
6247 (and a package name). */
6249 if (len > 2 && !PL_madskills &&
6250 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6252 if (ckWARN(WARN_BAREWORD)
6253 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
6254 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6255 "Bareword \"%s\" refers to nonexistent package",
6258 PL_tokenbuf[len] = '\0';
6264 /* Mustn't actually add anything to a symbol table.
6265 But also don't want to "initialise" any placeholder
6266 constants that might already be there into full
6267 blown PVGVs with attached PVCV. */
6268 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6269 GV_NOADD_NOINIT, SVt_PVCV);
6274 /* if we saw a global override before, get the right name */
6277 sv = newSVpvs("CORE::GLOBAL::");
6278 sv_catpv(sv,PL_tokenbuf);
6281 /* If len is 0, newSVpv does strlen(), which is correct.
6282 If len is non-zero, then it will be the true length,
6283 and so the scalar will be created correctly. */
6284 sv = newSVpv(PL_tokenbuf,len);
6287 if (PL_madskills && !PL_thistoken) {
6288 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6289 PL_thistoken = newSVpvn(start,s - start);
6290 PL_realtokenstart = s - SvPVX(PL_linestr);
6294 /* Presume this is going to be a bareword of some sort. */
6297 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6298 pl_yylval.opval->op_private = OPpCONST_BARE;
6299 /* UTF-8 package name? */
6300 if (UTF && !IN_BYTES &&
6301 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
6304 /* And if "Foo::", then that's what it certainly is. */
6311 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc(sv));
6312 const_op->op_private = OPpCONST_BARE;
6313 rv2cv_op = newCVREF(0, const_op);
6315 if (rv2cv_op->op_type == OP_RV2CV &&
6316 (rv2cv_op->op_flags & OPf_KIDS)) {
6317 OP *rv_op = cUNOPx(rv2cv_op)->op_first;
6318 switch (rv_op->op_type) {
6320 SV *sv = cSVOPx_sv(rv_op);
6321 if (SvROK(sv) && SvTYPE(SvRV(sv)) == SVt_PVCV)
6325 GV *gv = cGVOPx_gv(rv_op);
6326 CV *maybe_cv = GvCVu(gv);
6327 if (maybe_cv && SvTYPE((SV*)maybe_cv) == SVt_PVCV)
6333 /* See if it's the indirect object for a list operator. */
6335 if (PL_oldoldbufptr &&
6336 PL_oldoldbufptr < PL_bufptr &&
6337 (PL_oldoldbufptr == PL_last_lop
6338 || PL_oldoldbufptr == PL_last_uni) &&
6339 /* NO SKIPSPACE BEFORE HERE! */
6340 (PL_expect == XREF ||
6341 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6343 bool immediate_paren = *s == '(';
6345 /* (Now we can afford to cross potential line boundary.) */
6346 s = SKIPSPACE2(s,nextPL_nextwhite);
6348 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6351 /* Two barewords in a row may indicate method call. */
6353 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6354 (tmp = intuit_method(s, gv, cv))) {
6359 /* If not a declared subroutine, it's an indirect object. */
6360 /* (But it's an indir obj regardless for sort.) */
6361 /* Also, if "_" follows a filetest operator, it's a bareword */
6364 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6366 (PL_last_lop_op != OP_MAPSTART &&
6367 PL_last_lop_op != OP_GREPSTART))))
6368 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6369 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6372 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6377 PL_expect = XOPERATOR;
6380 s = SKIPSPACE2(s,nextPL_nextwhite);
6381 PL_nextwhite = nextPL_nextwhite;
6386 /* Is this a word before a => operator? */
6387 if (*s == '=' && s[1] == '>' && !pkgname) {
6390 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6391 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6392 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6396 /* If followed by a paren, it's certainly a subroutine. */
6401 while (SPACE_OR_TAB(*d))
6403 if (*d == ')' && (sv = cv_const_sv(cv))) {
6410 PL_nextwhite = PL_thiswhite;
6413 start_force(PL_curforce);
6415 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6416 PL_expect = XOPERATOR;
6419 PL_nextwhite = nextPL_nextwhite;
6420 curmad('X', PL_thistoken);
6421 PL_thistoken = newSVpvs("");
6430 /* If followed by var or block, call it a method (unless sub) */
6432 if ((*s == '$' || *s == '{') && !cv) {
6434 PL_last_lop = PL_oldbufptr;
6435 PL_last_lop_op = OP_METHOD;
6439 /* If followed by a bareword, see if it looks like indir obj. */
6442 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6443 && (tmp = intuit_method(s, gv, cv))) {
6448 /* Not a method, so call it a subroutine (if defined) */
6451 if (lastchar == '-')
6452 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6453 "Ambiguous use of -%s resolved as -&%s()",
6454 PL_tokenbuf, PL_tokenbuf);
6455 /* Check for a constant sub */
6456 if ((sv = cv_const_sv(cv))) {
6459 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6460 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6461 pl_yylval.opval->op_private = 0;
6465 op_free(pl_yylval.opval);
6466 pl_yylval.opval = rv2cv_op;
6467 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6468 PL_last_lop = PL_oldbufptr;
6469 PL_last_lop_op = OP_ENTERSUB;
6470 /* Is there a prototype? */
6478 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
6481 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
6483 while (*proto == ';')
6485 if (*proto == '&' && *s == '{') {
6487 sv_setpvs(PL_subname, "__ANON__");
6489 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6496 PL_nextwhite = PL_thiswhite;
6499 start_force(PL_curforce);
6500 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6503 PL_nextwhite = nextPL_nextwhite;
6504 curmad('X', PL_thistoken);
6505 PL_thistoken = newSVpvs("");
6512 /* Guess harder when madskills require "best effort". */
6513 if (PL_madskills && (!gv || !GvCVu(gv))) {
6514 int probable_sub = 0;
6515 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6517 else if (isALPHA(*s)) {
6521 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6522 if (!keyword(tmpbuf, tmplen, 0))
6525 while (d < PL_bufend && isSPACE(*d))
6527 if (*d == '=' && d[1] == '>')
6532 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
6533 op_free(pl_yylval.opval);
6534 pl_yylval.opval = rv2cv_op;
6535 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6536 PL_last_lop = PL_oldbufptr;
6537 PL_last_lop_op = OP_ENTERSUB;
6538 PL_nextwhite = PL_thiswhite;
6540 start_force(PL_curforce);
6541 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6543 PL_nextwhite = nextPL_nextwhite;
6544 curmad('X', PL_thistoken);
6545 PL_thistoken = newSVpvs("");
6550 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6557 /* Call it a bare word */
6559 if (PL_hints & HINT_STRICT_SUBS)
6560 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6563 /* after "print" and similar functions (corresponding to
6564 * "F? L" in opcode.pl), whatever wasn't already parsed as
6565 * a filehandle should be subject to "strict subs".
6566 * Likewise for the optional indirect-object argument to system
6567 * or exec, which can't be a bareword */
6568 if ((PL_last_lop_op == OP_PRINT
6569 || PL_last_lop_op == OP_PRTF
6570 || PL_last_lop_op == OP_SAY
6571 || PL_last_lop_op == OP_SYSTEM
6572 || PL_last_lop_op == OP_EXEC)
6573 && (PL_hints & HINT_STRICT_SUBS))
6574 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6575 if (lastchar != '-') {
6576 if (ckWARN(WARN_RESERVED)) {
6580 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
6581 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6589 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
6590 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6591 "Operator or semicolon missing before %c%s",
6592 lastchar, PL_tokenbuf);
6593 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6594 "Ambiguous use of %c resolved as operator %c",
6595 lastchar, lastchar);
6601 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6602 newSVpv(CopFILE(PL_curcop),0));
6606 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6607 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
6610 case KEY___PACKAGE__:
6611 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6613 ? newSVhek(HvNAME_HEK(PL_curstash))
6620 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6621 const char *pname = "main";
6622 if (PL_tokenbuf[2] == 'D')
6623 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
6624 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
6628 GvIOp(gv) = newIO();
6629 IoIFP(GvIOp(gv)) = PL_rsfp;
6630 #if defined(HAS_FCNTL) && defined(F_SETFD)
6632 const int fd = PerlIO_fileno(PL_rsfp);
6633 fcntl(fd,F_SETFD,fd >= 3);
6636 /* Mark this internal pseudo-handle as clean */
6637 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6638 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6639 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6641 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6642 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6643 /* if the script was opened in binmode, we need to revert
6644 * it to text mode for compatibility; but only iff it has CRs
6645 * XXX this is a questionable hack at best. */
6646 if (PL_bufend-PL_bufptr > 2
6647 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6650 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6651 loc = PerlIO_tell(PL_rsfp);
6652 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6655 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6657 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6658 #endif /* NETWARE */
6659 #ifdef PERLIO_IS_STDIO /* really? */
6660 # if defined(__BORLANDC__)
6661 /* XXX see note in do_binmode() */
6662 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
6666 PerlIO_seek(PL_rsfp, loc, 0);
6670 #ifdef PERLIO_LAYERS
6673 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6674 else if (PL_encoding) {
6681 XPUSHs(PL_encoding);
6683 call_method("name", G_SCALAR);
6687 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
6688 Perl_form(aTHX_ ":encoding(%"SVf")",
6697 if (PL_realtokenstart >= 0) {
6698 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6700 PL_endwhite = newSVpvs("");
6701 sv_catsv(PL_endwhite, PL_thiswhite);
6703 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
6704 PL_realtokenstart = -1;
6706 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
6722 if (PL_expect == XSTATE) {
6729 if (*s == ':' && s[1] == ':') {
6732 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6733 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
6734 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
6737 else if (tmp == KEY_require || tmp == KEY_do)
6738 /* that's a way to remember we saw "CORE::" */
6751 LOP(OP_ACCEPT,XTERM);
6757 LOP(OP_ATAN2,XTERM);
6763 LOP(OP_BINMODE,XTERM);
6766 LOP(OP_BLESS,XTERM);
6775 /* When 'use switch' is in effect, continue has a dual
6776 life as a control operator. */
6778 if (!FEATURE_IS_ENABLED("switch"))
6781 /* We have to disambiguate the two senses of
6782 "continue". If the next token is a '{' then
6783 treat it as the start of a continue block;
6784 otherwise treat it as a control operator.
6796 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
6813 if (!PL_cryptseen) {
6814 PL_cryptseen = TRUE;
6818 LOP(OP_CRYPT,XTERM);
6821 LOP(OP_CHMOD,XTERM);
6824 LOP(OP_CHOWN,XTERM);
6827 LOP(OP_CONNECT,XTERM);
6846 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6847 if (orig_keyword == KEY_do) {
6856 PL_hints |= HINT_BLOCK_SCOPE;
6866 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
6867 LOP(OP_DBMOPEN,XTERM);
6873 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6880 pl_yylval.ival = CopLINE(PL_curcop);
6896 if (*s == '{') { /* block eval */
6897 PL_expect = XTERMBLOCK;
6898 UNIBRACK(OP_ENTERTRY);
6900 else { /* string eval */
6902 UNIBRACK(OP_ENTEREVAL);
6917 case KEY_endhostent:
6923 case KEY_endservent:
6926 case KEY_endprotoent:
6937 pl_yylval.ival = CopLINE(PL_curcop);
6939 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6942 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6945 if ((PL_bufend - p) >= 3 &&
6946 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6948 else if ((PL_bufend - p) >= 4 &&
6949 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6952 if (isIDFIRST_lazy_if(p,UTF)) {
6953 p = scan_ident(p, PL_bufend,
6954 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6958 Perl_croak(aTHX_ "Missing $ on loop variable");
6960 s = SvPVX(PL_linestr) + soff;
6966 LOP(OP_FORMLINE,XTERM);
6972 LOP(OP_FCNTL,XTERM);
6978 LOP(OP_FLOCK,XTERM);
6987 LOP(OP_GREPSTART, XREF);
6990 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7005 case KEY_getpriority:
7006 LOP(OP_GETPRIORITY,XTERM);
7008 case KEY_getprotobyname:
7011 case KEY_getprotobynumber:
7012 LOP(OP_GPBYNUMBER,XTERM);
7014 case KEY_getprotoent:
7026 case KEY_getpeername:
7027 UNI(OP_GETPEERNAME);
7029 case KEY_gethostbyname:
7032 case KEY_gethostbyaddr:
7033 LOP(OP_GHBYADDR,XTERM);
7035 case KEY_gethostent:
7038 case KEY_getnetbyname:
7041 case KEY_getnetbyaddr:
7042 LOP(OP_GNBYADDR,XTERM);
7047 case KEY_getservbyname:
7048 LOP(OP_GSBYNAME,XTERM);
7050 case KEY_getservbyport:
7051 LOP(OP_GSBYPORT,XTERM);
7053 case KEY_getservent:
7056 case KEY_getsockname:
7057 UNI(OP_GETSOCKNAME);
7059 case KEY_getsockopt:
7060 LOP(OP_GSOCKOPT,XTERM);
7075 pl_yylval.ival = CopLINE(PL_curcop);
7085 pl_yylval.ival = CopLINE(PL_curcop);
7089 LOP(OP_INDEX,XTERM);
7095 LOP(OP_IOCTL,XTERM);
7107 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7139 LOP(OP_LISTEN,XTERM);
7148 s = scan_pat(s,OP_MATCH);
7149 TERM(sublex_start());
7152 LOP(OP_MAPSTART, XREF);
7155 LOP(OP_MKDIR,XTERM);
7158 LOP(OP_MSGCTL,XTERM);
7161 LOP(OP_MSGGET,XTERM);
7164 LOP(OP_MSGRCV,XTERM);
7167 LOP(OP_MSGSND,XTERM);
7172 PL_in_my = (U16)tmp;
7174 if (isIDFIRST_lazy_if(s,UTF)) {
7178 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7179 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7181 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7182 if (!PL_in_my_stash) {
7185 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7189 if (PL_madskills) { /* just add type to declarator token */
7190 sv_catsv(PL_thistoken, PL_nextwhite);
7192 sv_catpvn(PL_thistoken, start, s - start);
7200 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7207 s = tokenize_use(0, s);
7211 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7218 if (isIDFIRST_lazy_if(s,UTF)) {
7220 for (d = s; isALNUM_lazy_if(d,UTF);)
7222 for (t=d; isSPACE(*t);)
7224 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7226 && !(t[0] == '=' && t[1] == '>')
7228 int parms_len = (int)(d-s);
7229 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7230 "Precedence problem: open %.*s should be open(%.*s)",
7231 parms_len, s, parms_len, s);
7237 pl_yylval.ival = OP_OR;
7247 LOP(OP_OPEN_DIR,XTERM);
7250 checkcomma(s,PL_tokenbuf,"filehandle");
7254 checkcomma(s,PL_tokenbuf,"filehandle");
7273 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7275 s = force_strict_version(s);
7276 PL_lex_expect = XBLOCK;
7280 LOP(OP_PIPE_OP,XTERM);
7283 s = scan_str(s,!!PL_madskills,FALSE);
7286 pl_yylval.ival = OP_CONST;
7287 TERM(sublex_start());
7293 s = scan_str(s,!!PL_madskills,FALSE);
7296 PL_expect = XOPERATOR;
7298 if (SvCUR(PL_lex_stuff)) {
7301 d = SvPV_force(PL_lex_stuff, len);
7303 for (; isSPACE(*d) && len; --len, ++d)
7308 if (!warned && ckWARN(WARN_QW)) {
7309 for (; !isSPACE(*d) && len; --len, ++d) {
7311 Perl_warner(aTHX_ packWARN(WARN_QW),
7312 "Possible attempt to separate words with commas");
7315 else if (*d == '#') {
7316 Perl_warner(aTHX_ packWARN(WARN_QW),
7317 "Possible attempt to put comments in qw() list");
7323 for (; !isSPACE(*d) && len; --len, ++d)
7326 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7327 words = append_elem(OP_LIST, words,
7328 newSVOP(OP_CONST, 0, tokeq(sv)));
7332 start_force(PL_curforce);
7333 NEXTVAL_NEXTTOKE.opval = words;
7338 SvREFCNT_dec(PL_lex_stuff);
7339 PL_lex_stuff = NULL;
7345 s = scan_str(s,!!PL_madskills,FALSE);
7348 pl_yylval.ival = OP_STRINGIFY;
7349 if (SvIVX(PL_lex_stuff) == '\'')
7350 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
7351 TERM(sublex_start());
7354 s = scan_pat(s,OP_QR);
7355 TERM(sublex_start());
7358 s = scan_str(s,!!PL_madskills,FALSE);
7361 readpipe_override();
7362 TERM(sublex_start());
7370 s = force_version(s, FALSE);
7372 else if (*s != 'v' || !isDIGIT(s[1])
7373 || (s = force_version(s, TRUE), *s == 'v'))
7375 *PL_tokenbuf = '\0';
7376 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7377 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7378 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7380 yyerror("<> should be quotes");
7382 if (orig_keyword == KEY_require) {
7390 PL_last_uni = PL_oldbufptr;
7391 PL_last_lop_op = OP_REQUIRE;
7393 return REPORT( (int)REQUIRE );
7399 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7403 LOP(OP_RENAME,XTERM);
7412 LOP(OP_RINDEX,XTERM);
7421 UNIDOR(OP_READLINE);
7424 UNIDOR(OP_BACKTICK);
7433 LOP(OP_REVERSE,XTERM);
7436 UNIDOR(OP_READLINK);
7443 if (pl_yylval.opval)
7444 TERM(sublex_start());
7446 TOKEN(1); /* force error */
7449 checkcomma(s,PL_tokenbuf,"filehandle");
7459 LOP(OP_SELECT,XTERM);
7465 LOP(OP_SEMCTL,XTERM);
7468 LOP(OP_SEMGET,XTERM);
7471 LOP(OP_SEMOP,XTERM);
7477 LOP(OP_SETPGRP,XTERM);
7479 case KEY_setpriority:
7480 LOP(OP_SETPRIORITY,XTERM);
7482 case KEY_sethostent:
7488 case KEY_setservent:
7491 case KEY_setprotoent:
7501 LOP(OP_SEEKDIR,XTERM);
7503 case KEY_setsockopt:
7504 LOP(OP_SSOCKOPT,XTERM);
7510 LOP(OP_SHMCTL,XTERM);
7513 LOP(OP_SHMGET,XTERM);
7516 LOP(OP_SHMREAD,XTERM);
7519 LOP(OP_SHMWRITE,XTERM);
7522 LOP(OP_SHUTDOWN,XTERM);
7531 LOP(OP_SOCKET,XTERM);
7533 case KEY_socketpair:
7534 LOP(OP_SOCKPAIR,XTERM);
7537 checkcomma(s,PL_tokenbuf,"subroutine name");
7539 if (*s == ';' || *s == ')') /* probably a close */
7540 Perl_croak(aTHX_ "sort is now a reserved word");
7542 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7546 LOP(OP_SPLIT,XTERM);
7549 LOP(OP_SPRINTF,XTERM);
7552 LOP(OP_SPLICE,XTERM);
7567 LOP(OP_SUBSTR,XTERM);
7573 char tmpbuf[sizeof PL_tokenbuf];
7574 SSize_t tboffset = 0;
7575 expectation attrful;
7576 bool have_name, have_proto;
7577 const int key = tmp;
7582 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7583 SV *subtoken = newSVpvn(tstart, s - tstart);
7587 s = SKIPSPACE2(s,tmpwhite);
7592 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7593 (*s == ':' && s[1] == ':'))
7596 SV *nametoke = NULL;
7600 attrful = XATTRBLOCK;
7601 /* remember buffer pos'n for later force_word */
7602 tboffset = s - PL_oldbufptr;
7603 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7606 nametoke = newSVpvn(s, d - s);
7608 if (memchr(tmpbuf, ':', len))
7609 sv_setpvn(PL_subname, tmpbuf, len);
7611 sv_setsv(PL_subname,PL_curstname);
7612 sv_catpvs(PL_subname,"::");
7613 sv_catpvn(PL_subname,tmpbuf,len);
7620 CURMAD('X', nametoke);
7621 CURMAD('_', tmpwhite);
7622 (void) force_word(PL_oldbufptr + tboffset, WORD,
7625 s = SKIPSPACE2(d,tmpwhite);
7632 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7633 PL_expect = XTERMBLOCK;
7634 attrful = XATTRTERM;
7635 sv_setpvs(PL_subname,"?");
7639 if (key == KEY_format) {
7641 PL_lex_formbrack = PL_lex_brackets + 1;
7643 PL_thistoken = subtoken;
7647 (void) force_word(PL_oldbufptr + tboffset, WORD,
7653 /* Look for a prototype */
7656 bool bad_proto = FALSE;
7657 bool in_brackets = FALSE;
7658 char greedy_proto = ' ';
7659 bool proto_after_greedy_proto = FALSE;
7660 bool must_be_last = FALSE;
7661 bool underscore = FALSE;
7662 bool seen_underscore = FALSE;
7663 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
7665 s = scan_str(s,!!PL_madskills,FALSE);
7667 Perl_croak(aTHX_ "Prototype not terminated");
7668 /* strip spaces and check for bad characters */
7669 d = SvPVX(PL_lex_stuff);
7671 for (p = d; *p; ++p) {
7675 if (warnillegalproto) {
7677 proto_after_greedy_proto = TRUE;
7678 if (!strchr("$@%*;[]&\\_", *p)) {
7690 else if ( *p == ']' ) {
7691 in_brackets = FALSE;
7693 else if ( (*p == '@' || *p == '%') &&
7694 ( tmp < 2 || d[tmp-2] != '\\' ) &&
7696 must_be_last = TRUE;
7699 else if ( *p == '_' ) {
7700 underscore = seen_underscore = TRUE;
7707 if (proto_after_greedy_proto)
7708 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7709 "Prototype after '%c' for %"SVf" : %s",
7710 greedy_proto, SVfARG(PL_subname), d);
7712 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7713 "Illegal character %sin prototype for %"SVf" : %s",
7714 seen_underscore ? "after '_' " : "",
7715 SVfARG(PL_subname), d);
7716 SvCUR_set(PL_lex_stuff, tmp);
7721 CURMAD('q', PL_thisopen);
7722 CURMAD('_', tmpwhite);
7723 CURMAD('=', PL_thisstuff);
7724 CURMAD('Q', PL_thisclose);
7725 NEXTVAL_NEXTTOKE.opval =
7726 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7727 PL_lex_stuff = NULL;
7730 s = SKIPSPACE2(s,tmpwhite);
7738 if (*s == ':' && s[1] != ':')
7739 PL_expect = attrful;
7740 else if (*s != '{' && key == KEY_sub) {
7742 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
7743 else if (*s != ';' && *s != '}')
7744 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
7751 curmad('^', newSVpvs(""));
7752 CURMAD('_', tmpwhite);
7756 PL_thistoken = subtoken;
7759 NEXTVAL_NEXTTOKE.opval =
7760 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7761 PL_lex_stuff = NULL;
7767 sv_setpvs(PL_subname, "__ANON__");
7769 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7773 (void) force_word(PL_oldbufptr + tboffset, WORD,
7782 LOP(OP_SYSTEM,XREF);
7785 LOP(OP_SYMLINK,XTERM);
7788 LOP(OP_SYSCALL,XTERM);
7791 LOP(OP_SYSOPEN,XTERM);
7794 LOP(OP_SYSSEEK,XTERM);
7797 LOP(OP_SYSREAD,XTERM);
7800 LOP(OP_SYSWRITE,XTERM);
7804 TERM(sublex_start());
7825 LOP(OP_TRUNCATE,XTERM);
7837 pl_yylval.ival = CopLINE(PL_curcop);
7841 pl_yylval.ival = CopLINE(PL_curcop);
7845 LOP(OP_UNLINK,XTERM);
7851 LOP(OP_UNPACK,XTERM);
7854 LOP(OP_UTIME,XTERM);
7860 LOP(OP_UNSHIFT,XTERM);
7863 s = tokenize_use(1, s);
7873 pl_yylval.ival = CopLINE(PL_curcop);
7877 pl_yylval.ival = CopLINE(PL_curcop);
7881 PL_hints |= HINT_BLOCK_SCOPE;
7888 LOP(OP_WAITPID,XTERM);
7897 ctl_l[0] = toCTRL('L');
7899 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7902 /* Make sure $^L is defined */
7903 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7908 if (PL_expect == XOPERATOR)
7914 pl_yylval.ival = OP_XOR;
7919 TERM(sublex_start());
7924 #pragma segment Main
7928 S_pending_ident(pTHX)
7933 /* pit holds the identifier we read and pending_ident is reset */
7934 char pit = PL_pending_ident;
7935 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7936 /* All routes through this function want to know if there is a colon. */
7937 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7938 PL_pending_ident = 0;
7940 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7941 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7942 "### Pending identifier '%s'\n", PL_tokenbuf); });
7944 /* if we're in a my(), we can't allow dynamics here.
7945 $foo'bar has already been turned into $foo::bar, so
7946 just check for colons.
7948 if it's a legal name, the OP is a PADANY.
7951 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7953 yyerror(Perl_form(aTHX_ "No package name allowed for "
7954 "variable %s in \"our\"",
7956 tmp = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7960 yyerror(Perl_form(aTHX_ PL_no_myglob,
7961 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7963 pl_yylval.opval = newOP(OP_PADANY, 0);
7964 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7970 build the ops for accesses to a my() variable.
7972 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7973 then used in a comparison. This catches most, but not
7974 all cases. For instance, it catches
7975 sort { my($a); $a <=> $b }
7977 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7978 (although why you'd do that is anyone's guess).
7983 tmp = pad_findmy(PL_tokenbuf, tokenbuf_len, 0);
7984 if (tmp != NOT_IN_PAD) {
7985 /* might be an "our" variable" */
7986 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7987 /* build ops for a bareword */
7988 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7989 HEK * const stashname = HvNAME_HEK(stash);
7990 SV * const sym = newSVhek(stashname);
7991 sv_catpvs(sym, "::");
7992 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7993 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7994 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7997 ? (GV_ADDMULTI | GV_ADDINEVAL)
8000 ((PL_tokenbuf[0] == '$') ? SVt_PV
8001 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8006 /* if it's a sort block and they're naming $a or $b */
8007 if (PL_last_lop_op == OP_SORT &&
8008 PL_tokenbuf[0] == '$' &&
8009 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
8012 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
8013 d < PL_bufend && *d != '\n';
8016 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
8017 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
8023 pl_yylval.opval = newOP(OP_PADANY, 0);
8024 pl_yylval.opval->op_targ = tmp;
8030 Whine if they've said @foo in a doublequoted string,
8031 and @foo isn't a variable we can find in the symbol
8034 if (ckWARN(WARN_AMBIGUOUS) &&
8035 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8036 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
8038 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8039 /* DO NOT warn for @- and @+ */
8040 && !( PL_tokenbuf[2] == '\0' &&
8041 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8044 /* Downgraded from fatal to warning 20000522 mjd */
8045 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8046 "Possible unintended interpolation of %s in string",
8051 /* build ops for a bareword */
8052 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
8054 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8055 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8056 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD,
8057 ((PL_tokenbuf[0] == '$') ? SVt_PV
8058 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8064 * The following code was generated by perl_keyword.pl.
8068 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
8072 PERL_ARGS_ASSERT_KEYWORD;
8076 case 1: /* 5 tokens of length 1 */
8108 case 2: /* 18 tokens of length 2 */
8254 case 3: /* 29 tokens of length 3 */
8258 if (name[1] == 'N' &&
8321 if (name[1] == 'i' &&
8353 if (name[1] == 'o' &&
8362 if (name[1] == 'e' &&
8371 if (name[1] == 'n' &&
8380 if (name[1] == 'o' &&
8389 if (name[1] == 'a' &&
8398 if (name[1] == 'o' &&
8460 if (name[1] == 'e' &&
8474 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
8500 if (name[1] == 'i' &&
8509 if (name[1] == 's' &&
8518 if (name[1] == 'e' &&
8527 if (name[1] == 'o' &&
8539 case 4: /* 41 tokens of length 4 */
8543 if (name[1] == 'O' &&
8553 if (name[1] == 'N' &&
8563 if (name[1] == 'i' &&
8573 if (name[1] == 'h' &&
8583 if (name[1] == 'u' &&
8596 if (name[2] == 'c' &&
8605 if (name[2] == 's' &&
8614 if (name[2] == 'a' &&
8650 if (name[1] == 'o' &&
8663 if (name[2] == 't' &&
8672 if (name[2] == 'o' &&
8681 if (name[2] == 't' &&
8690 if (name[2] == 'e' &&
8703 if (name[1] == 'o' &&
8716 if (name[2] == 'y' &&
8725 if (name[2] == 'l' &&
8741 if (name[2] == 's' &&
8750 if (name[2] == 'n' &&
8759 if (name[2] == 'c' &&
8772 if (name[1] == 'e' &&
8782 if (name[1] == 'p' &&
8795 if (name[2] == 'c' &&
8804 if (name[2] == 'p' &&
8813 if (name[2] == 's' &&
8829 if (name[2] == 'n' &&
8899 if (name[2] == 'r' &&
8908 if (name[2] == 'r' &&
8917 if (name[2] == 'a' &&
8933 if (name[2] == 'l' &&
8995 if (name[2] == 'e' &&
8998 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
9011 case 5: /* 39 tokens of length 5 */
9015 if (name[1] == 'E' &&
9026 if (name[1] == 'H' &&
9040 if (name[2] == 'a' &&
9050 if (name[2] == 'a' &&
9067 if (name[2] == 'e' &&
9077 if (name[2] == 'e' &&
9081 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
9097 if (name[3] == 'i' &&
9106 if (name[3] == 'o' &&
9142 if (name[2] == 'o' &&
9152 if (name[2] == 'y' &&
9166 if (name[1] == 'l' &&
9180 if (name[2] == 'n' &&
9190 if (name[2] == 'o' &&
9204 if (name[1] == 'i' &&
9209 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
9218 if (name[2] == 'd' &&
9228 if (name[2] == 'c' &&
9245 if (name[2] == 'c' &&
9255 if (name[2] == 't' &&
9269 if (name[1] == 'k' &&
9280 if (name[1] == 'r' &&
9294 if (name[2] == 's' &&
9304 if (name[2] == 'd' &&
9321 if (name[2] == 'm' &&
9331 if (name[2] == 'i' &&
9341 if (name[2] == 'e' &&
9351 if (name[2] == 'l' &&
9361 if (name[2] == 'a' &&
9374 if (name[3] == 't' &&
9377 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
9383 if (name[3] == 'd' &&
9400 if (name[1] == 'i' &&
9414 if (name[2] == 'a' &&
9427 if (name[3] == 'e' &&
9462 if (name[2] == 'i' &&
9479 if (name[2] == 'i' &&
9489 if (name[2] == 'i' &&
9506 case 6: /* 33 tokens of length 6 */
9510 if (name[1] == 'c' &&
9525 if (name[2] == 'l' &&
9536 if (name[2] == 'r' &&
9551 if (name[1] == 'e' &&
9566 if (name[2] == 's' &&
9571 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
9577 if (name[2] == 'i' &&
9595 if (name[2] == 'l' &&
9606 if (name[2] == 'r' &&
9621 if (name[1] == 'm' &&
9636 if (name[2] == 'n' &&
9647 if (name[2] == 's' &&
9662 if (name[1] == 's' &&
9668 if (name[4] == 't' &&
9677 if (name[4] == 'e' &&
9686 if (name[4] == 'c' &&
9695 if (name[4] == 'n' &&
9711 if (name[1] == 'r' &&
9729 if (name[3] == 'a' &&
9739 if (name[3] == 'u' &&
9753 if (name[2] == 'n' &&
9771 if (name[2] == 'a' &&
9785 if (name[3] == 'e' &&
9798 if (name[4] == 't' &&
9807 if (name[4] == 'e' &&
9829 if (name[4] == 't' &&
9838 if (name[4] == 'e' &&
9854 if (name[2] == 'c' &&
9865 if (name[2] == 'l' &&
9876 if (name[2] == 'b' &&
9887 if (name[2] == 's' &&
9910 if (name[4] == 's' &&
9919 if (name[4] == 'n' &&
9932 if (name[3] == 'a' &&
9949 if (name[1] == 'a' &&
9964 case 7: /* 29 tokens of length 7 */
9968 if (name[1] == 'E' &&
9981 if (name[1] == '_' &&
9994 if (name[1] == 'i' &&
10001 return -KEY_binmode;
10007 if (name[1] == 'o' &&
10014 return -KEY_connect;
10023 if (name[2] == 'm' &&
10029 return -KEY_dbmopen;
10035 if (name[2] == 'f')
10040 if (name[4] == 'u' &&
10044 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
10050 if (name[4] == 'n' &&
10054 return KEY_defined;
10071 if (name[1] == 'o' &&
10078 return KEY_foreach;
10084 if (name[1] == 'e' &&
10091 if (name[5] == 'r' &&
10094 return -KEY_getpgrp;
10100 if (name[5] == 'i' &&
10103 return -KEY_getppid;
10116 if (name[1] == 'c' &&
10123 return -KEY_lcfirst;
10129 if (name[1] == 'p' &&
10136 return -KEY_opendir;
10142 if (name[1] == 'a' &&
10149 return KEY_package;
10155 if (name[1] == 'e')
10160 if (name[3] == 'd' &&
10165 return -KEY_readdir;
10171 if (name[3] == 'u' &&
10176 return KEY_require;
10182 if (name[3] == 'e' &&
10187 return -KEY_reverse;
10206 if (name[3] == 'k' &&
10211 return -KEY_seekdir;
10217 if (name[3] == 'p' &&
10222 return -KEY_setpgrp;
10232 if (name[2] == 'm' &&
10238 return -KEY_shmread;
10244 if (name[2] == 'r' &&
10250 return -KEY_sprintf;
10259 if (name[3] == 'l' &&
10264 return -KEY_symlink;
10273 if (name[4] == 'a' &&
10277 return -KEY_syscall;
10283 if (name[4] == 'p' &&
10287 return -KEY_sysopen;
10293 if (name[4] == 'e' &&
10297 return -KEY_sysread;
10303 if (name[4] == 'e' &&
10307 return -KEY_sysseek;
10325 if (name[1] == 'e' &&
10332 return -KEY_telldir;
10341 if (name[2] == 'f' &&
10347 return -KEY_ucfirst;
10353 if (name[2] == 's' &&
10359 return -KEY_unshift;
10369 if (name[1] == 'a' &&
10376 return -KEY_waitpid;
10385 case 8: /* 26 tokens of length 8 */
10389 if (name[1] == 'U' &&
10397 return KEY_AUTOLOAD;
10403 if (name[1] == '_')
10408 if (name[3] == 'A' &&
10414 return KEY___DATA__;
10420 if (name[3] == 'I' &&
10426 return -KEY___FILE__;
10432 if (name[3] == 'I' &&
10438 return -KEY___LINE__;
10454 if (name[2] == 'o' &&
10461 return -KEY_closedir;
10467 if (name[2] == 'n' &&
10474 return -KEY_continue;
10484 if (name[1] == 'b' &&
10492 return -KEY_dbmclose;
10498 if (name[1] == 'n' &&
10504 if (name[4] == 'r' &&
10509 return -KEY_endgrent;
10515 if (name[4] == 'w' &&
10520 return -KEY_endpwent;
10533 if (name[1] == 'o' &&
10541 return -KEY_formline;
10547 if (name[1] == 'e' &&
10553 if (name[4] == 'r')
10558 if (name[6] == 'n' &&
10561 return -KEY_getgrent;
10567 if (name[6] == 'i' &&
10570 return -KEY_getgrgid;
10576 if (name[6] == 'a' &&
10579 return -KEY_getgrnam;
10592 if (name[4] == 'o' &&
10597 return -KEY_getlogin;
10603 if (name[4] == 'w')
10608 if (name[6] == 'n' &&
10611 return -KEY_getpwent;
10617 if (name[6] == 'a' &&
10620 return -KEY_getpwnam;
10626 if (name[6] == 'i' &&
10629 return -KEY_getpwuid;
10649 if (name[1] == 'e' &&
10656 if (name[5] == 'i' &&
10663 return -KEY_readline;
10668 return -KEY_readlink;
10679 if (name[5] == 'i' &&
10683 return -KEY_readpipe;
10699 if (name[2] == 't')
10704 if (name[4] == 'r' &&
10709 return -KEY_setgrent;
10715 if (name[4] == 'w' &&
10720 return -KEY_setpwent;
10736 if (name[3] == 'w' &&
10742 return -KEY_shmwrite;
10748 if (name[3] == 't' &&
10754 return -KEY_shutdown;
10764 if (name[2] == 's' &&
10771 return -KEY_syswrite;
10781 if (name[1] == 'r' &&
10789 return -KEY_truncate;
10798 case 9: /* 9 tokens of length 9 */
10802 if (name[1] == 'N' &&
10811 return KEY_UNITCHECK;
10817 if (name[1] == 'n' &&
10826 return -KEY_endnetent;
10832 if (name[1] == 'e' &&
10841 return -KEY_getnetent;
10847 if (name[1] == 'o' &&
10856 return -KEY_localtime;
10862 if (name[1] == 'r' &&
10871 return KEY_prototype;
10877 if (name[1] == 'u' &&
10886 return -KEY_quotemeta;
10892 if (name[1] == 'e' &&
10901 return -KEY_rewinddir;
10907 if (name[1] == 'e' &&
10916 return -KEY_setnetent;
10922 if (name[1] == 'a' &&
10931 return -KEY_wantarray;
10940 case 10: /* 9 tokens of length 10 */
10944 if (name[1] == 'n' &&
10950 if (name[4] == 'o' &&
10957 return -KEY_endhostent;
10963 if (name[4] == 'e' &&
10970 return -KEY_endservent;
10983 if (name[1] == 'e' &&
10989 if (name[4] == 'o' &&
10996 return -KEY_gethostent;
11005 if (name[5] == 'r' &&
11011 return -KEY_getservent;
11017 if (name[5] == 'c' &&
11023 return -KEY_getsockopt;
11043 if (name[2] == 't')
11048 if (name[4] == 'o' &&
11055 return -KEY_sethostent;
11064 if (name[5] == 'r' &&
11070 return -KEY_setservent;
11076 if (name[5] == 'c' &&
11082 return -KEY_setsockopt;
11099 if (name[2] == 'c' &&
11108 return -KEY_socketpair;
11121 case 11: /* 8 tokens of length 11 */
11125 if (name[1] == '_' &&
11135 { /* __PACKAGE__ */
11136 return -KEY___PACKAGE__;
11142 if (name[1] == 'n' &&
11152 { /* endprotoent */
11153 return -KEY_endprotoent;
11159 if (name[1] == 'e' &&
11168 if (name[5] == 'e' &&
11174 { /* getpeername */
11175 return -KEY_getpeername;
11184 if (name[6] == 'o' &&
11189 { /* getpriority */
11190 return -KEY_getpriority;
11196 if (name[6] == 't' &&
11201 { /* getprotoent */
11202 return -KEY_getprotoent;
11216 if (name[4] == 'o' &&
11223 { /* getsockname */
11224 return -KEY_getsockname;
11237 if (name[1] == 'e' &&
11245 if (name[6] == 'o' &&
11250 { /* setpriority */
11251 return -KEY_setpriority;
11257 if (name[6] == 't' &&
11262 { /* setprotoent */
11263 return -KEY_setprotoent;
11279 case 12: /* 2 tokens of length 12 */
11280 if (name[0] == 'g' &&
11292 if (name[9] == 'd' &&
11295 { /* getnetbyaddr */
11296 return -KEY_getnetbyaddr;
11302 if (name[9] == 'a' &&
11305 { /* getnetbyname */
11306 return -KEY_getnetbyname;
11318 case 13: /* 4 tokens of length 13 */
11319 if (name[0] == 'g' &&
11326 if (name[4] == 'o' &&
11335 if (name[10] == 'd' &&
11338 { /* gethostbyaddr */
11339 return -KEY_gethostbyaddr;
11345 if (name[10] == 'a' &&
11348 { /* gethostbyname */
11349 return -KEY_gethostbyname;
11362 if (name[4] == 'e' &&
11371 if (name[10] == 'a' &&
11374 { /* getservbyname */
11375 return -KEY_getservbyname;
11381 if (name[10] == 'o' &&
11384 { /* getservbyport */
11385 return -KEY_getservbyport;
11404 case 14: /* 1 tokens of length 14 */
11405 if (name[0] == 'g' &&
11419 { /* getprotobyname */
11420 return -KEY_getprotobyname;
11425 case 16: /* 1 tokens of length 16 */
11426 if (name[0] == 'g' &&
11442 { /* getprotobynumber */
11443 return -KEY_getprotobynumber;
11457 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
11461 PERL_ARGS_ASSERT_CHECKCOMMA;
11463 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
11464 if (ckWARN(WARN_SYNTAX)) {
11467 for (w = s+2; *w && level; w++) {
11470 else if (*w == ')')
11473 while (isSPACE(*w))
11475 /* the list of chars below is for end of statements or
11476 * block / parens, boolean operators (&&, ||, //) and branch
11477 * constructs (or, and, if, until, unless, while, err, for).
11478 * Not a very solid hack... */
11479 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
11480 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11481 "%s (...) interpreted as function",name);
11484 while (s < PL_bufend && isSPACE(*s))
11488 while (s < PL_bufend && isSPACE(*s))
11490 if (isIDFIRST_lazy_if(s,UTF)) {
11491 const char * const w = s++;
11492 while (isALNUM_lazy_if(s,UTF))
11494 while (s < PL_bufend && isSPACE(*s))
11498 if (keyword(w, s - w, 0))
11501 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
11502 if (gv && GvCVu(gv))
11504 Perl_croak(aTHX_ "No comma allowed after %s", what);
11509 /* Either returns sv, or mortalizes sv and returns a new SV*.
11510 Best used as sv=new_constant(..., sv, ...).
11511 If s, pv are NULL, calls subroutine with one argument,
11512 and type is used with error messages only. */
11515 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
11516 SV *sv, SV *pv, const char *type, STRLEN typelen)
11519 HV * const table = GvHV(PL_hintgv); /* ^H */
11523 const char *why1 = "", *why2 = "", *why3 = "";
11525 PERL_ARGS_ASSERT_NEW_CONSTANT;
11527 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
11530 why2 = (const char *)
11531 (strEQ(key,"charnames")
11532 ? "(possibly a missing \"use charnames ...\")"
11534 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
11535 (type ? type: "undef"), why2);
11537 /* This is convoluted and evil ("goto considered harmful")
11538 * but I do not understand the intricacies of all the different
11539 * failure modes of %^H in here. The goal here is to make
11540 * the most probable error message user-friendly. --jhi */
11545 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
11546 (type ? type: "undef"), why1, why2, why3);
11548 yyerror(SvPVX_const(msg));
11553 /* charnames doesn't work well if there have been errors found */
11554 if (PL_error_count > 0 && strEQ(key,"charnames"))
11555 return &PL_sv_undef;
11557 cvp = hv_fetch(table, key, keylen, FALSE);
11558 if (!cvp || !SvOK(*cvp)) {
11561 why3 = "} is not defined";
11564 sv_2mortal(sv); /* Parent created it permanently */
11567 pv = newSVpvn_flags(s, len, SVs_TEMP);
11569 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
11571 typesv = &PL_sv_undef;
11573 PUSHSTACKi(PERLSI_OVERLOAD);
11585 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
11589 /* Check the eval first */
11590 if (!PL_in_eval && SvTRUE(ERRSV)) {
11591 sv_catpvs(ERRSV, "Propagated");
11592 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
11594 res = SvREFCNT_inc_simple(sv);
11598 SvREFCNT_inc_simple_void(res);
11607 why1 = "Call to &{$^H{";
11609 why3 = "}} did not return a defined value";
11617 /* Returns a NUL terminated string, with the length of the string written to
11621 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
11624 register char *d = dest;
11625 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11627 PERL_ARGS_ASSERT_SCAN_WORD;
11631 Perl_croak(aTHX_ ident_too_long);
11632 if (isALNUM(*s)) /* UTF handled below */
11634 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
11639 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
11643 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11644 char *t = s + UTF8SKIP(s);
11646 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11650 Perl_croak(aTHX_ ident_too_long);
11651 Copy(s, d, len, char);
11664 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
11667 char *bracket = NULL;
11669 register char *d = dest;
11670 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11672 PERL_ARGS_ASSERT_SCAN_IDENT;
11677 while (isDIGIT(*s)) {
11679 Perl_croak(aTHX_ ident_too_long);
11686 Perl_croak(aTHX_ ident_too_long);
11687 if (isALNUM(*s)) /* UTF handled below */
11689 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
11694 else if (*s == ':' && s[1] == ':') {
11698 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11699 char *t = s + UTF8SKIP(s);
11700 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11702 if (d + (t - s) > e)
11703 Perl_croak(aTHX_ ident_too_long);
11704 Copy(s, d, t - s, char);
11715 if (PL_lex_state != LEX_NORMAL)
11716 PL_lex_state = LEX_INTERPENDMAYBE;
11719 if (*s == '$' && s[1] &&
11720 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
11733 if (*d == '^' && *s && isCONTROLVAR(*s)) {
11738 if (isSPACE(s[-1])) {
11740 const char ch = *s++;
11741 if (!SPACE_OR_TAB(ch)) {
11747 if (isIDFIRST_lazy_if(d,UTF)) {
11751 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
11752 end += UTF8SKIP(end);
11753 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
11754 end += UTF8SKIP(end);
11756 Copy(s, d, end - s, char);
11761 while ((isALNUM(*s) || *s == ':') && d < e)
11764 Perl_croak(aTHX_ ident_too_long);
11767 while (s < send && SPACE_OR_TAB(*s))
11769 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
11770 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
11771 const char * const brack =
11773 ((*s == '[') ? "[...]" : "{...}");
11774 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11775 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
11776 funny, dest, brack, funny, dest, brack);
11779 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
11783 /* Handle extended ${^Foo} variables
11784 * 1999-02-27 mjd-perl-patch@plover.com */
11785 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
11789 while (isALNUM(*s) && d < e) {
11793 Perl_croak(aTHX_ ident_too_long);
11798 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
11799 PL_lex_state = LEX_INTERPEND;
11802 if (PL_lex_state == LEX_NORMAL) {
11803 if (ckWARN(WARN_AMBIGUOUS) &&
11804 (keyword(dest, d - dest, 0)
11805 || get_cvn_flags(dest, d - dest, 0)))
11809 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11810 "Ambiguous use of %c{%s} resolved to %c%s",
11811 funny, dest, funny, dest);
11816 s = bracket; /* let the parser handle it */
11820 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
11821 PL_lex_state = LEX_INTERPEND;
11826 S_pmflag(U32 pmfl, const char ch) {
11828 CASE_STD_PMMOD_FLAGS_PARSE_SET(&pmfl);
11829 case GLOBAL_PAT_MOD: pmfl |= PMf_GLOBAL; break;
11830 case CONTINUE_PAT_MOD: pmfl |= PMf_CONTINUE; break;
11831 case ONCE_PAT_MOD: pmfl |= PMf_KEEP; break;
11832 case KEEPCOPY_PAT_MOD: pmfl |= PMf_KEEPCOPY; break;
11833 case NONDESTRUCT_PAT_MOD: pmfl |= PMf_NONDESTRUCT; break;
11839 S_scan_pat(pTHX_ char *start, I32 type)
11843 char *s = scan_str(start,!!PL_madskills,FALSE);
11844 const char * const valid_flags =
11845 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
11850 PERL_ARGS_ASSERT_SCAN_PAT;
11853 const char * const delimiter = skipspace(start);
11857 ? "Search pattern not terminated or ternary operator parsed as search pattern"
11858 : "Search pattern not terminated" ));
11861 pm = (PMOP*)newPMOP(type, 0);
11862 if (PL_multi_open == '?') {
11863 /* This is the only point in the code that sets PMf_ONCE: */
11864 pm->op_pmflags |= PMf_ONCE;
11866 /* Hence it's safe to do this bit of PMOP book-keeping here, which
11867 allows us to restrict the list needed by reset to just the ??
11869 assert(type != OP_TRANS);
11871 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11874 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11877 elements = mg->mg_len / sizeof(PMOP**);
11878 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11879 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11880 mg->mg_len = elements * sizeof(PMOP**);
11881 PmopSTASH_set(pm,PL_curstash);
11887 while (*s && strchr(valid_flags, *s))
11888 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11890 if (PL_madskills && modstart != s) {
11891 SV* tmptoken = newSVpvn(modstart, s - modstart);
11892 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11895 /* issue a warning if /c is specified,but /g is not */
11896 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11898 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11899 "Use of /c modifier is meaningless without /g" );
11902 PL_lex_op = (OP*)pm;
11903 pl_yylval.ival = OP_MATCH;
11908 S_scan_subst(pTHX_ char *start)
11919 PERL_ARGS_ASSERT_SCAN_SUBST;
11921 pl_yylval.ival = OP_NULL;
11923 s = scan_str(start,!!PL_madskills,FALSE);
11926 Perl_croak(aTHX_ "Substitution pattern not terminated");
11928 if (s[-1] == PL_multi_open)
11931 if (PL_madskills) {
11932 CURMAD('q', PL_thisopen);
11933 CURMAD('_', PL_thiswhite);
11934 CURMAD('E', PL_thisstuff);
11935 CURMAD('Q', PL_thisclose);
11936 PL_realtokenstart = s - SvPVX(PL_linestr);
11940 first_start = PL_multi_start;
11941 s = scan_str(s,!!PL_madskills,FALSE);
11943 if (PL_lex_stuff) {
11944 SvREFCNT_dec(PL_lex_stuff);
11945 PL_lex_stuff = NULL;
11947 Perl_croak(aTHX_ "Substitution replacement not terminated");
11949 PL_multi_start = first_start; /* so whole substitution is taken together */
11951 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11954 if (PL_madskills) {
11955 CURMAD('z', PL_thisopen);
11956 CURMAD('R', PL_thisstuff);
11957 CURMAD('Z', PL_thisclose);
11963 if (*s == EXEC_PAT_MOD) {
11967 else if (strchr(S_PAT_MODS, *s))
11968 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11974 if (PL_madskills) {
11976 curmad('m', newSVpvn(modstart, s - modstart));
11977 append_madprops(PL_thismad, (OP*)pm, 0);
11981 if ((pm->op_pmflags & PMf_CONTINUE)) {
11982 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
11986 SV * const repl = newSVpvs("");
11988 PL_sublex_info.super_bufptr = s;
11989 PL_sublex_info.super_bufend = PL_bufend;
11991 pm->op_pmflags |= PMf_EVAL;
11994 sv_catpvs(repl, "eval ");
11996 sv_catpvs(repl, "do ");
11998 sv_catpvs(repl, "{");
11999 sv_catsv(repl, PL_lex_repl);
12000 if (strchr(SvPVX(PL_lex_repl), '#'))
12001 sv_catpvs(repl, "\n");
12002 sv_catpvs(repl, "}");
12004 SvREFCNT_dec(PL_lex_repl);
12005 PL_lex_repl = repl;
12008 PL_lex_op = (OP*)pm;
12009 pl_yylval.ival = OP_SUBST;
12014 S_scan_trans(pTHX_ char *start)
12027 PERL_ARGS_ASSERT_SCAN_TRANS;
12029 pl_yylval.ival = OP_NULL;
12031 s = scan_str(start,!!PL_madskills,FALSE);
12033 Perl_croak(aTHX_ "Transliteration pattern not terminated");
12035 if (s[-1] == PL_multi_open)
12038 if (PL_madskills) {
12039 CURMAD('q', PL_thisopen);
12040 CURMAD('_', PL_thiswhite);
12041 CURMAD('E', PL_thisstuff);
12042 CURMAD('Q', PL_thisclose);
12043 PL_realtokenstart = s - SvPVX(PL_linestr);
12047 s = scan_str(s,!!PL_madskills,FALSE);
12049 if (PL_lex_stuff) {
12050 SvREFCNT_dec(PL_lex_stuff);
12051 PL_lex_stuff = NULL;
12053 Perl_croak(aTHX_ "Transliteration replacement not terminated");
12055 if (PL_madskills) {
12056 CURMAD('z', PL_thisopen);
12057 CURMAD('R', PL_thisstuff);
12058 CURMAD('Z', PL_thisclose);
12061 complement = del = squash = 0;
12068 complement = OPpTRANS_COMPLEMENT;
12071 del = OPpTRANS_DELETE;
12074 squash = OPpTRANS_SQUASH;
12083 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
12084 o = newPVOP(OP_TRANS, 0, (char*)tbl);
12085 o->op_private &= ~OPpTRANS_ALL;
12086 o->op_private |= del|squash|complement|
12087 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
12088 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
12091 pl_yylval.ival = OP_TRANS;
12094 if (PL_madskills) {
12096 curmad('m', newSVpvn(modstart, s - modstart));
12097 append_madprops(PL_thismad, o, 0);
12106 S_scan_heredoc(pTHX_ register char *s)
12110 I32 op_type = OP_SCALAR;
12114 const char *found_newline;
12118 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
12120 I32 stuffstart = s - SvPVX(PL_linestr);
12123 PL_realtokenstart = -1;
12126 PERL_ARGS_ASSERT_SCAN_HEREDOC;
12130 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
12134 while (SPACE_OR_TAB(*peek))
12136 if (*peek == '`' || *peek == '\'' || *peek =='"') {
12139 s = delimcpy(d, e, s, PL_bufend, term, &len);
12149 if (!isALNUM_lazy_if(s,UTF))
12150 deprecate("bare << to mean <<\"\"");
12151 for (; isALNUM_lazy_if(s,UTF); s++) {
12156 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
12157 Perl_croak(aTHX_ "Delimiter for here document is too long");
12160 len = d - PL_tokenbuf;
12163 if (PL_madskills) {
12164 tstart = PL_tokenbuf + !outer;
12165 PL_thisclose = newSVpvn(tstart, len - !outer);
12166 tstart = SvPVX(PL_linestr) + stuffstart;
12167 PL_thisopen = newSVpvn(tstart, s - tstart);
12168 stuffstart = s - SvPVX(PL_linestr);
12171 #ifndef PERL_STRICT_CR
12172 d = strchr(s, '\r');
12174 char * const olds = s;
12176 while (s < PL_bufend) {
12182 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
12191 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12198 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
12199 herewas = newSVpvn(s,PL_bufend-s);
12203 herewas = newSVpvn(s-1,found_newline-s+1);
12206 herewas = newSVpvn(s,found_newline-s);
12210 if (PL_madskills) {
12211 tstart = SvPVX(PL_linestr) + stuffstart;
12213 sv_catpvn(PL_thisstuff, tstart, s - tstart);
12215 PL_thisstuff = newSVpvn(tstart, s - tstart);
12218 s += SvCUR(herewas);
12221 stuffstart = s - SvPVX(PL_linestr);
12227 tmpstr = newSV_type(SVt_PVIV);
12228 SvGROW(tmpstr, 80);
12229 if (term == '\'') {
12230 op_type = OP_CONST;
12231 SvIV_set(tmpstr, -1);
12233 else if (term == '`') {
12234 op_type = OP_BACKTICK;
12235 SvIV_set(tmpstr, '\\');
12239 PL_multi_start = CopLINE(PL_curcop);
12240 PL_multi_open = PL_multi_close = '<';
12241 term = *PL_tokenbuf;
12242 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
12243 char * const bufptr = PL_sublex_info.super_bufptr;
12244 char * const bufend = PL_sublex_info.super_bufend;
12245 char * const olds = s - SvCUR(herewas);
12246 s = strchr(bufptr, '\n');
12250 while (s < bufend &&
12251 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12253 CopLINE_inc(PL_curcop);
12256 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12257 missingterm(PL_tokenbuf);
12259 sv_setpvn(herewas,bufptr,d-bufptr+1);
12260 sv_setpvn(tmpstr,d+1,s-d);
12262 sv_catpvn(herewas,s,bufend-s);
12263 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
12270 while (s < PL_bufend &&
12271 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12273 CopLINE_inc(PL_curcop);
12275 if (s >= PL_bufend) {
12276 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12277 missingterm(PL_tokenbuf);
12279 sv_setpvn(tmpstr,d+1,s-d);
12281 if (PL_madskills) {
12283 sv_catpvn(PL_thisstuff, d + 1, s - d);
12285 PL_thisstuff = newSVpvn(d + 1, s - d);
12286 stuffstart = s - SvPVX(PL_linestr);
12290 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
12292 sv_catpvn(herewas,s,PL_bufend-s);
12293 sv_setsv(PL_linestr,herewas);
12294 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
12295 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12296 PL_last_lop = PL_last_uni = NULL;
12299 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
12300 while (s >= PL_bufend) { /* multiple line string? */
12302 if (PL_madskills) {
12303 tstart = SvPVX(PL_linestr) + stuffstart;
12305 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12307 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12311 CopLINE_inc(PL_curcop);
12312 if (!outer || !lex_next_chunk(0)) {
12313 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12314 missingterm(PL_tokenbuf);
12316 CopLINE_dec(PL_curcop);
12319 stuffstart = s - SvPVX(PL_linestr);
12321 CopLINE_inc(PL_curcop);
12322 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12323 PL_last_lop = PL_last_uni = NULL;
12324 #ifndef PERL_STRICT_CR
12325 if (PL_bufend - PL_linestart >= 2) {
12326 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
12327 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
12329 PL_bufend[-2] = '\n';
12331 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12333 else if (PL_bufend[-1] == '\r')
12334 PL_bufend[-1] = '\n';
12336 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
12337 PL_bufend[-1] = '\n';
12339 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
12340 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
12341 *(SvPVX(PL_linestr) + off ) = ' ';
12342 sv_catsv(PL_linestr,herewas);
12343 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12344 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
12348 sv_catsv(tmpstr,PL_linestr);
12353 PL_multi_end = CopLINE(PL_curcop);
12354 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
12355 SvPV_shrink_to_cur(tmpstr);
12357 SvREFCNT_dec(herewas);
12359 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
12361 else if (PL_encoding)
12362 sv_recode_to_utf8(tmpstr, PL_encoding);
12364 PL_lex_stuff = tmpstr;
12365 pl_yylval.ival = op_type;
12369 /* scan_inputsymbol
12370 takes: current position in input buffer
12371 returns: new position in input buffer
12372 side-effects: pl_yylval and lex_op are set.
12377 <FH> read from filehandle
12378 <pkg::FH> read from package qualified filehandle
12379 <pkg'FH> read from package qualified filehandle
12380 <$fh> read from filehandle in $fh
12381 <*.h> filename glob
12386 S_scan_inputsymbol(pTHX_ char *start)
12389 register char *s = start; /* current position in buffer */
12392 char *d = PL_tokenbuf; /* start of temp holding space */
12393 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
12395 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
12397 end = strchr(s, '\n');
12400 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
12402 /* die if we didn't have space for the contents of the <>,
12403 or if it didn't end, or if we see a newline
12406 if (len >= (I32)sizeof PL_tokenbuf)
12407 Perl_croak(aTHX_ "Excessively long <> operator");
12409 Perl_croak(aTHX_ "Unterminated <> operator");
12414 Remember, only scalar variables are interpreted as filehandles by
12415 this code. Anything more complex (e.g., <$fh{$num}>) will be
12416 treated as a glob() call.
12417 This code makes use of the fact that except for the $ at the front,
12418 a scalar variable and a filehandle look the same.
12420 if (*d == '$' && d[1]) d++;
12422 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
12423 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
12426 /* If we've tried to read what we allow filehandles to look like, and
12427 there's still text left, then it must be a glob() and not a getline.
12428 Use scan_str to pull out the stuff between the <> and treat it
12429 as nothing more than a string.
12432 if (d - PL_tokenbuf != len) {
12433 pl_yylval.ival = OP_GLOB;
12434 s = scan_str(start,!!PL_madskills,FALSE);
12436 Perl_croak(aTHX_ "Glob not terminated");
12440 bool readline_overriden = FALSE;
12443 /* we're in a filehandle read situation */
12446 /* turn <> into <ARGV> */
12448 Copy("ARGV",d,5,char);
12450 /* Check whether readline() is overriden */
12451 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
12453 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
12455 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
12456 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
12457 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
12458 readline_overriden = TRUE;
12460 /* if <$fh>, create the ops to turn the variable into a
12464 /* try to find it in the pad for this block, otherwise find
12465 add symbol table ops
12467 const PADOFFSET tmp = pad_findmy(d, len, 0);
12468 if (tmp != NOT_IN_PAD) {
12469 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
12470 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
12471 HEK * const stashname = HvNAME_HEK(stash);
12472 SV * const sym = sv_2mortal(newSVhek(stashname));
12473 sv_catpvs(sym, "::");
12474 sv_catpv(sym, d+1);
12479 OP * const o = newOP(OP_PADSV, 0);
12481 PL_lex_op = readline_overriden
12482 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12483 append_elem(OP_LIST, o,
12484 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
12485 : (OP*)newUNOP(OP_READLINE, 0, o);
12494 ? (GV_ADDMULTI | GV_ADDINEVAL)
12497 PL_lex_op = readline_overriden
12498 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12499 append_elem(OP_LIST,
12500 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
12501 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12502 : (OP*)newUNOP(OP_READLINE, 0,
12503 newUNOP(OP_RV2SV, 0,
12504 newGVOP(OP_GV, 0, gv)));
12506 if (!readline_overriden)
12507 PL_lex_op->op_flags |= OPf_SPECIAL;
12508 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
12509 pl_yylval.ival = OP_NULL;
12512 /* If it's none of the above, it must be a literal filehandle
12513 (<Foo::BAR> or <FOO>) so build a simple readline OP */
12515 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
12516 PL_lex_op = readline_overriden
12517 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12518 append_elem(OP_LIST,
12519 newGVOP(OP_GV, 0, gv),
12520 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12521 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
12522 pl_yylval.ival = OP_NULL;
12531 takes: start position in buffer
12532 keep_quoted preserve \ on the embedded delimiter(s)
12533 keep_delims preserve the delimiters around the string
12534 returns: position to continue reading from buffer
12535 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
12536 updates the read buffer.
12538 This subroutine pulls a string out of the input. It is called for:
12539 q single quotes q(literal text)
12540 ' single quotes 'literal text'
12541 qq double quotes qq(interpolate $here please)
12542 " double quotes "interpolate $here please"
12543 qx backticks qx(/bin/ls -l)
12544 ` backticks `/bin/ls -l`
12545 qw quote words @EXPORT_OK = qw( func() $spam )
12546 m// regexp match m/this/
12547 s/// regexp substitute s/this/that/
12548 tr/// string transliterate tr/this/that/
12549 y/// string transliterate y/this/that/
12550 ($*@) sub prototypes sub foo ($)
12551 (stuff) sub attr parameters sub foo : attr(stuff)
12552 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
12554 In most of these cases (all but <>, patterns and transliterate)
12555 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
12556 calls scan_str(). s/// makes yylex() call scan_subst() which calls
12557 scan_str(). tr/// and y/// make yylex() call scan_trans() which
12560 It skips whitespace before the string starts, and treats the first
12561 character as the delimiter. If the delimiter is one of ([{< then
12562 the corresponding "close" character )]}> is used as the closing
12563 delimiter. It allows quoting of delimiters, and if the string has
12564 balanced delimiters ([{<>}]) it allows nesting.
12566 On success, the SV with the resulting string is put into lex_stuff or,
12567 if that is already non-NULL, into lex_repl. The second case occurs only
12568 when parsing the RHS of the special constructs s/// and tr/// (y///).
12569 For convenience, the terminating delimiter character is stuffed into
12574 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
12577 SV *sv; /* scalar value: string */
12578 const char *tmps; /* temp string, used for delimiter matching */
12579 register char *s = start; /* current position in the buffer */
12580 register char term; /* terminating character */
12581 register char *to; /* current position in the sv's data */
12582 I32 brackets = 1; /* bracket nesting level */
12583 bool has_utf8 = FALSE; /* is there any utf8 content? */
12584 I32 termcode; /* terminating char. code */
12585 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
12586 STRLEN termlen; /* length of terminating string */
12587 int last_off = 0; /* last position for nesting bracket */
12593 PERL_ARGS_ASSERT_SCAN_STR;
12595 /* skip space before the delimiter */
12601 if (PL_realtokenstart >= 0) {
12602 stuffstart = PL_realtokenstart;
12603 PL_realtokenstart = -1;
12606 stuffstart = start - SvPVX(PL_linestr);
12608 /* mark where we are, in case we need to report errors */
12611 /* after skipping whitespace, the next character is the terminator */
12614 termcode = termstr[0] = term;
12618 termcode = utf8_to_uvchr((U8*)s, &termlen);
12619 Copy(s, termstr, termlen, U8);
12620 if (!UTF8_IS_INVARIANT(term))
12624 /* mark where we are */
12625 PL_multi_start = CopLINE(PL_curcop);
12626 PL_multi_open = term;
12628 /* find corresponding closing delimiter */
12629 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
12630 termcode = termstr[0] = term = tmps[5];
12632 PL_multi_close = term;
12634 /* create a new SV to hold the contents. 79 is the SV's initial length.
12635 What a random number. */
12636 sv = newSV_type(SVt_PVIV);
12638 SvIV_set(sv, termcode);
12639 (void)SvPOK_only(sv); /* validate pointer */
12641 /* move past delimiter and try to read a complete string */
12643 sv_catpvn(sv, s, termlen);
12646 tstart = SvPVX(PL_linestr) + stuffstart;
12647 if (!PL_thisopen && !keep_delims) {
12648 PL_thisopen = newSVpvn(tstart, s - tstart);
12649 stuffstart = s - SvPVX(PL_linestr);
12653 if (PL_encoding && !UTF) {
12657 int offset = s - SvPVX_const(PL_linestr);
12658 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
12659 &offset, (char*)termstr, termlen);
12660 const char * const ns = SvPVX_const(PL_linestr) + offset;
12661 char * const svlast = SvEND(sv) - 1;
12663 for (; s < ns; s++) {
12664 if (*s == '\n' && !PL_rsfp)
12665 CopLINE_inc(PL_curcop);
12668 goto read_more_line;
12670 /* handle quoted delimiters */
12671 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
12673 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
12675 if ((svlast-1 - t) % 2) {
12676 if (!keep_quoted) {
12677 *(svlast-1) = term;
12679 SvCUR_set(sv, SvCUR(sv) - 1);
12684 if (PL_multi_open == PL_multi_close) {
12690 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
12691 /* At here, all closes are "was quoted" one,
12692 so we don't check PL_multi_close. */
12694 if (!keep_quoted && *(t+1) == PL_multi_open)
12699 else if (*t == PL_multi_open)
12707 SvCUR_set(sv, w - SvPVX_const(sv));
12709 last_off = w - SvPVX(sv);
12710 if (--brackets <= 0)
12715 if (!keep_delims) {
12716 SvCUR_set(sv, SvCUR(sv) - 1);
12722 /* extend sv if need be */
12723 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
12724 /* set 'to' to the next character in the sv's string */
12725 to = SvPVX(sv)+SvCUR(sv);
12727 /* if open delimiter is the close delimiter read unbridle */
12728 if (PL_multi_open == PL_multi_close) {
12729 for (; s < PL_bufend; s++,to++) {
12730 /* embedded newlines increment the current line number */
12731 if (*s == '\n' && !PL_rsfp)
12732 CopLINE_inc(PL_curcop);
12733 /* handle quoted delimiters */
12734 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
12735 if (!keep_quoted && s[1] == term)
12737 /* any other quotes are simply copied straight through */
12741 /* terminate when run out of buffer (the for() condition), or
12742 have found the terminator */
12743 else if (*s == term) {
12746 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
12749 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12755 /* if the terminator isn't the same as the start character (e.g.,
12756 matched brackets), we have to allow more in the quoting, and
12757 be prepared for nested brackets.
12760 /* read until we run out of string, or we find the terminator */
12761 for (; s < PL_bufend; s++,to++) {
12762 /* embedded newlines increment the line count */
12763 if (*s == '\n' && !PL_rsfp)
12764 CopLINE_inc(PL_curcop);
12765 /* backslashes can escape the open or closing characters */
12766 if (*s == '\\' && s+1 < PL_bufend) {
12767 if (!keep_quoted &&
12768 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
12773 /* allow nested opens and closes */
12774 else if (*s == PL_multi_close && --brackets <= 0)
12776 else if (*s == PL_multi_open)
12778 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12783 /* terminate the copied string and update the sv's end-of-string */
12785 SvCUR_set(sv, to - SvPVX_const(sv));
12788 * this next chunk reads more into the buffer if we're not done yet
12792 break; /* handle case where we are done yet :-) */
12794 #ifndef PERL_STRICT_CR
12795 if (to - SvPVX_const(sv) >= 2) {
12796 if ((to[-2] == '\r' && to[-1] == '\n') ||
12797 (to[-2] == '\n' && to[-1] == '\r'))
12801 SvCUR_set(sv, to - SvPVX_const(sv));
12803 else if (to[-1] == '\r')
12806 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
12811 /* if we're out of file, or a read fails, bail and reset the current
12812 line marker so we can report where the unterminated string began
12815 if (PL_madskills) {
12816 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12818 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12820 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12823 CopLINE_inc(PL_curcop);
12824 PL_bufptr = PL_bufend;
12825 if (!lex_next_chunk(0)) {
12827 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12836 /* at this point, we have successfully read the delimited string */
12838 if (!PL_encoding || UTF) {
12840 if (PL_madskills) {
12841 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12842 const int len = s - tstart;
12844 sv_catpvn(PL_thisstuff, tstart, len);
12846 PL_thisstuff = newSVpvn(tstart, len);
12847 if (!PL_thisclose && !keep_delims)
12848 PL_thisclose = newSVpvn(s,termlen);
12853 sv_catpvn(sv, s, termlen);
12858 if (PL_madskills) {
12859 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12860 const int len = s - tstart - termlen;
12862 sv_catpvn(PL_thisstuff, tstart, len);
12864 PL_thisstuff = newSVpvn(tstart, len);
12865 if (!PL_thisclose && !keep_delims)
12866 PL_thisclose = newSVpvn(s - termlen,termlen);
12870 if (has_utf8 || PL_encoding)
12873 PL_multi_end = CopLINE(PL_curcop);
12875 /* if we allocated too much space, give some back */
12876 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12877 SvLEN_set(sv, SvCUR(sv) + 1);
12878 SvPV_renew(sv, SvLEN(sv));
12881 /* decide whether this is the first or second quoted string we've read
12894 takes: pointer to position in buffer
12895 returns: pointer to new position in buffer
12896 side-effects: builds ops for the constant in pl_yylval.op
12898 Read a number in any of the formats that Perl accepts:
12900 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12901 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12904 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12906 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12909 If it reads a number without a decimal point or an exponent, it will
12910 try converting the number to an integer and see if it can do so
12911 without loss of precision.
12915 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12918 register const char *s = start; /* current position in buffer */
12919 register char *d; /* destination in temp buffer */
12920 register char *e; /* end of temp buffer */
12921 NV nv; /* number read, as a double */
12922 SV *sv = NULL; /* place to put the converted number */
12923 bool floatit; /* boolean: int or float? */
12924 const char *lastub = NULL; /* position of last underbar */
12925 static char const number_too_long[] = "Number too long";
12927 PERL_ARGS_ASSERT_SCAN_NUM;
12929 /* We use the first character to decide what type of number this is */
12933 Perl_croak(aTHX_ "panic: scan_num");
12935 /* if it starts with a 0, it could be an octal number, a decimal in
12936 0.13 disguise, or a hexadecimal number, or a binary number. */
12940 u holds the "number so far"
12941 shift the power of 2 of the base
12942 (hex == 4, octal == 3, binary == 1)
12943 overflowed was the number more than we can hold?
12945 Shift is used when we add a digit. It also serves as an "are
12946 we in octal/hex/binary?" indicator to disallow hex characters
12947 when in octal mode.
12952 bool overflowed = FALSE;
12953 bool just_zero = TRUE; /* just plain 0 or binary number? */
12954 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12955 static const char* const bases[5] =
12956 { "", "binary", "", "octal", "hexadecimal" };
12957 static const char* const Bases[5] =
12958 { "", "Binary", "", "Octal", "Hexadecimal" };
12959 static const char* const maxima[5] =
12961 "0b11111111111111111111111111111111",
12965 const char *base, *Base, *max;
12967 /* check for hex */
12972 } else if (s[1] == 'b') {
12977 /* check for a decimal in disguise */
12978 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
12980 /* so it must be octal */
12987 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12988 "Misplaced _ in number");
12992 base = bases[shift];
12993 Base = Bases[shift];
12994 max = maxima[shift];
12996 /* read the rest of the number */
12998 /* x is used in the overflow test,
12999 b is the digit we're adding on. */
13004 /* if we don't mention it, we're done */
13008 /* _ are ignored -- but warned about if consecutive */
13010 if (lastub && s == lastub + 1)
13011 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13012 "Misplaced _ in number");
13016 /* 8 and 9 are not octal */
13017 case '8': case '9':
13019 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
13023 case '2': case '3': case '4':
13024 case '5': case '6': case '7':
13026 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
13029 case '0': case '1':
13030 b = *s++ & 15; /* ASCII digit -> value of digit */
13034 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
13035 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
13036 /* make sure they said 0x */
13039 b = (*s++ & 7) + 9;
13041 /* Prepare to put the digit we have onto the end
13042 of the number so far. We check for overflows.
13048 x = u << shift; /* make room for the digit */
13050 if ((x >> shift) != u
13051 && !(PL_hints & HINT_NEW_BINARY)) {
13054 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13055 "Integer overflow in %s number",
13058 u = x | b; /* add the digit to the end */
13061 n *= nvshift[shift];
13062 /* If an NV has not enough bits in its
13063 * mantissa to represent an UV this summing of
13064 * small low-order numbers is a waste of time
13065 * (because the NV cannot preserve the
13066 * low-order bits anyway): we could just
13067 * remember when did we overflow and in the
13068 * end just multiply n by the right
13076 /* if we get here, we had success: make a scalar value from
13081 /* final misplaced underbar check */
13082 if (s[-1] == '_') {
13083 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13087 if (n > 4294967295.0)
13088 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13089 "%s number > %s non-portable",
13095 if (u > 0xffffffff)
13096 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13097 "%s number > %s non-portable",
13102 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
13103 sv = new_constant(start, s - start, "integer",
13104 sv, NULL, NULL, 0);
13105 else if (PL_hints & HINT_NEW_BINARY)
13106 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
13111 handle decimal numbers.
13112 we're also sent here when we read a 0 as the first digit
13114 case '1': case '2': case '3': case '4': case '5':
13115 case '6': case '7': case '8': case '9': case '.':
13118 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
13121 /* read next group of digits and _ and copy into d */
13122 while (isDIGIT(*s) || *s == '_') {
13123 /* skip underscores, checking for misplaced ones
13127 if (lastub && s == lastub + 1)
13128 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13129 "Misplaced _ in number");
13133 /* check for end of fixed-length buffer */
13135 Perl_croak(aTHX_ number_too_long);
13136 /* if we're ok, copy the character */
13141 /* final misplaced underbar check */
13142 if (lastub && s == lastub + 1) {
13143 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13146 /* read a decimal portion if there is one. avoid
13147 3..5 being interpreted as the number 3. followed
13150 if (*s == '.' && s[1] != '.') {
13155 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13156 "Misplaced _ in number");
13160 /* copy, ignoring underbars, until we run out of digits.
13162 for (; isDIGIT(*s) || *s == '_'; s++) {
13163 /* fixed length buffer check */
13165 Perl_croak(aTHX_ number_too_long);
13167 if (lastub && s == lastub + 1)
13168 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13169 "Misplaced _ in number");
13175 /* fractional part ending in underbar? */
13176 if (s[-1] == '_') {
13177 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13178 "Misplaced _ in number");
13180 if (*s == '.' && isDIGIT(s[1])) {
13181 /* oops, it's really a v-string, but without the "v" */
13187 /* read exponent part, if present */
13188 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
13192 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
13193 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
13195 /* stray preinitial _ */
13197 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13198 "Misplaced _ in number");
13202 /* allow positive or negative exponent */
13203 if (*s == '+' || *s == '-')
13206 /* stray initial _ */
13208 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13209 "Misplaced _ in number");
13213 /* read digits of exponent */
13214 while (isDIGIT(*s) || *s == '_') {
13217 Perl_croak(aTHX_ number_too_long);
13221 if (((lastub && s == lastub + 1) ||
13222 (!isDIGIT(s[1]) && s[1] != '_')))
13223 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13224 "Misplaced _ in number");
13232 We try to do an integer conversion first if no characters
13233 indicating "float" have been found.
13238 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
13240 if (flags == IS_NUMBER_IN_UV) {
13242 sv = newSViv(uv); /* Prefer IVs over UVs. */
13245 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
13246 if (uv <= (UV) IV_MIN)
13247 sv = newSViv(-(IV)uv);
13254 /* terminate the string */
13256 nv = Atof(PL_tokenbuf);
13261 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
13262 const char *const key = floatit ? "float" : "integer";
13263 const STRLEN keylen = floatit ? 5 : 7;
13264 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
13265 key, keylen, sv, NULL, NULL, 0);
13269 /* if it starts with a v, it could be a v-string */
13272 sv = newSV(5); /* preallocate storage space */
13273 s = scan_vstring(s, PL_bufend, sv);
13277 /* make the op for the constant and return */
13280 lvalp->opval = newSVOP(OP_CONST, 0, sv);
13282 lvalp->opval = NULL;
13288 S_scan_formline(pTHX_ register char *s)
13291 register char *eol;
13293 SV * const stuff = newSVpvs("");
13294 bool needargs = FALSE;
13295 bool eofmt = FALSE;
13297 char *tokenstart = s;
13298 SV* savewhite = NULL;
13300 if (PL_madskills) {
13301 savewhite = PL_thiswhite;
13306 PERL_ARGS_ASSERT_SCAN_FORMLINE;
13308 while (!needargs) {
13311 #ifdef PERL_STRICT_CR
13312 while (SPACE_OR_TAB(*t))
13315 while (SPACE_OR_TAB(*t) || *t == '\r')
13318 if (*t == '\n' || t == PL_bufend) {
13323 if (PL_in_eval && !PL_rsfp) {
13324 eol = (char *) memchr(s,'\n',PL_bufend-s);
13329 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
13331 for (t = s; t < eol; t++) {
13332 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
13334 goto enough; /* ~~ must be first line in formline */
13336 if (*t == '@' || *t == '^')
13340 sv_catpvn(stuff, s, eol-s);
13341 #ifndef PERL_STRICT_CR
13342 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
13343 char *end = SvPVX(stuff) + SvCUR(stuff);
13346 SvCUR_set(stuff, SvCUR(stuff) - 1);
13357 if (PL_madskills) {
13359 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
13361 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
13364 PL_bufptr = PL_bufend;
13365 CopLINE_inc(PL_curcop);
13366 got_some = lex_next_chunk(0);
13367 CopLINE_dec(PL_curcop);
13370 tokenstart = PL_bufptr;
13378 if (SvCUR(stuff)) {
13381 PL_lex_state = LEX_NORMAL;
13382 start_force(PL_curforce);
13383 NEXTVAL_NEXTTOKE.ival = 0;
13387 PL_lex_state = LEX_FORMLINE;
13389 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
13391 else if (PL_encoding)
13392 sv_recode_to_utf8(stuff, PL_encoding);
13394 start_force(PL_curforce);
13395 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
13397 start_force(PL_curforce);
13398 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
13402 SvREFCNT_dec(stuff);
13404 PL_lex_formbrack = 0;
13408 if (PL_madskills) {
13410 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
13412 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
13413 PL_thiswhite = savewhite;
13420 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
13423 const I32 oldsavestack_ix = PL_savestack_ix;
13424 CV* const outsidecv = PL_compcv;
13427 assert(SvTYPE(PL_compcv) == SVt_PVCV);
13429 SAVEI32(PL_subline);
13430 save_item(PL_subname);
13431 SAVESPTR(PL_compcv);
13433 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
13434 CvFLAGS(PL_compcv) |= flags;
13436 PL_subline = CopLINE(PL_curcop);
13437 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
13438 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
13439 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
13441 return oldsavestack_ix;
13445 #pragma segment Perl_yylex
13448 S_yywarn(pTHX_ const char *const s)
13452 PERL_ARGS_ASSERT_YYWARN;
13454 PL_in_eval |= EVAL_WARNONLY;
13456 PL_in_eval &= ~EVAL_WARNONLY;
13461 Perl_yyerror(pTHX_ const char *const s)
13464 const char *where = NULL;
13465 const char *context = NULL;
13468 int yychar = PL_parser->yychar;
13470 PERL_ARGS_ASSERT_YYERROR;
13472 if (!yychar || (yychar == ';' && !PL_rsfp))
13474 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
13475 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
13476 PL_oldbufptr != PL_bufptr) {
13479 The code below is removed for NetWare because it abends/crashes on NetWare
13480 when the script has error such as not having the closing quotes like:
13481 if ($var eq "value)
13482 Checking of white spaces is anyway done in NetWare code.
13485 while (isSPACE(*PL_oldoldbufptr))
13488 context = PL_oldoldbufptr;
13489 contlen = PL_bufptr - PL_oldoldbufptr;
13491 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
13492 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
13495 The code below is removed for NetWare because it abends/crashes on NetWare
13496 when the script has error such as not having the closing quotes like:
13497 if ($var eq "value)
13498 Checking of white spaces is anyway done in NetWare code.
13501 while (isSPACE(*PL_oldbufptr))
13504 context = PL_oldbufptr;
13505 contlen = PL_bufptr - PL_oldbufptr;
13507 else if (yychar > 255)
13508 where = "next token ???";
13509 else if (yychar == -2) { /* YYEMPTY */
13510 if (PL_lex_state == LEX_NORMAL ||
13511 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
13512 where = "at end of line";
13513 else if (PL_lex_inpat)
13514 where = "within pattern";
13516 where = "within string";
13519 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
13521 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
13522 else if (isPRINT_LC(yychar)) {
13523 const char string = yychar;
13524 sv_catpvn(where_sv, &string, 1);
13527 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
13528 where = SvPVX_const(where_sv);
13530 msg = sv_2mortal(newSVpv(s, 0));
13531 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
13532 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
13534 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
13536 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
13537 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
13538 Perl_sv_catpvf(aTHX_ msg,
13539 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
13540 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
13543 if (PL_in_eval & EVAL_WARNONLY) {
13544 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
13548 if (PL_error_count >= 10) {
13549 if (PL_in_eval && SvCUR(ERRSV))
13550 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
13551 SVfARG(ERRSV), OutCopFILE(PL_curcop));
13553 Perl_croak(aTHX_ "%s has too many errors.\n",
13554 OutCopFILE(PL_curcop));
13557 PL_in_my_stash = NULL;
13561 #pragma segment Main
13565 S_swallow_bom(pTHX_ U8 *s)
13568 const STRLEN slen = SvCUR(PL_linestr);
13570 PERL_ARGS_ASSERT_SWALLOW_BOM;
13574 if (s[1] == 0xFE) {
13575 /* UTF-16 little-endian? (or UTF-32LE?) */
13576 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
13577 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
13578 #ifndef PERL_NO_UTF16_FILTER
13579 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
13581 if (PL_bufend > (char*)s) {
13582 s = add_utf16_textfilter(s, TRUE);
13585 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13590 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
13591 #ifndef PERL_NO_UTF16_FILTER
13592 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
13594 if (PL_bufend > (char *)s) {
13595 s = add_utf16_textfilter(s, FALSE);
13598 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13603 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
13604 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13605 s += 3; /* UTF-8 */
13611 if (s[2] == 0xFE && s[3] == 0xFF) {
13612 /* UTF-32 big-endian */
13613 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
13616 else if (s[2] == 0 && s[3] != 0) {
13619 * are a good indicator of UTF-16BE. */
13620 #ifndef PERL_NO_UTF16_FILTER
13621 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
13622 s = add_utf16_textfilter(s, FALSE);
13624 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13630 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
13631 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13632 s += 4; /* UTF-8 */
13638 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
13641 * are a good indicator of UTF-16LE. */
13642 #ifndef PERL_NO_UTF16_FILTER
13643 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
13644 s = add_utf16_textfilter(s, TRUE);
13646 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13654 #ifndef PERL_NO_UTF16_FILTER
13656 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
13659 SV *const filter = FILTER_DATA(idx);
13660 /* We re-use this each time round, throwing the contents away before we
13662 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
13663 SV *const utf8_buffer = filter;
13664 IV status = IoPAGE(filter);
13665 const bool reverse = cBOOL(IoLINES(filter));
13668 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
13670 /* As we're automatically added, at the lowest level, and hence only called
13671 from this file, we can be sure that we're not called in block mode. Hence
13672 don't bother writing code to deal with block mode. */
13674 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
13677 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
13679 DEBUG_P(PerlIO_printf(Perl_debug_log,
13680 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13681 FPTR2DPTR(void *, S_utf16_textfilter),
13682 reverse ? 'l' : 'b', idx, maxlen, status,
13683 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13690 /* First, look in our buffer of existing UTF-8 data: */
13691 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
13695 } else if (status == 0) {
13697 IoPAGE(filter) = 0;
13698 nl = SvEND(utf8_buffer);
13701 STRLEN got = nl - SvPVX(utf8_buffer);
13702 /* Did we have anything to append? */
13704 sv_catpvn(sv, SvPVX(utf8_buffer), got);
13705 /* Everything else in this code works just fine if SVp_POK isn't
13706 set. This, however, needs it, and we need it to work, else
13707 we loop infinitely because the buffer is never consumed. */
13708 sv_chop(utf8_buffer, nl);
13712 /* OK, not a complete line there, so need to read some more UTF-16.
13713 Read an extra octect if the buffer currently has an odd number. */
13717 if (SvCUR(utf16_buffer) >= 2) {
13718 /* Location of the high octet of the last complete code point.
13719 Gosh, UTF-16 is a pain. All the benefits of variable length,
13720 *coupled* with all the benefits of partial reads and
13722 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
13723 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
13725 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
13729 /* We have the first half of a surrogate. Read more. */
13730 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
13733 status = FILTER_READ(idx + 1, utf16_buffer,
13734 160 + (SvCUR(utf16_buffer) & 1));
13735 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
13736 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
13739 IoPAGE(filter) = status;
13744 chars = SvCUR(utf16_buffer) >> 1;
13745 have = SvCUR(utf8_buffer);
13746 SvGROW(utf8_buffer, have + chars * 3 + 1);
13749 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
13750 (U8*)SvPVX_const(utf8_buffer) + have,
13751 chars * 2, &newlen);
13753 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
13754 (U8*)SvPVX_const(utf8_buffer) + have,
13755 chars * 2, &newlen);
13757 SvCUR_set(utf8_buffer, have + newlen);
13760 /* No need to keep this SV "well-formed" with a '\0' after the end, as
13761 it's private to us, and utf16_to_utf8{,reversed} take a
13762 (pointer,length) pair, rather than a NUL-terminated string. */
13763 if(SvCUR(utf16_buffer) & 1) {
13764 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
13765 SvCUR_set(utf16_buffer, 1);
13767 SvCUR_set(utf16_buffer, 0);
13770 DEBUG_P(PerlIO_printf(Perl_debug_log,
13771 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13773 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13774 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
13779 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
13781 SV *filter = filter_add(S_utf16_textfilter, NULL);
13783 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
13785 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
13786 sv_setpvs(filter, "");
13787 IoLINES(filter) = reversed;
13788 IoPAGE(filter) = 1; /* Not EOF */
13790 /* Sadly, we have to return a valid pointer, come what may, so we have to
13791 ignore any error return from this. */
13792 SvCUR_set(PL_linestr, 0);
13793 if (FILTER_READ(0, PL_linestr, 0)) {
13794 SvUTF8_on(PL_linestr);
13796 SvUTF8_on(PL_linestr);
13798 PL_bufend = SvEND(PL_linestr);
13799 return (U8*)SvPVX(PL_linestr);
13804 Returns a pointer to the next character after the parsed
13805 vstring, as well as updating the passed in sv.
13807 Function must be called like
13810 s = scan_vstring(s,e,sv);
13812 where s and e are the start and end of the string.
13813 The sv should already be large enough to store the vstring
13814 passed in, for performance reasons.
13819 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
13822 const char *pos = s;
13823 const char *start = s;
13825 PERL_ARGS_ASSERT_SCAN_VSTRING;
13827 if (*pos == 'v') pos++; /* get past 'v' */
13828 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13830 if ( *pos != '.') {
13831 /* this may not be a v-string if followed by => */
13832 const char *next = pos;
13833 while (next < e && isSPACE(*next))
13835 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
13836 /* return string not v-string */
13837 sv_setpvn(sv,(char *)s,pos-s);
13838 return (char *)pos;
13842 if (!isALPHA(*pos)) {
13843 U8 tmpbuf[UTF8_MAXBYTES+1];
13846 s++; /* get past 'v' */
13851 /* this is atoi() that tolerates underscores */
13854 const char *end = pos;
13856 while (--end >= s) {
13858 const UV orev = rev;
13859 rev += (*end - '0') * mult;
13862 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13863 "Integer overflow in decimal number");
13867 if (rev > 0x7FFFFFFF)
13868 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
13870 /* Append native character for the rev point */
13871 tmpend = uvchr_to_utf8(tmpbuf, rev);
13872 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
13873 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
13875 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
13881 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13885 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
13892 Perl_keyword_plugin_standard(pTHX_
13893 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
13895 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
13896 PERL_UNUSED_CONTEXT;
13897 PERL_UNUSED_ARG(keyword_ptr);
13898 PERL_UNUSED_ARG(keyword_len);
13899 PERL_UNUSED_ARG(op_ptr);
13900 return KEYWORD_PLUGIN_DECLINE;
13905 * c-indentation-style: bsd
13906 * c-basic-offset: 4
13907 * indent-tabs-mode: t
13910 * ex: set ts=8 sts=4 sw=4 noet: