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) {
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 = SKIPSPACE1(s), (*s != ';' && *s !='}' ))) {
2173 yyerror(errstr); /* version required */
2178 if (PL_madskills && !version) {
2179 sv_free(PL_nextwhite); /* let next token collect whitespace */
2181 s = SvPVX(PL_linestr) + startoff;
2184 /* NOTE: The parser sees the package name and the VERSION swapped */
2185 start_force(PL_curforce);
2186 NEXTVAL_NEXTTOKE.opval = version;
2194 * Tokenize a quoted string passed in as an SV. It finds the next
2195 * chunk, up to end of string or a backslash. It may make a new
2196 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2201 S_tokeq(pTHX_ SV *sv)
2205 register char *send;
2210 PERL_ARGS_ASSERT_TOKEQ;
2215 s = SvPV_force(sv, len);
2216 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2219 while (s < send && *s != '\\')
2224 if ( PL_hints & HINT_NEW_STRING ) {
2225 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2229 if (s + 1 < send && (s[1] == '\\'))
2230 s++; /* all that, just for this */
2235 SvCUR_set(sv, d - SvPVX_const(sv));
2237 if ( PL_hints & HINT_NEW_STRING )
2238 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2243 * Now come three functions related to double-quote context,
2244 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2245 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2246 * interact with PL_lex_state, and create fake ( ... ) argument lists
2247 * to handle functions and concatenation.
2248 * They assume that whoever calls them will be setting up a fake
2249 * join call, because each subthing puts a ',' after it. This lets
2252 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2254 * (I'm not sure whether the spurious commas at the end of lcfirst's
2255 * arguments and join's arguments are created or not).
2260 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2262 * Pattern matching will set PL_lex_op to the pattern-matching op to
2263 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2265 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2267 * Everything else becomes a FUNC.
2269 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2270 * had an OP_CONST or OP_READLINE). This just sets us up for a
2271 * call to S_sublex_push().
2275 S_sublex_start(pTHX)
2278 register const I32 op_type = pl_yylval.ival;
2280 if (op_type == OP_NULL) {
2281 pl_yylval.opval = PL_lex_op;
2285 if (op_type == OP_CONST || op_type == OP_READLINE) {
2286 SV *sv = tokeq(PL_lex_stuff);
2288 if (SvTYPE(sv) == SVt_PVIV) {
2289 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2291 const char * const p = SvPV_const(sv, len);
2292 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2296 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2297 PL_lex_stuff = NULL;
2298 /* Allow <FH> // "foo" */
2299 if (op_type == OP_READLINE)
2300 PL_expect = XTERMORDORDOR;
2303 else if (op_type == OP_BACKTICK && PL_lex_op) {
2304 /* readpipe() vas overriden */
2305 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2306 pl_yylval.opval = PL_lex_op;
2308 PL_lex_stuff = NULL;
2312 PL_sublex_info.super_state = PL_lex_state;
2313 PL_sublex_info.sub_inwhat = (U16)op_type;
2314 PL_sublex_info.sub_op = PL_lex_op;
2315 PL_lex_state = LEX_INTERPPUSH;
2319 pl_yylval.opval = PL_lex_op;
2329 * Create a new scope to save the lexing state. The scope will be
2330 * ended in S_sublex_done. Returns a '(', starting the function arguments
2331 * to the uc, lc, etc. found before.
2332 * Sets PL_lex_state to LEX_INTERPCONCAT.
2341 PL_lex_state = PL_sublex_info.super_state;
2342 SAVEBOOL(PL_lex_dojoin);
2343 SAVEI32(PL_lex_brackets);
2344 SAVEI32(PL_lex_casemods);
2345 SAVEI32(PL_lex_starts);
2346 SAVEI8(PL_lex_state);
2347 SAVEVPTR(PL_lex_inpat);
2348 SAVEI16(PL_lex_inwhat);
2349 SAVECOPLINE(PL_curcop);
2350 SAVEPPTR(PL_bufptr);
2351 SAVEPPTR(PL_bufend);
2352 SAVEPPTR(PL_oldbufptr);
2353 SAVEPPTR(PL_oldoldbufptr);
2354 SAVEPPTR(PL_last_lop);
2355 SAVEPPTR(PL_last_uni);
2356 SAVEPPTR(PL_linestart);
2357 SAVESPTR(PL_linestr);
2358 SAVEGENERICPV(PL_lex_brackstack);
2359 SAVEGENERICPV(PL_lex_casestack);
2361 PL_linestr = PL_lex_stuff;
2362 PL_lex_stuff = NULL;
2364 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2365 = SvPVX(PL_linestr);
2366 PL_bufend += SvCUR(PL_linestr);
2367 PL_last_lop = PL_last_uni = NULL;
2368 SAVEFREESV(PL_linestr);
2370 PL_lex_dojoin = FALSE;
2371 PL_lex_brackets = 0;
2372 Newx(PL_lex_brackstack, 120, char);
2373 Newx(PL_lex_casestack, 12, char);
2374 PL_lex_casemods = 0;
2375 *PL_lex_casestack = '\0';
2377 PL_lex_state = LEX_INTERPCONCAT;
2378 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2380 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2381 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2382 PL_lex_inpat = PL_sublex_info.sub_op;
2384 PL_lex_inpat = NULL;
2391 * Restores lexer state after a S_sublex_push.
2398 if (!PL_lex_starts++) {
2399 SV * const sv = newSVpvs("");
2400 if (SvUTF8(PL_linestr))
2402 PL_expect = XOPERATOR;
2403 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2407 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2408 PL_lex_state = LEX_INTERPCASEMOD;
2412 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2413 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2414 PL_linestr = PL_lex_repl;
2416 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2417 PL_bufend += SvCUR(PL_linestr);
2418 PL_last_lop = PL_last_uni = NULL;
2419 SAVEFREESV(PL_linestr);
2420 PL_lex_dojoin = FALSE;
2421 PL_lex_brackets = 0;
2422 PL_lex_casemods = 0;
2423 *PL_lex_casestack = '\0';
2425 if (SvEVALED(PL_lex_repl)) {
2426 PL_lex_state = LEX_INTERPNORMAL;
2428 /* we don't clear PL_lex_repl here, so that we can check later
2429 whether this is an evalled subst; that means we rely on the
2430 logic to ensure sublex_done() is called again only via the
2431 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2434 PL_lex_state = LEX_INTERPCONCAT;
2444 PL_endwhite = newSVpvs("");
2445 sv_catsv(PL_endwhite, PL_thiswhite);
2449 sv_setpvs(PL_thistoken,"");
2451 PL_realtokenstart = -1;
2455 PL_bufend = SvPVX(PL_linestr);
2456 PL_bufend += SvCUR(PL_linestr);
2457 PL_expect = XOPERATOR;
2458 PL_sublex_info.sub_inwhat = 0;
2466 Extracts a pattern, double-quoted string, or transliteration. This
2469 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2470 processing a pattern (PL_lex_inpat is true), a transliteration
2471 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2473 Returns a pointer to the character scanned up to. If this is
2474 advanced from the start pointer supplied (i.e. if anything was
2475 successfully parsed), will leave an OP for the substring scanned
2476 in pl_yylval. Caller must intuit reason for not parsing further
2477 by looking at the next characters herself.
2481 constants: \N{NAME} only
2482 case and quoting: \U \Q \E
2483 stops on @ and $, but not for $ as tail anchor
2485 In transliterations:
2486 characters are VERY literal, except for - not at the start or end
2487 of the string, which indicates a range. If the range is in bytes,
2488 scan_const expands the range to the full set of intermediate
2489 characters. If the range is in utf8, the hyphen is replaced with
2490 a certain range mark which will be handled by pmtrans() in op.c.
2492 In double-quoted strings:
2494 double-quoted style: \r and \n
2495 constants: \x31, etc.
2496 deprecated backrefs: \1 (in substitution replacements)
2497 case and quoting: \U \Q \E
2500 scan_const does *not* construct ops to handle interpolated strings.
2501 It stops processing as soon as it finds an embedded $ or @ variable
2502 and leaves it to the caller to work out what's going on.
2504 embedded arrays (whether in pattern or not) could be:
2505 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2507 $ in double-quoted strings must be the symbol of an embedded scalar.
2509 $ in pattern could be $foo or could be tail anchor. Assumption:
2510 it's a tail anchor if $ is the last thing in the string, or if it's
2511 followed by one of "()| \r\n\t"
2513 \1 (backreferences) are turned into $1
2515 The structure of the code is
2516 while (there's a character to process) {
2517 handle transliteration ranges
2518 skip regexp comments /(?#comment)/ and codes /(?{code})/
2519 skip #-initiated comments in //x patterns
2520 check for embedded arrays
2521 check for embedded scalars
2523 deprecate \1 in substitution replacements
2524 handle string-changing backslashes \l \U \Q \E, etc.
2525 switch (what was escaped) {
2526 handle \- in a transliteration (becomes a literal -)
2527 if a pattern and not \N{, go treat as regular character
2528 handle \132 (octal characters)
2529 handle \x15 and \x{1234} (hex characters)
2530 handle \N{name} (named characters, also \N{3,5} in a pattern)
2531 handle \cV (control characters)
2532 handle printf-style backslashes (\f, \r, \n, etc)
2535 } (end if backslash)
2536 handle regular character
2537 } (end while character to read)
2542 S_scan_const(pTHX_ char *start)
2545 register char *send = PL_bufend; /* end of the constant */
2546 SV *sv = newSV(send - start); /* sv for the constant. See
2547 note below on sizing. */
2548 register char *s = start; /* start of the constant */
2549 register char *d = SvPVX(sv); /* destination for copies */
2550 bool dorange = FALSE; /* are we in a translit range? */
2551 bool didrange = FALSE; /* did we just finish a range? */
2552 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
2553 I32 this_utf8 = UTF; /* Is the source string assumed
2554 to be UTF8? But, this can
2555 show as true when the source
2556 isn't utf8, as for example
2557 when it is entirely composed
2560 /* Note on sizing: The scanned constant is placed into sv, which is
2561 * initialized by newSV() assuming one byte of output for every byte of
2562 * input. This routine expects newSV() to allocate an extra byte for a
2563 * trailing NUL, which this routine will append if it gets to the end of
2564 * the input. There may be more bytes of input than output (eg., \N{LATIN
2565 * CAPITAL LETTER A}), or more output than input if the constant ends up
2566 * recoded to utf8, but each time a construct is found that might increase
2567 * the needed size, SvGROW() is called. Its size parameter each time is
2568 * based on the best guess estimate at the time, namely the length used so
2569 * far, plus the length the current construct will occupy, plus room for
2570 * the trailing NUL, plus one byte for every input byte still unscanned */
2574 UV literal_endpoint = 0;
2575 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2578 PERL_ARGS_ASSERT_SCAN_CONST;
2580 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2581 /* If we are doing a trans and we know we want UTF8 set expectation */
2582 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2583 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2587 while (s < send || dorange) {
2589 /* get transliterations out of the way (they're most literal) */
2590 if (PL_lex_inwhat == OP_TRANS) {
2591 /* expand a range A-Z to the full set of characters. AIE! */
2593 I32 i; /* current expanded character */
2594 I32 min; /* first character in range */
2595 I32 max; /* last character in range */
2606 char * const c = (char*)utf8_hop((U8*)d, -1);
2610 *c = (char)UTF_TO_NATIVE(0xff);
2611 /* mark the range as done, and continue */
2617 i = d - SvPVX_const(sv); /* remember current offset */
2620 SvLEN(sv) + (has_utf8 ?
2621 (512 - UTF_CONTINUATION_MARK +
2624 /* How many two-byte within 0..255: 128 in UTF-8,
2625 * 96 in UTF-8-mod. */
2627 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2629 d = SvPVX(sv) + i; /* refresh d after realloc */
2633 for (j = 0; j <= 1; j++) {
2634 char * const c = (char*)utf8_hop((U8*)d, -1);
2635 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2641 max = (U8)0xff; /* only to \xff */
2642 uvmax = uv; /* \x{100} to uvmax */
2644 d = c; /* eat endpoint chars */
2649 d -= 2; /* eat the first char and the - */
2650 min = (U8)*d; /* first char in range */
2651 max = (U8)d[1]; /* last char in range */
2658 "Invalid range \"%c-%c\" in transliteration operator",
2659 (char)min, (char)max);
2663 if (literal_endpoint == 2 &&
2664 ((isLOWER(min) && isLOWER(max)) ||
2665 (isUPPER(min) && isUPPER(max)))) {
2667 for (i = min; i <= max; i++)
2669 *d++ = NATIVE_TO_NEED(has_utf8,i);
2671 for (i = min; i <= max; i++)
2673 *d++ = NATIVE_TO_NEED(has_utf8,i);
2678 for (i = min; i <= max; i++)
2681 const U8 ch = (U8)NATIVE_TO_UTF(i);
2682 if (UNI_IS_INVARIANT(ch))
2685 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2686 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2695 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2697 *d++ = (char)UTF_TO_NATIVE(0xff);
2699 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2703 /* mark the range as done, and continue */
2707 literal_endpoint = 0;
2712 /* range begins (ignore - as first or last char) */
2713 else if (*s == '-' && s+1 < send && s != start) {
2715 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2722 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2732 literal_endpoint = 0;
2733 native_range = TRUE;
2738 /* if we get here, we're not doing a transliteration */
2740 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2741 except for the last char, which will be done separately. */
2742 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2744 while (s+1 < send && *s != ')')
2745 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2747 else if (s[2] == '{' /* This should match regcomp.c */
2748 || (s[2] == '?' && s[3] == '{'))
2751 char *regparse = s + (s[2] == '{' ? 3 : 4);
2754 while (count && (c = *regparse)) {
2755 if (c == '\\' && regparse[1])
2763 if (*regparse != ')')
2764 regparse--; /* Leave one char for continuation. */
2765 while (s < regparse)
2766 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2770 /* likewise skip #-initiated comments in //x patterns */
2771 else if (*s == '#' && PL_lex_inpat &&
2772 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2773 while (s+1 < send && *s != '\n')
2774 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2777 /* check for embedded arrays
2778 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2780 else if (*s == '@' && s[1]) {
2781 if (isALNUM_lazy_if(s+1,UTF))
2783 if (strchr(":'{$", s[1]))
2785 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2786 break; /* in regexp, neither @+ nor @- are interpolated */
2789 /* check for embedded scalars. only stop if we're sure it's a
2792 else if (*s == '$') {
2793 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2795 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2797 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2798 "Possible unintended interpolation of $\\ in regex");
2800 break; /* in regexp, $ might be tail anchor */
2804 /* End of else if chain - OP_TRANS rejoin rest */
2807 if (*s == '\\' && s+1 < send) {
2808 char* e; /* Can be used for ending '}', etc. */
2812 /* deprecate \1 in strings and substitution replacements */
2813 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2814 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2816 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2821 /* string-change backslash escapes */
2822 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2826 /* In a pattern, process \N, but skip any other backslash escapes.
2827 * This is because we don't want to translate an escape sequence
2828 * into a meta symbol and have the regex compiler use the meta
2829 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2830 * in spite of this, we do have to process \N here while the proper
2831 * charnames handler is in scope. See bugs #56444 and #62056.
2832 * There is a complication because \N in a pattern may also stand
2833 * for 'match a non-nl', and not mean a charname, in which case its
2834 * processing should be deferred to the regex compiler. To be a
2835 * charname it must be followed immediately by a '{', and not look
2836 * like \N followed by a curly quantifier, i.e., not something like
2837 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2839 else if (PL_lex_inpat
2842 || regcurly(s + 1)))
2844 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2845 goto default_action;
2850 /* quoted - in transliterations */
2852 if (PL_lex_inwhat == OP_TRANS) {
2859 if ((isALPHA(*s) || isDIGIT(*s)))
2860 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2861 "Unrecognized escape \\%c passed through",
2863 /* default action is to copy the quoted character */
2864 goto default_action;
2867 /* eg. \132 indicates the octal constant 0x132 */
2868 case '0': case '1': case '2': case '3':
2869 case '4': case '5': case '6': case '7':
2873 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2876 goto NUM_ESCAPE_INSERT;
2878 /* eg. \x24 indicates the hex constant 0x24 */
2882 char* const e = strchr(s, '}');
2883 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2884 PERL_SCAN_DISALLOW_PREFIX;
2889 yyerror("Missing right brace on \\x{}");
2893 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2899 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2900 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2906 /* Insert oct or hex escaped character. There will always be
2907 * enough room in sv since such escapes will be longer than any
2908 * UTF-8 sequence they can end up as, except if they force us
2909 * to recode the rest of the string into utf8 */
2911 /* Here uv is the ordinal of the next character being added in
2912 * unicode (converted from native). */
2913 if (!UNI_IS_INVARIANT(uv)) {
2914 if (!has_utf8 && uv > 255) {
2915 /* Might need to recode whatever we have accumulated so
2916 * far if it contains any chars variant in utf8 or
2919 SvCUR_set(sv, d - SvPVX_const(sv));
2922 /* See Note on sizing above. */
2923 sv_utf8_upgrade_flags_grow(sv,
2924 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2925 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2926 d = SvPVX(sv) + SvCUR(sv);
2931 d = (char*)uvuni_to_utf8((U8*)d, uv);
2932 if (PL_lex_inwhat == OP_TRANS &&
2933 PL_sublex_info.sub_op) {
2934 PL_sublex_info.sub_op->op_private |=
2935 (PL_lex_repl ? OPpTRANS_FROM_UTF
2939 if (uv > 255 && !dorange)
2940 native_range = FALSE;
2953 /* In a non-pattern \N must be a named character, like \N{LATIN
2954 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
2955 * mean to match a non-newline. For non-patterns, named
2956 * characters are converted to their string equivalents. In
2957 * patterns, named characters are not converted to their
2958 * ultimate forms for the same reasons that other escapes
2959 * aren't. Instead, they are converted to the \N{U+...} form
2960 * to get the value from the charnames that is in effect right
2961 * now, while preserving the fact that it was a named character
2962 * so that the regex compiler knows this */
2964 /* This section of code doesn't generally use the
2965 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
2966 * a close examination of this macro and determined it is a
2967 * no-op except on utfebcdic variant characters. Every
2968 * character generated by this that would normally need to be
2969 * enclosed by this macro is invariant, so the macro is not
2970 * needed, and would complicate use of copy(). There are other
2971 * parts of this file where the macro is used inconsistently,
2972 * but are saved by it being a no-op */
2974 /* The structure of this section of code (besides checking for
2975 * errors and upgrading to utf8) is:
2976 * Further disambiguate between the two meanings of \N, and if
2977 * not a charname, go process it elsewhere
2978 * If of form \N{U+...}, pass it through if a pattern;
2979 * otherwise convert to utf8
2980 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
2981 * pattern; otherwise convert to utf8 */
2983 /* Here, s points to the 'N'; the test below is guaranteed to
2984 * succeed if we are being called on a pattern as we already
2985 * know from a test above that the next character is a '{'.
2986 * On a non-pattern \N must mean 'named sequence, which
2987 * requires braces */
2990 yyerror("Missing braces on \\N{}");
2995 /* If there is no matching '}', it is an error. */
2996 if (! (e = strchr(s, '}'))) {
2997 if (! PL_lex_inpat) {
2998 yyerror("Missing right brace on \\N{}");
3000 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3005 /* Here it looks like a named character */
3009 /* XXX This block is temporary code. \N{} implies that the
3010 * pattern is to have Unicode semantics, and therefore
3011 * currently has to be encoded in utf8. By putting it in
3012 * utf8 now, we save a whole pass in the regular expression
3013 * compiler. Once that code is changed so Unicode
3014 * semantics doesn't necessarily have to be in utf8, this
3015 * block should be removed */
3017 SvCUR_set(sv, d - SvPVX_const(sv));
3020 /* See Note on sizing above. */
3021 sv_utf8_upgrade_flags_grow(sv,
3022 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3023 /* 5 = '\N{' + cur char + NUL */
3024 (STRLEN)(send - s) + 5);
3025 d = SvPVX(sv) + SvCUR(sv);
3030 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3031 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3032 | PERL_SCAN_DISALLOW_PREFIX;
3035 /* For \N{U+...}, the '...' is a unicode value even on
3036 * EBCDIC machines */
3037 s += 2; /* Skip to next char after the 'U+' */
3039 uv = grok_hex(s, &len, &flags, NULL);
3040 if (len == 0 || len != (STRLEN)(e - s)) {
3041 yyerror("Invalid hexadecimal number in \\N{U+...}");
3048 /* Pass through to the regex compiler unchanged. The
3049 * reason we evaluated the number above is to make sure
3050 * there wasn't a syntax error. */
3051 s -= 5; /* Include the '\N{U+' */
3052 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3055 else { /* Not a pattern: convert the hex to string */
3057 /* If destination is not in utf8, unconditionally
3058 * recode it to be so. This is because \N{} implies
3059 * Unicode semantics, and scalars have to be in utf8
3060 * to guarantee those semantics */
3062 SvCUR_set(sv, d - SvPVX_const(sv));
3065 /* See Note on sizing above. */
3066 sv_utf8_upgrade_flags_grow(
3068 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3069 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3070 d = SvPVX(sv) + SvCUR(sv);
3074 /* Add the string to the output */
3075 if (UNI_IS_INVARIANT(uv)) {
3078 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3081 else { /* Here is \N{NAME} but not \N{U+...}. */
3083 SV *res; /* result from charnames */
3084 const char *str; /* the string in 'res' */
3085 STRLEN len; /* its length */
3087 /* Get the value for NAME */
3088 res = newSVpvn(s, e - s);
3089 res = new_constant( NULL, 0, "charnames",
3090 /* includes all of: \N{...} */
3091 res, NULL, s - 3, e - s + 4 );
3093 /* Most likely res will be in utf8 already since the
3094 * standard charnames uses pack U, but a custom translator
3095 * can leave it otherwise, so make sure. XXX This can be
3096 * revisited to not have charnames use utf8 for characters
3097 * that don't need it when regexes don't have to be in utf8
3098 * for Unicode semantics. If doing so, remember EBCDIC */
3099 sv_utf8_upgrade(res);
3100 str = SvPV_const(res, len);
3102 /* Don't accept malformed input */
3103 if (! is_utf8_string((U8 *) str, len)) {
3104 yyerror("Malformed UTF-8 returned by \\N");
3106 else if (PL_lex_inpat) {
3108 if (! len) { /* The name resolved to an empty string */
3109 Copy("\\N{}", d, 4, char);
3113 /* In order to not lose information for the regex
3114 * compiler, pass the result in the specially made
3115 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3116 * the code points in hex of each character
3117 * returned by charnames */
3119 const char *str_end = str + len;
3120 STRLEN char_length; /* cur char's byte length */
3121 STRLEN output_length; /* and the number of bytes
3122 after this is translated
3124 const STRLEN off = d - SvPVX_const(sv);
3126 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3127 * max('U+', '.'); and 1 for NUL */
3128 char hex_string[2 * UTF8_MAXBYTES + 5];
3130 /* Get the first character of the result. */
3131 U32 uv = utf8n_to_uvuni((U8 *) str,
3136 /* The call to is_utf8_string() above hopefully
3137 * guarantees that there won't be an error. But
3138 * it's easy here to make sure. The function just
3139 * above warns and returns 0 if invalid utf8, but
3140 * it can also return 0 if the input is validly a
3141 * NUL. Disambiguate */
3142 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3143 uv = UNICODE_REPLACEMENT;
3146 /* Convert first code point to hex, including the
3147 * boiler plate before it */
3148 sprintf(hex_string, "\\N{U+%X", (unsigned int) uv);
3149 output_length = strlen(hex_string);
3151 /* Make sure there is enough space to hold it */
3152 d = off + SvGROW(sv, off
3154 + (STRLEN)(send - e)
3155 + 2); /* '}' + NUL */
3157 Copy(hex_string, d, output_length, char);
3160 /* For each subsequent character, append dot and
3161 * its ordinal in hex */
3162 while ((str += char_length) < str_end) {
3163 const STRLEN off = d - SvPVX_const(sv);
3164 U32 uv = utf8n_to_uvuni((U8 *) str,
3168 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3169 uv = UNICODE_REPLACEMENT;
3172 sprintf(hex_string, ".%X", (unsigned int) uv);
3173 output_length = strlen(hex_string);
3175 d = off + SvGROW(sv, off
3177 + (STRLEN)(send - e)
3178 + 2); /* '}' + NUL */
3179 Copy(hex_string, d, output_length, char);
3183 *d++ = '}'; /* Done. Add the trailing brace */
3186 else { /* Here, not in a pattern. Convert the name to a
3189 /* If destination is not in utf8, unconditionally
3190 * recode it to be so. This is because \N{} implies
3191 * Unicode semantics, and scalars have to be in utf8
3192 * to guarantee those semantics */
3194 SvCUR_set(sv, d - SvPVX_const(sv));
3197 /* See Note on sizing above. */
3198 sv_utf8_upgrade_flags_grow(sv,
3199 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3200 len + (STRLEN)(send - s) + 1);
3201 d = SvPVX(sv) + SvCUR(sv);
3203 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3205 /* See Note on sizing above. (NOTE: SvCUR() is not
3206 * set correctly here). */
3207 const STRLEN off = d - SvPVX_const(sv);
3208 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3210 Copy(str, d, len, char);
3215 /* Deprecate non-approved name syntax */
3216 if (ckWARN_d(WARN_DEPRECATED)) {
3217 bool problematic = FALSE;
3220 /* For non-ut8 input, look to see that the first
3221 * character is an alpha, then loop through the rest
3222 * checking that each is a continuation */
3224 if (! isALPHAU(*i)) problematic = TRUE;
3225 else for (i = s + 1; i < e; i++) {
3226 if (isCHARNAME_CONT(*i)) continue;
3232 /* Similarly for utf8. For invariants can check
3233 * directly. We accept anything above the latin1
3234 * range because it is immaterial to Perl if it is
3235 * correct or not, and is expensive to check. But
3236 * it is fairly easy in the latin1 range to convert
3237 * the variants into a single character and check
3239 if (UTF8_IS_INVARIANT(*i)) {
3240 if (! isALPHAU(*i)) problematic = TRUE;
3241 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3242 if (! isALPHAU(UNI_TO_NATIVE(UTF8_ACCUMULATE(*i,
3248 if (! problematic) for (i = s + UTF8SKIP(s);
3252 if (UTF8_IS_INVARIANT(*i)) {
3253 if (isCHARNAME_CONT(*i)) continue;
3254 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3256 } else if (isCHARNAME_CONT(
3258 UTF8_ACCUMULATE(*i, *(i+1)))))
3267 /* The e-i passed to the final %.*s makes sure that
3268 * should the trailing NUL be missing that this
3269 * print won't run off the end of the string */
3270 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3271 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s", i - s + 1, s, e - i, i + 1);
3274 } /* End \N{NAME} */
3277 native_range = FALSE; /* \N{} is defined to be Unicode */
3279 s = e + 1; /* Point to just after the '}' */
3282 /* \c is a control character */
3291 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
3294 yyerror("Missing control char name in \\c");
3298 /* printf-style backslashes, formfeeds, newlines, etc */
3300 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3303 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3306 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3309 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3312 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3315 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3318 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3324 } /* end if (backslash) */
3331 /* If we started with encoded form, or already know we want it,
3332 then encode the next character */
3333 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3337 /* One might think that it is wasted effort in the case of the
3338 * source being utf8 (this_utf8 == TRUE) to take the next character
3339 * in the source, convert it to an unsigned value, and then convert
3340 * it back again. But the source has not been validated here. The
3341 * routine that does the conversion checks for errors like
3344 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3345 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3347 SvCUR_set(sv, d - SvPVX_const(sv));
3350 /* See Note on sizing above. */
3351 sv_utf8_upgrade_flags_grow(sv,
3352 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3353 need + (STRLEN)(send - s) + 1);
3354 d = SvPVX(sv) + SvCUR(sv);
3356 } else if (need > len) {
3357 /* encoded value larger than old, may need extra space (NOTE:
3358 * SvCUR() is not set correctly here). See Note on sizing
3360 const STRLEN off = d - SvPVX_const(sv);
3361 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3365 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3367 if (uv > 255 && !dorange)
3368 native_range = FALSE;
3372 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3374 } /* while loop to process each character */
3376 /* terminate the string and set up the sv */
3378 SvCUR_set(sv, d - SvPVX_const(sv));
3379 if (SvCUR(sv) >= SvLEN(sv))
3380 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3383 if (PL_encoding && !has_utf8) {
3384 sv_recode_to_utf8(sv, PL_encoding);
3390 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3391 PL_sublex_info.sub_op->op_private |=
3392 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3396 /* shrink the sv if we allocated more than we used */
3397 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3398 SvPV_shrink_to_cur(sv);
3401 /* return the substring (via pl_yylval) only if we parsed anything */
3402 if (s > PL_bufptr) {
3403 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3404 const char *const key = PL_lex_inpat ? "qr" : "q";
3405 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3409 if (PL_lex_inwhat == OP_TRANS) {
3412 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3420 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3423 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3430 * Returns TRUE if there's more to the expression (e.g., a subscript),
3433 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3435 * ->[ and ->{ return TRUE
3436 * { and [ outside a pattern are always subscripts, so return TRUE
3437 * if we're outside a pattern and it's not { or [, then return FALSE
3438 * if we're in a pattern and the first char is a {
3439 * {4,5} (any digits around the comma) returns FALSE
3440 * if we're in a pattern and the first char is a [
3442 * [SOMETHING] has a funky algorithm to decide whether it's a
3443 * character class or not. It has to deal with things like
3444 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3445 * anything else returns TRUE
3448 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3451 S_intuit_more(pTHX_ register char *s)
3455 PERL_ARGS_ASSERT_INTUIT_MORE;
3457 if (PL_lex_brackets)
3459 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3461 if (*s != '{' && *s != '[')
3466 /* In a pattern, so maybe we have {n,m}. */
3483 /* On the other hand, maybe we have a character class */
3486 if (*s == ']' || *s == '^')
3489 /* this is terrifying, and it works */
3490 int weight = 2; /* let's weigh the evidence */
3492 unsigned char un_char = 255, last_un_char;
3493 const char * const send = strchr(s,']');
3494 char tmpbuf[sizeof PL_tokenbuf * 4];
3496 if (!send) /* has to be an expression */
3499 Zero(seen,256,char);
3502 else if (isDIGIT(*s)) {
3504 if (isDIGIT(s[1]) && s[2] == ']')
3510 for (; s < send; s++) {
3511 last_un_char = un_char;
3512 un_char = (unsigned char)*s;
3517 weight -= seen[un_char] * 10;
3518 if (isALNUM_lazy_if(s+1,UTF)) {
3520 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3521 len = (int)strlen(tmpbuf);
3522 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
3527 else if (*s == '$' && s[1] &&
3528 strchr("[#!%*<>()-=",s[1])) {
3529 if (/*{*/ strchr("])} =",s[2]))
3538 if (strchr("wds]",s[1]))
3540 else if (seen[(U8)'\''] || seen[(U8)'"'])
3542 else if (strchr("rnftbxcav",s[1]))
3544 else if (isDIGIT(s[1])) {
3546 while (s[1] && isDIGIT(s[1]))
3556 if (strchr("aA01! ",last_un_char))
3558 if (strchr("zZ79~",s[1]))
3560 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3561 weight -= 5; /* cope with negative subscript */
3564 if (!isALNUM(last_un_char)
3565 && !(last_un_char == '$' || last_un_char == '@'
3566 || last_un_char == '&')
3567 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3572 if (keyword(tmpbuf, d - tmpbuf, 0))
3575 if (un_char == last_un_char + 1)
3577 weight -= seen[un_char];
3582 if (weight >= 0) /* probably a character class */
3592 * Does all the checking to disambiguate
3594 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3595 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3597 * First argument is the stuff after the first token, e.g. "bar".
3599 * Not a method if bar is a filehandle.
3600 * Not a method if foo is a subroutine prototyped to take a filehandle.
3601 * Not a method if it's really "Foo $bar"
3602 * Method if it's "foo $bar"
3603 * Not a method if it's really "print foo $bar"
3604 * Method if it's really "foo package::" (interpreted as package->foo)
3605 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3606 * Not a method if bar is a filehandle or package, but is quoted with
3611 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3614 char *s = start + (*start == '$');
3615 char tmpbuf[sizeof PL_tokenbuf];
3622 PERL_ARGS_ASSERT_INTUIT_METHOD;
3625 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3629 const char *proto = SvPVX_const(cv);
3640 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3641 /* start is the beginning of the possible filehandle/object,
3642 * and s is the end of it
3643 * tmpbuf is a copy of it
3646 if (*start == '$') {
3647 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3648 isUPPER(*PL_tokenbuf))
3651 len = start - SvPVX(PL_linestr);
3655 start = SvPVX(PL_linestr) + len;
3659 return *s == '(' ? FUNCMETH : METHOD;
3661 if (!keyword(tmpbuf, len, 0)) {
3662 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3666 soff = s - SvPVX(PL_linestr);
3670 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
3671 if (indirgv && GvCVu(indirgv))
3673 /* filehandle or package name makes it a method */
3674 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
3676 soff = s - SvPVX(PL_linestr);
3679 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3680 return 0; /* no assumptions -- "=>" quotes bearword */
3682 start_force(PL_curforce);
3683 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3684 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3685 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3687 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3692 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3694 return *s == '(' ? FUNCMETH : METHOD;
3700 /* Encoded script support. filter_add() effectively inserts a
3701 * 'pre-processing' function into the current source input stream.
3702 * Note that the filter function only applies to the current source file
3703 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3705 * The datasv parameter (which may be NULL) can be used to pass
3706 * private data to this instance of the filter. The filter function
3707 * can recover the SV using the FILTER_DATA macro and use it to
3708 * store private buffers and state information.
3710 * The supplied datasv parameter is upgraded to a PVIO type
3711 * and the IoDIRP/IoANY field is used to store the function pointer,
3712 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3713 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3714 * private use must be set using malloc'd pointers.
3718 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3727 if (!PL_rsfp_filters)
3728 PL_rsfp_filters = newAV();
3731 SvUPGRADE(datasv, SVt_PVIO);
3732 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3733 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3734 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3735 FPTR2DPTR(void *, IoANY(datasv)),
3736 SvPV_nolen(datasv)));
3737 av_unshift(PL_rsfp_filters, 1);
3738 av_store(PL_rsfp_filters, 0, datasv) ;
3743 /* Delete most recently added instance of this filter function. */
3745 Perl_filter_del(pTHX_ filter_t funcp)
3750 PERL_ARGS_ASSERT_FILTER_DEL;
3753 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3754 FPTR2DPTR(void*, funcp)));
3756 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3758 /* if filter is on top of stack (usual case) just pop it off */
3759 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3760 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3761 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
3762 IoANY(datasv) = (void *)NULL;
3763 sv_free(av_pop(PL_rsfp_filters));
3767 /* we need to search for the correct entry and clear it */
3768 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3772 /* Invoke the idxth filter function for the current rsfp. */
3773 /* maxlen 0 = read one text line */
3775 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3780 /* This API is bad. It should have been using unsigned int for maxlen.
3781 Not sure if we want to change the API, but if not we should sanity
3782 check the value here. */
3783 const unsigned int correct_length
3792 PERL_ARGS_ASSERT_FILTER_READ;
3794 if (!PL_parser || !PL_rsfp_filters)
3796 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3797 /* Provide a default input filter to make life easy. */
3798 /* Note that we append to the line. This is handy. */
3799 DEBUG_P(PerlIO_printf(Perl_debug_log,
3800 "filter_read %d: from rsfp\n", idx));
3801 if (correct_length) {
3804 const int old_len = SvCUR(buf_sv);
3806 /* ensure buf_sv is large enough */
3807 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3808 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3809 correct_length)) <= 0) {
3810 if (PerlIO_error(PL_rsfp))
3811 return -1; /* error */
3813 return 0 ; /* end of file */
3815 SvCUR_set(buf_sv, old_len + len) ;
3816 SvPVX(buf_sv)[old_len + len] = '\0';
3819 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3820 if (PerlIO_error(PL_rsfp))
3821 return -1; /* error */
3823 return 0 ; /* end of file */
3826 return SvCUR(buf_sv);
3828 /* Skip this filter slot if filter has been deleted */
3829 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3830 DEBUG_P(PerlIO_printf(Perl_debug_log,
3831 "filter_read %d: skipped (filter deleted)\n",
3833 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
3835 /* Get function pointer hidden within datasv */
3836 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
3837 DEBUG_P(PerlIO_printf(Perl_debug_log,
3838 "filter_read %d: via function %p (%s)\n",
3839 idx, (void*)datasv, SvPV_nolen_const(datasv)));
3840 /* Call function. The function is expected to */
3841 /* call "FILTER_READ(idx+1, buf_sv)" first. */
3842 /* Return: <0:error, =0:eof, >0:not eof */
3843 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
3847 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
3851 PERL_ARGS_ASSERT_FILTER_GETS;
3853 #ifdef PERL_CR_FILTER
3854 if (!PL_rsfp_filters) {
3855 filter_add(S_cr_textfilter,NULL);
3858 if (PL_rsfp_filters) {
3860 SvCUR_set(sv, 0); /* start with empty line */
3861 if (FILTER_READ(0, sv, 0) > 0)
3862 return ( SvPVX(sv) ) ;
3867 return (sv_gets(sv, PL_rsfp, append));
3871 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3876 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3878 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3882 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3883 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3885 return GvHV(gv); /* Foo:: */
3888 /* use constant CLASS => 'MyClass' */
3889 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3890 if (gv && GvCV(gv)) {
3891 SV * const sv = cv_const_sv(GvCV(gv));
3893 pkgname = SvPV_const(sv, len);
3896 return gv_stashpvn(pkgname, len, 0);
3900 * S_readpipe_override
3901 * Check whether readpipe() is overriden, and generates the appropriate
3902 * optree, provided sublex_start() is called afterwards.
3905 S_readpipe_override(pTHX)
3908 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3909 pl_yylval.ival = OP_BACKTICK;
3911 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3913 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3914 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3915 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3917 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3918 append_elem(OP_LIST,
3919 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3920 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3927 * The intent of this yylex wrapper is to minimize the changes to the
3928 * tokener when we aren't interested in collecting madprops. It remains
3929 * to be seen how successful this strategy will be...
3936 char *s = PL_bufptr;
3938 /* make sure PL_thiswhite is initialized */
3942 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3943 if (PL_pending_ident)
3944 return S_pending_ident(aTHX);
3946 /* previous token ate up our whitespace? */
3947 if (!PL_lasttoke && PL_nextwhite) {
3948 PL_thiswhite = PL_nextwhite;
3952 /* isolate the token, and figure out where it is without whitespace */
3953 PL_realtokenstart = -1;
3957 assert(PL_curforce < 0);
3959 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3960 if (!PL_thistoken) {
3961 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3962 PL_thistoken = newSVpvs("");
3964 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3965 PL_thistoken = newSVpvn(tstart, s - tstart);
3968 if (PL_thismad) /* install head */
3969 CURMAD('X', PL_thistoken);
3972 /* last whitespace of a sublex? */
3973 if (optype == ')' && PL_endwhite) {
3974 CURMAD('X', PL_endwhite);
3979 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3980 if (!PL_thiswhite && !PL_endwhite && !optype) {
3981 sv_free(PL_thistoken);
3986 /* put off final whitespace till peg */
3987 if (optype == ';' && !PL_rsfp) {
3988 PL_nextwhite = PL_thiswhite;
3991 else if (PL_thisopen) {
3992 CURMAD('q', PL_thisopen);
3994 sv_free(PL_thistoken);
3998 /* Store actual token text as madprop X */
3999 CURMAD('X', PL_thistoken);
4003 /* add preceding whitespace as madprop _ */
4004 CURMAD('_', PL_thiswhite);
4008 /* add quoted material as madprop = */
4009 CURMAD('=', PL_thisstuff);
4013 /* add terminating quote as madprop Q */
4014 CURMAD('Q', PL_thisclose);
4018 /* special processing based on optype */
4022 /* opval doesn't need a TOKEN since it can already store mp */
4032 if (pl_yylval.opval)
4033 append_madprops(PL_thismad, pl_yylval.opval, 0);
4041 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4050 /* remember any fake bracket that lexer is about to discard */
4051 if (PL_lex_brackets == 1 &&
4052 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4055 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4058 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4059 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4062 break; /* don't bother looking for trailing comment */
4071 /* attach a trailing comment to its statement instead of next token */
4075 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4077 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4079 if (*s == '\n' || *s == '#') {
4080 while (s < PL_bufend && *s != '\n')
4084 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4085 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4102 /* Create new token struct. Note: opvals return early above. */
4103 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4110 S_tokenize_use(pTHX_ int is_use, char *s) {
4113 PERL_ARGS_ASSERT_TOKENIZE_USE;
4115 if (PL_expect != XSTATE)
4116 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4117 is_use ? "use" : "no"));
4119 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4120 s = force_version(s, TRUE);
4121 if (*s == ';' || *s == '}'
4122 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4123 start_force(PL_curforce);
4124 NEXTVAL_NEXTTOKE.opval = NULL;
4127 else if (*s == 'v') {
4128 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4129 s = force_version(s, FALSE);
4133 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4134 s = force_version(s, FALSE);
4136 pl_yylval.ival = is_use;
4140 static const char* const exp_name[] =
4141 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4142 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4149 Works out what to call the token just pulled out of the input
4150 stream. The yacc parser takes care of taking the ops we return and
4151 stitching them into a tree.
4157 if read an identifier
4158 if we're in a my declaration
4159 croak if they tried to say my($foo::bar)
4160 build the ops for a my() declaration
4161 if it's an access to a my() variable
4162 are we in a sort block?
4163 croak if my($a); $a <=> $b
4164 build ops for access to a my() variable
4165 if in a dq string, and they've said @foo and we can't find @foo
4167 build ops for a bareword
4168 if we already built the token before, use it.
4173 #pragma segment Perl_yylex
4179 register char *s = PL_bufptr;
4185 /* orig_keyword, gvp, and gv are initialized here because
4186 * jump to the label just_a_word_zero can bypass their
4187 * initialization later. */
4188 I32 orig_keyword = 0;
4193 SV* tmp = newSVpvs("");
4194 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4195 (IV)CopLINE(PL_curcop),
4196 lex_state_names[PL_lex_state],
4197 exp_name[PL_expect],
4198 pv_display(tmp, s, strlen(s), 0, 60));
4201 /* check if there's an identifier for us to look at */
4202 if (PL_pending_ident)
4203 return REPORT(S_pending_ident(aTHX));
4205 /* no identifier pending identification */
4207 switch (PL_lex_state) {
4209 case LEX_NORMAL: /* Some compilers will produce faster */
4210 case LEX_INTERPNORMAL: /* code if we comment these out. */
4214 /* when we've already built the next token, just pull it out of the queue */
4218 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4220 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4221 PL_nexttoke[PL_lasttoke].next_mad = 0;
4222 if (PL_thismad && PL_thismad->mad_key == '_') {
4223 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4224 PL_thismad->mad_val = 0;
4225 mad_free(PL_thismad);
4230 PL_lex_state = PL_lex_defer;
4231 PL_expect = PL_lex_expect;
4232 PL_lex_defer = LEX_NORMAL;
4233 if (!PL_nexttoke[PL_lasttoke].next_type)
4238 pl_yylval = PL_nextval[PL_nexttoke];
4240 PL_lex_state = PL_lex_defer;
4241 PL_expect = PL_lex_expect;
4242 PL_lex_defer = LEX_NORMAL;
4246 /* FIXME - can these be merged? */
4247 return(PL_nexttoke[PL_lasttoke].next_type);
4249 return REPORT(PL_nexttype[PL_nexttoke]);
4252 /* interpolated case modifiers like \L \U, including \Q and \E.
4253 when we get here, PL_bufptr is at the \
4255 case LEX_INTERPCASEMOD:
4257 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4258 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
4260 /* handle \E or end of string */
4261 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4263 if (PL_lex_casemods) {
4264 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4265 PL_lex_casestack[PL_lex_casemods] = '\0';
4267 if (PL_bufptr != PL_bufend
4268 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
4270 PL_lex_state = LEX_INTERPCONCAT;
4273 PL_thistoken = newSVpvs("\\E");
4279 while (PL_bufptr != PL_bufend &&
4280 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4282 PL_thiswhite = newSVpvs("");
4283 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4287 if (PL_bufptr != PL_bufend)
4290 PL_lex_state = LEX_INTERPCONCAT;
4294 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4295 "### Saw case modifier\n"); });
4297 if (s[1] == '\\' && s[2] == 'E') {
4300 PL_thiswhite = newSVpvs("");
4301 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4304 PL_lex_state = LEX_INTERPCONCAT;
4309 if (!PL_madskills) /* when just compiling don't need correct */
4310 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4311 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4312 if ((*s == 'L' || *s == 'U') &&
4313 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4314 PL_lex_casestack[--PL_lex_casemods] = '\0';
4317 if (PL_lex_casemods > 10)
4318 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4319 PL_lex_casestack[PL_lex_casemods++] = *s;
4320 PL_lex_casestack[PL_lex_casemods] = '\0';
4321 PL_lex_state = LEX_INTERPCONCAT;
4322 start_force(PL_curforce);
4323 NEXTVAL_NEXTTOKE.ival = 0;
4325 start_force(PL_curforce);
4327 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4329 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4331 NEXTVAL_NEXTTOKE.ival = OP_LC;
4333 NEXTVAL_NEXTTOKE.ival = OP_UC;
4335 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4337 Perl_croak(aTHX_ "panic: yylex");
4339 SV* const tmpsv = newSVpvs("\\ ");
4340 /* replace the space with the character we want to escape
4342 SvPVX(tmpsv)[1] = *s;
4348 if (PL_lex_starts) {
4354 sv_free(PL_thistoken);
4355 PL_thistoken = newSVpvs("");
4358 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4359 if (PL_lex_casemods == 1 && PL_lex_inpat)
4368 case LEX_INTERPPUSH:
4369 return REPORT(sublex_push());
4371 case LEX_INTERPSTART:
4372 if (PL_bufptr == PL_bufend)
4373 return REPORT(sublex_done());
4374 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4375 "### Interpolated variable\n"); });
4377 PL_lex_dojoin = (*PL_bufptr == '@');
4378 PL_lex_state = LEX_INTERPNORMAL;
4379 if (PL_lex_dojoin) {
4380 start_force(PL_curforce);
4381 NEXTVAL_NEXTTOKE.ival = 0;
4383 start_force(PL_curforce);
4384 force_ident("\"", '$');
4385 start_force(PL_curforce);
4386 NEXTVAL_NEXTTOKE.ival = 0;
4388 start_force(PL_curforce);
4389 NEXTVAL_NEXTTOKE.ival = 0;
4391 start_force(PL_curforce);
4392 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4395 if (PL_lex_starts++) {
4400 sv_free(PL_thistoken);
4401 PL_thistoken = newSVpvs("");
4404 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4405 if (!PL_lex_casemods && PL_lex_inpat)
4412 case LEX_INTERPENDMAYBE:
4413 if (intuit_more(PL_bufptr)) {
4414 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4420 if (PL_lex_dojoin) {
4421 PL_lex_dojoin = FALSE;
4422 PL_lex_state = LEX_INTERPCONCAT;
4426 sv_free(PL_thistoken);
4427 PL_thistoken = newSVpvs("");
4432 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4433 && SvEVALED(PL_lex_repl))
4435 if (PL_bufptr != PL_bufend)
4436 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4440 case LEX_INTERPCONCAT:
4442 if (PL_lex_brackets)
4443 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4445 if (PL_bufptr == PL_bufend)
4446 return REPORT(sublex_done());
4448 if (SvIVX(PL_linestr) == '\'') {
4449 SV *sv = newSVsv(PL_linestr);
4452 else if ( PL_hints & HINT_NEW_RE )
4453 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4454 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4458 s = scan_const(PL_bufptr);
4460 PL_lex_state = LEX_INTERPCASEMOD;
4462 PL_lex_state = LEX_INTERPSTART;
4465 if (s != PL_bufptr) {
4466 start_force(PL_curforce);
4468 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4470 NEXTVAL_NEXTTOKE = pl_yylval;
4473 if (PL_lex_starts++) {
4477 sv_free(PL_thistoken);
4478 PL_thistoken = newSVpvs("");
4481 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4482 if (!PL_lex_casemods && PL_lex_inpat)
4495 PL_lex_state = LEX_NORMAL;
4496 s = scan_formline(PL_bufptr);
4497 if (!PL_lex_formbrack)
4503 PL_oldoldbufptr = PL_oldbufptr;
4509 sv_free(PL_thistoken);
4512 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4516 if (isIDFIRST_lazy_if(s,UTF))
4519 unsigned char c = *s;
4520 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4521 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4522 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4527 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4531 goto fake_eof; /* emulate EOF on ^D or ^Z */
4540 if (PL_lex_brackets) {
4541 yyerror((const char *)
4543 ? "Format not terminated"
4544 : "Missing right curly or square bracket"));
4546 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4547 "### Tokener got EOF\n");
4551 if (s++ < PL_bufend)
4552 goto retry; /* ignore stray nulls */
4555 if (!PL_in_eval && !PL_preambled) {
4556 PL_preambled = TRUE;
4562 /* Generate a string of Perl code to load the debugger.
4563 * If PERL5DB is set, it will return the contents of that,
4564 * otherwise a compile-time require of perl5db.pl. */
4566 const char * const pdb = PerlEnv_getenv("PERL5DB");
4569 sv_setpv(PL_linestr, pdb);
4570 sv_catpvs(PL_linestr,";");
4572 SETERRNO(0,SS_NORMAL);
4573 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4576 sv_setpvs(PL_linestr,"");
4577 if (PL_preambleav) {
4578 SV **svp = AvARRAY(PL_preambleav);
4579 SV **const end = svp + AvFILLp(PL_preambleav);
4581 sv_catsv(PL_linestr, *svp);
4583 sv_catpvs(PL_linestr, ";");
4585 sv_free(MUTABLE_SV(PL_preambleav));
4586 PL_preambleav = NULL;
4589 sv_catpvs(PL_linestr,
4590 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4591 if (PL_minus_n || PL_minus_p) {
4592 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4594 sv_catpvs(PL_linestr,"chomp;");
4597 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4598 || *PL_splitstr == '"')
4599 && strchr(PL_splitstr + 1, *PL_splitstr))
4600 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4602 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4603 bytes can be used as quoting characters. :-) */
4604 const char *splits = PL_splitstr;
4605 sv_catpvs(PL_linestr, "our @F=split(q\0");
4608 if (*splits == '\\')
4609 sv_catpvn(PL_linestr, splits, 1);
4610 sv_catpvn(PL_linestr, splits, 1);
4611 } while (*splits++);
4612 /* This loop will embed the trailing NUL of
4613 PL_linestr as the last thing it does before
4615 sv_catpvs(PL_linestr, ");");
4619 sv_catpvs(PL_linestr,"our @F=split(' ');");
4622 sv_catpvs(PL_linestr, "\n");
4623 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4624 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4625 PL_last_lop = PL_last_uni = NULL;
4626 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4627 update_debugger_info(PL_linestr, NULL, 0);
4632 bof = PL_rsfp ? TRUE : FALSE;
4635 fake_eof = LEX_FAKE_EOF;
4637 PL_bufptr = PL_bufend;
4638 CopLINE_inc(PL_curcop);
4639 if (!lex_next_chunk(fake_eof)) {
4640 CopLINE_dec(PL_curcop);
4642 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4644 CopLINE_dec(PL_curcop);
4647 PL_realtokenstart = -1;
4650 /* If it looks like the start of a BOM or raw UTF-16,
4651 * check if it in fact is. */
4652 if (bof && PL_rsfp &&
4657 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
4659 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4660 s = swallow_bom((U8*)s);
4664 /* Incest with pod. */
4667 sv_catsv(PL_thiswhite, PL_linestr);
4669 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4670 sv_setpvs(PL_linestr, "");
4671 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4672 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4673 PL_last_lop = PL_last_uni = NULL;
4674 PL_doextract = FALSE;
4679 } while (PL_doextract);
4680 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4681 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4682 PL_last_lop = PL_last_uni = NULL;
4683 if (CopLINE(PL_curcop) == 1) {
4684 while (s < PL_bufend && isSPACE(*s))
4686 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4690 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4694 if (*s == '#' && *(s+1) == '!')
4696 #ifdef ALTERNATE_SHEBANG
4698 static char const as[] = ALTERNATE_SHEBANG;
4699 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4700 d = s + (sizeof(as) - 1);
4702 #endif /* ALTERNATE_SHEBANG */
4711 while (*d && !isSPACE(*d))
4715 #ifdef ARG_ZERO_IS_SCRIPT
4716 if (ipathend > ipath) {
4718 * HP-UX (at least) sets argv[0] to the script name,
4719 * which makes $^X incorrect. And Digital UNIX and Linux,
4720 * at least, set argv[0] to the basename of the Perl
4721 * interpreter. So, having found "#!", we'll set it right.
4723 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4725 assert(SvPOK(x) || SvGMAGICAL(x));
4726 if (sv_eq(x, CopFILESV(PL_curcop))) {
4727 sv_setpvn(x, ipath, ipathend - ipath);
4733 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4734 const char * const lstart = SvPV_const(x,llen);
4736 bstart += blen - llen;
4737 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4738 sv_setpvn(x, ipath, ipathend - ipath);
4743 TAINT_NOT; /* $^X is always tainted, but that's OK */
4745 #endif /* ARG_ZERO_IS_SCRIPT */
4750 d = instr(s,"perl -");
4752 d = instr(s,"perl");
4754 /* avoid getting into infinite loops when shebang
4755 * line contains "Perl" rather than "perl" */
4757 for (d = ipathend-4; d >= ipath; --d) {
4758 if ((*d == 'p' || *d == 'P')
4759 && !ibcmp(d, "perl", 4))
4769 #ifdef ALTERNATE_SHEBANG
4771 * If the ALTERNATE_SHEBANG on this system starts with a
4772 * character that can be part of a Perl expression, then if
4773 * we see it but not "perl", we're probably looking at the
4774 * start of Perl code, not a request to hand off to some
4775 * other interpreter. Similarly, if "perl" is there, but
4776 * not in the first 'word' of the line, we assume the line
4777 * contains the start of the Perl program.
4779 if (d && *s != '#') {
4780 const char *c = ipath;
4781 while (*c && !strchr("; \t\r\n\f\v#", *c))
4784 d = NULL; /* "perl" not in first word; ignore */
4786 *s = '#'; /* Don't try to parse shebang line */
4788 #endif /* ALTERNATE_SHEBANG */
4793 !instr(s,"indir") &&
4794 instr(PL_origargv[0],"perl"))
4801 while (s < PL_bufend && isSPACE(*s))
4803 if (s < PL_bufend) {
4804 Newx(newargv,PL_origargc+3,char*);
4806 while (s < PL_bufend && !isSPACE(*s))
4809 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4812 newargv = PL_origargv;
4815 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4817 Perl_croak(aTHX_ "Can't exec %s", ipath);
4820 while (*d && !isSPACE(*d))
4822 while (SPACE_OR_TAB(*d))
4826 const bool switches_done = PL_doswitches;
4827 const U32 oldpdb = PL_perldb;
4828 const bool oldn = PL_minus_n;
4829 const bool oldp = PL_minus_p;
4833 bool baduni = FALSE;
4835 const char *d2 = d1 + 1;
4836 if (parse_unicode_opts((const char **)&d2)
4840 if (baduni || *d1 == 'M' || *d1 == 'm') {
4841 const char * const m = d1;
4842 while (*d1 && !isSPACE(*d1))
4844 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4847 d1 = moreswitches(d1);
4849 if (PL_doswitches && !switches_done) {
4850 int argc = PL_origargc;
4851 char **argv = PL_origargv;
4854 } while (argc && argv[0][0] == '-' && argv[0][1]);
4855 init_argv_symbols(argc,argv);
4857 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4858 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4859 /* if we have already added "LINE: while (<>) {",
4860 we must not do it again */
4862 sv_setpvs(PL_linestr, "");
4863 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4864 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4865 PL_last_lop = PL_last_uni = NULL;
4866 PL_preambled = FALSE;
4867 if (PERLDB_LINE || PERLDB_SAVESRC)
4868 (void)gv_fetchfile(PL_origfilename);
4875 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4877 PL_lex_state = LEX_FORMLINE;
4882 #ifdef PERL_STRICT_CR
4883 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4885 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4887 case ' ': case '\t': case '\f': case 013:
4889 PL_realtokenstart = -1;
4891 PL_thiswhite = newSVpvs("");
4892 sv_catpvn(PL_thiswhite, s, 1);
4899 PL_realtokenstart = -1;
4903 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4904 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4905 /* handle eval qq[#line 1 "foo"\n ...] */
4906 CopLINE_dec(PL_curcop);
4909 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4911 if (!PL_in_eval || PL_rsfp)
4916 while (d < PL_bufend && *d != '\n')
4920 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4921 Perl_croak(aTHX_ "panic: input overflow");
4924 PL_thiswhite = newSVpvn(s, d - s);
4929 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4931 PL_lex_state = LEX_FORMLINE;
4937 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4938 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4941 TOKEN(PEG); /* make sure any #! line is accessible */
4946 /* if (PL_madskills && PL_lex_formbrack) { */
4948 while (d < PL_bufend && *d != '\n')
4952 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4953 Perl_croak(aTHX_ "panic: input overflow");
4954 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4956 PL_thiswhite = newSVpvs("");
4957 if (CopLINE(PL_curcop) == 1) {
4958 sv_setpvs(PL_thiswhite, "");
4961 sv_catpvn(PL_thiswhite, s, d - s);
4975 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4983 while (s < PL_bufend && SPACE_OR_TAB(*s))
4986 if (strnEQ(s,"=>",2)) {
4987 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4988 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4989 OPERATOR('-'); /* unary minus */
4991 PL_last_uni = PL_oldbufptr;
4993 case 'r': ftst = OP_FTEREAD; break;
4994 case 'w': ftst = OP_FTEWRITE; break;
4995 case 'x': ftst = OP_FTEEXEC; break;
4996 case 'o': ftst = OP_FTEOWNED; break;
4997 case 'R': ftst = OP_FTRREAD; break;
4998 case 'W': ftst = OP_FTRWRITE; break;
4999 case 'X': ftst = OP_FTREXEC; break;
5000 case 'O': ftst = OP_FTROWNED; break;
5001 case 'e': ftst = OP_FTIS; break;
5002 case 'z': ftst = OP_FTZERO; break;
5003 case 's': ftst = OP_FTSIZE; break;
5004 case 'f': ftst = OP_FTFILE; break;
5005 case 'd': ftst = OP_FTDIR; break;
5006 case 'l': ftst = OP_FTLINK; break;
5007 case 'p': ftst = OP_FTPIPE; break;
5008 case 'S': ftst = OP_FTSOCK; break;
5009 case 'u': ftst = OP_FTSUID; break;
5010 case 'g': ftst = OP_FTSGID; break;
5011 case 'k': ftst = OP_FTSVTX; break;
5012 case 'b': ftst = OP_FTBLK; break;
5013 case 'c': ftst = OP_FTCHR; break;
5014 case 't': ftst = OP_FTTTY; break;
5015 case 'T': ftst = OP_FTTEXT; break;
5016 case 'B': ftst = OP_FTBINARY; break;
5017 case 'M': case 'A': case 'C':
5018 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5020 case 'M': ftst = OP_FTMTIME; break;
5021 case 'A': ftst = OP_FTATIME; break;
5022 case 'C': ftst = OP_FTCTIME; break;
5030 PL_last_lop_op = (OPCODE)ftst;
5031 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5032 "### Saw file test %c\n", (int)tmp);
5037 /* Assume it was a minus followed by a one-letter named
5038 * subroutine call (or a -bareword), then. */
5039 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5040 "### '-%c' looked like a file test but was not\n",
5047 const char tmp = *s++;
5050 if (PL_expect == XOPERATOR)
5055 else if (*s == '>') {
5058 if (isIDFIRST_lazy_if(s,UTF)) {
5059 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5067 if (PL_expect == XOPERATOR)
5070 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5072 OPERATOR('-'); /* unary minus */
5078 const char tmp = *s++;
5081 if (PL_expect == XOPERATOR)
5086 if (PL_expect == XOPERATOR)
5089 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5096 if (PL_expect != XOPERATOR) {
5097 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5098 PL_expect = XOPERATOR;
5099 force_ident(PL_tokenbuf, '*');
5112 if (PL_expect == XOPERATOR) {
5116 PL_tokenbuf[0] = '%';
5117 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5118 sizeof PL_tokenbuf - 1, FALSE);
5119 if (!PL_tokenbuf[1]) {
5122 PL_pending_ident = '%';
5131 const char tmp = *s++;
5136 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5143 const char tmp = *s++;
5149 goto just_a_word_zero_gv;
5152 switch (PL_expect) {
5158 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5160 PL_bufptr = s; /* update in case we back off */
5162 deprecate(":= for an empty attribute list");
5169 PL_expect = XTERMBLOCK;
5172 stuffstart = s - SvPVX(PL_linestr) - 1;
5176 while (isIDFIRST_lazy_if(s,UTF)) {
5179 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5180 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5181 if (tmp < 0) tmp = -tmp;
5196 sv = newSVpvn(s, len);
5198 d = scan_str(d,TRUE,TRUE);
5200 /* MUST advance bufptr here to avoid bogus
5201 "at end of line" context messages from yyerror().
5203 PL_bufptr = s + len;
5204 yyerror("Unterminated attribute parameter in attribute list");
5208 return REPORT(0); /* EOF indicator */
5212 sv_catsv(sv, PL_lex_stuff);
5213 attrs = append_elem(OP_LIST, attrs,
5214 newSVOP(OP_CONST, 0, sv));
5215 SvREFCNT_dec(PL_lex_stuff);
5216 PL_lex_stuff = NULL;
5219 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5221 if (PL_in_my == KEY_our) {
5222 deprecate(":unique");
5225 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5228 /* NOTE: any CV attrs applied here need to be part of
5229 the CVf_BUILTIN_ATTRS define in cv.h! */
5230 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5232 CvLVALUE_on(PL_compcv);
5234 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5236 deprecate(":locked");
5238 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5240 CvMETHOD_on(PL_compcv);
5242 /* After we've set the flags, it could be argued that
5243 we don't need to do the attributes.pm-based setting
5244 process, and shouldn't bother appending recognized
5245 flags. To experiment with that, uncomment the
5246 following "else". (Note that's already been
5247 uncommented. That keeps the above-applied built-in
5248 attributes from being intercepted (and possibly
5249 rejected) by a package's attribute routines, but is
5250 justified by the performance win for the common case
5251 of applying only built-in attributes.) */
5253 attrs = append_elem(OP_LIST, attrs,
5254 newSVOP(OP_CONST, 0,
5258 if (*s == ':' && s[1] != ':')
5261 break; /* require real whitespace or :'s */
5262 /* XXX losing whitespace on sequential attributes here */
5266 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5267 if (*s != ';' && *s != '}' && *s != tmp
5268 && (tmp != '=' || *s != ')')) {
5269 const char q = ((*s == '\'') ? '"' : '\'');
5270 /* If here for an expression, and parsed no attrs, back
5272 if (tmp == '=' && !attrs) {
5276 /* MUST advance bufptr here to avoid bogus "at end of line"
5277 context messages from yyerror().
5280 yyerror( (const char *)
5282 ? Perl_form(aTHX_ "Invalid separator character "
5283 "%c%c%c in attribute list", q, *s, q)
5284 : "Unterminated attribute list" ) );
5292 start_force(PL_curforce);
5293 NEXTVAL_NEXTTOKE.opval = attrs;
5294 CURMAD('_', PL_nextwhite);
5299 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5300 (s - SvPVX(PL_linestr)) - stuffstart);
5308 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5309 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5317 const char tmp = *s++;
5322 const char tmp = *s++;
5330 if (PL_lex_brackets <= 0)
5331 yyerror("Unmatched right square bracket");
5334 if (PL_lex_state == LEX_INTERPNORMAL) {
5335 if (PL_lex_brackets == 0) {
5336 if (*s == '-' && s[1] == '>')
5337 PL_lex_state = LEX_INTERPENDMAYBE;
5338 else if (*s != '[' && *s != '{')
5339 PL_lex_state = LEX_INTERPEND;
5346 if (PL_lex_brackets > 100) {
5347 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5349 switch (PL_expect) {
5351 if (PL_lex_formbrack) {
5355 if (PL_oldoldbufptr == PL_last_lop)
5356 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5358 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5359 OPERATOR(HASHBRACK);
5361 while (s < PL_bufend && SPACE_OR_TAB(*s))
5364 PL_tokenbuf[0] = '\0';
5365 if (d < PL_bufend && *d == '-') {
5366 PL_tokenbuf[0] = '-';
5368 while (d < PL_bufend && SPACE_OR_TAB(*d))
5371 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5372 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5374 while (d < PL_bufend && SPACE_OR_TAB(*d))
5377 const char minus = (PL_tokenbuf[0] == '-');
5378 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5386 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5391 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5396 if (PL_oldoldbufptr == PL_last_lop)
5397 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5399 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5402 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5404 /* This hack is to get the ${} in the message. */
5406 yyerror("syntax error");
5409 OPERATOR(HASHBRACK);
5411 /* This hack serves to disambiguate a pair of curlies
5412 * as being a block or an anon hash. Normally, expectation
5413 * determines that, but in cases where we're not in a
5414 * position to expect anything in particular (like inside
5415 * eval"") we have to resolve the ambiguity. This code
5416 * covers the case where the first term in the curlies is a
5417 * quoted string. Most other cases need to be explicitly
5418 * disambiguated by prepending a "+" before the opening
5419 * curly in order to force resolution as an anon hash.
5421 * XXX should probably propagate the outer expectation
5422 * into eval"" to rely less on this hack, but that could
5423 * potentially break current behavior of eval"".
5427 if (*s == '\'' || *s == '"' || *s == '`') {
5428 /* common case: get past first string, handling escapes */
5429 for (t++; t < PL_bufend && *t != *s;)
5430 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5434 else if (*s == 'q') {
5437 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5440 /* skip q//-like construct */
5442 char open, close, term;
5445 while (t < PL_bufend && isSPACE(*t))
5447 /* check for q => */
5448 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5449 OPERATOR(HASHBRACK);
5453 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5457 for (t++; t < PL_bufend; t++) {
5458 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5460 else if (*t == open)
5464 for (t++; t < PL_bufend; t++) {
5465 if (*t == '\\' && t+1 < PL_bufend)
5467 else if (*t == close && --brackets <= 0)
5469 else if (*t == open)
5476 /* skip plain q word */
5477 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5480 else if (isALNUM_lazy_if(t,UTF)) {
5482 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5485 while (t < PL_bufend && isSPACE(*t))
5487 /* if comma follows first term, call it an anon hash */
5488 /* XXX it could be a comma expression with loop modifiers */
5489 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5490 || (*t == '=' && t[1] == '>')))
5491 OPERATOR(HASHBRACK);
5492 if (PL_expect == XREF)
5495 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5501 pl_yylval.ival = CopLINE(PL_curcop);
5502 if (isSPACE(*s) || *s == '#')
5503 PL_copline = NOLINE; /* invalidate current command line number */
5508 if (PL_lex_brackets <= 0)
5509 yyerror("Unmatched right curly bracket");
5511 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5512 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5513 PL_lex_formbrack = 0;
5514 if (PL_lex_state == LEX_INTERPNORMAL) {
5515 if (PL_lex_brackets == 0) {
5516 if (PL_expect & XFAKEBRACK) {
5517 PL_expect &= XENUMMASK;
5518 PL_lex_state = LEX_INTERPEND;
5523 PL_thiswhite = newSVpvs("");
5524 sv_catpvs(PL_thiswhite,"}");
5527 return yylex(); /* ignore fake brackets */
5529 if (*s == '-' && s[1] == '>')
5530 PL_lex_state = LEX_INTERPENDMAYBE;
5531 else if (*s != '[' && *s != '{')
5532 PL_lex_state = LEX_INTERPEND;
5535 if (PL_expect & XFAKEBRACK) {
5536 PL_expect &= XENUMMASK;
5538 return yylex(); /* ignore fake brackets */
5540 start_force(PL_curforce);
5542 curmad('X', newSVpvn(s-1,1));
5543 CURMAD('_', PL_thiswhite);
5548 PL_thistoken = newSVpvs("");
5556 if (PL_expect == XOPERATOR) {
5557 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5558 && isIDFIRST_lazy_if(s,UTF))
5560 CopLINE_dec(PL_curcop);
5561 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5562 CopLINE_inc(PL_curcop);
5567 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5569 PL_expect = XOPERATOR;
5570 force_ident(PL_tokenbuf, '&');
5574 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5586 const char tmp = *s++;
5593 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5594 && strchr("+-*/%.^&|<",tmp))
5595 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5596 "Reversed %c= operator",(int)tmp);
5598 if (PL_expect == XSTATE && isALPHA(tmp) &&
5599 (s == PL_linestart+1 || s[-2] == '\n') )
5601 if (PL_in_eval && !PL_rsfp) {
5606 if (strnEQ(s,"=cut",4)) {
5622 PL_thiswhite = newSVpvs("");
5623 sv_catpvn(PL_thiswhite, PL_linestart,
5624 PL_bufend - PL_linestart);
5628 PL_doextract = TRUE;
5632 if (PL_lex_brackets < PL_lex_formbrack) {
5634 #ifdef PERL_STRICT_CR
5635 while (SPACE_OR_TAB(*t))
5637 while (SPACE_OR_TAB(*t) || *t == '\r')
5640 if (*t == '\n' || *t == '#') {
5651 const char tmp = *s++;
5653 /* was this !=~ where !~ was meant?
5654 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5656 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5657 const char *t = s+1;
5659 while (t < PL_bufend && isSPACE(*t))
5662 if (*t == '/' || *t == '?' ||
5663 ((*t == 'm' || *t == 's' || *t == 'y')
5664 && !isALNUM(t[1])) ||
5665 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5666 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5667 "!=~ should be !~");
5677 if (PL_expect != XOPERATOR) {
5678 if (s[1] != '<' && !strchr(s,'>'))
5681 s = scan_heredoc(s);
5683 s = scan_inputsymbol(s);
5684 TERM(sublex_start());
5690 SHop(OP_LEFT_SHIFT);
5704 const char tmp = *s++;
5706 SHop(OP_RIGHT_SHIFT);
5707 else if (tmp == '=')
5716 if (PL_expect == XOPERATOR) {
5717 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5718 return deprecate_commaless_var_list();
5722 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
5723 PL_tokenbuf[0] = '@';
5724 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
5725 sizeof PL_tokenbuf - 1, FALSE);
5726 if (PL_expect == XOPERATOR)
5727 no_op("Array length", s);
5728 if (!PL_tokenbuf[1])
5730 PL_expect = XOPERATOR;
5731 PL_pending_ident = '#';
5735 PL_tokenbuf[0] = '$';
5736 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5737 sizeof PL_tokenbuf - 1, FALSE);
5738 if (PL_expect == XOPERATOR)
5740 if (!PL_tokenbuf[1]) {
5742 yyerror("Final $ should be \\$ or $name");
5746 /* This kludge not intended to be bulletproof. */
5747 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
5748 pl_yylval.opval = newSVOP(OP_CONST, 0,
5749 newSViv(CopARYBASE_get(&PL_compiling)));
5750 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
5756 const char tmp = *s;
5757 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
5760 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
5761 && intuit_more(s)) {
5763 PL_tokenbuf[0] = '@';
5764 if (ckWARN(WARN_SYNTAX)) {
5767 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
5770 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5771 while (t < PL_bufend && *t != ']')
5773 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5774 "Multidimensional syntax %.*s not supported",
5775 (int)((t - PL_bufptr) + 1), PL_bufptr);
5779 else if (*s == '{') {
5781 PL_tokenbuf[0] = '%';
5782 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
5783 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
5785 char tmpbuf[sizeof PL_tokenbuf];
5788 } while (isSPACE(*t));
5789 if (isIDFIRST_lazy_if(t,UTF)) {
5791 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
5795 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
5796 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5797 "You need to quote \"%s\"",
5804 PL_expect = XOPERATOR;
5805 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
5806 const bool islop = (PL_last_lop == PL_oldoldbufptr);
5807 if (!islop || PL_last_lop_op == OP_GREPSTART)
5808 PL_expect = XOPERATOR;
5809 else if (strchr("$@\"'`q", *s))
5810 PL_expect = XTERM; /* e.g. print $fh "foo" */
5811 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
5812 PL_expect = XTERM; /* e.g. print $fh &sub */
5813 else if (isIDFIRST_lazy_if(s,UTF)) {
5814 char tmpbuf[sizeof PL_tokenbuf];
5816 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5817 if ((t2 = keyword(tmpbuf, len, 0))) {
5818 /* binary operators exclude handle interpretations */
5830 PL_expect = XTERM; /* e.g. print $fh length() */
5835 PL_expect = XTERM; /* e.g. print $fh subr() */
5838 else if (isDIGIT(*s))
5839 PL_expect = XTERM; /* e.g. print $fh 3 */
5840 else if (*s == '.' && isDIGIT(s[1]))
5841 PL_expect = XTERM; /* e.g. print $fh .3 */
5842 else if ((*s == '?' || *s == '-' || *s == '+')
5843 && !isSPACE(s[1]) && s[1] != '=')
5844 PL_expect = XTERM; /* e.g. print $fh -1 */
5845 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5847 PL_expect = XTERM; /* e.g. print $fh /.../
5848 XXX except DORDOR operator
5850 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5852 PL_expect = XTERM; /* print $fh <<"EOF" */
5855 PL_pending_ident = '$';
5859 if (PL_expect == XOPERATOR)
5861 PL_tokenbuf[0] = '@';
5862 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5863 if (!PL_tokenbuf[1]) {
5866 if (PL_lex_state == LEX_NORMAL)
5868 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5870 PL_tokenbuf[0] = '%';
5872 /* Warn about @ where they meant $. */
5873 if (*s == '[' || *s == '{') {
5874 if (ckWARN(WARN_SYNTAX)) {
5875 const char *t = s + 1;
5876 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5878 if (*t == '}' || *t == ']') {
5880 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5881 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5882 "Scalar value %.*s better written as $%.*s",
5883 (int)(t-PL_bufptr), PL_bufptr,
5884 (int)(t-PL_bufptr-1), PL_bufptr+1);
5889 PL_pending_ident = '@';
5892 case '/': /* may be division, defined-or, or pattern */
5893 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5897 case '?': /* may either be conditional or pattern */
5898 if (PL_expect == XOPERATOR) {
5906 /* A // operator. */
5916 /* Disable warning on "study /blah/" */
5917 if (PL_oldoldbufptr == PL_last_uni
5918 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5919 || memNE(PL_last_uni, "study", 5)
5920 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5923 s = scan_pat(s,OP_MATCH);
5924 TERM(sublex_start());
5928 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5929 #ifdef PERL_STRICT_CR
5932 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5934 && (s == PL_linestart || s[-1] == '\n') )
5936 PL_lex_formbrack = 0;
5940 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5944 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5950 pl_yylval.ival = OPf_SPECIAL;
5959 case '0': case '1': case '2': case '3': case '4':
5960 case '5': case '6': case '7': case '8': case '9':
5961 s = scan_num(s, &pl_yylval);
5962 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5963 if (PL_expect == XOPERATOR)
5968 s = scan_str(s,!!PL_madskills,FALSE);
5969 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5970 if (PL_expect == XOPERATOR) {
5971 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5972 return deprecate_commaless_var_list();
5979 pl_yylval.ival = OP_CONST;
5980 TERM(sublex_start());
5983 s = scan_str(s,!!PL_madskills,FALSE);
5984 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5985 if (PL_expect == XOPERATOR) {
5986 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5987 return deprecate_commaless_var_list();
5994 pl_yylval.ival = OP_CONST;
5995 /* FIXME. I think that this can be const if char *d is replaced by
5996 more localised variables. */
5997 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5998 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5999 pl_yylval.ival = OP_STRINGIFY;
6003 TERM(sublex_start());
6006 s = scan_str(s,!!PL_madskills,FALSE);
6007 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6008 if (PL_expect == XOPERATOR)
6009 no_op("Backticks",s);
6012 readpipe_override();
6013 TERM(sublex_start());
6017 if (PL_lex_inwhat && isDIGIT(*s))
6018 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6020 if (PL_expect == XOPERATOR)
6021 no_op("Backslash",s);
6025 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6026 char *start = s + 2;
6027 while (isDIGIT(*start) || *start == '_')
6029 if (*start == '.' && isDIGIT(start[1])) {
6030 s = scan_num(s, &pl_yylval);
6033 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6034 else if (!isALPHA(*start) && (PL_expect == XTERM
6035 || PL_expect == XREF || PL_expect == XSTATE
6036 || PL_expect == XTERMORDORDOR)) {
6037 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
6039 s = scan_num(s, &pl_yylval);
6046 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6089 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6091 /* Some keywords can be followed by any delimiter, including ':' */
6092 anydelim = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
6093 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
6094 (PL_tokenbuf[0] == 'q' &&
6095 strchr("qwxr", PL_tokenbuf[1])))));
6097 /* x::* is just a word, unless x is "CORE" */
6098 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6102 while (d < PL_bufend && isSPACE(*d))
6103 d++; /* no comments skipped here, or s### is misparsed */
6105 /* Is this a word before a => operator? */
6106 if (*d == '=' && d[1] == '>') {
6109 = (OP*)newSVOP(OP_CONST, 0,
6110 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6111 pl_yylval.opval->op_private = OPpCONST_BARE;
6115 /* Check for plugged-in keyword */
6119 char *saved_bufptr = PL_bufptr;
6121 result = CALL_FPTR(PL_keyword_plugin)(aTHX_ PL_tokenbuf, len, &o);
6123 if (result == KEYWORD_PLUGIN_DECLINE) {
6124 /* not a plugged-in keyword */
6125 PL_bufptr = saved_bufptr;
6126 } else if (result == KEYWORD_PLUGIN_STMT) {
6127 pl_yylval.opval = o;
6130 return REPORT(PLUGSTMT);
6131 } else if (result == KEYWORD_PLUGIN_EXPR) {
6132 pl_yylval.opval = o;
6134 PL_expect = XOPERATOR;
6135 return REPORT(PLUGEXPR);
6137 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6142 /* Check for built-in keyword */
6143 tmp = keyword(PL_tokenbuf, len, 0);
6145 /* Is this a label? */
6146 if (!anydelim && PL_expect == XSTATE
6147 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6149 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
6154 if (tmp < 0) { /* second-class keyword? */
6155 GV *ogv = NULL; /* override (winner) */
6156 GV *hgv = NULL; /* hidden (loser) */
6157 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6159 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
6162 if (GvIMPORTED_CV(gv))
6164 else if (! CvMETHOD(cv))
6168 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
6169 (gv = *gvp) && isGV_with_GP(gv) &&
6170 GvCVu(gv) && GvIMPORTED_CV(gv))
6177 tmp = 0; /* overridden by import or by GLOBAL */
6180 && -tmp==KEY_lock /* XXX generalizable kludge */
6183 tmp = 0; /* any sub overrides "weak" keyword */
6185 else { /* no override */
6187 if (tmp == KEY_dump) {
6188 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6189 "dump() better written as CORE::dump()");
6193 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6194 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6195 "Ambiguous call resolved as CORE::%s(), %s",
6196 GvENAME(hgv), "qualify as such or use &");
6203 default: /* not a keyword */
6204 /* Trade off - by using this evil construction we can pull the
6205 variable gv into the block labelled keylookup. If not, then
6206 we have to give it function scope so that the goto from the
6207 earlier ':' case doesn't bypass the initialisation. */
6209 just_a_word_zero_gv:
6217 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6221 SV *nextPL_nextwhite = 0;
6225 /* Get the rest if it looks like a package qualifier */
6227 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6229 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6232 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
6233 *s == '\'' ? "'" : "::");
6238 if (PL_expect == XOPERATOR) {
6239 if (PL_bufptr == PL_linestart) {
6240 CopLINE_dec(PL_curcop);
6241 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6242 CopLINE_inc(PL_curcop);
6245 no_op("Bareword",s);
6248 /* Look for a subroutine with this name in current package,
6249 unless name is "Foo::", in which case Foo is a bearword
6250 (and a package name). */
6252 if (len > 2 && !PL_madskills &&
6253 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6255 if (ckWARN(WARN_BAREWORD)
6256 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
6257 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6258 "Bareword \"%s\" refers to nonexistent package",
6261 PL_tokenbuf[len] = '\0';
6267 /* Mustn't actually add anything to a symbol table.
6268 But also don't want to "initialise" any placeholder
6269 constants that might already be there into full
6270 blown PVGVs with attached PVCV. */
6271 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6272 GV_NOADD_NOINIT, SVt_PVCV);
6277 /* if we saw a global override before, get the right name */
6280 sv = newSVpvs("CORE::GLOBAL::");
6281 sv_catpv(sv,PL_tokenbuf);
6284 /* If len is 0, newSVpv does strlen(), which is correct.
6285 If len is non-zero, then it will be the true length,
6286 and so the scalar will be created correctly. */
6287 sv = newSVpv(PL_tokenbuf,len);
6290 if (PL_madskills && !PL_thistoken) {
6291 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6292 PL_thistoken = newSVpvn(start,s - start);
6293 PL_realtokenstart = s - SvPVX(PL_linestr);
6297 /* Presume this is going to be a bareword of some sort. */
6300 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6301 pl_yylval.opval->op_private = OPpCONST_BARE;
6302 /* UTF-8 package name? */
6303 if (UTF && !IN_BYTES &&
6304 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
6307 /* And if "Foo::", then that's what it certainly is. */
6314 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc(sv));
6315 const_op->op_private = OPpCONST_BARE;
6316 rv2cv_op = newCVREF(0, const_op);
6318 if (rv2cv_op->op_type == OP_RV2CV &&
6319 (rv2cv_op->op_flags & OPf_KIDS)) {
6320 OP *rv_op = cUNOPx(rv2cv_op)->op_first;
6321 switch (rv_op->op_type) {
6323 SV *sv = cSVOPx_sv(rv_op);
6324 if (SvROK(sv) && SvTYPE(SvRV(sv)) == SVt_PVCV)
6328 GV *gv = cGVOPx_gv(rv_op);
6329 CV *maybe_cv = GvCVu(gv);
6330 if (maybe_cv && SvTYPE((SV*)maybe_cv) == SVt_PVCV)
6336 /* See if it's the indirect object for a list operator. */
6338 if (PL_oldoldbufptr &&
6339 PL_oldoldbufptr < PL_bufptr &&
6340 (PL_oldoldbufptr == PL_last_lop
6341 || PL_oldoldbufptr == PL_last_uni) &&
6342 /* NO SKIPSPACE BEFORE HERE! */
6343 (PL_expect == XREF ||
6344 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6346 bool immediate_paren = *s == '(';
6348 /* (Now we can afford to cross potential line boundary.) */
6349 s = SKIPSPACE2(s,nextPL_nextwhite);
6351 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6354 /* Two barewords in a row may indicate method call. */
6356 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6357 (tmp = intuit_method(s, gv, cv))) {
6362 /* If not a declared subroutine, it's an indirect object. */
6363 /* (But it's an indir obj regardless for sort.) */
6364 /* Also, if "_" follows a filetest operator, it's a bareword */
6367 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6369 (PL_last_lop_op != OP_MAPSTART &&
6370 PL_last_lop_op != OP_GREPSTART))))
6371 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6372 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6375 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6380 PL_expect = XOPERATOR;
6383 s = SKIPSPACE2(s,nextPL_nextwhite);
6384 PL_nextwhite = nextPL_nextwhite;
6389 /* Is this a word before a => operator? */
6390 if (*s == '=' && s[1] == '>' && !pkgname) {
6393 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6394 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6395 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6399 /* If followed by a paren, it's certainly a subroutine. */
6404 while (SPACE_OR_TAB(*d))
6406 if (*d == ')' && (sv = cv_const_sv(cv))) {
6413 PL_nextwhite = PL_thiswhite;
6416 start_force(PL_curforce);
6418 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6419 PL_expect = XOPERATOR;
6422 PL_nextwhite = nextPL_nextwhite;
6423 curmad('X', PL_thistoken);
6424 PL_thistoken = newSVpvs("");
6433 /* If followed by var or block, call it a method (unless sub) */
6435 if ((*s == '$' || *s == '{') && !cv) {
6437 PL_last_lop = PL_oldbufptr;
6438 PL_last_lop_op = OP_METHOD;
6442 /* If followed by a bareword, see if it looks like indir obj. */
6445 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6446 && (tmp = intuit_method(s, gv, cv))) {
6451 /* Not a method, so call it a subroutine (if defined) */
6454 if (lastchar == '-')
6455 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6456 "Ambiguous use of -%s resolved as -&%s()",
6457 PL_tokenbuf, PL_tokenbuf);
6458 /* Check for a constant sub */
6459 if ((sv = cv_const_sv(cv))) {
6462 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6463 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6464 pl_yylval.opval->op_private = 0;
6468 op_free(pl_yylval.opval);
6469 pl_yylval.opval = rv2cv_op;
6470 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6471 PL_last_lop = PL_oldbufptr;
6472 PL_last_lop_op = OP_ENTERSUB;
6473 /* Is there a prototype? */
6481 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
6484 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
6486 while (*proto == ';')
6488 if (*proto == '&' && *s == '{') {
6490 sv_setpvs(PL_subname, "__ANON__");
6492 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6499 PL_nextwhite = PL_thiswhite;
6502 start_force(PL_curforce);
6503 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6506 PL_nextwhite = nextPL_nextwhite;
6507 curmad('X', PL_thistoken);
6508 PL_thistoken = newSVpvs("");
6515 /* Guess harder when madskills require "best effort". */
6516 if (PL_madskills && (!gv || !GvCVu(gv))) {
6517 int probable_sub = 0;
6518 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6520 else if (isALPHA(*s)) {
6524 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6525 if (!keyword(tmpbuf, tmplen, 0))
6528 while (d < PL_bufend && isSPACE(*d))
6530 if (*d == '=' && d[1] == '>')
6535 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
6536 op_free(pl_yylval.opval);
6537 pl_yylval.opval = rv2cv_op;
6538 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6539 PL_last_lop = PL_oldbufptr;
6540 PL_last_lop_op = OP_ENTERSUB;
6541 PL_nextwhite = PL_thiswhite;
6543 start_force(PL_curforce);
6544 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6546 PL_nextwhite = nextPL_nextwhite;
6547 curmad('X', PL_thistoken);
6548 PL_thistoken = newSVpvs("");
6553 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6560 /* Call it a bare word */
6562 if (PL_hints & HINT_STRICT_SUBS)
6563 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6566 /* after "print" and similar functions (corresponding to
6567 * "F? L" in opcode.pl), whatever wasn't already parsed as
6568 * a filehandle should be subject to "strict subs".
6569 * Likewise for the optional indirect-object argument to system
6570 * or exec, which can't be a bareword */
6571 if ((PL_last_lop_op == OP_PRINT
6572 || PL_last_lop_op == OP_PRTF
6573 || PL_last_lop_op == OP_SAY
6574 || PL_last_lop_op == OP_SYSTEM
6575 || PL_last_lop_op == OP_EXEC)
6576 && (PL_hints & HINT_STRICT_SUBS))
6577 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6578 if (lastchar != '-') {
6579 if (ckWARN(WARN_RESERVED)) {
6583 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
6584 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6592 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
6593 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6594 "Operator or semicolon missing before %c%s",
6595 lastchar, PL_tokenbuf);
6596 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6597 "Ambiguous use of %c resolved as operator %c",
6598 lastchar, lastchar);
6604 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6605 newSVpv(CopFILE(PL_curcop),0));
6609 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6610 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
6613 case KEY___PACKAGE__:
6614 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6616 ? newSVhek(HvNAME_HEK(PL_curstash))
6623 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6624 const char *pname = "main";
6625 if (PL_tokenbuf[2] == 'D')
6626 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
6627 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
6631 GvIOp(gv) = newIO();
6632 IoIFP(GvIOp(gv)) = PL_rsfp;
6633 #if defined(HAS_FCNTL) && defined(F_SETFD)
6635 const int fd = PerlIO_fileno(PL_rsfp);
6636 fcntl(fd,F_SETFD,fd >= 3);
6639 /* Mark this internal pseudo-handle as clean */
6640 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6641 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6642 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6644 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6645 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6646 /* if the script was opened in binmode, we need to revert
6647 * it to text mode for compatibility; but only iff it has CRs
6648 * XXX this is a questionable hack at best. */
6649 if (PL_bufend-PL_bufptr > 2
6650 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6653 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6654 loc = PerlIO_tell(PL_rsfp);
6655 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6658 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6660 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6661 #endif /* NETWARE */
6662 #ifdef PERLIO_IS_STDIO /* really? */
6663 # if defined(__BORLANDC__)
6664 /* XXX see note in do_binmode() */
6665 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
6669 PerlIO_seek(PL_rsfp, loc, 0);
6673 #ifdef PERLIO_LAYERS
6676 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6677 else if (PL_encoding) {
6684 XPUSHs(PL_encoding);
6686 call_method("name", G_SCALAR);
6690 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
6691 Perl_form(aTHX_ ":encoding(%"SVf")",
6700 if (PL_realtokenstart >= 0) {
6701 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6703 PL_endwhite = newSVpvs("");
6704 sv_catsv(PL_endwhite, PL_thiswhite);
6706 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
6707 PL_realtokenstart = -1;
6709 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
6725 if (PL_expect == XSTATE) {
6732 if (*s == ':' && s[1] == ':') {
6735 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6736 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
6737 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
6740 else if (tmp == KEY_require || tmp == KEY_do)
6741 /* that's a way to remember we saw "CORE::" */
6754 LOP(OP_ACCEPT,XTERM);
6760 LOP(OP_ATAN2,XTERM);
6766 LOP(OP_BINMODE,XTERM);
6769 LOP(OP_BLESS,XTERM);
6778 /* When 'use switch' is in effect, continue has a dual
6779 life as a control operator. */
6781 if (!FEATURE_IS_ENABLED("switch"))
6784 /* We have to disambiguate the two senses of
6785 "continue". If the next token is a '{' then
6786 treat it as the start of a continue block;
6787 otherwise treat it as a control operator.
6799 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
6816 if (!PL_cryptseen) {
6817 PL_cryptseen = TRUE;
6821 LOP(OP_CRYPT,XTERM);
6824 LOP(OP_CHMOD,XTERM);
6827 LOP(OP_CHOWN,XTERM);
6830 LOP(OP_CONNECT,XTERM);
6849 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6850 if (orig_keyword == KEY_do) {
6859 PL_hints |= HINT_BLOCK_SCOPE;
6869 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
6870 LOP(OP_DBMOPEN,XTERM);
6876 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6883 pl_yylval.ival = CopLINE(PL_curcop);
6899 if (*s == '{') { /* block eval */
6900 PL_expect = XTERMBLOCK;
6901 UNIBRACK(OP_ENTERTRY);
6903 else { /* string eval */
6905 UNIBRACK(OP_ENTEREVAL);
6920 case KEY_endhostent:
6926 case KEY_endservent:
6929 case KEY_endprotoent:
6940 pl_yylval.ival = CopLINE(PL_curcop);
6942 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6945 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6948 if ((PL_bufend - p) >= 3 &&
6949 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6951 else if ((PL_bufend - p) >= 4 &&
6952 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6955 if (isIDFIRST_lazy_if(p,UTF)) {
6956 p = scan_ident(p, PL_bufend,
6957 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6961 Perl_croak(aTHX_ "Missing $ on loop variable");
6963 s = SvPVX(PL_linestr) + soff;
6969 LOP(OP_FORMLINE,XTERM);
6975 LOP(OP_FCNTL,XTERM);
6981 LOP(OP_FLOCK,XTERM);
6990 LOP(OP_GREPSTART, XREF);
6993 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7008 case KEY_getpriority:
7009 LOP(OP_GETPRIORITY,XTERM);
7011 case KEY_getprotobyname:
7014 case KEY_getprotobynumber:
7015 LOP(OP_GPBYNUMBER,XTERM);
7017 case KEY_getprotoent:
7029 case KEY_getpeername:
7030 UNI(OP_GETPEERNAME);
7032 case KEY_gethostbyname:
7035 case KEY_gethostbyaddr:
7036 LOP(OP_GHBYADDR,XTERM);
7038 case KEY_gethostent:
7041 case KEY_getnetbyname:
7044 case KEY_getnetbyaddr:
7045 LOP(OP_GNBYADDR,XTERM);
7050 case KEY_getservbyname:
7051 LOP(OP_GSBYNAME,XTERM);
7053 case KEY_getservbyport:
7054 LOP(OP_GSBYPORT,XTERM);
7056 case KEY_getservent:
7059 case KEY_getsockname:
7060 UNI(OP_GETSOCKNAME);
7062 case KEY_getsockopt:
7063 LOP(OP_GSOCKOPT,XTERM);
7078 pl_yylval.ival = CopLINE(PL_curcop);
7088 pl_yylval.ival = CopLINE(PL_curcop);
7092 LOP(OP_INDEX,XTERM);
7098 LOP(OP_IOCTL,XTERM);
7110 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7142 LOP(OP_LISTEN,XTERM);
7151 s = scan_pat(s,OP_MATCH);
7152 TERM(sublex_start());
7155 LOP(OP_MAPSTART, XREF);
7158 LOP(OP_MKDIR,XTERM);
7161 LOP(OP_MSGCTL,XTERM);
7164 LOP(OP_MSGGET,XTERM);
7167 LOP(OP_MSGRCV,XTERM);
7170 LOP(OP_MSGSND,XTERM);
7175 PL_in_my = (U16)tmp;
7177 if (isIDFIRST_lazy_if(s,UTF)) {
7181 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7182 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7184 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7185 if (!PL_in_my_stash) {
7188 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7192 if (PL_madskills) { /* just add type to declarator token */
7193 sv_catsv(PL_thistoken, PL_nextwhite);
7195 sv_catpvn(PL_thistoken, start, s - start);
7203 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7210 s = tokenize_use(0, s);
7214 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7221 if (isIDFIRST_lazy_if(s,UTF)) {
7223 for (d = s; isALNUM_lazy_if(d,UTF);)
7225 for (t=d; isSPACE(*t);)
7227 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7229 && !(t[0] == '=' && t[1] == '>')
7231 int parms_len = (int)(d-s);
7232 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7233 "Precedence problem: open %.*s should be open(%.*s)",
7234 parms_len, s, parms_len, s);
7240 pl_yylval.ival = OP_OR;
7250 LOP(OP_OPEN_DIR,XTERM);
7253 checkcomma(s,PL_tokenbuf,"filehandle");
7257 checkcomma(s,PL_tokenbuf,"filehandle");
7276 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7278 s = force_strict_version(s);
7282 LOP(OP_PIPE_OP,XTERM);
7285 s = scan_str(s,!!PL_madskills,FALSE);
7288 pl_yylval.ival = OP_CONST;
7289 TERM(sublex_start());
7295 s = scan_str(s,!!PL_madskills,FALSE);
7298 PL_expect = XOPERATOR;
7300 if (SvCUR(PL_lex_stuff)) {
7303 d = SvPV_force(PL_lex_stuff, len);
7305 for (; isSPACE(*d) && len; --len, ++d)
7310 if (!warned && ckWARN(WARN_QW)) {
7311 for (; !isSPACE(*d) && len; --len, ++d) {
7313 Perl_warner(aTHX_ packWARN(WARN_QW),
7314 "Possible attempt to separate words with commas");
7317 else if (*d == '#') {
7318 Perl_warner(aTHX_ packWARN(WARN_QW),
7319 "Possible attempt to put comments in qw() list");
7325 for (; !isSPACE(*d) && len; --len, ++d)
7328 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7329 words = append_elem(OP_LIST, words,
7330 newSVOP(OP_CONST, 0, tokeq(sv)));
7334 start_force(PL_curforce);
7335 NEXTVAL_NEXTTOKE.opval = words;
7340 SvREFCNT_dec(PL_lex_stuff);
7341 PL_lex_stuff = NULL;
7347 s = scan_str(s,!!PL_madskills,FALSE);
7350 pl_yylval.ival = OP_STRINGIFY;
7351 if (SvIVX(PL_lex_stuff) == '\'')
7352 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
7353 TERM(sublex_start());
7356 s = scan_pat(s,OP_QR);
7357 TERM(sublex_start());
7360 s = scan_str(s,!!PL_madskills,FALSE);
7363 readpipe_override();
7364 TERM(sublex_start());
7372 s = force_version(s, FALSE);
7374 else if (*s != 'v' || !isDIGIT(s[1])
7375 || (s = force_version(s, TRUE), *s == 'v'))
7377 *PL_tokenbuf = '\0';
7378 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7379 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7380 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7382 yyerror("<> should be quotes");
7384 if (orig_keyword == KEY_require) {
7392 PL_last_uni = PL_oldbufptr;
7393 PL_last_lop_op = OP_REQUIRE;
7395 return REPORT( (int)REQUIRE );
7401 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7405 LOP(OP_RENAME,XTERM);
7414 LOP(OP_RINDEX,XTERM);
7423 UNIDOR(OP_READLINE);
7426 UNIDOR(OP_BACKTICK);
7435 LOP(OP_REVERSE,XTERM);
7438 UNIDOR(OP_READLINK);
7445 if (pl_yylval.opval)
7446 TERM(sublex_start());
7448 TOKEN(1); /* force error */
7451 checkcomma(s,PL_tokenbuf,"filehandle");
7461 LOP(OP_SELECT,XTERM);
7467 LOP(OP_SEMCTL,XTERM);
7470 LOP(OP_SEMGET,XTERM);
7473 LOP(OP_SEMOP,XTERM);
7479 LOP(OP_SETPGRP,XTERM);
7481 case KEY_setpriority:
7482 LOP(OP_SETPRIORITY,XTERM);
7484 case KEY_sethostent:
7490 case KEY_setservent:
7493 case KEY_setprotoent:
7503 LOP(OP_SEEKDIR,XTERM);
7505 case KEY_setsockopt:
7506 LOP(OP_SSOCKOPT,XTERM);
7512 LOP(OP_SHMCTL,XTERM);
7515 LOP(OP_SHMGET,XTERM);
7518 LOP(OP_SHMREAD,XTERM);
7521 LOP(OP_SHMWRITE,XTERM);
7524 LOP(OP_SHUTDOWN,XTERM);
7533 LOP(OP_SOCKET,XTERM);
7535 case KEY_socketpair:
7536 LOP(OP_SOCKPAIR,XTERM);
7539 checkcomma(s,PL_tokenbuf,"subroutine name");
7541 if (*s == ';' || *s == ')') /* probably a close */
7542 Perl_croak(aTHX_ "sort is now a reserved word");
7544 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7548 LOP(OP_SPLIT,XTERM);
7551 LOP(OP_SPRINTF,XTERM);
7554 LOP(OP_SPLICE,XTERM);
7569 LOP(OP_SUBSTR,XTERM);
7575 char tmpbuf[sizeof PL_tokenbuf];
7576 SSize_t tboffset = 0;
7577 expectation attrful;
7578 bool have_name, have_proto;
7579 const int key = tmp;
7584 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7585 SV *subtoken = newSVpvn(tstart, s - tstart);
7589 s = SKIPSPACE2(s,tmpwhite);
7594 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7595 (*s == ':' && s[1] == ':'))
7598 SV *nametoke = NULL;
7602 attrful = XATTRBLOCK;
7603 /* remember buffer pos'n for later force_word */
7604 tboffset = s - PL_oldbufptr;
7605 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7608 nametoke = newSVpvn(s, d - s);
7610 if (memchr(tmpbuf, ':', len))
7611 sv_setpvn(PL_subname, tmpbuf, len);
7613 sv_setsv(PL_subname,PL_curstname);
7614 sv_catpvs(PL_subname,"::");
7615 sv_catpvn(PL_subname,tmpbuf,len);
7622 CURMAD('X', nametoke);
7623 CURMAD('_', tmpwhite);
7624 (void) force_word(PL_oldbufptr + tboffset, WORD,
7627 s = SKIPSPACE2(d,tmpwhite);
7634 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7635 PL_expect = XTERMBLOCK;
7636 attrful = XATTRTERM;
7637 sv_setpvs(PL_subname,"?");
7641 if (key == KEY_format) {
7643 PL_lex_formbrack = PL_lex_brackets + 1;
7645 PL_thistoken = subtoken;
7649 (void) force_word(PL_oldbufptr + tboffset, WORD,
7655 /* Look for a prototype */
7658 bool bad_proto = FALSE;
7659 bool in_brackets = FALSE;
7660 char greedy_proto = ' ';
7661 bool proto_after_greedy_proto = FALSE;
7662 bool must_be_last = FALSE;
7663 bool underscore = FALSE;
7664 bool seen_underscore = FALSE;
7665 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
7667 s = scan_str(s,!!PL_madskills,FALSE);
7669 Perl_croak(aTHX_ "Prototype not terminated");
7670 /* strip spaces and check for bad characters */
7671 d = SvPVX(PL_lex_stuff);
7673 for (p = d; *p; ++p) {
7677 if (warnillegalproto) {
7679 proto_after_greedy_proto = TRUE;
7680 if (!strchr("$@%*;[]&\\_", *p)) {
7692 else if ( *p == ']' ) {
7693 in_brackets = FALSE;
7695 else if ( (*p == '@' || *p == '%') &&
7696 ( tmp < 2 || d[tmp-2] != '\\' ) &&
7698 must_be_last = TRUE;
7701 else if ( *p == '_' ) {
7702 underscore = seen_underscore = TRUE;
7709 if (proto_after_greedy_proto)
7710 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7711 "Prototype after '%c' for %"SVf" : %s",
7712 greedy_proto, SVfARG(PL_subname), d);
7714 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7715 "Illegal character %sin prototype for %"SVf" : %s",
7716 seen_underscore ? "after '_' " : "",
7717 SVfARG(PL_subname), d);
7718 SvCUR_set(PL_lex_stuff, tmp);
7723 CURMAD('q', PL_thisopen);
7724 CURMAD('_', tmpwhite);
7725 CURMAD('=', PL_thisstuff);
7726 CURMAD('Q', PL_thisclose);
7727 NEXTVAL_NEXTTOKE.opval =
7728 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7729 PL_lex_stuff = NULL;
7732 s = SKIPSPACE2(s,tmpwhite);
7740 if (*s == ':' && s[1] != ':')
7741 PL_expect = attrful;
7742 else if (*s != '{' && key == KEY_sub) {
7744 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
7745 else if (*s != ';' && *s != '}')
7746 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
7753 curmad('^', newSVpvs(""));
7754 CURMAD('_', tmpwhite);
7758 PL_thistoken = subtoken;
7761 NEXTVAL_NEXTTOKE.opval =
7762 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7763 PL_lex_stuff = NULL;
7769 sv_setpvs(PL_subname, "__ANON__");
7771 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7775 (void) force_word(PL_oldbufptr + tboffset, WORD,
7784 LOP(OP_SYSTEM,XREF);
7787 LOP(OP_SYMLINK,XTERM);
7790 LOP(OP_SYSCALL,XTERM);
7793 LOP(OP_SYSOPEN,XTERM);
7796 LOP(OP_SYSSEEK,XTERM);
7799 LOP(OP_SYSREAD,XTERM);
7802 LOP(OP_SYSWRITE,XTERM);
7806 TERM(sublex_start());
7827 LOP(OP_TRUNCATE,XTERM);
7839 pl_yylval.ival = CopLINE(PL_curcop);
7843 pl_yylval.ival = CopLINE(PL_curcop);
7847 LOP(OP_UNLINK,XTERM);
7853 LOP(OP_UNPACK,XTERM);
7856 LOP(OP_UTIME,XTERM);
7862 LOP(OP_UNSHIFT,XTERM);
7865 s = tokenize_use(1, s);
7875 pl_yylval.ival = CopLINE(PL_curcop);
7879 pl_yylval.ival = CopLINE(PL_curcop);
7883 PL_hints |= HINT_BLOCK_SCOPE;
7890 LOP(OP_WAITPID,XTERM);
7899 ctl_l[0] = toCTRL('L');
7901 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7904 /* Make sure $^L is defined */
7905 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7910 if (PL_expect == XOPERATOR)
7916 pl_yylval.ival = OP_XOR;
7921 TERM(sublex_start());
7926 #pragma segment Main
7930 S_pending_ident(pTHX)
7935 /* pit holds the identifier we read and pending_ident is reset */
7936 char pit = PL_pending_ident;
7937 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7938 /* All routes through this function want to know if there is a colon. */
7939 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7940 PL_pending_ident = 0;
7942 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7943 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7944 "### Pending identifier '%s'\n", PL_tokenbuf); });
7946 /* if we're in a my(), we can't allow dynamics here.
7947 $foo'bar has already been turned into $foo::bar, so
7948 just check for colons.
7950 if it's a legal name, the OP is a PADANY.
7953 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7955 yyerror(Perl_form(aTHX_ "No package name allowed for "
7956 "variable %s in \"our\"",
7958 tmp = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7962 yyerror(Perl_form(aTHX_ PL_no_myglob,
7963 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7965 pl_yylval.opval = newOP(OP_PADANY, 0);
7966 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7972 build the ops for accesses to a my() variable.
7974 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7975 then used in a comparison. This catches most, but not
7976 all cases. For instance, it catches
7977 sort { my($a); $a <=> $b }
7979 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7980 (although why you'd do that is anyone's guess).
7985 tmp = pad_findmy(PL_tokenbuf, tokenbuf_len, 0);
7986 if (tmp != NOT_IN_PAD) {
7987 /* might be an "our" variable" */
7988 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7989 /* build ops for a bareword */
7990 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7991 HEK * const stashname = HvNAME_HEK(stash);
7992 SV * const sym = newSVhek(stashname);
7993 sv_catpvs(sym, "::");
7994 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7995 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7996 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7999 ? (GV_ADDMULTI | GV_ADDINEVAL)
8002 ((PL_tokenbuf[0] == '$') ? SVt_PV
8003 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8008 /* if it's a sort block and they're naming $a or $b */
8009 if (PL_last_lop_op == OP_SORT &&
8010 PL_tokenbuf[0] == '$' &&
8011 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
8014 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
8015 d < PL_bufend && *d != '\n';
8018 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
8019 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
8025 pl_yylval.opval = newOP(OP_PADANY, 0);
8026 pl_yylval.opval->op_targ = tmp;
8032 Whine if they've said @foo in a doublequoted string,
8033 and @foo isn't a variable we can find in the symbol
8036 if (ckWARN(WARN_AMBIGUOUS) &&
8037 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8038 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
8040 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8041 /* DO NOT warn for @- and @+ */
8042 && !( PL_tokenbuf[2] == '\0' &&
8043 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8046 /* Downgraded from fatal to warning 20000522 mjd */
8047 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8048 "Possible unintended interpolation of %s in string",
8053 /* build ops for a bareword */
8054 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
8056 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8058 PL_tokenbuf + 1, tokenbuf_len - 1,
8059 /* If the identifier refers to a stash, don't autovivify it.
8060 * Change 24660 had the side effect of causing symbol table
8061 * hashes to always be defined, even if they were freshly
8062 * created and the only reference in the entire program was
8063 * the single statement with the defined %foo::bar:: test.
8064 * It appears that all code in the wild doing this actually
8065 * wants to know whether sub-packages have been loaded, so
8066 * by avoiding auto-vivifying symbol tables, we ensure that
8067 * defined %foo::bar:: continues to be false, and the existing
8068 * tests still give the expected answers, even though what
8069 * they're actually testing has now changed subtly.
8071 (*PL_tokenbuf == '%'
8072 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
8075 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
8076 ((PL_tokenbuf[0] == '$') ? SVt_PV
8077 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8083 * The following code was generated by perl_keyword.pl.
8087 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
8091 PERL_ARGS_ASSERT_KEYWORD;
8095 case 1: /* 5 tokens of length 1 */
8127 case 2: /* 18 tokens of length 2 */
8273 case 3: /* 29 tokens of length 3 */
8277 if (name[1] == 'N' &&
8340 if (name[1] == 'i' &&
8372 if (name[1] == 'o' &&
8381 if (name[1] == 'e' &&
8390 if (name[1] == 'n' &&
8399 if (name[1] == 'o' &&
8408 if (name[1] == 'a' &&
8417 if (name[1] == 'o' &&
8479 if (name[1] == 'e' &&
8493 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
8519 if (name[1] == 'i' &&
8528 if (name[1] == 's' &&
8537 if (name[1] == 'e' &&
8546 if (name[1] == 'o' &&
8558 case 4: /* 41 tokens of length 4 */
8562 if (name[1] == 'O' &&
8572 if (name[1] == 'N' &&
8582 if (name[1] == 'i' &&
8592 if (name[1] == 'h' &&
8602 if (name[1] == 'u' &&
8615 if (name[2] == 'c' &&
8624 if (name[2] == 's' &&
8633 if (name[2] == 'a' &&
8669 if (name[1] == 'o' &&
8682 if (name[2] == 't' &&
8691 if (name[2] == 'o' &&
8700 if (name[2] == 't' &&
8709 if (name[2] == 'e' &&
8722 if (name[1] == 'o' &&
8735 if (name[2] == 'y' &&
8744 if (name[2] == 'l' &&
8760 if (name[2] == 's' &&
8769 if (name[2] == 'n' &&
8778 if (name[2] == 'c' &&
8791 if (name[1] == 'e' &&
8801 if (name[1] == 'p' &&
8814 if (name[2] == 'c' &&
8823 if (name[2] == 'p' &&
8832 if (name[2] == 's' &&
8848 if (name[2] == 'n' &&
8918 if (name[2] == 'r' &&
8927 if (name[2] == 'r' &&
8936 if (name[2] == 'a' &&
8952 if (name[2] == 'l' &&
9014 if (name[2] == 'e' &&
9017 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
9030 case 5: /* 39 tokens of length 5 */
9034 if (name[1] == 'E' &&
9045 if (name[1] == 'H' &&
9059 if (name[2] == 'a' &&
9069 if (name[2] == 'a' &&
9086 if (name[2] == 'e' &&
9096 if (name[2] == 'e' &&
9100 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
9116 if (name[3] == 'i' &&
9125 if (name[3] == 'o' &&
9161 if (name[2] == 'o' &&
9171 if (name[2] == 'y' &&
9185 if (name[1] == 'l' &&
9199 if (name[2] == 'n' &&
9209 if (name[2] == 'o' &&
9223 if (name[1] == 'i' &&
9228 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
9237 if (name[2] == 'd' &&
9247 if (name[2] == 'c' &&
9264 if (name[2] == 'c' &&
9274 if (name[2] == 't' &&
9288 if (name[1] == 'k' &&
9299 if (name[1] == 'r' &&
9313 if (name[2] == 's' &&
9323 if (name[2] == 'd' &&
9340 if (name[2] == 'm' &&
9350 if (name[2] == 'i' &&
9360 if (name[2] == 'e' &&
9370 if (name[2] == 'l' &&
9380 if (name[2] == 'a' &&
9393 if (name[3] == 't' &&
9396 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
9402 if (name[3] == 'd' &&
9419 if (name[1] == 'i' &&
9433 if (name[2] == 'a' &&
9446 if (name[3] == 'e' &&
9481 if (name[2] == 'i' &&
9498 if (name[2] == 'i' &&
9508 if (name[2] == 'i' &&
9525 case 6: /* 33 tokens of length 6 */
9529 if (name[1] == 'c' &&
9544 if (name[2] == 'l' &&
9555 if (name[2] == 'r' &&
9570 if (name[1] == 'e' &&
9585 if (name[2] == 's' &&
9590 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
9596 if (name[2] == 'i' &&
9614 if (name[2] == 'l' &&
9625 if (name[2] == 'r' &&
9640 if (name[1] == 'm' &&
9655 if (name[2] == 'n' &&
9666 if (name[2] == 's' &&
9681 if (name[1] == 's' &&
9687 if (name[4] == 't' &&
9696 if (name[4] == 'e' &&
9705 if (name[4] == 'c' &&
9714 if (name[4] == 'n' &&
9730 if (name[1] == 'r' &&
9748 if (name[3] == 'a' &&
9758 if (name[3] == 'u' &&
9772 if (name[2] == 'n' &&
9790 if (name[2] == 'a' &&
9804 if (name[3] == 'e' &&
9817 if (name[4] == 't' &&
9826 if (name[4] == 'e' &&
9848 if (name[4] == 't' &&
9857 if (name[4] == 'e' &&
9873 if (name[2] == 'c' &&
9884 if (name[2] == 'l' &&
9895 if (name[2] == 'b' &&
9906 if (name[2] == 's' &&
9929 if (name[4] == 's' &&
9938 if (name[4] == 'n' &&
9951 if (name[3] == 'a' &&
9968 if (name[1] == 'a' &&
9983 case 7: /* 29 tokens of length 7 */
9987 if (name[1] == 'E' &&
10000 if (name[1] == '_' &&
10007 return KEY___END__;
10013 if (name[1] == 'i' &&
10020 return -KEY_binmode;
10026 if (name[1] == 'o' &&
10033 return -KEY_connect;
10042 if (name[2] == 'm' &&
10048 return -KEY_dbmopen;
10054 if (name[2] == 'f')
10059 if (name[4] == 'u' &&
10063 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
10069 if (name[4] == 'n' &&
10073 return KEY_defined;
10090 if (name[1] == 'o' &&
10097 return KEY_foreach;
10103 if (name[1] == 'e' &&
10110 if (name[5] == 'r' &&
10113 return -KEY_getpgrp;
10119 if (name[5] == 'i' &&
10122 return -KEY_getppid;
10135 if (name[1] == 'c' &&
10142 return -KEY_lcfirst;
10148 if (name[1] == 'p' &&
10155 return -KEY_opendir;
10161 if (name[1] == 'a' &&
10168 return KEY_package;
10174 if (name[1] == 'e')
10179 if (name[3] == 'd' &&
10184 return -KEY_readdir;
10190 if (name[3] == 'u' &&
10195 return KEY_require;
10201 if (name[3] == 'e' &&
10206 return -KEY_reverse;
10225 if (name[3] == 'k' &&
10230 return -KEY_seekdir;
10236 if (name[3] == 'p' &&
10241 return -KEY_setpgrp;
10251 if (name[2] == 'm' &&
10257 return -KEY_shmread;
10263 if (name[2] == 'r' &&
10269 return -KEY_sprintf;
10278 if (name[3] == 'l' &&
10283 return -KEY_symlink;
10292 if (name[4] == 'a' &&
10296 return -KEY_syscall;
10302 if (name[4] == 'p' &&
10306 return -KEY_sysopen;
10312 if (name[4] == 'e' &&
10316 return -KEY_sysread;
10322 if (name[4] == 'e' &&
10326 return -KEY_sysseek;
10344 if (name[1] == 'e' &&
10351 return -KEY_telldir;
10360 if (name[2] == 'f' &&
10366 return -KEY_ucfirst;
10372 if (name[2] == 's' &&
10378 return -KEY_unshift;
10388 if (name[1] == 'a' &&
10395 return -KEY_waitpid;
10404 case 8: /* 26 tokens of length 8 */
10408 if (name[1] == 'U' &&
10416 return KEY_AUTOLOAD;
10422 if (name[1] == '_')
10427 if (name[3] == 'A' &&
10433 return KEY___DATA__;
10439 if (name[3] == 'I' &&
10445 return -KEY___FILE__;
10451 if (name[3] == 'I' &&
10457 return -KEY___LINE__;
10473 if (name[2] == 'o' &&
10480 return -KEY_closedir;
10486 if (name[2] == 'n' &&
10493 return -KEY_continue;
10503 if (name[1] == 'b' &&
10511 return -KEY_dbmclose;
10517 if (name[1] == 'n' &&
10523 if (name[4] == 'r' &&
10528 return -KEY_endgrent;
10534 if (name[4] == 'w' &&
10539 return -KEY_endpwent;
10552 if (name[1] == 'o' &&
10560 return -KEY_formline;
10566 if (name[1] == 'e' &&
10572 if (name[4] == 'r')
10577 if (name[6] == 'n' &&
10580 return -KEY_getgrent;
10586 if (name[6] == 'i' &&
10589 return -KEY_getgrgid;
10595 if (name[6] == 'a' &&
10598 return -KEY_getgrnam;
10611 if (name[4] == 'o' &&
10616 return -KEY_getlogin;
10622 if (name[4] == 'w')
10627 if (name[6] == 'n' &&
10630 return -KEY_getpwent;
10636 if (name[6] == 'a' &&
10639 return -KEY_getpwnam;
10645 if (name[6] == 'i' &&
10648 return -KEY_getpwuid;
10668 if (name[1] == 'e' &&
10675 if (name[5] == 'i' &&
10682 return -KEY_readline;
10687 return -KEY_readlink;
10698 if (name[5] == 'i' &&
10702 return -KEY_readpipe;
10718 if (name[2] == 't')
10723 if (name[4] == 'r' &&
10728 return -KEY_setgrent;
10734 if (name[4] == 'w' &&
10739 return -KEY_setpwent;
10755 if (name[3] == 'w' &&
10761 return -KEY_shmwrite;
10767 if (name[3] == 't' &&
10773 return -KEY_shutdown;
10783 if (name[2] == 's' &&
10790 return -KEY_syswrite;
10800 if (name[1] == 'r' &&
10808 return -KEY_truncate;
10817 case 9: /* 9 tokens of length 9 */
10821 if (name[1] == 'N' &&
10830 return KEY_UNITCHECK;
10836 if (name[1] == 'n' &&
10845 return -KEY_endnetent;
10851 if (name[1] == 'e' &&
10860 return -KEY_getnetent;
10866 if (name[1] == 'o' &&
10875 return -KEY_localtime;
10881 if (name[1] == 'r' &&
10890 return KEY_prototype;
10896 if (name[1] == 'u' &&
10905 return -KEY_quotemeta;
10911 if (name[1] == 'e' &&
10920 return -KEY_rewinddir;
10926 if (name[1] == 'e' &&
10935 return -KEY_setnetent;
10941 if (name[1] == 'a' &&
10950 return -KEY_wantarray;
10959 case 10: /* 9 tokens of length 10 */
10963 if (name[1] == 'n' &&
10969 if (name[4] == 'o' &&
10976 return -KEY_endhostent;
10982 if (name[4] == 'e' &&
10989 return -KEY_endservent;
11002 if (name[1] == 'e' &&
11008 if (name[4] == 'o' &&
11015 return -KEY_gethostent;
11024 if (name[5] == 'r' &&
11030 return -KEY_getservent;
11036 if (name[5] == 'c' &&
11042 return -KEY_getsockopt;
11062 if (name[2] == 't')
11067 if (name[4] == 'o' &&
11074 return -KEY_sethostent;
11083 if (name[5] == 'r' &&
11089 return -KEY_setservent;
11095 if (name[5] == 'c' &&
11101 return -KEY_setsockopt;
11118 if (name[2] == 'c' &&
11127 return -KEY_socketpair;
11140 case 11: /* 8 tokens of length 11 */
11144 if (name[1] == '_' &&
11154 { /* __PACKAGE__ */
11155 return -KEY___PACKAGE__;
11161 if (name[1] == 'n' &&
11171 { /* endprotoent */
11172 return -KEY_endprotoent;
11178 if (name[1] == 'e' &&
11187 if (name[5] == 'e' &&
11193 { /* getpeername */
11194 return -KEY_getpeername;
11203 if (name[6] == 'o' &&
11208 { /* getpriority */
11209 return -KEY_getpriority;
11215 if (name[6] == 't' &&
11220 { /* getprotoent */
11221 return -KEY_getprotoent;
11235 if (name[4] == 'o' &&
11242 { /* getsockname */
11243 return -KEY_getsockname;
11256 if (name[1] == 'e' &&
11264 if (name[6] == 'o' &&
11269 { /* setpriority */
11270 return -KEY_setpriority;
11276 if (name[6] == 't' &&
11281 { /* setprotoent */
11282 return -KEY_setprotoent;
11298 case 12: /* 2 tokens of length 12 */
11299 if (name[0] == 'g' &&
11311 if (name[9] == 'd' &&
11314 { /* getnetbyaddr */
11315 return -KEY_getnetbyaddr;
11321 if (name[9] == 'a' &&
11324 { /* getnetbyname */
11325 return -KEY_getnetbyname;
11337 case 13: /* 4 tokens of length 13 */
11338 if (name[0] == 'g' &&
11345 if (name[4] == 'o' &&
11354 if (name[10] == 'd' &&
11357 { /* gethostbyaddr */
11358 return -KEY_gethostbyaddr;
11364 if (name[10] == 'a' &&
11367 { /* gethostbyname */
11368 return -KEY_gethostbyname;
11381 if (name[4] == 'e' &&
11390 if (name[10] == 'a' &&
11393 { /* getservbyname */
11394 return -KEY_getservbyname;
11400 if (name[10] == 'o' &&
11403 { /* getservbyport */
11404 return -KEY_getservbyport;
11423 case 14: /* 1 tokens of length 14 */
11424 if (name[0] == 'g' &&
11438 { /* getprotobyname */
11439 return -KEY_getprotobyname;
11444 case 16: /* 1 tokens of length 16 */
11445 if (name[0] == 'g' &&
11461 { /* getprotobynumber */
11462 return -KEY_getprotobynumber;
11476 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
11480 PERL_ARGS_ASSERT_CHECKCOMMA;
11482 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
11483 if (ckWARN(WARN_SYNTAX)) {
11486 for (w = s+2; *w && level; w++) {
11489 else if (*w == ')')
11492 while (isSPACE(*w))
11494 /* the list of chars below is for end of statements or
11495 * block / parens, boolean operators (&&, ||, //) and branch
11496 * constructs (or, and, if, until, unless, while, err, for).
11497 * Not a very solid hack... */
11498 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
11499 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11500 "%s (...) interpreted as function",name);
11503 while (s < PL_bufend && isSPACE(*s))
11507 while (s < PL_bufend && isSPACE(*s))
11509 if (isIDFIRST_lazy_if(s,UTF)) {
11510 const char * const w = s++;
11511 while (isALNUM_lazy_if(s,UTF))
11513 while (s < PL_bufend && isSPACE(*s))
11517 if (keyword(w, s - w, 0))
11520 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
11521 if (gv && GvCVu(gv))
11523 Perl_croak(aTHX_ "No comma allowed after %s", what);
11528 /* Either returns sv, or mortalizes sv and returns a new SV*.
11529 Best used as sv=new_constant(..., sv, ...).
11530 If s, pv are NULL, calls subroutine with one argument,
11531 and type is used with error messages only. */
11534 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
11535 SV *sv, SV *pv, const char *type, STRLEN typelen)
11538 HV * const table = GvHV(PL_hintgv); /* ^H */
11542 const char *why1 = "", *why2 = "", *why3 = "";
11544 PERL_ARGS_ASSERT_NEW_CONSTANT;
11546 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
11549 why2 = (const char *)
11550 (strEQ(key,"charnames")
11551 ? "(possibly a missing \"use charnames ...\")"
11553 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
11554 (type ? type: "undef"), why2);
11556 /* This is convoluted and evil ("goto considered harmful")
11557 * but I do not understand the intricacies of all the different
11558 * failure modes of %^H in here. The goal here is to make
11559 * the most probable error message user-friendly. --jhi */
11564 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
11565 (type ? type: "undef"), why1, why2, why3);
11567 yyerror(SvPVX_const(msg));
11572 /* charnames doesn't work well if there have been errors found */
11573 if (PL_error_count > 0 && strEQ(key,"charnames"))
11574 return &PL_sv_undef;
11576 cvp = hv_fetch(table, key, keylen, FALSE);
11577 if (!cvp || !SvOK(*cvp)) {
11580 why3 = "} is not defined";
11583 sv_2mortal(sv); /* Parent created it permanently */
11586 pv = newSVpvn_flags(s, len, SVs_TEMP);
11588 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
11590 typesv = &PL_sv_undef;
11592 PUSHSTACKi(PERLSI_OVERLOAD);
11604 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
11608 /* Check the eval first */
11609 if (!PL_in_eval && SvTRUE(ERRSV)) {
11610 sv_catpvs(ERRSV, "Propagated");
11611 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
11613 res = SvREFCNT_inc_simple(sv);
11617 SvREFCNT_inc_simple_void(res);
11626 why1 = "Call to &{$^H{";
11628 why3 = "}} did not return a defined value";
11636 /* Returns a NUL terminated string, with the length of the string written to
11640 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
11643 register char *d = dest;
11644 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11646 PERL_ARGS_ASSERT_SCAN_WORD;
11650 Perl_croak(aTHX_ ident_too_long);
11651 if (isALNUM(*s)) /* UTF handled below */
11653 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
11658 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
11662 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11663 char *t = s + UTF8SKIP(s);
11665 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11669 Perl_croak(aTHX_ ident_too_long);
11670 Copy(s, d, len, char);
11683 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
11686 char *bracket = NULL;
11688 register char *d = dest;
11689 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11691 PERL_ARGS_ASSERT_SCAN_IDENT;
11696 while (isDIGIT(*s)) {
11698 Perl_croak(aTHX_ ident_too_long);
11705 Perl_croak(aTHX_ ident_too_long);
11706 if (isALNUM(*s)) /* UTF handled below */
11708 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
11713 else if (*s == ':' && s[1] == ':') {
11717 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11718 char *t = s + UTF8SKIP(s);
11719 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11721 if (d + (t - s) > e)
11722 Perl_croak(aTHX_ ident_too_long);
11723 Copy(s, d, t - s, char);
11734 if (PL_lex_state != LEX_NORMAL)
11735 PL_lex_state = LEX_INTERPENDMAYBE;
11738 if (*s == '$' && s[1] &&
11739 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
11752 if (*d == '^' && *s && isCONTROLVAR(*s)) {
11757 if (isSPACE(s[-1])) {
11759 const char ch = *s++;
11760 if (!SPACE_OR_TAB(ch)) {
11766 if (isIDFIRST_lazy_if(d,UTF)) {
11770 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
11771 end += UTF8SKIP(end);
11772 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
11773 end += UTF8SKIP(end);
11775 Copy(s, d, end - s, char);
11780 while ((isALNUM(*s) || *s == ':') && d < e)
11783 Perl_croak(aTHX_ ident_too_long);
11786 while (s < send && SPACE_OR_TAB(*s))
11788 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
11789 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
11790 const char * const brack =
11792 ((*s == '[') ? "[...]" : "{...}");
11793 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11794 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
11795 funny, dest, brack, funny, dest, brack);
11798 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
11802 /* Handle extended ${^Foo} variables
11803 * 1999-02-27 mjd-perl-patch@plover.com */
11804 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
11808 while (isALNUM(*s) && d < e) {
11812 Perl_croak(aTHX_ ident_too_long);
11817 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
11818 PL_lex_state = LEX_INTERPEND;
11821 if (PL_lex_state == LEX_NORMAL) {
11822 if (ckWARN(WARN_AMBIGUOUS) &&
11823 (keyword(dest, d - dest, 0)
11824 || get_cvn_flags(dest, d - dest, 0)))
11828 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11829 "Ambiguous use of %c{%s} resolved to %c%s",
11830 funny, dest, funny, dest);
11835 s = bracket; /* let the parser handle it */
11839 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
11840 PL_lex_state = LEX_INTERPEND;
11845 S_pmflag(U32 pmfl, const char ch) {
11847 CASE_STD_PMMOD_FLAGS_PARSE_SET(&pmfl);
11848 case GLOBAL_PAT_MOD: pmfl |= PMf_GLOBAL; break;
11849 case CONTINUE_PAT_MOD: pmfl |= PMf_CONTINUE; break;
11850 case ONCE_PAT_MOD: pmfl |= PMf_KEEP; break;
11851 case KEEPCOPY_PAT_MOD: pmfl |= PMf_KEEPCOPY; break;
11857 Perl_pmflag(pTHX_ U32* pmfl, int ch)
11859 PERL_ARGS_ASSERT_PMFLAG;
11861 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
11862 "Perl_pmflag() is deprecated, and will be removed from the XS API");
11865 *pmfl = S_pmflag(*pmfl, (char)ch);
11870 S_scan_pat(pTHX_ char *start, I32 type)
11874 char *s = scan_str(start,!!PL_madskills,FALSE);
11875 const char * const valid_flags =
11876 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
11881 PERL_ARGS_ASSERT_SCAN_PAT;
11884 const char * const delimiter = skipspace(start);
11888 ? "Search pattern not terminated or ternary operator parsed as search pattern"
11889 : "Search pattern not terminated" ));
11892 pm = (PMOP*)newPMOP(type, 0);
11893 if (PL_multi_open == '?') {
11894 /* This is the only point in the code that sets PMf_ONCE: */
11895 pm->op_pmflags |= PMf_ONCE;
11897 /* Hence it's safe to do this bit of PMOP book-keeping here, which
11898 allows us to restrict the list needed by reset to just the ??
11900 assert(type != OP_TRANS);
11902 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11905 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11908 elements = mg->mg_len / sizeof(PMOP**);
11909 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11910 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11911 mg->mg_len = elements * sizeof(PMOP**);
11912 PmopSTASH_set(pm,PL_curstash);
11918 while (*s && strchr(valid_flags, *s))
11919 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11921 if (PL_madskills && modstart != s) {
11922 SV* tmptoken = newSVpvn(modstart, s - modstart);
11923 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11926 /* issue a warning if /c is specified,but /g is not */
11927 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11929 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11930 "Use of /c modifier is meaningless without /g" );
11933 PL_lex_op = (OP*)pm;
11934 pl_yylval.ival = OP_MATCH;
11939 S_scan_subst(pTHX_ char *start)
11950 PERL_ARGS_ASSERT_SCAN_SUBST;
11952 pl_yylval.ival = OP_NULL;
11954 s = scan_str(start,!!PL_madskills,FALSE);
11957 Perl_croak(aTHX_ "Substitution pattern not terminated");
11959 if (s[-1] == PL_multi_open)
11962 if (PL_madskills) {
11963 CURMAD('q', PL_thisopen);
11964 CURMAD('_', PL_thiswhite);
11965 CURMAD('E', PL_thisstuff);
11966 CURMAD('Q', PL_thisclose);
11967 PL_realtokenstart = s - SvPVX(PL_linestr);
11971 first_start = PL_multi_start;
11972 s = scan_str(s,!!PL_madskills,FALSE);
11974 if (PL_lex_stuff) {
11975 SvREFCNT_dec(PL_lex_stuff);
11976 PL_lex_stuff = NULL;
11978 Perl_croak(aTHX_ "Substitution replacement not terminated");
11980 PL_multi_start = first_start; /* so whole substitution is taken together */
11982 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11985 if (PL_madskills) {
11986 CURMAD('z', PL_thisopen);
11987 CURMAD('R', PL_thisstuff);
11988 CURMAD('Z', PL_thisclose);
11994 if (*s == EXEC_PAT_MOD) {
11998 else if (strchr(S_PAT_MODS, *s))
11999 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
12005 if (PL_madskills) {
12007 curmad('m', newSVpvn(modstart, s - modstart));
12008 append_madprops(PL_thismad, (OP*)pm, 0);
12012 if ((pm->op_pmflags & PMf_CONTINUE)) {
12013 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
12017 SV * const repl = newSVpvs("");
12019 PL_sublex_info.super_bufptr = s;
12020 PL_sublex_info.super_bufend = PL_bufend;
12022 pm->op_pmflags |= PMf_EVAL;
12025 sv_catpvs(repl, "eval ");
12027 sv_catpvs(repl, "do ");
12029 sv_catpvs(repl, "{");
12030 sv_catsv(repl, PL_lex_repl);
12031 if (strchr(SvPVX(PL_lex_repl), '#'))
12032 sv_catpvs(repl, "\n");
12033 sv_catpvs(repl, "}");
12035 SvREFCNT_dec(PL_lex_repl);
12036 PL_lex_repl = repl;
12039 PL_lex_op = (OP*)pm;
12040 pl_yylval.ival = OP_SUBST;
12045 S_scan_trans(pTHX_ char *start)
12058 PERL_ARGS_ASSERT_SCAN_TRANS;
12060 pl_yylval.ival = OP_NULL;
12062 s = scan_str(start,!!PL_madskills,FALSE);
12064 Perl_croak(aTHX_ "Transliteration pattern not terminated");
12066 if (s[-1] == PL_multi_open)
12069 if (PL_madskills) {
12070 CURMAD('q', PL_thisopen);
12071 CURMAD('_', PL_thiswhite);
12072 CURMAD('E', PL_thisstuff);
12073 CURMAD('Q', PL_thisclose);
12074 PL_realtokenstart = s - SvPVX(PL_linestr);
12078 s = scan_str(s,!!PL_madskills,FALSE);
12080 if (PL_lex_stuff) {
12081 SvREFCNT_dec(PL_lex_stuff);
12082 PL_lex_stuff = NULL;
12084 Perl_croak(aTHX_ "Transliteration replacement not terminated");
12086 if (PL_madskills) {
12087 CURMAD('z', PL_thisopen);
12088 CURMAD('R', PL_thisstuff);
12089 CURMAD('Z', PL_thisclose);
12092 complement = del = squash = 0;
12099 complement = OPpTRANS_COMPLEMENT;
12102 del = OPpTRANS_DELETE;
12105 squash = OPpTRANS_SQUASH;
12114 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
12115 o = newPVOP(OP_TRANS, 0, (char*)tbl);
12116 o->op_private &= ~OPpTRANS_ALL;
12117 o->op_private |= del|squash|complement|
12118 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
12119 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
12122 pl_yylval.ival = OP_TRANS;
12125 if (PL_madskills) {
12127 curmad('m', newSVpvn(modstart, s - modstart));
12128 append_madprops(PL_thismad, o, 0);
12137 S_scan_heredoc(pTHX_ register char *s)
12141 I32 op_type = OP_SCALAR;
12145 const char *found_newline;
12149 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
12151 I32 stuffstart = s - SvPVX(PL_linestr);
12154 PL_realtokenstart = -1;
12157 PERL_ARGS_ASSERT_SCAN_HEREDOC;
12161 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
12165 while (SPACE_OR_TAB(*peek))
12167 if (*peek == '`' || *peek == '\'' || *peek =='"') {
12170 s = delimcpy(d, e, s, PL_bufend, term, &len);
12180 if (!isALNUM_lazy_if(s,UTF))
12181 deprecate("bare << to mean <<\"\"");
12182 for (; isALNUM_lazy_if(s,UTF); s++) {
12187 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
12188 Perl_croak(aTHX_ "Delimiter for here document is too long");
12191 len = d - PL_tokenbuf;
12194 if (PL_madskills) {
12195 tstart = PL_tokenbuf + !outer;
12196 PL_thisclose = newSVpvn(tstart, len - !outer);
12197 tstart = SvPVX(PL_linestr) + stuffstart;
12198 PL_thisopen = newSVpvn(tstart, s - tstart);
12199 stuffstart = s - SvPVX(PL_linestr);
12202 #ifndef PERL_STRICT_CR
12203 d = strchr(s, '\r');
12205 char * const olds = s;
12207 while (s < PL_bufend) {
12213 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
12222 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12229 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
12230 herewas = newSVpvn(s,PL_bufend-s);
12234 herewas = newSVpvn(s-1,found_newline-s+1);
12237 herewas = newSVpvn(s,found_newline-s);
12241 if (PL_madskills) {
12242 tstart = SvPVX(PL_linestr) + stuffstart;
12244 sv_catpvn(PL_thisstuff, tstart, s - tstart);
12246 PL_thisstuff = newSVpvn(tstart, s - tstart);
12249 s += SvCUR(herewas);
12252 stuffstart = s - SvPVX(PL_linestr);
12258 tmpstr = newSV_type(SVt_PVIV);
12259 SvGROW(tmpstr, 80);
12260 if (term == '\'') {
12261 op_type = OP_CONST;
12262 SvIV_set(tmpstr, -1);
12264 else if (term == '`') {
12265 op_type = OP_BACKTICK;
12266 SvIV_set(tmpstr, '\\');
12270 PL_multi_start = CopLINE(PL_curcop);
12271 PL_multi_open = PL_multi_close = '<';
12272 term = *PL_tokenbuf;
12273 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
12274 char * const bufptr = PL_sublex_info.super_bufptr;
12275 char * const bufend = PL_sublex_info.super_bufend;
12276 char * const olds = s - SvCUR(herewas);
12277 s = strchr(bufptr, '\n');
12281 while (s < bufend &&
12282 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12284 CopLINE_inc(PL_curcop);
12287 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12288 missingterm(PL_tokenbuf);
12290 sv_setpvn(herewas,bufptr,d-bufptr+1);
12291 sv_setpvn(tmpstr,d+1,s-d);
12293 sv_catpvn(herewas,s,bufend-s);
12294 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
12301 while (s < PL_bufend &&
12302 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12304 CopLINE_inc(PL_curcop);
12306 if (s >= PL_bufend) {
12307 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12308 missingterm(PL_tokenbuf);
12310 sv_setpvn(tmpstr,d+1,s-d);
12312 if (PL_madskills) {
12314 sv_catpvn(PL_thisstuff, d + 1, s - d);
12316 PL_thisstuff = newSVpvn(d + 1, s - d);
12317 stuffstart = s - SvPVX(PL_linestr);
12321 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
12323 sv_catpvn(herewas,s,PL_bufend-s);
12324 sv_setsv(PL_linestr,herewas);
12325 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
12326 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12327 PL_last_lop = PL_last_uni = NULL;
12330 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
12331 while (s >= PL_bufend) { /* multiple line string? */
12333 if (PL_madskills) {
12334 tstart = SvPVX(PL_linestr) + stuffstart;
12336 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12338 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12342 CopLINE_inc(PL_curcop);
12343 if (!outer || !lex_next_chunk(0)) {
12344 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12345 missingterm(PL_tokenbuf);
12347 CopLINE_dec(PL_curcop);
12350 stuffstart = s - SvPVX(PL_linestr);
12352 CopLINE_inc(PL_curcop);
12353 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12354 PL_last_lop = PL_last_uni = NULL;
12355 #ifndef PERL_STRICT_CR
12356 if (PL_bufend - PL_linestart >= 2) {
12357 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
12358 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
12360 PL_bufend[-2] = '\n';
12362 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12364 else if (PL_bufend[-1] == '\r')
12365 PL_bufend[-1] = '\n';
12367 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
12368 PL_bufend[-1] = '\n';
12370 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
12371 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
12372 *(SvPVX(PL_linestr) + off ) = ' ';
12373 sv_catsv(PL_linestr,herewas);
12374 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12375 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
12379 sv_catsv(tmpstr,PL_linestr);
12384 PL_multi_end = CopLINE(PL_curcop);
12385 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
12386 SvPV_shrink_to_cur(tmpstr);
12388 SvREFCNT_dec(herewas);
12390 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
12392 else if (PL_encoding)
12393 sv_recode_to_utf8(tmpstr, PL_encoding);
12395 PL_lex_stuff = tmpstr;
12396 pl_yylval.ival = op_type;
12400 /* scan_inputsymbol
12401 takes: current position in input buffer
12402 returns: new position in input buffer
12403 side-effects: pl_yylval and lex_op are set.
12408 <FH> read from filehandle
12409 <pkg::FH> read from package qualified filehandle
12410 <pkg'FH> read from package qualified filehandle
12411 <$fh> read from filehandle in $fh
12412 <*.h> filename glob
12417 S_scan_inputsymbol(pTHX_ char *start)
12420 register char *s = start; /* current position in buffer */
12423 char *d = PL_tokenbuf; /* start of temp holding space */
12424 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
12426 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
12428 end = strchr(s, '\n');
12431 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
12433 /* die if we didn't have space for the contents of the <>,
12434 or if it didn't end, or if we see a newline
12437 if (len >= (I32)sizeof PL_tokenbuf)
12438 Perl_croak(aTHX_ "Excessively long <> operator");
12440 Perl_croak(aTHX_ "Unterminated <> operator");
12445 Remember, only scalar variables are interpreted as filehandles by
12446 this code. Anything more complex (e.g., <$fh{$num}>) will be
12447 treated as a glob() call.
12448 This code makes use of the fact that except for the $ at the front,
12449 a scalar variable and a filehandle look the same.
12451 if (*d == '$' && d[1]) d++;
12453 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
12454 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
12457 /* If we've tried to read what we allow filehandles to look like, and
12458 there's still text left, then it must be a glob() and not a getline.
12459 Use scan_str to pull out the stuff between the <> and treat it
12460 as nothing more than a string.
12463 if (d - PL_tokenbuf != len) {
12464 pl_yylval.ival = OP_GLOB;
12465 s = scan_str(start,!!PL_madskills,FALSE);
12467 Perl_croak(aTHX_ "Glob not terminated");
12471 bool readline_overriden = FALSE;
12474 /* we're in a filehandle read situation */
12477 /* turn <> into <ARGV> */
12479 Copy("ARGV",d,5,char);
12481 /* Check whether readline() is overriden */
12482 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
12484 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
12486 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
12487 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
12488 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
12489 readline_overriden = TRUE;
12491 /* if <$fh>, create the ops to turn the variable into a
12495 /* try to find it in the pad for this block, otherwise find
12496 add symbol table ops
12498 const PADOFFSET tmp = pad_findmy(d, len, 0);
12499 if (tmp != NOT_IN_PAD) {
12500 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
12501 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
12502 HEK * const stashname = HvNAME_HEK(stash);
12503 SV * const sym = sv_2mortal(newSVhek(stashname));
12504 sv_catpvs(sym, "::");
12505 sv_catpv(sym, d+1);
12510 OP * const o = newOP(OP_PADSV, 0);
12512 PL_lex_op = readline_overriden
12513 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12514 append_elem(OP_LIST, o,
12515 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
12516 : (OP*)newUNOP(OP_READLINE, 0, o);
12525 ? (GV_ADDMULTI | GV_ADDINEVAL)
12528 PL_lex_op = readline_overriden
12529 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12530 append_elem(OP_LIST,
12531 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
12532 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12533 : (OP*)newUNOP(OP_READLINE, 0,
12534 newUNOP(OP_RV2SV, 0,
12535 newGVOP(OP_GV, 0, gv)));
12537 if (!readline_overriden)
12538 PL_lex_op->op_flags |= OPf_SPECIAL;
12539 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
12540 pl_yylval.ival = OP_NULL;
12543 /* If it's none of the above, it must be a literal filehandle
12544 (<Foo::BAR> or <FOO>) so build a simple readline OP */
12546 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
12547 PL_lex_op = readline_overriden
12548 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12549 append_elem(OP_LIST,
12550 newGVOP(OP_GV, 0, gv),
12551 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12552 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
12553 pl_yylval.ival = OP_NULL;
12562 takes: start position in buffer
12563 keep_quoted preserve \ on the embedded delimiter(s)
12564 keep_delims preserve the delimiters around the string
12565 returns: position to continue reading from buffer
12566 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
12567 updates the read buffer.
12569 This subroutine pulls a string out of the input. It is called for:
12570 q single quotes q(literal text)
12571 ' single quotes 'literal text'
12572 qq double quotes qq(interpolate $here please)
12573 " double quotes "interpolate $here please"
12574 qx backticks qx(/bin/ls -l)
12575 ` backticks `/bin/ls -l`
12576 qw quote words @EXPORT_OK = qw( func() $spam )
12577 m// regexp match m/this/
12578 s/// regexp substitute s/this/that/
12579 tr/// string transliterate tr/this/that/
12580 y/// string transliterate y/this/that/
12581 ($*@) sub prototypes sub foo ($)
12582 (stuff) sub attr parameters sub foo : attr(stuff)
12583 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
12585 In most of these cases (all but <>, patterns and transliterate)
12586 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
12587 calls scan_str(). s/// makes yylex() call scan_subst() which calls
12588 scan_str(). tr/// and y/// make yylex() call scan_trans() which
12591 It skips whitespace before the string starts, and treats the first
12592 character as the delimiter. If the delimiter is one of ([{< then
12593 the corresponding "close" character )]}> is used as the closing
12594 delimiter. It allows quoting of delimiters, and if the string has
12595 balanced delimiters ([{<>}]) it allows nesting.
12597 On success, the SV with the resulting string is put into lex_stuff or,
12598 if that is already non-NULL, into lex_repl. The second case occurs only
12599 when parsing the RHS of the special constructs s/// and tr/// (y///).
12600 For convenience, the terminating delimiter character is stuffed into
12605 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
12608 SV *sv; /* scalar value: string */
12609 const char *tmps; /* temp string, used for delimiter matching */
12610 register char *s = start; /* current position in the buffer */
12611 register char term; /* terminating character */
12612 register char *to; /* current position in the sv's data */
12613 I32 brackets = 1; /* bracket nesting level */
12614 bool has_utf8 = FALSE; /* is there any utf8 content? */
12615 I32 termcode; /* terminating char. code */
12616 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
12617 STRLEN termlen; /* length of terminating string */
12618 int last_off = 0; /* last position for nesting bracket */
12624 PERL_ARGS_ASSERT_SCAN_STR;
12626 /* skip space before the delimiter */
12632 if (PL_realtokenstart >= 0) {
12633 stuffstart = PL_realtokenstart;
12634 PL_realtokenstart = -1;
12637 stuffstart = start - SvPVX(PL_linestr);
12639 /* mark where we are, in case we need to report errors */
12642 /* after skipping whitespace, the next character is the terminator */
12645 termcode = termstr[0] = term;
12649 termcode = utf8_to_uvchr((U8*)s, &termlen);
12650 Copy(s, termstr, termlen, U8);
12651 if (!UTF8_IS_INVARIANT(term))
12655 /* mark where we are */
12656 PL_multi_start = CopLINE(PL_curcop);
12657 PL_multi_open = term;
12659 /* find corresponding closing delimiter */
12660 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
12661 termcode = termstr[0] = term = tmps[5];
12663 PL_multi_close = term;
12665 /* create a new SV to hold the contents. 79 is the SV's initial length.
12666 What a random number. */
12667 sv = newSV_type(SVt_PVIV);
12669 SvIV_set(sv, termcode);
12670 (void)SvPOK_only(sv); /* validate pointer */
12672 /* move past delimiter and try to read a complete string */
12674 sv_catpvn(sv, s, termlen);
12677 tstart = SvPVX(PL_linestr) + stuffstart;
12678 if (!PL_thisopen && !keep_delims) {
12679 PL_thisopen = newSVpvn(tstart, s - tstart);
12680 stuffstart = s - SvPVX(PL_linestr);
12684 if (PL_encoding && !UTF) {
12688 int offset = s - SvPVX_const(PL_linestr);
12689 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
12690 &offset, (char*)termstr, termlen);
12691 const char * const ns = SvPVX_const(PL_linestr) + offset;
12692 char * const svlast = SvEND(sv) - 1;
12694 for (; s < ns; s++) {
12695 if (*s == '\n' && !PL_rsfp)
12696 CopLINE_inc(PL_curcop);
12699 goto read_more_line;
12701 /* handle quoted delimiters */
12702 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
12704 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
12706 if ((svlast-1 - t) % 2) {
12707 if (!keep_quoted) {
12708 *(svlast-1) = term;
12710 SvCUR_set(sv, SvCUR(sv) - 1);
12715 if (PL_multi_open == PL_multi_close) {
12721 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
12722 /* At here, all closes are "was quoted" one,
12723 so we don't check PL_multi_close. */
12725 if (!keep_quoted && *(t+1) == PL_multi_open)
12730 else if (*t == PL_multi_open)
12738 SvCUR_set(sv, w - SvPVX_const(sv));
12740 last_off = w - SvPVX(sv);
12741 if (--brackets <= 0)
12746 if (!keep_delims) {
12747 SvCUR_set(sv, SvCUR(sv) - 1);
12753 /* extend sv if need be */
12754 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
12755 /* set 'to' to the next character in the sv's string */
12756 to = SvPVX(sv)+SvCUR(sv);
12758 /* if open delimiter is the close delimiter read unbridle */
12759 if (PL_multi_open == PL_multi_close) {
12760 for (; s < PL_bufend; s++,to++) {
12761 /* embedded newlines increment the current line number */
12762 if (*s == '\n' && !PL_rsfp)
12763 CopLINE_inc(PL_curcop);
12764 /* handle quoted delimiters */
12765 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
12766 if (!keep_quoted && s[1] == term)
12768 /* any other quotes are simply copied straight through */
12772 /* terminate when run out of buffer (the for() condition), or
12773 have found the terminator */
12774 else if (*s == term) {
12777 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
12780 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12786 /* if the terminator isn't the same as the start character (e.g.,
12787 matched brackets), we have to allow more in the quoting, and
12788 be prepared for nested brackets.
12791 /* read until we run out of string, or we find the terminator */
12792 for (; s < PL_bufend; s++,to++) {
12793 /* embedded newlines increment the line count */
12794 if (*s == '\n' && !PL_rsfp)
12795 CopLINE_inc(PL_curcop);
12796 /* backslashes can escape the open or closing characters */
12797 if (*s == '\\' && s+1 < PL_bufend) {
12798 if (!keep_quoted &&
12799 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
12804 /* allow nested opens and closes */
12805 else if (*s == PL_multi_close && --brackets <= 0)
12807 else if (*s == PL_multi_open)
12809 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12814 /* terminate the copied string and update the sv's end-of-string */
12816 SvCUR_set(sv, to - SvPVX_const(sv));
12819 * this next chunk reads more into the buffer if we're not done yet
12823 break; /* handle case where we are done yet :-) */
12825 #ifndef PERL_STRICT_CR
12826 if (to - SvPVX_const(sv) >= 2) {
12827 if ((to[-2] == '\r' && to[-1] == '\n') ||
12828 (to[-2] == '\n' && to[-1] == '\r'))
12832 SvCUR_set(sv, to - SvPVX_const(sv));
12834 else if (to[-1] == '\r')
12837 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
12842 /* if we're out of file, or a read fails, bail and reset the current
12843 line marker so we can report where the unterminated string began
12846 if (PL_madskills) {
12847 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12849 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12851 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12854 CopLINE_inc(PL_curcop);
12855 PL_bufptr = PL_bufend;
12856 if (!lex_next_chunk(0)) {
12858 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12867 /* at this point, we have successfully read the delimited string */
12869 if (!PL_encoding || UTF) {
12871 if (PL_madskills) {
12872 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12873 const int len = s - tstart;
12875 sv_catpvn(PL_thisstuff, tstart, len);
12877 PL_thisstuff = newSVpvn(tstart, len);
12878 if (!PL_thisclose && !keep_delims)
12879 PL_thisclose = newSVpvn(s,termlen);
12884 sv_catpvn(sv, s, termlen);
12889 if (PL_madskills) {
12890 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12891 const int len = s - tstart - termlen;
12893 sv_catpvn(PL_thisstuff, tstart, len);
12895 PL_thisstuff = newSVpvn(tstart, len);
12896 if (!PL_thisclose && !keep_delims)
12897 PL_thisclose = newSVpvn(s - termlen,termlen);
12901 if (has_utf8 || PL_encoding)
12904 PL_multi_end = CopLINE(PL_curcop);
12906 /* if we allocated too much space, give some back */
12907 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12908 SvLEN_set(sv, SvCUR(sv) + 1);
12909 SvPV_renew(sv, SvLEN(sv));
12912 /* decide whether this is the first or second quoted string we've read
12925 takes: pointer to position in buffer
12926 returns: pointer to new position in buffer
12927 side-effects: builds ops for the constant in pl_yylval.op
12929 Read a number in any of the formats that Perl accepts:
12931 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12932 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12935 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12937 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12940 If it reads a number without a decimal point or an exponent, it will
12941 try converting the number to an integer and see if it can do so
12942 without loss of precision.
12946 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12949 register const char *s = start; /* current position in buffer */
12950 register char *d; /* destination in temp buffer */
12951 register char *e; /* end of temp buffer */
12952 NV nv; /* number read, as a double */
12953 SV *sv = NULL; /* place to put the converted number */
12954 bool floatit; /* boolean: int or float? */
12955 const char *lastub = NULL; /* position of last underbar */
12956 static char const number_too_long[] = "Number too long";
12958 PERL_ARGS_ASSERT_SCAN_NUM;
12960 /* We use the first character to decide what type of number this is */
12964 Perl_croak(aTHX_ "panic: scan_num");
12966 /* if it starts with a 0, it could be an octal number, a decimal in
12967 0.13 disguise, or a hexadecimal number, or a binary number. */
12971 u holds the "number so far"
12972 shift the power of 2 of the base
12973 (hex == 4, octal == 3, binary == 1)
12974 overflowed was the number more than we can hold?
12976 Shift is used when we add a digit. It also serves as an "are
12977 we in octal/hex/binary?" indicator to disallow hex characters
12978 when in octal mode.
12983 bool overflowed = FALSE;
12984 bool just_zero = TRUE; /* just plain 0 or binary number? */
12985 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12986 static const char* const bases[5] =
12987 { "", "binary", "", "octal", "hexadecimal" };
12988 static const char* const Bases[5] =
12989 { "", "Binary", "", "Octal", "Hexadecimal" };
12990 static const char* const maxima[5] =
12992 "0b11111111111111111111111111111111",
12996 const char *base, *Base, *max;
12998 /* check for hex */
13003 } else if (s[1] == 'b') {
13008 /* check for a decimal in disguise */
13009 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
13011 /* so it must be octal */
13018 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13019 "Misplaced _ in number");
13023 base = bases[shift];
13024 Base = Bases[shift];
13025 max = maxima[shift];
13027 /* read the rest of the number */
13029 /* x is used in the overflow test,
13030 b is the digit we're adding on. */
13035 /* if we don't mention it, we're done */
13039 /* _ are ignored -- but warned about if consecutive */
13041 if (lastub && s == lastub + 1)
13042 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13043 "Misplaced _ in number");
13047 /* 8 and 9 are not octal */
13048 case '8': case '9':
13050 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
13054 case '2': case '3': case '4':
13055 case '5': case '6': case '7':
13057 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
13060 case '0': case '1':
13061 b = *s++ & 15; /* ASCII digit -> value of digit */
13065 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
13066 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
13067 /* make sure they said 0x */
13070 b = (*s++ & 7) + 9;
13072 /* Prepare to put the digit we have onto the end
13073 of the number so far. We check for overflows.
13079 x = u << shift; /* make room for the digit */
13081 if ((x >> shift) != u
13082 && !(PL_hints & HINT_NEW_BINARY)) {
13085 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13086 "Integer overflow in %s number",
13089 u = x | b; /* add the digit to the end */
13092 n *= nvshift[shift];
13093 /* If an NV has not enough bits in its
13094 * mantissa to represent an UV this summing of
13095 * small low-order numbers is a waste of time
13096 * (because the NV cannot preserve the
13097 * low-order bits anyway): we could just
13098 * remember when did we overflow and in the
13099 * end just multiply n by the right
13107 /* if we get here, we had success: make a scalar value from
13112 /* final misplaced underbar check */
13113 if (s[-1] == '_') {
13114 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13117 /* Dot here is historically concat, not a radix point.
13118 Deprecate that; it's confusing, and gets in the way of
13119 hex(ish) fractions... but '..' is OK. */
13122 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
13123 "Dot after %s literal is concatenation", base);
13128 if (n > 4294967295.0)
13129 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13130 "%s number > %s non-portable",
13136 if (u > 0xffffffff)
13137 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13138 "%s number > %s non-portable",
13143 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
13144 sv = new_constant(start, s - start, "integer",
13145 sv, NULL, NULL, 0);
13146 else if (PL_hints & HINT_NEW_BINARY)
13147 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
13152 handle decimal numbers.
13153 we're also sent here when we read a 0 as the first digit
13155 case '1': case '2': case '3': case '4': case '5':
13156 case '6': case '7': case '8': case '9': case '.':
13159 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
13162 /* read next group of digits and _ and copy into d */
13163 while (isDIGIT(*s) || *s == '_') {
13164 /* skip underscores, checking for misplaced ones
13168 if (lastub && s == lastub + 1)
13169 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13170 "Misplaced _ in number");
13174 /* check for end of fixed-length buffer */
13176 Perl_croak(aTHX_ number_too_long);
13177 /* if we're ok, copy the character */
13182 /* final misplaced underbar check */
13183 if (lastub && s == lastub + 1) {
13184 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13187 /* read a decimal portion if there is one. avoid
13188 3..5 being interpreted as the number 3. followed
13191 if (*s == '.' && s[1] != '.') {
13196 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13197 "Misplaced _ in number");
13201 /* copy, ignoring underbars, until we run out of digits.
13203 for (; isDIGIT(*s) || *s == '_'; s++) {
13204 /* fixed length buffer check */
13206 Perl_croak(aTHX_ number_too_long);
13208 if (lastub && s == lastub + 1)
13209 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13210 "Misplaced _ in number");
13216 /* fractional part ending in underbar? */
13217 if (s[-1] == '_') {
13218 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13219 "Misplaced _ in number");
13221 if (*s == '.' && isDIGIT(s[1])) {
13222 /* oops, it's really a v-string, but without the "v" */
13228 /* read exponent part, if present */
13229 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
13233 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
13234 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
13236 /* stray preinitial _ */
13238 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13239 "Misplaced _ in number");
13243 /* allow positive or negative exponent */
13244 if (*s == '+' || *s == '-')
13247 /* stray initial _ */
13249 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13250 "Misplaced _ in number");
13254 /* read digits of exponent */
13255 while (isDIGIT(*s) || *s == '_') {
13258 Perl_croak(aTHX_ number_too_long);
13262 if (((lastub && s == lastub + 1) ||
13263 (!isDIGIT(s[1]) && s[1] != '_')))
13264 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13265 "Misplaced _ in number");
13272 /* make an sv from the string */
13276 We try to do an integer conversion first if no characters
13277 indicating "float" have been found.
13282 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
13284 if (flags == IS_NUMBER_IN_UV) {
13286 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
13289 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
13290 if (uv <= (UV) IV_MIN)
13291 sv_setiv(sv, -(IV)uv);
13298 /* terminate the string */
13300 nv = Atof(PL_tokenbuf);
13305 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
13306 const char *const key = floatit ? "float" : "integer";
13307 const STRLEN keylen = floatit ? 5 : 7;
13308 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
13309 key, keylen, sv, NULL, NULL, 0);
13313 /* if it starts with a v, it could be a v-string */
13316 sv = newSV(5); /* preallocate storage space */
13317 s = scan_vstring(s, PL_bufend, sv);
13321 /* make the op for the constant and return */
13324 lvalp->opval = newSVOP(OP_CONST, 0, sv);
13326 lvalp->opval = NULL;
13332 S_scan_formline(pTHX_ register char *s)
13335 register char *eol;
13337 SV * const stuff = newSVpvs("");
13338 bool needargs = FALSE;
13339 bool eofmt = FALSE;
13341 char *tokenstart = s;
13342 SV* savewhite = NULL;
13344 if (PL_madskills) {
13345 savewhite = PL_thiswhite;
13350 PERL_ARGS_ASSERT_SCAN_FORMLINE;
13352 while (!needargs) {
13355 #ifdef PERL_STRICT_CR
13356 while (SPACE_OR_TAB(*t))
13359 while (SPACE_OR_TAB(*t) || *t == '\r')
13362 if (*t == '\n' || t == PL_bufend) {
13367 if (PL_in_eval && !PL_rsfp) {
13368 eol = (char *) memchr(s,'\n',PL_bufend-s);
13373 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
13375 for (t = s; t < eol; t++) {
13376 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
13378 goto enough; /* ~~ must be first line in formline */
13380 if (*t == '@' || *t == '^')
13384 sv_catpvn(stuff, s, eol-s);
13385 #ifndef PERL_STRICT_CR
13386 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
13387 char *end = SvPVX(stuff) + SvCUR(stuff);
13390 SvCUR_set(stuff, SvCUR(stuff) - 1);
13401 if (PL_madskills) {
13403 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
13405 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
13408 PL_bufptr = PL_bufend;
13409 CopLINE_inc(PL_curcop);
13410 got_some = lex_next_chunk(0);
13411 CopLINE_dec(PL_curcop);
13414 tokenstart = PL_bufptr;
13422 if (SvCUR(stuff)) {
13425 PL_lex_state = LEX_NORMAL;
13426 start_force(PL_curforce);
13427 NEXTVAL_NEXTTOKE.ival = 0;
13431 PL_lex_state = LEX_FORMLINE;
13433 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
13435 else if (PL_encoding)
13436 sv_recode_to_utf8(stuff, PL_encoding);
13438 start_force(PL_curforce);
13439 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
13441 start_force(PL_curforce);
13442 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
13446 SvREFCNT_dec(stuff);
13448 PL_lex_formbrack = 0;
13452 if (PL_madskills) {
13454 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
13456 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
13457 PL_thiswhite = savewhite;
13464 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
13467 const I32 oldsavestack_ix = PL_savestack_ix;
13468 CV* const outsidecv = PL_compcv;
13471 assert(SvTYPE(PL_compcv) == SVt_PVCV);
13473 SAVEI32(PL_subline);
13474 save_item(PL_subname);
13475 SAVESPTR(PL_compcv);
13477 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
13478 CvFLAGS(PL_compcv) |= flags;
13480 PL_subline = CopLINE(PL_curcop);
13481 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
13482 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
13483 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
13485 return oldsavestack_ix;
13489 #pragma segment Perl_yylex
13492 S_yywarn(pTHX_ const char *const s)
13496 PERL_ARGS_ASSERT_YYWARN;
13498 PL_in_eval |= EVAL_WARNONLY;
13500 PL_in_eval &= ~EVAL_WARNONLY;
13505 Perl_yyerror(pTHX_ const char *const s)
13508 const char *where = NULL;
13509 const char *context = NULL;
13512 int yychar = PL_parser->yychar;
13514 PERL_ARGS_ASSERT_YYERROR;
13516 if (!yychar || (yychar == ';' && !PL_rsfp))
13518 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
13519 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
13520 PL_oldbufptr != PL_bufptr) {
13523 The code below is removed for NetWare because it abends/crashes on NetWare
13524 when the script has error such as not having the closing quotes like:
13525 if ($var eq "value)
13526 Checking of white spaces is anyway done in NetWare code.
13529 while (isSPACE(*PL_oldoldbufptr))
13532 context = PL_oldoldbufptr;
13533 contlen = PL_bufptr - PL_oldoldbufptr;
13535 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
13536 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
13539 The code below is removed for NetWare because it abends/crashes on NetWare
13540 when the script has error such as not having the closing quotes like:
13541 if ($var eq "value)
13542 Checking of white spaces is anyway done in NetWare code.
13545 while (isSPACE(*PL_oldbufptr))
13548 context = PL_oldbufptr;
13549 contlen = PL_bufptr - PL_oldbufptr;
13551 else if (yychar > 255)
13552 where = "next token ???";
13553 else if (yychar == -2) { /* YYEMPTY */
13554 if (PL_lex_state == LEX_NORMAL ||
13555 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
13556 where = "at end of line";
13557 else if (PL_lex_inpat)
13558 where = "within pattern";
13560 where = "within string";
13563 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
13565 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
13566 else if (isPRINT_LC(yychar)) {
13567 const char string = yychar;
13568 sv_catpvn(where_sv, &string, 1);
13571 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
13572 where = SvPVX_const(where_sv);
13574 msg = sv_2mortal(newSVpv(s, 0));
13575 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
13576 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
13578 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
13580 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
13581 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
13582 Perl_sv_catpvf(aTHX_ msg,
13583 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
13584 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
13587 if (PL_in_eval & EVAL_WARNONLY) {
13588 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
13592 if (PL_error_count >= 10) {
13593 if (PL_in_eval && SvCUR(ERRSV))
13594 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
13595 SVfARG(ERRSV), OutCopFILE(PL_curcop));
13597 Perl_croak(aTHX_ "%s has too many errors.\n",
13598 OutCopFILE(PL_curcop));
13601 PL_in_my_stash = NULL;
13605 #pragma segment Main
13609 S_swallow_bom(pTHX_ U8 *s)
13612 const STRLEN slen = SvCUR(PL_linestr);
13614 PERL_ARGS_ASSERT_SWALLOW_BOM;
13618 if (s[1] == 0xFE) {
13619 /* UTF-16 little-endian? (or UTF-32LE?) */
13620 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
13621 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
13622 #ifndef PERL_NO_UTF16_FILTER
13623 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
13625 if (PL_bufend > (char*)s) {
13626 s = add_utf16_textfilter(s, TRUE);
13629 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13634 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
13635 #ifndef PERL_NO_UTF16_FILTER
13636 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
13638 if (PL_bufend > (char *)s) {
13639 s = add_utf16_textfilter(s, FALSE);
13642 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13647 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
13648 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13649 s += 3; /* UTF-8 */
13655 if (s[2] == 0xFE && s[3] == 0xFF) {
13656 /* UTF-32 big-endian */
13657 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
13660 else if (s[2] == 0 && s[3] != 0) {
13663 * are a good indicator of UTF-16BE. */
13664 #ifndef PERL_NO_UTF16_FILTER
13665 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
13666 s = add_utf16_textfilter(s, FALSE);
13668 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13674 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
13675 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13676 s += 4; /* UTF-8 */
13682 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
13685 * are a good indicator of UTF-16LE. */
13686 #ifndef PERL_NO_UTF16_FILTER
13687 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
13688 s = add_utf16_textfilter(s, TRUE);
13690 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13698 #ifndef PERL_NO_UTF16_FILTER
13700 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
13703 SV *const filter = FILTER_DATA(idx);
13704 /* We re-use this each time round, throwing the contents away before we
13706 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
13707 SV *const utf8_buffer = filter;
13708 IV status = IoPAGE(filter);
13709 const bool reverse = cBOOL(IoLINES(filter));
13712 /* As we're automatically added, at the lowest level, and hence only called
13713 from this file, we can be sure that we're not called in block mode. Hence
13714 don't bother writing code to deal with block mode. */
13716 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
13719 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
13721 DEBUG_P(PerlIO_printf(Perl_debug_log,
13722 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13723 FPTR2DPTR(void *, S_utf16_textfilter),
13724 reverse ? 'l' : 'b', idx, maxlen, status,
13725 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13732 /* First, look in our buffer of existing UTF-8 data: */
13733 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
13737 } else if (status == 0) {
13739 IoPAGE(filter) = 0;
13740 nl = SvEND(utf8_buffer);
13743 STRLEN got = nl - SvPVX(utf8_buffer);
13744 /* Did we have anything to append? */
13746 sv_catpvn(sv, SvPVX(utf8_buffer), got);
13747 /* Everything else in this code works just fine if SVp_POK isn't
13748 set. This, however, needs it, and we need it to work, else
13749 we loop infinitely because the buffer is never consumed. */
13750 sv_chop(utf8_buffer, nl);
13754 /* OK, not a complete line there, so need to read some more UTF-16.
13755 Read an extra octect if the buffer currently has an odd number. */
13759 if (SvCUR(utf16_buffer) >= 2) {
13760 /* Location of the high octet of the last complete code point.
13761 Gosh, UTF-16 is a pain. All the benefits of variable length,
13762 *coupled* with all the benefits of partial reads and
13764 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
13765 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
13767 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
13771 /* We have the first half of a surrogate. Read more. */
13772 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
13775 status = FILTER_READ(idx + 1, utf16_buffer,
13776 160 + (SvCUR(utf16_buffer) & 1));
13777 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
13778 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
13781 IoPAGE(filter) = status;
13786 chars = SvCUR(utf16_buffer) >> 1;
13787 have = SvCUR(utf8_buffer);
13788 SvGROW(utf8_buffer, have + chars * 3 + 1);
13791 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
13792 (U8*)SvPVX_const(utf8_buffer) + have,
13793 chars * 2, &newlen);
13795 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
13796 (U8*)SvPVX_const(utf8_buffer) + have,
13797 chars * 2, &newlen);
13799 SvCUR_set(utf8_buffer, have + newlen);
13802 /* No need to keep this SV "well-formed" with a '\0' after the end, as
13803 it's private to us, and utf16_to_utf8{,reversed} take a
13804 (pointer,length) pair, rather than a NUL-terminated string. */
13805 if(SvCUR(utf16_buffer) & 1) {
13806 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
13807 SvCUR_set(utf16_buffer, 1);
13809 SvCUR_set(utf16_buffer, 0);
13812 DEBUG_P(PerlIO_printf(Perl_debug_log,
13813 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13815 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13816 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
13821 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
13823 SV *filter = filter_add(S_utf16_textfilter, NULL);
13825 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
13826 sv_setpvs(filter, "");
13827 IoLINES(filter) = reversed;
13828 IoPAGE(filter) = 1; /* Not EOF */
13830 /* Sadly, we have to return a valid pointer, come what may, so we have to
13831 ignore any error return from this. */
13832 SvCUR_set(PL_linestr, 0);
13833 if (FILTER_READ(0, PL_linestr, 0)) {
13834 SvUTF8_on(PL_linestr);
13836 SvUTF8_on(PL_linestr);
13838 PL_bufend = SvEND(PL_linestr);
13839 return (U8*)SvPVX(PL_linestr);
13844 Returns a pointer to the next character after the parsed
13845 vstring, as well as updating the passed in sv.
13847 Function must be called like
13850 s = scan_vstring(s,e,sv);
13852 where s and e are the start and end of the string.
13853 The sv should already be large enough to store the vstring
13854 passed in, for performance reasons.
13859 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
13862 const char *pos = s;
13863 const char *start = s;
13865 PERL_ARGS_ASSERT_SCAN_VSTRING;
13867 if (*pos == 'v') pos++; /* get past 'v' */
13868 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13870 if ( *pos != '.') {
13871 /* this may not be a v-string if followed by => */
13872 const char *next = pos;
13873 while (next < e && isSPACE(*next))
13875 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
13876 /* return string not v-string */
13877 sv_setpvn(sv,(char *)s,pos-s);
13878 return (char *)pos;
13882 if (!isALPHA(*pos)) {
13883 U8 tmpbuf[UTF8_MAXBYTES+1];
13886 s++; /* get past 'v' */
13891 /* this is atoi() that tolerates underscores */
13894 const char *end = pos;
13896 while (--end >= s) {
13898 const UV orev = rev;
13899 rev += (*end - '0') * mult;
13902 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13903 "Integer overflow in decimal number");
13907 if (rev > 0x7FFFFFFF)
13908 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
13910 /* Append native character for the rev point */
13911 tmpend = uvchr_to_utf8(tmpbuf, rev);
13912 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
13913 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
13915 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
13921 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13925 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
13932 Perl_keyword_plugin_standard(pTHX_
13933 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
13935 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
13936 PERL_UNUSED_CONTEXT;
13937 PERL_UNUSED_ARG(keyword_ptr);
13938 PERL_UNUSED_ARG(keyword_len);
13939 PERL_UNUSED_ARG(op_ptr);
13940 return KEYWORD_PLUGIN_DECLINE;
13945 * c-indentation-style: bsd
13946 * c-basic-offset: 4
13947 * indent-tabs-mode: t
13950 * ex: set ts=8 sts=4 sw=4 noet: