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] != ';' || !SvPOK(line)) {
718 /* avoid tie/overload weirdness */
719 parser->linestr = newSVpvn_flags(s, len, SvUTF8(line));
721 sv_catpvs(parser->linestr, "\n;");
724 SvREFCNT_inc_simple_void_NN(line);
725 parser->linestr = line;
727 parser->oldoldbufptr =
730 parser->linestart = SvPVX(parser->linestr);
731 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
732 parser->last_lop = parser->last_uni = NULL;
736 /* delete a parser object */
739 Perl_parser_free(pTHX_ const yy_parser *parser)
741 PERL_ARGS_ASSERT_PARSER_FREE;
743 PL_curcop = parser->saved_curcop;
744 SvREFCNT_dec(parser->linestr);
746 if (parser->rsfp == PerlIO_stdin())
747 PerlIO_clearerr(parser->rsfp);
748 else if (parser->rsfp && (!parser->old_parser ||
749 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
750 PerlIO_close(parser->rsfp);
751 SvREFCNT_dec(parser->rsfp_filters);
753 Safefree(parser->stack);
754 Safefree(parser->lex_brackstack);
755 Safefree(parser->lex_casestack);
756 PL_parser = parser->old_parser;
763 * Finalizer for lexing operations. Must be called when the parser is
764 * done with the lexer.
771 PL_doextract = FALSE;
775 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
777 Buffer scalar containing the chunk currently under consideration of the
778 text currently being lexed. This is always a plain string scalar (for
779 which C<SvPOK> is true). It is not intended to be used as a scalar by
780 normal scalar means; instead refer to the buffer directly by the pointer
781 variables described below.
783 The lexer maintains various C<char*> pointers to things in the
784 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
785 reallocated, all of these pointers must be updated. Don't attempt to
786 do this manually, but rather use L</lex_grow_linestr> if you need to
787 reallocate the buffer.
789 The content of the text chunk in the buffer is commonly exactly one
790 complete line of input, up to and including a newline terminator,
791 but there are situations where it is otherwise. The octets of the
792 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
793 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
794 flag on this scalar, which may disagree with it.
796 For direct examination of the buffer, the variable
797 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
798 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
799 of these pointers is usually preferable to examination of the scalar
800 through normal scalar means.
802 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
804 Direct pointer to the end of the chunk of text currently being lexed, the
805 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
806 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
807 always located at the end of the buffer, and does not count as part of
808 the buffer's contents.
810 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
812 Points to the current position of lexing inside the lexer buffer.
813 Characters around this point may be freely examined, within
814 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
815 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
816 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
818 Lexing code (whether in the Perl core or not) moves this pointer past
819 the characters that it consumes. It is also expected to perform some
820 bookkeeping whenever a newline character is consumed. This movement
821 can be more conveniently performed by the function L</lex_read_to>,
822 which handles newlines appropriately.
824 Interpretation of the buffer's octets can be abstracted out by
825 using the slightly higher-level functions L</lex_peek_unichar> and
826 L</lex_read_unichar>.
828 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
830 Points to the start of the current line inside the lexer buffer.
831 This is useful for indicating at which column an error occurred, and
832 not much else. This must be updated by any lexing code that consumes
833 a newline; the function L</lex_read_to> handles this detail.
839 =for apidoc Amx|bool|lex_bufutf8
841 Indicates whether the octets in the lexer buffer
842 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
843 of Unicode characters. If not, they should be interpreted as Latin-1
844 characters. This is analogous to the C<SvUTF8> flag for scalars.
846 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
847 contains valid UTF-8. Lexing code must be robust in the face of invalid
850 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
851 is significant, but not the whole story regarding the input character
852 encoding. Normally, when a file is being read, the scalar contains octets
853 and its C<SvUTF8> flag is off, but the octets should be interpreted as
854 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
855 however, the scalar may have the C<SvUTF8> flag on, and in this case its
856 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
857 is in effect. This logic may change in the future; use this function
858 instead of implementing the logic yourself.
864 Perl_lex_bufutf8(pTHX)
870 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
872 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
873 at least I<len> octets (including terminating NUL). Returns a
874 pointer to the reallocated buffer. This is necessary before making
875 any direct modification of the buffer that would increase its length.
876 L</lex_stuff_pvn> provides a more convenient way to insert text into
879 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
880 this function updates all of the lexer's variables that point directly
887 Perl_lex_grow_linestr(pTHX_ STRLEN len)
891 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
892 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
893 linestr = PL_parser->linestr;
894 buf = SvPVX(linestr);
895 if (len <= SvLEN(linestr))
897 bufend_pos = PL_parser->bufend - buf;
898 bufptr_pos = PL_parser->bufptr - buf;
899 oldbufptr_pos = PL_parser->oldbufptr - buf;
900 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
901 linestart_pos = PL_parser->linestart - buf;
902 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
903 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
904 buf = sv_grow(linestr, len);
905 PL_parser->bufend = buf + bufend_pos;
906 PL_parser->bufptr = buf + bufptr_pos;
907 PL_parser->oldbufptr = buf + oldbufptr_pos;
908 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
909 PL_parser->linestart = buf + linestart_pos;
910 if (PL_parser->last_uni)
911 PL_parser->last_uni = buf + last_uni_pos;
912 if (PL_parser->last_lop)
913 PL_parser->last_lop = buf + last_lop_pos;
918 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
920 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
921 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
922 reallocating the buffer if necessary. This means that lexing code that
923 runs later will see the characters as if they had appeared in the input.
924 It is not recommended to do this as part of normal parsing, and most
925 uses of this facility run the risk of the inserted characters being
926 interpreted in an unintended manner.
928 The string to be inserted is represented by I<len> octets starting
929 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
930 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
931 The characters are recoded for the lexer buffer, according to how the
932 buffer is currently being interpreted (L</lex_bufutf8>). If a string
933 to be interpreted is available as a Perl scalar, the L</lex_stuff_sv>
934 function is more convenient.
940 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
944 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
945 if (flags & ~(LEX_STUFF_UTF8))
946 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
948 if (flags & LEX_STUFF_UTF8) {
952 const char *p, *e = pv+len;
953 for (p = pv; p != e; p++)
954 highhalf += !!(((U8)*p) & 0x80);
957 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
958 bufptr = PL_parser->bufptr;
959 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
960 SvCUR_set(PL_parser->linestr,
961 SvCUR(PL_parser->linestr) + len+highhalf);
962 PL_parser->bufend += len+highhalf;
963 for (p = pv; p != e; p++) {
966 *bufptr++ = (char)(0xc0 | (c >> 6));
967 *bufptr++ = (char)(0x80 | (c & 0x3f));
974 if (flags & LEX_STUFF_UTF8) {
976 const char *p, *e = pv+len;
977 for (p = pv; p != e; p++) {
980 Perl_croak(aTHX_ "Lexing code attempted to stuff "
981 "non-Latin-1 character into Latin-1 input");
982 } else if (c >= 0xc2 && p+1 != e &&
983 (((U8)p[1]) & 0xc0) == 0x80) {
986 } else if (c >= 0x80) {
987 /* malformed UTF-8 */
989 SAVESPTR(PL_warnhook);
990 PL_warnhook = PERL_WARNHOOK_FATAL;
991 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
997 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
998 bufptr = PL_parser->bufptr;
999 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1000 SvCUR_set(PL_parser->linestr,
1001 SvCUR(PL_parser->linestr) + len-highhalf);
1002 PL_parser->bufend += len-highhalf;
1003 for (p = pv; p != e; p++) {
1006 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1009 *bufptr++ = (char)c;
1014 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1015 bufptr = PL_parser->bufptr;
1016 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1017 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1018 PL_parser->bufend += len;
1019 Copy(pv, bufptr, len, char);
1025 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1027 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1028 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1029 reallocating the buffer if necessary. This means that lexing code that
1030 runs later will see the characters as if they had appeared in the input.
1031 It is not recommended to do this as part of normal parsing, and most
1032 uses of this facility run the risk of the inserted characters being
1033 interpreted in an unintended manner.
1035 The string to be inserted is the string value of I<sv>. The characters
1036 are recoded for the lexer buffer, according to how the buffer is currently
1037 being interpreted (L</lex_bufutf8>). If a string to be interpreted is
1038 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1039 need to construct a scalar.
1045 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1049 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1051 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1053 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1057 =for apidoc Amx|void|lex_unstuff|char *ptr
1059 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1060 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1061 This hides the discarded text from any lexing code that runs later,
1062 as if the text had never appeared.
1064 This is not the normal way to consume lexed text. For that, use
1071 Perl_lex_unstuff(pTHX_ char *ptr)
1075 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1076 buf = PL_parser->bufptr;
1078 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1081 bufend = PL_parser->bufend;
1083 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1084 unstuff_len = ptr - buf;
1085 Move(ptr, buf, bufend+1-ptr, char);
1086 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1087 PL_parser->bufend = bufend - unstuff_len;
1091 =for apidoc Amx|void|lex_read_to|char *ptr
1093 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1094 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1095 performing the correct bookkeeping whenever a newline character is passed.
1096 This is the normal way to consume lexed text.
1098 Interpretation of the buffer's octets can be abstracted out by
1099 using the slightly higher-level functions L</lex_peek_unichar> and
1100 L</lex_read_unichar>.
1106 Perl_lex_read_to(pTHX_ char *ptr)
1109 PERL_ARGS_ASSERT_LEX_READ_TO;
1110 s = PL_parser->bufptr;
1111 if (ptr < s || ptr > PL_parser->bufend)
1112 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1113 for (; s != ptr; s++)
1115 CopLINE_inc(PL_curcop);
1116 PL_parser->linestart = s+1;
1118 PL_parser->bufptr = ptr;
1122 =for apidoc Amx|void|lex_discard_to|char *ptr
1124 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1125 up to I<ptr>. The remaining content of the buffer will be moved, and
1126 all pointers into the buffer updated appropriately. I<ptr> must not
1127 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1128 it is not permitted to discard text that has yet to be lexed.
1130 Normally it is not necessarily to do this directly, because it suffices to
1131 use the implicit discarding behaviour of L</lex_next_chunk> and things
1132 based on it. However, if a token stretches across multiple lines,
1133 and the lexing code has kept multiple lines of text in the buffer for
1134 that purpose, then after completion of the token it would be wise to
1135 explicitly discard the now-unneeded earlier lines, to avoid future
1136 multi-line tokens growing the buffer without bound.
1142 Perl_lex_discard_to(pTHX_ char *ptr)
1146 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1147 buf = SvPVX(PL_parser->linestr);
1149 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1152 if (ptr > PL_parser->bufptr)
1153 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1154 discard_len = ptr - buf;
1155 if (PL_parser->oldbufptr < ptr)
1156 PL_parser->oldbufptr = ptr;
1157 if (PL_parser->oldoldbufptr < ptr)
1158 PL_parser->oldoldbufptr = ptr;
1159 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1160 PL_parser->last_uni = NULL;
1161 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1162 PL_parser->last_lop = NULL;
1163 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1164 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1165 PL_parser->bufend -= discard_len;
1166 PL_parser->bufptr -= discard_len;
1167 PL_parser->oldbufptr -= discard_len;
1168 PL_parser->oldoldbufptr -= discard_len;
1169 if (PL_parser->last_uni)
1170 PL_parser->last_uni -= discard_len;
1171 if (PL_parser->last_lop)
1172 PL_parser->last_lop -= discard_len;
1176 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1178 Reads in the next chunk of text to be lexed, appending it to
1179 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1180 looked to the end of the current chunk and wants to know more. It is
1181 usual, but not necessary, for lexing to have consumed the entirety of
1182 the current chunk at this time.
1184 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1185 chunk (i.e., the current chunk has been entirely consumed), normally the
1186 current chunk will be discarded at the same time that the new chunk is
1187 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1188 will not be discarded. If the current chunk has not been entirely
1189 consumed, then it will not be discarded regardless of the flag.
1191 Returns true if some new text was added to the buffer, or false if the
1192 buffer has reached the end of the input text.
1197 #define LEX_FAKE_EOF 0x80000000
1200 Perl_lex_next_chunk(pTHX_ U32 flags)
1204 STRLEN old_bufend_pos, new_bufend_pos;
1205 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1206 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1207 bool got_some_for_debugger = 0;
1209 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1210 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1211 linestr = PL_parser->linestr;
1212 buf = SvPVX(linestr);
1213 if (!(flags & LEX_KEEP_PREVIOUS) &&
1214 PL_parser->bufptr == PL_parser->bufend) {
1215 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1217 if (PL_parser->last_uni != PL_parser->bufend)
1218 PL_parser->last_uni = NULL;
1219 if (PL_parser->last_lop != PL_parser->bufend)
1220 PL_parser->last_lop = NULL;
1221 last_uni_pos = last_lop_pos = 0;
1225 old_bufend_pos = PL_parser->bufend - buf;
1226 bufptr_pos = PL_parser->bufptr - buf;
1227 oldbufptr_pos = PL_parser->oldbufptr - buf;
1228 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1229 linestart_pos = PL_parser->linestart - buf;
1230 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1231 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1233 if (flags & LEX_FAKE_EOF) {
1235 } else if (!PL_parser->rsfp) {
1237 } else if (filter_gets(linestr, old_bufend_pos)) {
1239 got_some_for_debugger = 1;
1241 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1242 sv_setpvs(linestr, "");
1244 /* End of real input. Close filehandle (unless it was STDIN),
1245 * then add implicit termination.
1247 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1248 PerlIO_clearerr(PL_parser->rsfp);
1249 else if (PL_parser->rsfp)
1250 (void)PerlIO_close(PL_parser->rsfp);
1251 PL_parser->rsfp = NULL;
1252 PL_doextract = FALSE;
1254 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1257 if (!PL_in_eval && PL_minus_p) {
1259 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1260 PL_minus_n = PL_minus_p = 0;
1261 } else if (!PL_in_eval && PL_minus_n) {
1262 sv_catpvs(linestr, /*{*/";}");
1265 sv_catpvs(linestr, ";");
1268 buf = SvPVX(linestr);
1269 new_bufend_pos = SvCUR(linestr);
1270 PL_parser->bufend = buf + new_bufend_pos;
1271 PL_parser->bufptr = buf + bufptr_pos;
1272 PL_parser->oldbufptr = buf + oldbufptr_pos;
1273 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1274 PL_parser->linestart = buf + linestart_pos;
1275 if (PL_parser->last_uni)
1276 PL_parser->last_uni = buf + last_uni_pos;
1277 if (PL_parser->last_lop)
1278 PL_parser->last_lop = buf + last_lop_pos;
1279 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1280 PL_curstash != PL_debstash) {
1281 /* debugger active and we're not compiling the debugger code,
1282 * so store the line into the debugger's array of lines
1284 update_debugger_info(NULL, buf+old_bufend_pos,
1285 new_bufend_pos-old_bufend_pos);
1291 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1293 Looks ahead one (Unicode) character in the text currently being lexed.
1294 Returns the codepoint (unsigned integer value) of the next character,
1295 or -1 if lexing has reached the end of the input text. To consume the
1296 peeked character, use L</lex_read_unichar>.
1298 If the next character is in (or extends into) the next chunk of input
1299 text, the next chunk will be read in. Normally the current chunk will be
1300 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1301 then the current chunk will not be discarded.
1303 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1304 is encountered, an exception is generated.
1310 Perl_lex_peek_unichar(pTHX_ U32 flags)
1314 if (flags & ~(LEX_KEEP_PREVIOUS))
1315 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1316 s = PL_parser->bufptr;
1317 bufend = PL_parser->bufend;
1323 if (!lex_next_chunk(flags))
1325 s = PL_parser->bufptr;
1326 bufend = PL_parser->bufend;
1332 len = PL_utf8skip[head];
1333 while ((STRLEN)(bufend-s) < len) {
1334 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1336 s = PL_parser->bufptr;
1337 bufend = PL_parser->bufend;
1340 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1341 if (retlen == (STRLEN)-1) {
1342 /* malformed UTF-8 */
1344 SAVESPTR(PL_warnhook);
1345 PL_warnhook = PERL_WARNHOOK_FATAL;
1346 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1352 if (!lex_next_chunk(flags))
1354 s = PL_parser->bufptr;
1361 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1363 Reads the next (Unicode) character in the text currently being lexed.
1364 Returns the codepoint (unsigned integer value) of the character read,
1365 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1366 if lexing has reached the end of the input text. To non-destructively
1367 examine the next character, use L</lex_peek_unichar> instead.
1369 If the next character is in (or extends into) the next chunk of input
1370 text, the next chunk will be read in. Normally the current chunk will be
1371 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1372 then the current chunk will not be discarded.
1374 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1375 is encountered, an exception is generated.
1381 Perl_lex_read_unichar(pTHX_ U32 flags)
1384 if (flags & ~(LEX_KEEP_PREVIOUS))
1385 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1386 c = lex_peek_unichar(flags);
1389 CopLINE_inc(PL_curcop);
1390 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1396 =for apidoc Amx|void|lex_read_space|U32 flags
1398 Reads optional spaces, in Perl style, in the text currently being
1399 lexed. The spaces may include ordinary whitespace characters and
1400 Perl-style comments. C<#line> directives are processed if encountered.
1401 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1402 at a non-space character (or the end of the input text).
1404 If spaces extend into the next chunk of input text, the next chunk will
1405 be read in. Normally the current chunk will be discarded at the same
1406 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1407 chunk will not be discarded.
1412 #define LEX_NO_NEXT_CHUNK 0x80000000
1415 Perl_lex_read_space(pTHX_ U32 flags)
1418 bool need_incline = 0;
1419 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1420 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1423 sv_free(PL_skipwhite);
1424 PL_skipwhite = NULL;
1427 PL_skipwhite = newSVpvs("");
1428 #endif /* PERL_MAD */
1429 s = PL_parser->bufptr;
1430 bufend = PL_parser->bufend;
1436 } while (!(c == '\n' || (c == 0 && s == bufend)));
1437 } else if (c == '\n') {
1439 PL_parser->linestart = s;
1444 } else if (isSPACE(c)) {
1446 } else if (c == 0 && s == bufend) {
1450 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1451 #endif /* PERL_MAD */
1452 if (flags & LEX_NO_NEXT_CHUNK)
1454 PL_parser->bufptr = s;
1455 CopLINE_inc(PL_curcop);
1456 got_more = lex_next_chunk(flags);
1457 CopLINE_dec(PL_curcop);
1458 s = PL_parser->bufptr;
1459 bufend = PL_parser->bufend;
1462 if (need_incline && PL_parser->rsfp) {
1472 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1473 #endif /* PERL_MAD */
1474 PL_parser->bufptr = s;
1479 * This subroutine has nothing to do with tilting, whether at windmills
1480 * or pinball tables. Its name is short for "increment line". It
1481 * increments the current line number in CopLINE(PL_curcop) and checks
1482 * to see whether the line starts with a comment of the form
1483 * # line 500 "foo.pm"
1484 * If so, it sets the current line number and file to the values in the comment.
1488 S_incline(pTHX_ const char *s)
1495 PERL_ARGS_ASSERT_INCLINE;
1497 CopLINE_inc(PL_curcop);
1500 while (SPACE_OR_TAB(*s))
1502 if (strnEQ(s, "line", 4))
1506 if (SPACE_OR_TAB(*s))
1510 while (SPACE_OR_TAB(*s))
1518 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1520 while (SPACE_OR_TAB(*s))
1522 if (*s == '"' && (t = strchr(s+1, '"'))) {
1528 while (!isSPACE(*t))
1532 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1534 if (*e != '\n' && *e != '\0')
1535 return; /* false alarm */
1538 const STRLEN len = t - s;
1539 #ifndef USE_ITHREADS
1540 SV *const temp_sv = CopFILESV(PL_curcop);
1545 cf = SvPVX(temp_sv);
1546 tmplen = SvCUR(temp_sv);
1552 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
1553 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1554 * to *{"::_<newfilename"} */
1555 /* However, the long form of evals is only turned on by the
1556 debugger - usually they're "(eval %lu)" */
1560 STRLEN tmplen2 = len;
1561 if (tmplen + 2 <= sizeof smallbuf)
1564 Newx(tmpbuf, tmplen + 2, char);
1567 memcpy(tmpbuf + 2, cf, tmplen);
1569 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1574 if (tmplen2 + 2 <= sizeof smallbuf)
1577 Newx(tmpbuf2, tmplen2 + 2, char);
1579 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1580 /* Either they malloc'd it, or we malloc'd it,
1581 so no prefix is present in ours. */
1586 memcpy(tmpbuf2 + 2, s, tmplen2);
1589 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1591 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1592 /* adjust ${"::_<newfilename"} to store the new file name */
1593 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1594 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1595 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1598 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1600 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1603 CopFILE_free(PL_curcop);
1604 CopFILE_setn(PL_curcop, s, len);
1606 CopLINE_set(PL_curcop, atoi(n)-1);
1610 /* skip space before PL_thistoken */
1613 S_skipspace0(pTHX_ register char *s)
1615 PERL_ARGS_ASSERT_SKIPSPACE0;
1622 PL_thiswhite = newSVpvs("");
1623 sv_catsv(PL_thiswhite, PL_skipwhite);
1624 sv_free(PL_skipwhite);
1627 PL_realtokenstart = s - SvPVX(PL_linestr);
1631 /* skip space after PL_thistoken */
1634 S_skipspace1(pTHX_ register char *s)
1636 const char *start = s;
1637 I32 startoff = start - SvPVX(PL_linestr);
1639 PERL_ARGS_ASSERT_SKIPSPACE1;
1644 start = SvPVX(PL_linestr) + startoff;
1645 if (!PL_thistoken && PL_realtokenstart >= 0) {
1646 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1647 PL_thistoken = newSVpvn(tstart, start - tstart);
1649 PL_realtokenstart = -1;
1652 PL_nextwhite = newSVpvs("");
1653 sv_catsv(PL_nextwhite, PL_skipwhite);
1654 sv_free(PL_skipwhite);
1661 S_skipspace2(pTHX_ register char *s, SV **svp)
1664 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1665 const I32 startoff = s - SvPVX(PL_linestr);
1667 PERL_ARGS_ASSERT_SKIPSPACE2;
1670 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1671 if (!PL_madskills || !svp)
1673 start = SvPVX(PL_linestr) + startoff;
1674 if (!PL_thistoken && PL_realtokenstart >= 0) {
1675 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1676 PL_thistoken = newSVpvn(tstart, start - tstart);
1677 PL_realtokenstart = -1;
1681 *svp = newSVpvs("");
1682 sv_setsv(*svp, PL_skipwhite);
1683 sv_free(PL_skipwhite);
1692 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1694 AV *av = CopFILEAVx(PL_curcop);
1696 SV * const sv = newSV_type(SVt_PVMG);
1698 sv_setsv(sv, orig_sv);
1700 sv_setpvn(sv, buf, len);
1703 av_store(av, (I32)CopLINE(PL_curcop), sv);
1709 * Called to gobble the appropriate amount and type of whitespace.
1710 * Skips comments as well.
1714 S_skipspace(pTHX_ register char *s)
1718 #endif /* PERL_MAD */
1719 PERL_ARGS_ASSERT_SKIPSPACE;
1722 sv_free(PL_skipwhite);
1723 PL_skipwhite = NULL;
1725 #endif /* PERL_MAD */
1726 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1727 while (s < PL_bufend && SPACE_OR_TAB(*s))
1730 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1732 lex_read_space(LEX_KEEP_PREVIOUS |
1733 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1734 LEX_NO_NEXT_CHUNK : 0));
1736 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1737 if (PL_linestart > PL_bufptr)
1738 PL_bufptr = PL_linestart;
1743 PL_skipwhite = newSVpvn(start, s-start);
1744 #endif /* PERL_MAD */
1750 * Check the unary operators to ensure there's no ambiguity in how they're
1751 * used. An ambiguous piece of code would be:
1753 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1754 * the +5 is its argument.
1764 if (PL_oldoldbufptr != PL_last_uni)
1766 while (isSPACE(*PL_last_uni))
1769 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1771 if ((t = strchr(s, '(')) && t < PL_bufptr)
1774 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1775 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1776 (int)(s - PL_last_uni), PL_last_uni);
1780 * LOP : macro to build a list operator. Its behaviour has been replaced
1781 * with a subroutine, S_lop() for which LOP is just another name.
1784 #define LOP(f,x) return lop(f,x,s)
1788 * Build a list operator (or something that might be one). The rules:
1789 * - if we have a next token, then it's a list operator [why?]
1790 * - if the next thing is an opening paren, then it's a function
1791 * - else it's a list operator
1795 S_lop(pTHX_ I32 f, int x, char *s)
1799 PERL_ARGS_ASSERT_LOP;
1805 PL_last_lop = PL_oldbufptr;
1806 PL_last_lop_op = (OPCODE)f;
1809 return REPORT(LSTOP);
1812 return REPORT(LSTOP);
1815 return REPORT(FUNC);
1818 return REPORT(FUNC);
1820 return REPORT(LSTOP);
1826 * Sets up for an eventual force_next(). start_force(0) basically does
1827 * an unshift, while start_force(-1) does a push. yylex removes items
1832 S_start_force(pTHX_ int where)
1836 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1837 where = PL_lasttoke;
1838 assert(PL_curforce < 0 || PL_curforce == where);
1839 if (PL_curforce != where) {
1840 for (i = PL_lasttoke; i > where; --i) {
1841 PL_nexttoke[i] = PL_nexttoke[i-1];
1845 if (PL_curforce < 0) /* in case of duplicate start_force() */
1846 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1847 PL_curforce = where;
1850 curmad('^', newSVpvs(""));
1851 CURMAD('_', PL_nextwhite);
1856 S_curmad(pTHX_ char slot, SV *sv)
1862 if (PL_curforce < 0)
1863 where = &PL_thismad;
1865 where = &PL_nexttoke[PL_curforce].next_mad;
1871 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1873 else if (PL_encoding) {
1874 sv_recode_to_utf8(sv, PL_encoding);
1879 /* keep a slot open for the head of the list? */
1880 if (slot != '_' && *where && (*where)->mad_key == '^') {
1881 (*where)->mad_key = slot;
1882 sv_free(MUTABLE_SV(((*where)->mad_val)));
1883 (*where)->mad_val = (void*)sv;
1886 addmad(newMADsv(slot, sv), where, 0);
1889 # define start_force(where) NOOP
1890 # define curmad(slot, sv) NOOP
1895 * When the lexer realizes it knows the next token (for instance,
1896 * it is reordering tokens for the parser) then it can call S_force_next
1897 * to know what token to return the next time the lexer is called. Caller
1898 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1899 * and possibly PL_expect to ensure the lexer handles the token correctly.
1903 S_force_next(pTHX_ I32 type)
1908 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1909 tokereport(type, &NEXTVAL_NEXTTOKE);
1913 if (PL_curforce < 0)
1914 start_force(PL_lasttoke);
1915 PL_nexttoke[PL_curforce].next_type = type;
1916 if (PL_lex_state != LEX_KNOWNEXT)
1917 PL_lex_defer = PL_lex_state;
1918 PL_lex_state = LEX_KNOWNEXT;
1919 PL_lex_expect = PL_expect;
1922 PL_nexttype[PL_nexttoke] = type;
1924 if (PL_lex_state != LEX_KNOWNEXT) {
1925 PL_lex_defer = PL_lex_state;
1926 PL_lex_expect = PL_expect;
1927 PL_lex_state = LEX_KNOWNEXT;
1933 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1936 SV * const sv = newSVpvn_utf8(start, len,
1939 && !is_ascii_string((const U8*)start, len)
1940 && is_utf8_string((const U8*)start, len));
1946 * When the lexer knows the next thing is a word (for instance, it has
1947 * just seen -> and it knows that the next char is a word char, then
1948 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1952 * char *start : buffer position (must be within PL_linestr)
1953 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1954 * int check_keyword : if true, Perl checks to make sure the word isn't
1955 * a keyword (do this if the word is a label, e.g. goto FOO)
1956 * int allow_pack : if true, : characters will also be allowed (require,
1957 * use, etc. do this)
1958 * int allow_initial_tick : used by the "sub" lexer only.
1962 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1968 PERL_ARGS_ASSERT_FORCE_WORD;
1970 start = SKIPSPACE1(start);
1972 if (isIDFIRST_lazy_if(s,UTF) ||
1973 (allow_pack && *s == ':') ||
1974 (allow_initial_tick && *s == '\'') )
1976 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1977 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1979 start_force(PL_curforce);
1981 curmad('X', newSVpvn(start,s-start));
1982 if (token == METHOD) {
1987 PL_expect = XOPERATOR;
1991 curmad('g', newSVpvs( "forced" ));
1992 NEXTVAL_NEXTTOKE.opval
1993 = (OP*)newSVOP(OP_CONST,0,
1994 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1995 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2003 * Called when the lexer wants $foo *foo &foo etc, but the program
2004 * text only contains the "foo" portion. The first argument is a pointer
2005 * to the "foo", and the second argument is the type symbol to prefix.
2006 * Forces the next token to be a "WORD".
2007 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2011 S_force_ident(pTHX_ register const char *s, int kind)
2015 PERL_ARGS_ASSERT_FORCE_IDENT;
2018 const STRLEN len = strlen(s);
2019 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
2020 start_force(PL_curforce);
2021 NEXTVAL_NEXTTOKE.opval = o;
2024 o->op_private = OPpCONST_ENTERED;
2025 /* XXX see note in pp_entereval() for why we forgo typo
2026 warnings if the symbol must be introduced in an eval.
2028 gv_fetchpvn_flags(s, len,
2029 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2031 kind == '$' ? SVt_PV :
2032 kind == '@' ? SVt_PVAV :
2033 kind == '%' ? SVt_PVHV :
2041 Perl_str_to_version(pTHX_ SV *sv)
2046 const char *start = SvPV_const(sv,len);
2047 const char * const end = start + len;
2048 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2050 PERL_ARGS_ASSERT_STR_TO_VERSION;
2052 while (start < end) {
2056 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2061 retval += ((NV)n)/nshift;
2070 * Forces the next token to be a version number.
2071 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2072 * and if "guessing" is TRUE, then no new token is created (and the caller
2073 * must use an alternative parsing method).
2077 S_force_version(pTHX_ char *s, int guessing)
2083 I32 startoff = s - SvPVX(PL_linestr);
2086 PERL_ARGS_ASSERT_FORCE_VERSION;
2094 while (isDIGIT(*d) || *d == '_' || *d == '.')
2098 start_force(PL_curforce);
2099 curmad('X', newSVpvn(s,d-s));
2102 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2104 #ifdef USE_LOCALE_NUMERIC
2105 char *loc = setlocale(LC_NUMERIC, "C");
2107 s = scan_num(s, &pl_yylval);
2108 #ifdef USE_LOCALE_NUMERIC
2109 setlocale(LC_NUMERIC, loc);
2111 version = pl_yylval.opval;
2112 ver = cSVOPx(version)->op_sv;
2113 if (SvPOK(ver) && !SvNIOK(ver)) {
2114 SvUPGRADE(ver, SVt_PVNV);
2115 SvNV_set(ver, str_to_version(ver));
2116 SvNOK_on(ver); /* hint that it is a version */
2119 else if (guessing) {
2122 sv_free(PL_nextwhite); /* let next token collect whitespace */
2124 s = SvPVX(PL_linestr) + startoff;
2132 if (PL_madskills && !version) {
2133 sv_free(PL_nextwhite); /* let next token collect whitespace */
2135 s = SvPVX(PL_linestr) + startoff;
2138 /* NOTE: The parser sees the package name and the VERSION swapped */
2139 start_force(PL_curforce);
2140 NEXTVAL_NEXTTOKE.opval = version;
2147 * S_force_strict_version
2148 * Forces the next token to be a version number using strict syntax rules.
2152 S_force_strict_version(pTHX_ char *s)
2157 I32 startoff = s - SvPVX(PL_linestr);
2159 const char *errstr = NULL;
2161 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2163 while (isSPACE(*s)) /* leading whitespace */
2166 if (is_STRICT_VERSION(s,&errstr)) {
2168 s = (char *)scan_version(s, ver, 0);
2169 version = newSVOP(OP_CONST, 0, ver);
2171 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2172 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2176 yyerror(errstr); /* version required */
2181 if (PL_madskills && !version) {
2182 sv_free(PL_nextwhite); /* let next token collect whitespace */
2184 s = SvPVX(PL_linestr) + startoff;
2187 /* NOTE: The parser sees the package name and the VERSION swapped */
2188 start_force(PL_curforce);
2189 NEXTVAL_NEXTTOKE.opval = version;
2197 * Tokenize a quoted string passed in as an SV. It finds the next
2198 * chunk, up to end of string or a backslash. It may make a new
2199 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2204 S_tokeq(pTHX_ SV *sv)
2208 register char *send;
2213 PERL_ARGS_ASSERT_TOKEQ;
2218 s = SvPV_force(sv, len);
2219 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2222 while (s < send && *s != '\\')
2227 if ( PL_hints & HINT_NEW_STRING ) {
2228 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2232 if (s + 1 < send && (s[1] == '\\'))
2233 s++; /* all that, just for this */
2238 SvCUR_set(sv, d - SvPVX_const(sv));
2240 if ( PL_hints & HINT_NEW_STRING )
2241 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2246 * Now come three functions related to double-quote context,
2247 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2248 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2249 * interact with PL_lex_state, and create fake ( ... ) argument lists
2250 * to handle functions and concatenation.
2251 * They assume that whoever calls them will be setting up a fake
2252 * join call, because each subthing puts a ',' after it. This lets
2255 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2257 * (I'm not sure whether the spurious commas at the end of lcfirst's
2258 * arguments and join's arguments are created or not).
2263 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2265 * Pattern matching will set PL_lex_op to the pattern-matching op to
2266 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2268 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2270 * Everything else becomes a FUNC.
2272 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2273 * had an OP_CONST or OP_READLINE). This just sets us up for a
2274 * call to S_sublex_push().
2278 S_sublex_start(pTHX)
2281 register const I32 op_type = pl_yylval.ival;
2283 if (op_type == OP_NULL) {
2284 pl_yylval.opval = PL_lex_op;
2288 if (op_type == OP_CONST || op_type == OP_READLINE) {
2289 SV *sv = tokeq(PL_lex_stuff);
2291 if (SvTYPE(sv) == SVt_PVIV) {
2292 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2294 const char * const p = SvPV_const(sv, len);
2295 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2299 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2300 PL_lex_stuff = NULL;
2301 /* Allow <FH> // "foo" */
2302 if (op_type == OP_READLINE)
2303 PL_expect = XTERMORDORDOR;
2306 else if (op_type == OP_BACKTICK && PL_lex_op) {
2307 /* readpipe() vas overriden */
2308 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2309 pl_yylval.opval = PL_lex_op;
2311 PL_lex_stuff = NULL;
2315 PL_sublex_info.super_state = PL_lex_state;
2316 PL_sublex_info.sub_inwhat = (U16)op_type;
2317 PL_sublex_info.sub_op = PL_lex_op;
2318 PL_lex_state = LEX_INTERPPUSH;
2322 pl_yylval.opval = PL_lex_op;
2332 * Create a new scope to save the lexing state. The scope will be
2333 * ended in S_sublex_done. Returns a '(', starting the function arguments
2334 * to the uc, lc, etc. found before.
2335 * Sets PL_lex_state to LEX_INTERPCONCAT.
2344 PL_lex_state = PL_sublex_info.super_state;
2345 SAVEBOOL(PL_lex_dojoin);
2346 SAVEI32(PL_lex_brackets);
2347 SAVEI32(PL_lex_casemods);
2348 SAVEI32(PL_lex_starts);
2349 SAVEI8(PL_lex_state);
2350 SAVEVPTR(PL_lex_inpat);
2351 SAVEI16(PL_lex_inwhat);
2352 SAVECOPLINE(PL_curcop);
2353 SAVEPPTR(PL_bufptr);
2354 SAVEPPTR(PL_bufend);
2355 SAVEPPTR(PL_oldbufptr);
2356 SAVEPPTR(PL_oldoldbufptr);
2357 SAVEPPTR(PL_last_lop);
2358 SAVEPPTR(PL_last_uni);
2359 SAVEPPTR(PL_linestart);
2360 SAVESPTR(PL_linestr);
2361 SAVEGENERICPV(PL_lex_brackstack);
2362 SAVEGENERICPV(PL_lex_casestack);
2364 PL_linestr = PL_lex_stuff;
2365 PL_lex_stuff = NULL;
2367 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2368 = SvPVX(PL_linestr);
2369 PL_bufend += SvCUR(PL_linestr);
2370 PL_last_lop = PL_last_uni = NULL;
2371 SAVEFREESV(PL_linestr);
2373 PL_lex_dojoin = FALSE;
2374 PL_lex_brackets = 0;
2375 Newx(PL_lex_brackstack, 120, char);
2376 Newx(PL_lex_casestack, 12, char);
2377 PL_lex_casemods = 0;
2378 *PL_lex_casestack = '\0';
2380 PL_lex_state = LEX_INTERPCONCAT;
2381 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2383 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2384 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2385 PL_lex_inpat = PL_sublex_info.sub_op;
2387 PL_lex_inpat = NULL;
2394 * Restores lexer state after a S_sublex_push.
2401 if (!PL_lex_starts++) {
2402 SV * const sv = newSVpvs("");
2403 if (SvUTF8(PL_linestr))
2405 PL_expect = XOPERATOR;
2406 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2410 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2411 PL_lex_state = LEX_INTERPCASEMOD;
2415 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2416 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2417 PL_linestr = PL_lex_repl;
2419 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2420 PL_bufend += SvCUR(PL_linestr);
2421 PL_last_lop = PL_last_uni = NULL;
2422 SAVEFREESV(PL_linestr);
2423 PL_lex_dojoin = FALSE;
2424 PL_lex_brackets = 0;
2425 PL_lex_casemods = 0;
2426 *PL_lex_casestack = '\0';
2428 if (SvEVALED(PL_lex_repl)) {
2429 PL_lex_state = LEX_INTERPNORMAL;
2431 /* we don't clear PL_lex_repl here, so that we can check later
2432 whether this is an evalled subst; that means we rely on the
2433 logic to ensure sublex_done() is called again only via the
2434 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2437 PL_lex_state = LEX_INTERPCONCAT;
2447 PL_endwhite = newSVpvs("");
2448 sv_catsv(PL_endwhite, PL_thiswhite);
2452 sv_setpvs(PL_thistoken,"");
2454 PL_realtokenstart = -1;
2458 PL_bufend = SvPVX(PL_linestr);
2459 PL_bufend += SvCUR(PL_linestr);
2460 PL_expect = XOPERATOR;
2461 PL_sublex_info.sub_inwhat = 0;
2469 Extracts a pattern, double-quoted string, or transliteration. This
2472 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2473 processing a pattern (PL_lex_inpat is true), a transliteration
2474 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2476 Returns a pointer to the character scanned up to. If this is
2477 advanced from the start pointer supplied (i.e. if anything was
2478 successfully parsed), will leave an OP for the substring scanned
2479 in pl_yylval. Caller must intuit reason for not parsing further
2480 by looking at the next characters herself.
2484 constants: \N{NAME} only
2485 case and quoting: \U \Q \E
2486 stops on @ and $, but not for $ as tail anchor
2488 In transliterations:
2489 characters are VERY literal, except for - not at the start or end
2490 of the string, which indicates a range. If the range is in bytes,
2491 scan_const expands the range to the full set of intermediate
2492 characters. If the range is in utf8, the hyphen is replaced with
2493 a certain range mark which will be handled by pmtrans() in op.c.
2495 In double-quoted strings:
2497 double-quoted style: \r and \n
2498 constants: \x31, etc.
2499 deprecated backrefs: \1 (in substitution replacements)
2500 case and quoting: \U \Q \E
2503 scan_const does *not* construct ops to handle interpolated strings.
2504 It stops processing as soon as it finds an embedded $ or @ variable
2505 and leaves it to the caller to work out what's going on.
2507 embedded arrays (whether in pattern or not) could be:
2508 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2510 $ in double-quoted strings must be the symbol of an embedded scalar.
2512 $ in pattern could be $foo or could be tail anchor. Assumption:
2513 it's a tail anchor if $ is the last thing in the string, or if it's
2514 followed by one of "()| \r\n\t"
2516 \1 (backreferences) are turned into $1
2518 The structure of the code is
2519 while (there's a character to process) {
2520 handle transliteration ranges
2521 skip regexp comments /(?#comment)/ and codes /(?{code})/
2522 skip #-initiated comments in //x patterns
2523 check for embedded arrays
2524 check for embedded scalars
2526 deprecate \1 in substitution replacements
2527 handle string-changing backslashes \l \U \Q \E, etc.
2528 switch (what was escaped) {
2529 handle \- in a transliteration (becomes a literal -)
2530 if a pattern and not \N{, go treat as regular character
2531 handle \132 (octal characters)
2532 handle \x15 and \x{1234} (hex characters)
2533 handle \N{name} (named characters, also \N{3,5} in a pattern)
2534 handle \cV (control characters)
2535 handle printf-style backslashes (\f, \r, \n, etc)
2538 } (end if backslash)
2539 handle regular character
2540 } (end while character to read)
2545 S_scan_const(pTHX_ char *start)
2548 register char *send = PL_bufend; /* end of the constant */
2549 SV *sv = newSV(send - start); /* sv for the constant. See
2550 note below on sizing. */
2551 register char *s = start; /* start of the constant */
2552 register char *d = SvPVX(sv); /* destination for copies */
2553 bool dorange = FALSE; /* are we in a translit range? */
2554 bool didrange = FALSE; /* did we just finish a range? */
2555 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
2556 I32 this_utf8 = UTF; /* Is the source string assumed
2557 to be UTF8? But, this can
2558 show as true when the source
2559 isn't utf8, as for example
2560 when it is entirely composed
2563 /* Note on sizing: The scanned constant is placed into sv, which is
2564 * initialized by newSV() assuming one byte of output for every byte of
2565 * input. This routine expects newSV() to allocate an extra byte for a
2566 * trailing NUL, which this routine will append if it gets to the end of
2567 * the input. There may be more bytes of input than output (eg., \N{LATIN
2568 * CAPITAL LETTER A}), or more output than input if the constant ends up
2569 * recoded to utf8, but each time a construct is found that might increase
2570 * the needed size, SvGROW() is called. Its size parameter each time is
2571 * based on the best guess estimate at the time, namely the length used so
2572 * far, plus the length the current construct will occupy, plus room for
2573 * the trailing NUL, plus one byte for every input byte still unscanned */
2577 UV literal_endpoint = 0;
2578 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2581 PERL_ARGS_ASSERT_SCAN_CONST;
2583 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2584 /* If we are doing a trans and we know we want UTF8 set expectation */
2585 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2586 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2590 while (s < send || dorange) {
2592 /* get transliterations out of the way (they're most literal) */
2593 if (PL_lex_inwhat == OP_TRANS) {
2594 /* expand a range A-Z to the full set of characters. AIE! */
2596 I32 i; /* current expanded character */
2597 I32 min; /* first character in range */
2598 I32 max; /* last character in range */
2609 char * const c = (char*)utf8_hop((U8*)d, -1);
2613 *c = (char)UTF_TO_NATIVE(0xff);
2614 /* mark the range as done, and continue */
2620 i = d - SvPVX_const(sv); /* remember current offset */
2623 SvLEN(sv) + (has_utf8 ?
2624 (512 - UTF_CONTINUATION_MARK +
2627 /* How many two-byte within 0..255: 128 in UTF-8,
2628 * 96 in UTF-8-mod. */
2630 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2632 d = SvPVX(sv) + i; /* refresh d after realloc */
2636 for (j = 0; j <= 1; j++) {
2637 char * const c = (char*)utf8_hop((U8*)d, -1);
2638 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2644 max = (U8)0xff; /* only to \xff */
2645 uvmax = uv; /* \x{100} to uvmax */
2647 d = c; /* eat endpoint chars */
2652 d -= 2; /* eat the first char and the - */
2653 min = (U8)*d; /* first char in range */
2654 max = (U8)d[1]; /* last char in range */
2661 "Invalid range \"%c-%c\" in transliteration operator",
2662 (char)min, (char)max);
2666 if (literal_endpoint == 2 &&
2667 ((isLOWER(min) && isLOWER(max)) ||
2668 (isUPPER(min) && isUPPER(max)))) {
2670 for (i = min; i <= max; i++)
2672 *d++ = NATIVE_TO_NEED(has_utf8,i);
2674 for (i = min; i <= max; i++)
2676 *d++ = NATIVE_TO_NEED(has_utf8,i);
2681 for (i = min; i <= max; i++)
2684 const U8 ch = (U8)NATIVE_TO_UTF(i);
2685 if (UNI_IS_INVARIANT(ch))
2688 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2689 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2698 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2700 *d++ = (char)UTF_TO_NATIVE(0xff);
2702 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2706 /* mark the range as done, and continue */
2710 literal_endpoint = 0;
2715 /* range begins (ignore - as first or last char) */
2716 else if (*s == '-' && s+1 < send && s != start) {
2718 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2725 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2735 literal_endpoint = 0;
2736 native_range = TRUE;
2741 /* if we get here, we're not doing a transliteration */
2743 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2744 except for the last char, which will be done separately. */
2745 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2747 while (s+1 < send && *s != ')')
2748 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2750 else if (s[2] == '{' /* This should match regcomp.c */
2751 || (s[2] == '?' && s[3] == '{'))
2754 char *regparse = s + (s[2] == '{' ? 3 : 4);
2757 while (count && (c = *regparse)) {
2758 if (c == '\\' && regparse[1])
2766 if (*regparse != ')')
2767 regparse--; /* Leave one char for continuation. */
2768 while (s < regparse)
2769 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2773 /* likewise skip #-initiated comments in //x patterns */
2774 else if (*s == '#' && PL_lex_inpat &&
2775 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2776 while (s+1 < send && *s != '\n')
2777 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2780 /* check for embedded arrays
2781 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2783 else if (*s == '@' && s[1]) {
2784 if (isALNUM_lazy_if(s+1,UTF))
2786 if (strchr(":'{$", s[1]))
2788 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2789 break; /* in regexp, neither @+ nor @- are interpolated */
2792 /* check for embedded scalars. only stop if we're sure it's a
2795 else if (*s == '$') {
2796 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2798 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2800 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2801 "Possible unintended interpolation of $\\ in regex");
2803 break; /* in regexp, $ might be tail anchor */
2807 /* End of else if chain - OP_TRANS rejoin rest */
2810 if (*s == '\\' && s+1 < send) {
2811 char* e; /* Can be used for ending '}', etc. */
2815 /* warn on \1 - \9 in substitution replacements, but note that \11
2816 * is an octal; and \19 is \1 followed by '9' */
2817 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2818 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2820 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2825 /* string-change backslash escapes */
2826 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2830 /* In a pattern, process \N, but skip any other backslash escapes.
2831 * This is because we don't want to translate an escape sequence
2832 * into a meta symbol and have the regex compiler use the meta
2833 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2834 * in spite of this, we do have to process \N here while the proper
2835 * charnames handler is in scope. See bugs #56444 and #62056.
2836 * There is a complication because \N in a pattern may also stand
2837 * for 'match a non-nl', and not mean a charname, in which case its
2838 * processing should be deferred to the regex compiler. To be a
2839 * charname it must be followed immediately by a '{', and not look
2840 * like \N followed by a curly quantifier, i.e., not something like
2841 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2843 else if (PL_lex_inpat
2846 || regcurly(s + 1)))
2848 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2849 goto default_action;
2854 /* quoted - in transliterations */
2856 if (PL_lex_inwhat == OP_TRANS) {
2863 if ((isALPHA(*s) || isDIGIT(*s)))
2864 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2865 "Unrecognized escape \\%c passed through",
2867 /* default action is to copy the quoted character */
2868 goto default_action;
2871 /* eg. \132 indicates the octal constant 0132 */
2872 case '0': case '1': case '2': case '3':
2873 case '4': case '5': case '6': case '7':
2877 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2880 goto NUM_ESCAPE_INSERT;
2882 /* eg. \o{24} indicates the octal constant \024 */
2888 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
2894 goto NUM_ESCAPE_INSERT;
2897 /* eg. \x24 indicates the hex constant 0x24 */
2901 char* const e = strchr(s, '}');
2902 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2903 PERL_SCAN_DISALLOW_PREFIX;
2908 yyerror("Missing right brace on \\x{}");
2912 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2918 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2919 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2925 /* Insert oct or hex escaped character. There will always be
2926 * enough room in sv since such escapes will be longer than any
2927 * UTF-8 sequence they can end up as, except if they force us
2928 * to recode the rest of the string into utf8 */
2930 /* Here uv is the ordinal of the next character being added in
2931 * unicode (converted from native). */
2932 if (!UNI_IS_INVARIANT(uv)) {
2933 if (!has_utf8 && uv > 255) {
2934 /* Might need to recode whatever we have accumulated so
2935 * far if it contains any chars variant in utf8 or
2938 SvCUR_set(sv, d - SvPVX_const(sv));
2941 /* See Note on sizing above. */
2942 sv_utf8_upgrade_flags_grow(sv,
2943 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2944 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2945 d = SvPVX(sv) + SvCUR(sv);
2950 d = (char*)uvuni_to_utf8((U8*)d, uv);
2951 if (PL_lex_inwhat == OP_TRANS &&
2952 PL_sublex_info.sub_op) {
2953 PL_sublex_info.sub_op->op_private |=
2954 (PL_lex_repl ? OPpTRANS_FROM_UTF
2958 if (uv > 255 && !dorange)
2959 native_range = FALSE;
2972 /* In a non-pattern \N must be a named character, like \N{LATIN
2973 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
2974 * mean to match a non-newline. For non-patterns, named
2975 * characters are converted to their string equivalents. In
2976 * patterns, named characters are not converted to their
2977 * ultimate forms for the same reasons that other escapes
2978 * aren't. Instead, they are converted to the \N{U+...} form
2979 * to get the value from the charnames that is in effect right
2980 * now, while preserving the fact that it was a named character
2981 * so that the regex compiler knows this */
2983 /* This section of code doesn't generally use the
2984 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
2985 * a close examination of this macro and determined it is a
2986 * no-op except on utfebcdic variant characters. Every
2987 * character generated by this that would normally need to be
2988 * enclosed by this macro is invariant, so the macro is not
2989 * needed, and would complicate use of copy(). There are other
2990 * parts of this file where the macro is used inconsistently,
2991 * but are saved by it being a no-op */
2993 /* The structure of this section of code (besides checking for
2994 * errors and upgrading to utf8) is:
2995 * Further disambiguate between the two meanings of \N, and if
2996 * not a charname, go process it elsewhere
2997 * If of form \N{U+...}, pass it through if a pattern;
2998 * otherwise convert to utf8
2999 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3000 * pattern; otherwise convert to utf8 */
3002 /* Here, s points to the 'N'; the test below is guaranteed to
3003 * succeed if we are being called on a pattern as we already
3004 * know from a test above that the next character is a '{'.
3005 * On a non-pattern \N must mean 'named sequence, which
3006 * requires braces */
3009 yyerror("Missing braces on \\N{}");
3014 /* If there is no matching '}', it is an error. */
3015 if (! (e = strchr(s, '}'))) {
3016 if (! PL_lex_inpat) {
3017 yyerror("Missing right brace on \\N{}");
3019 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3024 /* Here it looks like a named character */
3028 /* XXX This block is temporary code. \N{} implies that the
3029 * pattern is to have Unicode semantics, and therefore
3030 * currently has to be encoded in utf8. By putting it in
3031 * utf8 now, we save a whole pass in the regular expression
3032 * compiler. Once that code is changed so Unicode
3033 * semantics doesn't necessarily have to be in utf8, this
3034 * block should be removed */
3036 SvCUR_set(sv, d - SvPVX_const(sv));
3039 /* See Note on sizing above. */
3040 sv_utf8_upgrade_flags_grow(sv,
3041 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3042 /* 5 = '\N{' + cur char + NUL */
3043 (STRLEN)(send - s) + 5);
3044 d = SvPVX(sv) + SvCUR(sv);
3049 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3050 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3051 | PERL_SCAN_DISALLOW_PREFIX;
3054 /* For \N{U+...}, the '...' is a unicode value even on
3055 * EBCDIC machines */
3056 s += 2; /* Skip to next char after the 'U+' */
3058 uv = grok_hex(s, &len, &flags, NULL);
3059 if (len == 0 || len != (STRLEN)(e - s)) {
3060 yyerror("Invalid hexadecimal number in \\N{U+...}");
3067 /* Pass through to the regex compiler unchanged. The
3068 * reason we evaluated the number above is to make sure
3069 * there wasn't a syntax error. */
3070 s -= 5; /* Include the '\N{U+' */
3071 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3074 else { /* Not a pattern: convert the hex to string */
3076 /* If destination is not in utf8, unconditionally
3077 * recode it to be so. This is because \N{} implies
3078 * Unicode semantics, and scalars have to be in utf8
3079 * to guarantee those semantics */
3081 SvCUR_set(sv, d - SvPVX_const(sv));
3084 /* See Note on sizing above. */
3085 sv_utf8_upgrade_flags_grow(
3087 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3088 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3089 d = SvPVX(sv) + SvCUR(sv);
3093 /* Add the string to the output */
3094 if (UNI_IS_INVARIANT(uv)) {
3097 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3100 else { /* Here is \N{NAME} but not \N{U+...}. */
3102 SV *res; /* result from charnames */
3103 const char *str; /* the string in 'res' */
3104 STRLEN len; /* its length */
3106 /* Get the value for NAME */
3107 res = newSVpvn(s, e - s);
3108 res = new_constant( NULL, 0, "charnames",
3109 /* includes all of: \N{...} */
3110 res, NULL, s - 3, e - s + 4 );
3112 /* Most likely res will be in utf8 already since the
3113 * standard charnames uses pack U, but a custom translator
3114 * can leave it otherwise, so make sure. XXX This can be
3115 * revisited to not have charnames use utf8 for characters
3116 * that don't need it when regexes don't have to be in utf8
3117 * for Unicode semantics. If doing so, remember EBCDIC */
3118 sv_utf8_upgrade(res);
3119 str = SvPV_const(res, len);
3121 /* Don't accept malformed input */
3122 if (! is_utf8_string((U8 *) str, len)) {
3123 yyerror("Malformed UTF-8 returned by \\N");
3125 else if (PL_lex_inpat) {
3127 if (! len) { /* The name resolved to an empty string */
3128 Copy("\\N{}", d, 4, char);
3132 /* In order to not lose information for the regex
3133 * compiler, pass the result in the specially made
3134 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3135 * the code points in hex of each character
3136 * returned by charnames */
3138 const char *str_end = str + len;
3139 STRLEN char_length; /* cur char's byte length */
3140 STRLEN output_length; /* and the number of bytes
3141 after this is translated
3143 const STRLEN off = d - SvPVX_const(sv);
3145 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3146 * max('U+', '.'); and 1 for NUL */
3147 char hex_string[2 * UTF8_MAXBYTES + 5];
3149 /* Get the first character of the result. */
3150 U32 uv = utf8n_to_uvuni((U8 *) str,
3155 /* The call to is_utf8_string() above hopefully
3156 * guarantees that there won't be an error. But
3157 * it's easy here to make sure. The function just
3158 * above warns and returns 0 if invalid utf8, but
3159 * it can also return 0 if the input is validly a
3160 * NUL. Disambiguate */
3161 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3162 uv = UNICODE_REPLACEMENT;
3165 /* Convert first code point to hex, including the
3166 * boiler plate before it */
3167 sprintf(hex_string, "\\N{U+%X", (unsigned int) uv);
3168 output_length = strlen(hex_string);
3170 /* Make sure there is enough space to hold it */
3171 d = off + SvGROW(sv, off
3173 + (STRLEN)(send - e)
3174 + 2); /* '}' + NUL */
3176 Copy(hex_string, d, output_length, char);
3179 /* For each subsequent character, append dot and
3180 * its ordinal in hex */
3181 while ((str += char_length) < str_end) {
3182 const STRLEN off = d - SvPVX_const(sv);
3183 U32 uv = utf8n_to_uvuni((U8 *) str,
3187 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3188 uv = UNICODE_REPLACEMENT;
3191 sprintf(hex_string, ".%X", (unsigned int) uv);
3192 output_length = strlen(hex_string);
3194 d = off + SvGROW(sv, off
3196 + (STRLEN)(send - e)
3197 + 2); /* '}' + NUL */
3198 Copy(hex_string, d, output_length, char);
3202 *d++ = '}'; /* Done. Add the trailing brace */
3205 else { /* Here, not in a pattern. Convert the name to a
3208 /* If destination is not in utf8, unconditionally
3209 * recode it to be so. This is because \N{} implies
3210 * Unicode semantics, and scalars have to be in utf8
3211 * to guarantee those semantics */
3213 SvCUR_set(sv, d - SvPVX_const(sv));
3216 /* See Note on sizing above. */
3217 sv_utf8_upgrade_flags_grow(sv,
3218 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3219 len + (STRLEN)(send - s) + 1);
3220 d = SvPVX(sv) + SvCUR(sv);
3222 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3224 /* See Note on sizing above. (NOTE: SvCUR() is not
3225 * set correctly here). */
3226 const STRLEN off = d - SvPVX_const(sv);
3227 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3229 Copy(str, d, len, char);
3234 /* Deprecate non-approved name syntax */
3235 if (ckWARN_d(WARN_DEPRECATED)) {
3236 bool problematic = FALSE;
3239 /* For non-ut8 input, look to see that the first
3240 * character is an alpha, then loop through the rest
3241 * checking that each is a continuation */
3243 if (! isALPHAU(*i)) problematic = TRUE;
3244 else for (i = s + 1; i < e; i++) {
3245 if (isCHARNAME_CONT(*i)) continue;
3251 /* Similarly for utf8. For invariants can check
3252 * directly. We accept anything above the latin1
3253 * range because it is immaterial to Perl if it is
3254 * correct or not, and is expensive to check. But
3255 * it is fairly easy in the latin1 range to convert
3256 * the variants into a single character and check
3258 if (UTF8_IS_INVARIANT(*i)) {
3259 if (! isALPHAU(*i)) problematic = TRUE;
3260 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3261 if (! isALPHAU(UNI_TO_NATIVE(UTF8_ACCUMULATE(*i,
3267 if (! problematic) for (i = s + UTF8SKIP(s);
3271 if (UTF8_IS_INVARIANT(*i)) {
3272 if (isCHARNAME_CONT(*i)) continue;
3273 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3275 } else if (isCHARNAME_CONT(
3277 UTF8_ACCUMULATE(*i, *(i+1)))))
3286 /* The e-i passed to the final %.*s makes sure that
3287 * should the trailing NUL be missing that this
3288 * print won't run off the end of the string */
3289 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3290 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s",
3291 (int)(i - s + 1), s, (int)(e - i), i + 1);
3294 } /* End \N{NAME} */
3297 native_range = FALSE; /* \N{} is defined to be Unicode */
3299 s = e + 1; /* Point to just after the '}' */
3302 /* \c is a control character */
3306 *d++ = grok_bslash_c(*s++, 1);
3309 yyerror("Missing control char name in \\c");
3313 /* printf-style backslashes, formfeeds, newlines, etc */
3315 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3318 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3321 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3324 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3327 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3330 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3333 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3339 } /* end if (backslash) */
3346 /* If we started with encoded form, or already know we want it,
3347 then encode the next character */
3348 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3352 /* One might think that it is wasted effort in the case of the
3353 * source being utf8 (this_utf8 == TRUE) to take the next character
3354 * in the source, convert it to an unsigned value, and then convert
3355 * it back again. But the source has not been validated here. The
3356 * routine that does the conversion checks for errors like
3359 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3360 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3362 SvCUR_set(sv, d - SvPVX_const(sv));
3365 /* See Note on sizing above. */
3366 sv_utf8_upgrade_flags_grow(sv,
3367 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3368 need + (STRLEN)(send - s) + 1);
3369 d = SvPVX(sv) + SvCUR(sv);
3371 } else if (need > len) {
3372 /* encoded value larger than old, may need extra space (NOTE:
3373 * SvCUR() is not set correctly here). See Note on sizing
3375 const STRLEN off = d - SvPVX_const(sv);
3376 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3380 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3382 if (uv > 255 && !dorange)
3383 native_range = FALSE;
3387 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3389 } /* while loop to process each character */
3391 /* terminate the string and set up the sv */
3393 SvCUR_set(sv, d - SvPVX_const(sv));
3394 if (SvCUR(sv) >= SvLEN(sv))
3395 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3398 if (PL_encoding && !has_utf8) {
3399 sv_recode_to_utf8(sv, PL_encoding);
3405 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3406 PL_sublex_info.sub_op->op_private |=
3407 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3411 /* shrink the sv if we allocated more than we used */
3412 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3413 SvPV_shrink_to_cur(sv);
3416 /* return the substring (via pl_yylval) only if we parsed anything */
3417 if (s > PL_bufptr) {
3418 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3419 const char *const key = PL_lex_inpat ? "qr" : "q";
3420 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3424 if (PL_lex_inwhat == OP_TRANS) {
3427 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3435 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3438 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3445 * Returns TRUE if there's more to the expression (e.g., a subscript),
3448 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3450 * ->[ and ->{ return TRUE
3451 * { and [ outside a pattern are always subscripts, so return TRUE
3452 * if we're outside a pattern and it's not { or [, then return FALSE
3453 * if we're in a pattern and the first char is a {
3454 * {4,5} (any digits around the comma) returns FALSE
3455 * if we're in a pattern and the first char is a [
3457 * [SOMETHING] has a funky algorithm to decide whether it's a
3458 * character class or not. It has to deal with things like
3459 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3460 * anything else returns TRUE
3463 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3466 S_intuit_more(pTHX_ register char *s)
3470 PERL_ARGS_ASSERT_INTUIT_MORE;
3472 if (PL_lex_brackets)
3474 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3476 if (*s != '{' && *s != '[')
3481 /* In a pattern, so maybe we have {n,m}. */
3498 /* On the other hand, maybe we have a character class */
3501 if (*s == ']' || *s == '^')
3504 /* this is terrifying, and it works */
3505 int weight = 2; /* let's weigh the evidence */
3507 unsigned char un_char = 255, last_un_char;
3508 const char * const send = strchr(s,']');
3509 char tmpbuf[sizeof PL_tokenbuf * 4];
3511 if (!send) /* has to be an expression */
3514 Zero(seen,256,char);
3517 else if (isDIGIT(*s)) {
3519 if (isDIGIT(s[1]) && s[2] == ']')
3525 for (; s < send; s++) {
3526 last_un_char = un_char;
3527 un_char = (unsigned char)*s;
3532 weight -= seen[un_char] * 10;
3533 if (isALNUM_lazy_if(s+1,UTF)) {
3535 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3536 len = (int)strlen(tmpbuf);
3537 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
3542 else if (*s == '$' && s[1] &&
3543 strchr("[#!%*<>()-=",s[1])) {
3544 if (/*{*/ strchr("])} =",s[2]))
3553 if (strchr("wds]",s[1]))
3555 else if (seen[(U8)'\''] || seen[(U8)'"'])
3557 else if (strchr("rnftbxcav",s[1]))
3559 else if (isDIGIT(s[1])) {
3561 while (s[1] && isDIGIT(s[1]))
3571 if (strchr("aA01! ",last_un_char))
3573 if (strchr("zZ79~",s[1]))
3575 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3576 weight -= 5; /* cope with negative subscript */
3579 if (!isALNUM(last_un_char)
3580 && !(last_un_char == '$' || last_un_char == '@'
3581 || last_un_char == '&')
3582 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3587 if (keyword(tmpbuf, d - tmpbuf, 0))
3590 if (un_char == last_un_char + 1)
3592 weight -= seen[un_char];
3597 if (weight >= 0) /* probably a character class */
3607 * Does all the checking to disambiguate
3609 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3610 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3612 * First argument is the stuff after the first token, e.g. "bar".
3614 * Not a method if bar is a filehandle.
3615 * Not a method if foo is a subroutine prototyped to take a filehandle.
3616 * Not a method if it's really "Foo $bar"
3617 * Method if it's "foo $bar"
3618 * Not a method if it's really "print foo $bar"
3619 * Method if it's really "foo package::" (interpreted as package->foo)
3620 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3621 * Not a method if bar is a filehandle or package, but is quoted with
3626 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3629 char *s = start + (*start == '$');
3630 char tmpbuf[sizeof PL_tokenbuf];
3637 PERL_ARGS_ASSERT_INTUIT_METHOD;
3640 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3644 const char *proto = SvPVX_const(cv);
3655 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3656 /* start is the beginning of the possible filehandle/object,
3657 * and s is the end of it
3658 * tmpbuf is a copy of it
3661 if (*start == '$') {
3662 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3663 isUPPER(*PL_tokenbuf))
3666 len = start - SvPVX(PL_linestr);
3670 start = SvPVX(PL_linestr) + len;
3674 return *s == '(' ? FUNCMETH : METHOD;
3676 if (!keyword(tmpbuf, len, 0)) {
3677 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3681 soff = s - SvPVX(PL_linestr);
3685 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
3686 if (indirgv && GvCVu(indirgv))
3688 /* filehandle or package name makes it a method */
3689 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
3691 soff = s - SvPVX(PL_linestr);
3694 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3695 return 0; /* no assumptions -- "=>" quotes bearword */
3697 start_force(PL_curforce);
3698 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3699 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3700 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3702 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3707 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3709 return *s == '(' ? FUNCMETH : METHOD;
3715 /* Encoded script support. filter_add() effectively inserts a
3716 * 'pre-processing' function into the current source input stream.
3717 * Note that the filter function only applies to the current source file
3718 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3720 * The datasv parameter (which may be NULL) can be used to pass
3721 * private data to this instance of the filter. The filter function
3722 * can recover the SV using the FILTER_DATA macro and use it to
3723 * store private buffers and state information.
3725 * The supplied datasv parameter is upgraded to a PVIO type
3726 * and the IoDIRP/IoANY field is used to store the function pointer,
3727 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3728 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3729 * private use must be set using malloc'd pointers.
3733 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3742 if (!PL_rsfp_filters)
3743 PL_rsfp_filters = newAV();
3746 SvUPGRADE(datasv, SVt_PVIO);
3747 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3748 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3749 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3750 FPTR2DPTR(void *, IoANY(datasv)),
3751 SvPV_nolen(datasv)));
3752 av_unshift(PL_rsfp_filters, 1);
3753 av_store(PL_rsfp_filters, 0, datasv) ;
3758 /* Delete most recently added instance of this filter function. */
3760 Perl_filter_del(pTHX_ filter_t funcp)
3765 PERL_ARGS_ASSERT_FILTER_DEL;
3768 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3769 FPTR2DPTR(void*, funcp)));
3771 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3773 /* if filter is on top of stack (usual case) just pop it off */
3774 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3775 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3776 sv_free(av_pop(PL_rsfp_filters));
3780 /* we need to search for the correct entry and clear it */
3781 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3785 /* Invoke the idxth filter function for the current rsfp. */
3786 /* maxlen 0 = read one text line */
3788 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3793 /* This API is bad. It should have been using unsigned int for maxlen.
3794 Not sure if we want to change the API, but if not we should sanity
3795 check the value here. */
3796 const unsigned int correct_length
3805 PERL_ARGS_ASSERT_FILTER_READ;
3807 if (!PL_parser || !PL_rsfp_filters)
3809 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3810 /* Provide a default input filter to make life easy. */
3811 /* Note that we append to the line. This is handy. */
3812 DEBUG_P(PerlIO_printf(Perl_debug_log,
3813 "filter_read %d: from rsfp\n", idx));
3814 if (correct_length) {
3817 const int old_len = SvCUR(buf_sv);
3819 /* ensure buf_sv is large enough */
3820 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3821 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3822 correct_length)) <= 0) {
3823 if (PerlIO_error(PL_rsfp))
3824 return -1; /* error */
3826 return 0 ; /* end of file */
3828 SvCUR_set(buf_sv, old_len + len) ;
3829 SvPVX(buf_sv)[old_len + len] = '\0';
3832 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3833 if (PerlIO_error(PL_rsfp))
3834 return -1; /* error */
3836 return 0 ; /* end of file */
3839 return SvCUR(buf_sv);
3841 /* Skip this filter slot if filter has been deleted */
3842 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3843 DEBUG_P(PerlIO_printf(Perl_debug_log,
3844 "filter_read %d: skipped (filter deleted)\n",
3846 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
3848 /* Get function pointer hidden within datasv */
3849 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
3850 DEBUG_P(PerlIO_printf(Perl_debug_log,
3851 "filter_read %d: via function %p (%s)\n",
3852 idx, (void*)datasv, SvPV_nolen_const(datasv)));
3853 /* Call function. The function is expected to */
3854 /* call "FILTER_READ(idx+1, buf_sv)" first. */
3855 /* Return: <0:error, =0:eof, >0:not eof */
3856 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
3860 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
3864 PERL_ARGS_ASSERT_FILTER_GETS;
3866 #ifdef PERL_CR_FILTER
3867 if (!PL_rsfp_filters) {
3868 filter_add(S_cr_textfilter,NULL);
3871 if (PL_rsfp_filters) {
3873 SvCUR_set(sv, 0); /* start with empty line */
3874 if (FILTER_READ(0, sv, 0) > 0)
3875 return ( SvPVX(sv) ) ;
3880 return (sv_gets(sv, PL_rsfp, append));
3884 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3889 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3891 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3895 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3896 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3898 return GvHV(gv); /* Foo:: */
3901 /* use constant CLASS => 'MyClass' */
3902 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3903 if (gv && GvCV(gv)) {
3904 SV * const sv = cv_const_sv(GvCV(gv));
3906 pkgname = SvPV_const(sv, len);
3909 return gv_stashpvn(pkgname, len, 0);
3913 * S_readpipe_override
3914 * Check whether readpipe() is overriden, and generates the appropriate
3915 * optree, provided sublex_start() is called afterwards.
3918 S_readpipe_override(pTHX)
3921 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3922 pl_yylval.ival = OP_BACKTICK;
3924 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3926 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3927 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3928 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3930 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3931 append_elem(OP_LIST,
3932 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3933 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3940 * The intent of this yylex wrapper is to minimize the changes to the
3941 * tokener when we aren't interested in collecting madprops. It remains
3942 * to be seen how successful this strategy will be...
3949 char *s = PL_bufptr;
3951 /* make sure PL_thiswhite is initialized */
3955 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3956 if (PL_pending_ident)
3957 return S_pending_ident(aTHX);
3959 /* previous token ate up our whitespace? */
3960 if (!PL_lasttoke && PL_nextwhite) {
3961 PL_thiswhite = PL_nextwhite;
3965 /* isolate the token, and figure out where it is without whitespace */
3966 PL_realtokenstart = -1;
3970 assert(PL_curforce < 0);
3972 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3973 if (!PL_thistoken) {
3974 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3975 PL_thistoken = newSVpvs("");
3977 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3978 PL_thistoken = newSVpvn(tstart, s - tstart);
3981 if (PL_thismad) /* install head */
3982 CURMAD('X', PL_thistoken);
3985 /* last whitespace of a sublex? */
3986 if (optype == ')' && PL_endwhite) {
3987 CURMAD('X', PL_endwhite);
3992 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3993 if (!PL_thiswhite && !PL_endwhite && !optype) {
3994 sv_free(PL_thistoken);
3999 /* put off final whitespace till peg */
4000 if (optype == ';' && !PL_rsfp) {
4001 PL_nextwhite = PL_thiswhite;
4004 else if (PL_thisopen) {
4005 CURMAD('q', PL_thisopen);
4007 sv_free(PL_thistoken);
4011 /* Store actual token text as madprop X */
4012 CURMAD('X', PL_thistoken);
4016 /* add preceding whitespace as madprop _ */
4017 CURMAD('_', PL_thiswhite);
4021 /* add quoted material as madprop = */
4022 CURMAD('=', PL_thisstuff);
4026 /* add terminating quote as madprop Q */
4027 CURMAD('Q', PL_thisclose);
4031 /* special processing based on optype */
4035 /* opval doesn't need a TOKEN since it can already store mp */
4045 if (pl_yylval.opval)
4046 append_madprops(PL_thismad, pl_yylval.opval, 0);
4054 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4063 /* remember any fake bracket that lexer is about to discard */
4064 if (PL_lex_brackets == 1 &&
4065 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4068 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4071 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4072 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4075 break; /* don't bother looking for trailing comment */
4084 /* attach a trailing comment to its statement instead of next token */
4088 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4090 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4092 if (*s == '\n' || *s == '#') {
4093 while (s < PL_bufend && *s != '\n')
4097 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4098 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4115 /* Create new token struct. Note: opvals return early above. */
4116 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4123 S_tokenize_use(pTHX_ int is_use, char *s) {
4126 PERL_ARGS_ASSERT_TOKENIZE_USE;
4128 if (PL_expect != XSTATE)
4129 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4130 is_use ? "use" : "no"));
4132 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4133 s = force_version(s, TRUE);
4134 if (*s == ';' || *s == '}'
4135 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4136 start_force(PL_curforce);
4137 NEXTVAL_NEXTTOKE.opval = NULL;
4140 else if (*s == 'v') {
4141 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4142 s = force_version(s, FALSE);
4146 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4147 s = force_version(s, FALSE);
4149 pl_yylval.ival = is_use;
4153 static const char* const exp_name[] =
4154 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4155 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4162 Works out what to call the token just pulled out of the input
4163 stream. The yacc parser takes care of taking the ops we return and
4164 stitching them into a tree.
4170 if read an identifier
4171 if we're in a my declaration
4172 croak if they tried to say my($foo::bar)
4173 build the ops for a my() declaration
4174 if it's an access to a my() variable
4175 are we in a sort block?
4176 croak if my($a); $a <=> $b
4177 build ops for access to a my() variable
4178 if in a dq string, and they've said @foo and we can't find @foo
4180 build ops for a bareword
4181 if we already built the token before, use it.
4186 #pragma segment Perl_yylex
4192 register char *s = PL_bufptr;
4198 /* orig_keyword, gvp, and gv are initialized here because
4199 * jump to the label just_a_word_zero can bypass their
4200 * initialization later. */
4201 I32 orig_keyword = 0;
4206 SV* tmp = newSVpvs("");
4207 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4208 (IV)CopLINE(PL_curcop),
4209 lex_state_names[PL_lex_state],
4210 exp_name[PL_expect],
4211 pv_display(tmp, s, strlen(s), 0, 60));
4214 /* check if there's an identifier for us to look at */
4215 if (PL_pending_ident)
4216 return REPORT(S_pending_ident(aTHX));
4218 /* no identifier pending identification */
4220 switch (PL_lex_state) {
4222 case LEX_NORMAL: /* Some compilers will produce faster */
4223 case LEX_INTERPNORMAL: /* code if we comment these out. */
4227 /* when we've already built the next token, just pull it out of the queue */
4231 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4233 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4234 PL_nexttoke[PL_lasttoke].next_mad = 0;
4235 if (PL_thismad && PL_thismad->mad_key == '_') {
4236 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4237 PL_thismad->mad_val = 0;
4238 mad_free(PL_thismad);
4243 PL_lex_state = PL_lex_defer;
4244 PL_expect = PL_lex_expect;
4245 PL_lex_defer = LEX_NORMAL;
4246 if (!PL_nexttoke[PL_lasttoke].next_type)
4251 pl_yylval = PL_nextval[PL_nexttoke];
4253 PL_lex_state = PL_lex_defer;
4254 PL_expect = PL_lex_expect;
4255 PL_lex_defer = LEX_NORMAL;
4259 /* FIXME - can these be merged? */
4260 return(PL_nexttoke[PL_lasttoke].next_type);
4262 return REPORT(PL_nexttype[PL_nexttoke]);
4265 /* interpolated case modifiers like \L \U, including \Q and \E.
4266 when we get here, PL_bufptr is at the \
4268 case LEX_INTERPCASEMOD:
4270 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4271 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
4273 /* handle \E or end of string */
4274 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4276 if (PL_lex_casemods) {
4277 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4278 PL_lex_casestack[PL_lex_casemods] = '\0';
4280 if (PL_bufptr != PL_bufend
4281 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
4283 PL_lex_state = LEX_INTERPCONCAT;
4286 PL_thistoken = newSVpvs("\\E");
4292 while (PL_bufptr != PL_bufend &&
4293 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4295 PL_thiswhite = newSVpvs("");
4296 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4300 if (PL_bufptr != PL_bufend)
4303 PL_lex_state = LEX_INTERPCONCAT;
4307 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4308 "### Saw case modifier\n"); });
4310 if (s[1] == '\\' && s[2] == 'E') {
4313 PL_thiswhite = newSVpvs("");
4314 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4317 PL_lex_state = LEX_INTERPCONCAT;
4322 if (!PL_madskills) /* when just compiling don't need correct */
4323 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4324 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4325 if ((*s == 'L' || *s == 'U') &&
4326 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4327 PL_lex_casestack[--PL_lex_casemods] = '\0';
4330 if (PL_lex_casemods > 10)
4331 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4332 PL_lex_casestack[PL_lex_casemods++] = *s;
4333 PL_lex_casestack[PL_lex_casemods] = '\0';
4334 PL_lex_state = LEX_INTERPCONCAT;
4335 start_force(PL_curforce);
4336 NEXTVAL_NEXTTOKE.ival = 0;
4338 start_force(PL_curforce);
4340 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4342 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4344 NEXTVAL_NEXTTOKE.ival = OP_LC;
4346 NEXTVAL_NEXTTOKE.ival = OP_UC;
4348 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4350 Perl_croak(aTHX_ "panic: yylex");
4352 SV* const tmpsv = newSVpvs("\\ ");
4353 /* replace the space with the character we want to escape
4355 SvPVX(tmpsv)[1] = *s;
4361 if (PL_lex_starts) {
4367 sv_free(PL_thistoken);
4368 PL_thistoken = newSVpvs("");
4371 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4372 if (PL_lex_casemods == 1 && PL_lex_inpat)
4381 case LEX_INTERPPUSH:
4382 return REPORT(sublex_push());
4384 case LEX_INTERPSTART:
4385 if (PL_bufptr == PL_bufend)
4386 return REPORT(sublex_done());
4387 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4388 "### Interpolated variable\n"); });
4390 PL_lex_dojoin = (*PL_bufptr == '@');
4391 PL_lex_state = LEX_INTERPNORMAL;
4392 if (PL_lex_dojoin) {
4393 start_force(PL_curforce);
4394 NEXTVAL_NEXTTOKE.ival = 0;
4396 start_force(PL_curforce);
4397 force_ident("\"", '$');
4398 start_force(PL_curforce);
4399 NEXTVAL_NEXTTOKE.ival = 0;
4401 start_force(PL_curforce);
4402 NEXTVAL_NEXTTOKE.ival = 0;
4404 start_force(PL_curforce);
4405 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4408 if (PL_lex_starts++) {
4413 sv_free(PL_thistoken);
4414 PL_thistoken = newSVpvs("");
4417 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4418 if (!PL_lex_casemods && PL_lex_inpat)
4425 case LEX_INTERPENDMAYBE:
4426 if (intuit_more(PL_bufptr)) {
4427 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4433 if (PL_lex_dojoin) {
4434 PL_lex_dojoin = FALSE;
4435 PL_lex_state = LEX_INTERPCONCAT;
4439 sv_free(PL_thistoken);
4440 PL_thistoken = newSVpvs("");
4445 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4446 && SvEVALED(PL_lex_repl))
4448 if (PL_bufptr != PL_bufend)
4449 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4453 case LEX_INTERPCONCAT:
4455 if (PL_lex_brackets)
4456 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4458 if (PL_bufptr == PL_bufend)
4459 return REPORT(sublex_done());
4461 if (SvIVX(PL_linestr) == '\'') {
4462 SV *sv = newSVsv(PL_linestr);
4465 else if ( PL_hints & HINT_NEW_RE )
4466 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4467 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4471 s = scan_const(PL_bufptr);
4473 PL_lex_state = LEX_INTERPCASEMOD;
4475 PL_lex_state = LEX_INTERPSTART;
4478 if (s != PL_bufptr) {
4479 start_force(PL_curforce);
4481 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4483 NEXTVAL_NEXTTOKE = pl_yylval;
4486 if (PL_lex_starts++) {
4490 sv_free(PL_thistoken);
4491 PL_thistoken = newSVpvs("");
4494 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4495 if (!PL_lex_casemods && PL_lex_inpat)
4508 PL_lex_state = LEX_NORMAL;
4509 s = scan_formline(PL_bufptr);
4510 if (!PL_lex_formbrack)
4516 PL_oldoldbufptr = PL_oldbufptr;
4522 sv_free(PL_thistoken);
4525 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4529 if (isIDFIRST_lazy_if(s,UTF))
4532 unsigned char c = *s;
4533 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4534 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4535 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4540 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4544 goto fake_eof; /* emulate EOF on ^D or ^Z */
4553 if (PL_lex_brackets) {
4554 yyerror((const char *)
4556 ? "Format not terminated"
4557 : "Missing right curly or square bracket"));
4559 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4560 "### Tokener got EOF\n");
4564 if (s++ < PL_bufend)
4565 goto retry; /* ignore stray nulls */
4568 if (!PL_in_eval && !PL_preambled) {
4569 PL_preambled = TRUE;
4575 /* Generate a string of Perl code to load the debugger.
4576 * If PERL5DB is set, it will return the contents of that,
4577 * otherwise a compile-time require of perl5db.pl. */
4579 const char * const pdb = PerlEnv_getenv("PERL5DB");
4582 sv_setpv(PL_linestr, pdb);
4583 sv_catpvs(PL_linestr,";");
4585 SETERRNO(0,SS_NORMAL);
4586 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4589 sv_setpvs(PL_linestr,"");
4590 if (PL_preambleav) {
4591 SV **svp = AvARRAY(PL_preambleav);
4592 SV **const end = svp + AvFILLp(PL_preambleav);
4594 sv_catsv(PL_linestr, *svp);
4596 sv_catpvs(PL_linestr, ";");
4598 sv_free(MUTABLE_SV(PL_preambleav));
4599 PL_preambleav = NULL;
4602 sv_catpvs(PL_linestr,
4603 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4604 if (PL_minus_n || PL_minus_p) {
4605 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4607 sv_catpvs(PL_linestr,"chomp;");
4610 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4611 || *PL_splitstr == '"')
4612 && strchr(PL_splitstr + 1, *PL_splitstr))
4613 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4615 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4616 bytes can be used as quoting characters. :-) */
4617 const char *splits = PL_splitstr;
4618 sv_catpvs(PL_linestr, "our @F=split(q\0");
4621 if (*splits == '\\')
4622 sv_catpvn(PL_linestr, splits, 1);
4623 sv_catpvn(PL_linestr, splits, 1);
4624 } while (*splits++);
4625 /* This loop will embed the trailing NUL of
4626 PL_linestr as the last thing it does before
4628 sv_catpvs(PL_linestr, ");");
4632 sv_catpvs(PL_linestr,"our @F=split(' ');");
4635 sv_catpvs(PL_linestr, "\n");
4636 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4637 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4638 PL_last_lop = PL_last_uni = NULL;
4639 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4640 update_debugger_info(PL_linestr, NULL, 0);
4645 bof = PL_rsfp ? TRUE : FALSE;
4648 fake_eof = LEX_FAKE_EOF;
4650 PL_bufptr = PL_bufend;
4651 CopLINE_inc(PL_curcop);
4652 if (!lex_next_chunk(fake_eof)) {
4653 CopLINE_dec(PL_curcop);
4655 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4657 CopLINE_dec(PL_curcop);
4660 PL_realtokenstart = -1;
4663 /* If it looks like the start of a BOM or raw UTF-16,
4664 * check if it in fact is. */
4665 if (bof && PL_rsfp &&
4670 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
4672 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4673 s = swallow_bom((U8*)s);
4677 /* Incest with pod. */
4680 sv_catsv(PL_thiswhite, PL_linestr);
4682 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4683 sv_setpvs(PL_linestr, "");
4684 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4685 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4686 PL_last_lop = PL_last_uni = NULL;
4687 PL_doextract = FALSE;
4692 } while (PL_doextract);
4693 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4694 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4695 PL_last_lop = PL_last_uni = NULL;
4696 if (CopLINE(PL_curcop) == 1) {
4697 while (s < PL_bufend && isSPACE(*s))
4699 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4703 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4707 if (*s == '#' && *(s+1) == '!')
4709 #ifdef ALTERNATE_SHEBANG
4711 static char const as[] = ALTERNATE_SHEBANG;
4712 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4713 d = s + (sizeof(as) - 1);
4715 #endif /* ALTERNATE_SHEBANG */
4724 while (*d && !isSPACE(*d))
4728 #ifdef ARG_ZERO_IS_SCRIPT
4729 if (ipathend > ipath) {
4731 * HP-UX (at least) sets argv[0] to the script name,
4732 * which makes $^X incorrect. And Digital UNIX and Linux,
4733 * at least, set argv[0] to the basename of the Perl
4734 * interpreter. So, having found "#!", we'll set it right.
4736 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4738 assert(SvPOK(x) || SvGMAGICAL(x));
4739 if (sv_eq(x, CopFILESV(PL_curcop))) {
4740 sv_setpvn(x, ipath, ipathend - ipath);
4746 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4747 const char * const lstart = SvPV_const(x,llen);
4749 bstart += blen - llen;
4750 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4751 sv_setpvn(x, ipath, ipathend - ipath);
4756 TAINT_NOT; /* $^X is always tainted, but that's OK */
4758 #endif /* ARG_ZERO_IS_SCRIPT */
4763 d = instr(s,"perl -");
4765 d = instr(s,"perl");
4767 /* avoid getting into infinite loops when shebang
4768 * line contains "Perl" rather than "perl" */
4770 for (d = ipathend-4; d >= ipath; --d) {
4771 if ((*d == 'p' || *d == 'P')
4772 && !ibcmp(d, "perl", 4))
4782 #ifdef ALTERNATE_SHEBANG
4784 * If the ALTERNATE_SHEBANG on this system starts with a
4785 * character that can be part of a Perl expression, then if
4786 * we see it but not "perl", we're probably looking at the
4787 * start of Perl code, not a request to hand off to some
4788 * other interpreter. Similarly, if "perl" is there, but
4789 * not in the first 'word' of the line, we assume the line
4790 * contains the start of the Perl program.
4792 if (d && *s != '#') {
4793 const char *c = ipath;
4794 while (*c && !strchr("; \t\r\n\f\v#", *c))
4797 d = NULL; /* "perl" not in first word; ignore */
4799 *s = '#'; /* Don't try to parse shebang line */
4801 #endif /* ALTERNATE_SHEBANG */
4806 !instr(s,"indir") &&
4807 instr(PL_origargv[0],"perl"))
4814 while (s < PL_bufend && isSPACE(*s))
4816 if (s < PL_bufend) {
4817 Newx(newargv,PL_origargc+3,char*);
4819 while (s < PL_bufend && !isSPACE(*s))
4822 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4825 newargv = PL_origargv;
4828 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4830 Perl_croak(aTHX_ "Can't exec %s", ipath);
4833 while (*d && !isSPACE(*d))
4835 while (SPACE_OR_TAB(*d))
4839 const bool switches_done = PL_doswitches;
4840 const U32 oldpdb = PL_perldb;
4841 const bool oldn = PL_minus_n;
4842 const bool oldp = PL_minus_p;
4846 bool baduni = FALSE;
4848 const char *d2 = d1 + 1;
4849 if (parse_unicode_opts((const char **)&d2)
4853 if (baduni || *d1 == 'M' || *d1 == 'm') {
4854 const char * const m = d1;
4855 while (*d1 && !isSPACE(*d1))
4857 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4860 d1 = moreswitches(d1);
4862 if (PL_doswitches && !switches_done) {
4863 int argc = PL_origargc;
4864 char **argv = PL_origargv;
4867 } while (argc && argv[0][0] == '-' && argv[0][1]);
4868 init_argv_symbols(argc,argv);
4870 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4871 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4872 /* if we have already added "LINE: while (<>) {",
4873 we must not do it again */
4875 sv_setpvs(PL_linestr, "");
4876 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4877 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4878 PL_last_lop = PL_last_uni = NULL;
4879 PL_preambled = FALSE;
4880 if (PERLDB_LINE || PERLDB_SAVESRC)
4881 (void)gv_fetchfile(PL_origfilename);
4888 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4890 PL_lex_state = LEX_FORMLINE;
4895 #ifdef PERL_STRICT_CR
4896 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4898 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4900 case ' ': case '\t': case '\f': case 013:
4902 PL_realtokenstart = -1;
4904 PL_thiswhite = newSVpvs("");
4905 sv_catpvn(PL_thiswhite, s, 1);
4912 PL_realtokenstart = -1;
4916 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4917 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4918 /* handle eval qq[#line 1 "foo"\n ...] */
4919 CopLINE_dec(PL_curcop);
4922 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4924 if (!PL_in_eval || PL_rsfp)
4929 while (d < PL_bufend && *d != '\n')
4933 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4934 Perl_croak(aTHX_ "panic: input overflow");
4937 PL_thiswhite = newSVpvn(s, d - s);
4942 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4944 PL_lex_state = LEX_FORMLINE;
4950 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4951 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4954 TOKEN(PEG); /* make sure any #! line is accessible */
4959 /* if (PL_madskills && PL_lex_formbrack) { */
4961 while (d < PL_bufend && *d != '\n')
4965 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4966 Perl_croak(aTHX_ "panic: input overflow");
4967 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4969 PL_thiswhite = newSVpvs("");
4970 if (CopLINE(PL_curcop) == 1) {
4971 sv_setpvs(PL_thiswhite, "");
4974 sv_catpvn(PL_thiswhite, s, d - s);
4988 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4996 while (s < PL_bufend && SPACE_OR_TAB(*s))
4999 if (strnEQ(s,"=>",2)) {
5000 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
5001 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5002 OPERATOR('-'); /* unary minus */
5004 PL_last_uni = PL_oldbufptr;
5006 case 'r': ftst = OP_FTEREAD; break;
5007 case 'w': ftst = OP_FTEWRITE; break;
5008 case 'x': ftst = OP_FTEEXEC; break;
5009 case 'o': ftst = OP_FTEOWNED; break;
5010 case 'R': ftst = OP_FTRREAD; break;
5011 case 'W': ftst = OP_FTRWRITE; break;
5012 case 'X': ftst = OP_FTREXEC; break;
5013 case 'O': ftst = OP_FTROWNED; break;
5014 case 'e': ftst = OP_FTIS; break;
5015 case 'z': ftst = OP_FTZERO; break;
5016 case 's': ftst = OP_FTSIZE; break;
5017 case 'f': ftst = OP_FTFILE; break;
5018 case 'd': ftst = OP_FTDIR; break;
5019 case 'l': ftst = OP_FTLINK; break;
5020 case 'p': ftst = OP_FTPIPE; break;
5021 case 'S': ftst = OP_FTSOCK; break;
5022 case 'u': ftst = OP_FTSUID; break;
5023 case 'g': ftst = OP_FTSGID; break;
5024 case 'k': ftst = OP_FTSVTX; break;
5025 case 'b': ftst = OP_FTBLK; break;
5026 case 'c': ftst = OP_FTCHR; break;
5027 case 't': ftst = OP_FTTTY; break;
5028 case 'T': ftst = OP_FTTEXT; break;
5029 case 'B': ftst = OP_FTBINARY; break;
5030 case 'M': case 'A': case 'C':
5031 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5033 case 'M': ftst = OP_FTMTIME; break;
5034 case 'A': ftst = OP_FTATIME; break;
5035 case 'C': ftst = OP_FTCTIME; break;
5043 PL_last_lop_op = (OPCODE)ftst;
5044 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5045 "### Saw file test %c\n", (int)tmp);
5050 /* Assume it was a minus followed by a one-letter named
5051 * subroutine call (or a -bareword), then. */
5052 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5053 "### '-%c' looked like a file test but was not\n",
5060 const char tmp = *s++;
5063 if (PL_expect == XOPERATOR)
5068 else if (*s == '>') {
5071 if (isIDFIRST_lazy_if(s,UTF)) {
5072 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5080 if (PL_expect == XOPERATOR)
5083 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5085 OPERATOR('-'); /* unary minus */
5091 const char tmp = *s++;
5094 if (PL_expect == XOPERATOR)
5099 if (PL_expect == XOPERATOR)
5102 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5109 if (PL_expect != XOPERATOR) {
5110 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5111 PL_expect = XOPERATOR;
5112 force_ident(PL_tokenbuf, '*');
5125 if (PL_expect == XOPERATOR) {
5129 PL_tokenbuf[0] = '%';
5130 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5131 sizeof PL_tokenbuf - 1, FALSE);
5132 if (!PL_tokenbuf[1]) {
5135 PL_pending_ident = '%';
5144 const char tmp = *s++;
5149 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5156 const char tmp = *s++;
5162 goto just_a_word_zero_gv;
5165 switch (PL_expect) {
5171 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5173 PL_bufptr = s; /* update in case we back off */
5175 deprecate(":= for an empty attribute list");
5182 PL_expect = XTERMBLOCK;
5185 stuffstart = s - SvPVX(PL_linestr) - 1;
5189 while (isIDFIRST_lazy_if(s,UTF)) {
5192 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5193 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5194 if (tmp < 0) tmp = -tmp;
5209 sv = newSVpvn(s, len);
5211 d = scan_str(d,TRUE,TRUE);
5213 /* MUST advance bufptr here to avoid bogus
5214 "at end of line" context messages from yyerror().
5216 PL_bufptr = s + len;
5217 yyerror("Unterminated attribute parameter in attribute list");
5221 return REPORT(0); /* EOF indicator */
5225 sv_catsv(sv, PL_lex_stuff);
5226 attrs = append_elem(OP_LIST, attrs,
5227 newSVOP(OP_CONST, 0, sv));
5228 SvREFCNT_dec(PL_lex_stuff);
5229 PL_lex_stuff = NULL;
5232 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5234 if (PL_in_my == KEY_our) {
5235 deprecate(":unique");
5238 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5241 /* NOTE: any CV attrs applied here need to be part of
5242 the CVf_BUILTIN_ATTRS define in cv.h! */
5243 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5245 CvLVALUE_on(PL_compcv);
5247 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5249 deprecate(":locked");
5251 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5253 CvMETHOD_on(PL_compcv);
5255 /* After we've set the flags, it could be argued that
5256 we don't need to do the attributes.pm-based setting
5257 process, and shouldn't bother appending recognized
5258 flags. To experiment with that, uncomment the
5259 following "else". (Note that's already been
5260 uncommented. That keeps the above-applied built-in
5261 attributes from being intercepted (and possibly
5262 rejected) by a package's attribute routines, but is
5263 justified by the performance win for the common case
5264 of applying only built-in attributes.) */
5266 attrs = append_elem(OP_LIST, attrs,
5267 newSVOP(OP_CONST, 0,
5271 if (*s == ':' && s[1] != ':')
5274 break; /* require real whitespace or :'s */
5275 /* XXX losing whitespace on sequential attributes here */
5279 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5280 if (*s != ';' && *s != '}' && *s != tmp
5281 && (tmp != '=' || *s != ')')) {
5282 const char q = ((*s == '\'') ? '"' : '\'');
5283 /* If here for an expression, and parsed no attrs, back
5285 if (tmp == '=' && !attrs) {
5289 /* MUST advance bufptr here to avoid bogus "at end of line"
5290 context messages from yyerror().
5293 yyerror( (const char *)
5295 ? Perl_form(aTHX_ "Invalid separator character "
5296 "%c%c%c in attribute list", q, *s, q)
5297 : "Unterminated attribute list" ) );
5305 start_force(PL_curforce);
5306 NEXTVAL_NEXTTOKE.opval = attrs;
5307 CURMAD('_', PL_nextwhite);
5312 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5313 (s - SvPVX(PL_linestr)) - stuffstart);
5321 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5322 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5330 const char tmp = *s++;
5335 const char tmp = *s++;
5343 if (PL_lex_brackets <= 0)
5344 yyerror("Unmatched right square bracket");
5347 if (PL_lex_state == LEX_INTERPNORMAL) {
5348 if (PL_lex_brackets == 0) {
5349 if (*s == '-' && s[1] == '>')
5350 PL_lex_state = LEX_INTERPENDMAYBE;
5351 else if (*s != '[' && *s != '{')
5352 PL_lex_state = LEX_INTERPEND;
5359 if (PL_lex_brackets > 100) {
5360 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5362 switch (PL_expect) {
5364 if (PL_lex_formbrack) {
5368 if (PL_oldoldbufptr == PL_last_lop)
5369 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5371 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5372 OPERATOR(HASHBRACK);
5374 while (s < PL_bufend && SPACE_OR_TAB(*s))
5377 PL_tokenbuf[0] = '\0';
5378 if (d < PL_bufend && *d == '-') {
5379 PL_tokenbuf[0] = '-';
5381 while (d < PL_bufend && SPACE_OR_TAB(*d))
5384 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5385 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5387 while (d < PL_bufend && SPACE_OR_TAB(*d))
5390 const char minus = (PL_tokenbuf[0] == '-');
5391 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5399 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5404 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5409 if (PL_oldoldbufptr == PL_last_lop)
5410 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5412 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5415 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5417 /* This hack is to get the ${} in the message. */
5419 yyerror("syntax error");
5422 OPERATOR(HASHBRACK);
5424 /* This hack serves to disambiguate a pair of curlies
5425 * as being a block or an anon hash. Normally, expectation
5426 * determines that, but in cases where we're not in a
5427 * position to expect anything in particular (like inside
5428 * eval"") we have to resolve the ambiguity. This code
5429 * covers the case where the first term in the curlies is a
5430 * quoted string. Most other cases need to be explicitly
5431 * disambiguated by prepending a "+" before the opening
5432 * curly in order to force resolution as an anon hash.
5434 * XXX should probably propagate the outer expectation
5435 * into eval"" to rely less on this hack, but that could
5436 * potentially break current behavior of eval"".
5440 if (*s == '\'' || *s == '"' || *s == '`') {
5441 /* common case: get past first string, handling escapes */
5442 for (t++; t < PL_bufend && *t != *s;)
5443 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5447 else if (*s == 'q') {
5450 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5453 /* skip q//-like construct */
5455 char open, close, term;
5458 while (t < PL_bufend && isSPACE(*t))
5460 /* check for q => */
5461 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5462 OPERATOR(HASHBRACK);
5466 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5470 for (t++; t < PL_bufend; t++) {
5471 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5473 else if (*t == open)
5477 for (t++; t < PL_bufend; t++) {
5478 if (*t == '\\' && t+1 < PL_bufend)
5480 else if (*t == close && --brackets <= 0)
5482 else if (*t == open)
5489 /* skip plain q word */
5490 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5493 else if (isALNUM_lazy_if(t,UTF)) {
5495 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5498 while (t < PL_bufend && isSPACE(*t))
5500 /* if comma follows first term, call it an anon hash */
5501 /* XXX it could be a comma expression with loop modifiers */
5502 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5503 || (*t == '=' && t[1] == '>')))
5504 OPERATOR(HASHBRACK);
5505 if (PL_expect == XREF)
5508 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5514 pl_yylval.ival = CopLINE(PL_curcop);
5515 if (isSPACE(*s) || *s == '#')
5516 PL_copline = NOLINE; /* invalidate current command line number */
5521 if (PL_lex_brackets <= 0)
5522 yyerror("Unmatched right curly bracket");
5524 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5525 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5526 PL_lex_formbrack = 0;
5527 if (PL_lex_state == LEX_INTERPNORMAL) {
5528 if (PL_lex_brackets == 0) {
5529 if (PL_expect & XFAKEBRACK) {
5530 PL_expect &= XENUMMASK;
5531 PL_lex_state = LEX_INTERPEND;
5536 PL_thiswhite = newSVpvs("");
5537 sv_catpvs(PL_thiswhite,"}");
5540 return yylex(); /* ignore fake brackets */
5542 if (*s == '-' && s[1] == '>')
5543 PL_lex_state = LEX_INTERPENDMAYBE;
5544 else if (*s != '[' && *s != '{')
5545 PL_lex_state = LEX_INTERPEND;
5548 if (PL_expect & XFAKEBRACK) {
5549 PL_expect &= XENUMMASK;
5551 return yylex(); /* ignore fake brackets */
5553 start_force(PL_curforce);
5555 curmad('X', newSVpvn(s-1,1));
5556 CURMAD('_', PL_thiswhite);
5561 PL_thistoken = newSVpvs("");
5569 if (PL_expect == XOPERATOR) {
5570 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5571 && isIDFIRST_lazy_if(s,UTF))
5573 CopLINE_dec(PL_curcop);
5574 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5575 CopLINE_inc(PL_curcop);
5580 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5582 PL_expect = XOPERATOR;
5583 force_ident(PL_tokenbuf, '&');
5587 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5599 const char tmp = *s++;
5606 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5607 && strchr("+-*/%.^&|<",tmp))
5608 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5609 "Reversed %c= operator",(int)tmp);
5611 if (PL_expect == XSTATE && isALPHA(tmp) &&
5612 (s == PL_linestart+1 || s[-2] == '\n') )
5614 if (PL_in_eval && !PL_rsfp) {
5619 if (strnEQ(s,"=cut",4)) {
5635 PL_thiswhite = newSVpvs("");
5636 sv_catpvn(PL_thiswhite, PL_linestart,
5637 PL_bufend - PL_linestart);
5641 PL_doextract = TRUE;
5645 if (PL_lex_brackets < PL_lex_formbrack) {
5647 #ifdef PERL_STRICT_CR
5648 while (SPACE_OR_TAB(*t))
5650 while (SPACE_OR_TAB(*t) || *t == '\r')
5653 if (*t == '\n' || *t == '#') {
5664 const char tmp = *s++;
5666 /* was this !=~ where !~ was meant?
5667 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5669 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5670 const char *t = s+1;
5672 while (t < PL_bufend && isSPACE(*t))
5675 if (*t == '/' || *t == '?' ||
5676 ((*t == 'm' || *t == 's' || *t == 'y')
5677 && !isALNUM(t[1])) ||
5678 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5679 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5680 "!=~ should be !~");
5690 if (PL_expect != XOPERATOR) {
5691 if (s[1] != '<' && !strchr(s,'>'))
5694 s = scan_heredoc(s);
5696 s = scan_inputsymbol(s);
5697 TERM(sublex_start());
5703 SHop(OP_LEFT_SHIFT);
5717 const char tmp = *s++;
5719 SHop(OP_RIGHT_SHIFT);
5720 else if (tmp == '=')
5729 if (PL_expect == XOPERATOR) {
5730 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5731 return deprecate_commaless_var_list();
5735 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
5736 PL_tokenbuf[0] = '@';
5737 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
5738 sizeof PL_tokenbuf - 1, FALSE);
5739 if (PL_expect == XOPERATOR)
5740 no_op("Array length", s);
5741 if (!PL_tokenbuf[1])
5743 PL_expect = XOPERATOR;
5744 PL_pending_ident = '#';
5748 PL_tokenbuf[0] = '$';
5749 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5750 sizeof PL_tokenbuf - 1, FALSE);
5751 if (PL_expect == XOPERATOR)
5753 if (!PL_tokenbuf[1]) {
5755 yyerror("Final $ should be \\$ or $name");
5759 /* This kludge not intended to be bulletproof. */
5760 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
5761 pl_yylval.opval = newSVOP(OP_CONST, 0,
5762 newSViv(CopARYBASE_get(&PL_compiling)));
5763 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
5769 const char tmp = *s;
5770 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
5773 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
5774 && intuit_more(s)) {
5776 PL_tokenbuf[0] = '@';
5777 if (ckWARN(WARN_SYNTAX)) {
5780 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
5783 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5784 while (t < PL_bufend && *t != ']')
5786 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5787 "Multidimensional syntax %.*s not supported",
5788 (int)((t - PL_bufptr) + 1), PL_bufptr);
5792 else if (*s == '{') {
5794 PL_tokenbuf[0] = '%';
5795 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
5796 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
5798 char tmpbuf[sizeof PL_tokenbuf];
5801 } while (isSPACE(*t));
5802 if (isIDFIRST_lazy_if(t,UTF)) {
5804 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
5808 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
5809 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5810 "You need to quote \"%s\"",
5817 PL_expect = XOPERATOR;
5818 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
5819 const bool islop = (PL_last_lop == PL_oldoldbufptr);
5820 if (!islop || PL_last_lop_op == OP_GREPSTART)
5821 PL_expect = XOPERATOR;
5822 else if (strchr("$@\"'`q", *s))
5823 PL_expect = XTERM; /* e.g. print $fh "foo" */
5824 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
5825 PL_expect = XTERM; /* e.g. print $fh &sub */
5826 else if (isIDFIRST_lazy_if(s,UTF)) {
5827 char tmpbuf[sizeof PL_tokenbuf];
5829 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5830 if ((t2 = keyword(tmpbuf, len, 0))) {
5831 /* binary operators exclude handle interpretations */
5843 PL_expect = XTERM; /* e.g. print $fh length() */
5848 PL_expect = XTERM; /* e.g. print $fh subr() */
5851 else if (isDIGIT(*s))
5852 PL_expect = XTERM; /* e.g. print $fh 3 */
5853 else if (*s == '.' && isDIGIT(s[1]))
5854 PL_expect = XTERM; /* e.g. print $fh .3 */
5855 else if ((*s == '?' || *s == '-' || *s == '+')
5856 && !isSPACE(s[1]) && s[1] != '=')
5857 PL_expect = XTERM; /* e.g. print $fh -1 */
5858 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5860 PL_expect = XTERM; /* e.g. print $fh /.../
5861 XXX except DORDOR operator
5863 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5865 PL_expect = XTERM; /* print $fh <<"EOF" */
5868 PL_pending_ident = '$';
5872 if (PL_expect == XOPERATOR)
5874 PL_tokenbuf[0] = '@';
5875 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5876 if (!PL_tokenbuf[1]) {
5879 if (PL_lex_state == LEX_NORMAL)
5881 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5883 PL_tokenbuf[0] = '%';
5885 /* Warn about @ where they meant $. */
5886 if (*s == '[' || *s == '{') {
5887 if (ckWARN(WARN_SYNTAX)) {
5888 const char *t = s + 1;
5889 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5891 if (*t == '}' || *t == ']') {
5893 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5894 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5895 "Scalar value %.*s better written as $%.*s",
5896 (int)(t-PL_bufptr), PL_bufptr,
5897 (int)(t-PL_bufptr-1), PL_bufptr+1);
5902 PL_pending_ident = '@';
5905 case '/': /* may be division, defined-or, or pattern */
5906 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5910 case '?': /* may either be conditional or pattern */
5911 if (PL_expect == XOPERATOR) {
5919 /* A // operator. */
5929 /* Disable warning on "study /blah/" */
5930 if (PL_oldoldbufptr == PL_last_uni
5931 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5932 || memNE(PL_last_uni, "study", 5)
5933 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5936 s = scan_pat(s,OP_MATCH);
5937 TERM(sublex_start());
5941 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5942 #ifdef PERL_STRICT_CR
5945 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5947 && (s == PL_linestart || s[-1] == '\n') )
5949 PL_lex_formbrack = 0;
5953 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5957 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5963 pl_yylval.ival = OPf_SPECIAL;
5972 case '0': case '1': case '2': case '3': case '4':
5973 case '5': case '6': case '7': case '8': case '9':
5974 s = scan_num(s, &pl_yylval);
5975 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5976 if (PL_expect == XOPERATOR)
5981 s = scan_str(s,!!PL_madskills,FALSE);
5982 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5983 if (PL_expect == XOPERATOR) {
5984 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5985 return deprecate_commaless_var_list();
5992 pl_yylval.ival = OP_CONST;
5993 TERM(sublex_start());
5996 s = scan_str(s,!!PL_madskills,FALSE);
5997 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5998 if (PL_expect == XOPERATOR) {
5999 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6000 return deprecate_commaless_var_list();
6007 pl_yylval.ival = OP_CONST;
6008 /* FIXME. I think that this can be const if char *d is replaced by
6009 more localised variables. */
6010 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6011 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6012 pl_yylval.ival = OP_STRINGIFY;
6016 TERM(sublex_start());
6019 s = scan_str(s,!!PL_madskills,FALSE);
6020 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6021 if (PL_expect == XOPERATOR)
6022 no_op("Backticks",s);
6025 readpipe_override();
6026 TERM(sublex_start());
6030 if (PL_lex_inwhat && isDIGIT(*s))
6031 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6033 if (PL_expect == XOPERATOR)
6034 no_op("Backslash",s);
6038 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6039 char *start = s + 2;
6040 while (isDIGIT(*start) || *start == '_')
6042 if (*start == '.' && isDIGIT(start[1])) {
6043 s = scan_num(s, &pl_yylval);
6046 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6047 else if (!isALPHA(*start) && (PL_expect == XTERM
6048 || PL_expect == XREF || PL_expect == XSTATE
6049 || PL_expect == XTERMORDORDOR)) {
6050 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
6052 s = scan_num(s, &pl_yylval);
6059 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6102 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6104 /* Some keywords can be followed by any delimiter, including ':' */
6105 anydelim = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
6106 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
6107 (PL_tokenbuf[0] == 'q' &&
6108 strchr("qwxr", PL_tokenbuf[1])))));
6110 /* x::* is just a word, unless x is "CORE" */
6111 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6115 while (d < PL_bufend && isSPACE(*d))
6116 d++; /* no comments skipped here, or s### is misparsed */
6118 /* Is this a word before a => operator? */
6119 if (*d == '=' && d[1] == '>') {
6122 = (OP*)newSVOP(OP_CONST, 0,
6123 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6124 pl_yylval.opval->op_private = OPpCONST_BARE;
6128 /* Check for plugged-in keyword */
6132 char *saved_bufptr = PL_bufptr;
6134 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6136 if (result == KEYWORD_PLUGIN_DECLINE) {
6137 /* not a plugged-in keyword */
6138 PL_bufptr = saved_bufptr;
6139 } else if (result == KEYWORD_PLUGIN_STMT) {
6140 pl_yylval.opval = o;
6143 return REPORT(PLUGSTMT);
6144 } else if (result == KEYWORD_PLUGIN_EXPR) {
6145 pl_yylval.opval = o;
6147 PL_expect = XOPERATOR;
6148 return REPORT(PLUGEXPR);
6150 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6155 /* Check for built-in keyword */
6156 tmp = keyword(PL_tokenbuf, len, 0);
6158 /* Is this a label? */
6159 if (!anydelim && PL_expect == XSTATE
6160 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6162 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
6167 if (tmp < 0) { /* second-class keyword? */
6168 GV *ogv = NULL; /* override (winner) */
6169 GV *hgv = NULL; /* hidden (loser) */
6170 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6172 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
6175 if (GvIMPORTED_CV(gv))
6177 else if (! CvMETHOD(cv))
6181 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
6182 (gv = *gvp) && isGV_with_GP(gv) &&
6183 GvCVu(gv) && GvIMPORTED_CV(gv))
6190 tmp = 0; /* overridden by import or by GLOBAL */
6193 && -tmp==KEY_lock /* XXX generalizable kludge */
6196 tmp = 0; /* any sub overrides "weak" keyword */
6198 else { /* no override */
6200 if (tmp == KEY_dump) {
6201 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6202 "dump() better written as CORE::dump()");
6206 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6207 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6208 "Ambiguous call resolved as CORE::%s(), "
6209 "qualify as such or use &",
6217 default: /* not a keyword */
6218 /* Trade off - by using this evil construction we can pull the
6219 variable gv into the block labelled keylookup. If not, then
6220 we have to give it function scope so that the goto from the
6221 earlier ':' case doesn't bypass the initialisation. */
6223 just_a_word_zero_gv:
6231 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6235 SV *nextPL_nextwhite = 0;
6239 /* Get the rest if it looks like a package qualifier */
6241 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6243 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6246 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
6247 *s == '\'' ? "'" : "::");
6252 if (PL_expect == XOPERATOR) {
6253 if (PL_bufptr == PL_linestart) {
6254 CopLINE_dec(PL_curcop);
6255 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6256 CopLINE_inc(PL_curcop);
6259 no_op("Bareword",s);
6262 /* Look for a subroutine with this name in current package,
6263 unless name is "Foo::", in which case Foo is a bearword
6264 (and a package name). */
6266 if (len > 2 && !PL_madskills &&
6267 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6269 if (ckWARN(WARN_BAREWORD)
6270 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
6271 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6272 "Bareword \"%s\" refers to nonexistent package",
6275 PL_tokenbuf[len] = '\0';
6281 /* Mustn't actually add anything to a symbol table.
6282 But also don't want to "initialise" any placeholder
6283 constants that might already be there into full
6284 blown PVGVs with attached PVCV. */
6285 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6286 GV_NOADD_NOINIT, SVt_PVCV);
6291 /* if we saw a global override before, get the right name */
6294 sv = newSVpvs("CORE::GLOBAL::");
6295 sv_catpv(sv,PL_tokenbuf);
6298 /* If len is 0, newSVpv does strlen(), which is correct.
6299 If len is non-zero, then it will be the true length,
6300 and so the scalar will be created correctly. */
6301 sv = newSVpv(PL_tokenbuf,len);
6304 if (PL_madskills && !PL_thistoken) {
6305 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6306 PL_thistoken = newSVpvn(start,s - start);
6307 PL_realtokenstart = s - SvPVX(PL_linestr);
6311 /* Presume this is going to be a bareword of some sort. */
6314 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6315 pl_yylval.opval->op_private = OPpCONST_BARE;
6316 /* UTF-8 package name? */
6317 if (UTF && !IN_BYTES &&
6318 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
6321 /* And if "Foo::", then that's what it certainly is. */
6328 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc(sv));
6329 const_op->op_private = OPpCONST_BARE;
6330 rv2cv_op = newCVREF(0, const_op);
6332 if (rv2cv_op->op_type == OP_RV2CV &&
6333 (rv2cv_op->op_flags & OPf_KIDS)) {
6334 OP *rv_op = cUNOPx(rv2cv_op)->op_first;
6335 switch (rv_op->op_type) {
6337 SV *sv = cSVOPx_sv(rv_op);
6338 if (SvROK(sv) && SvTYPE(SvRV(sv)) == SVt_PVCV)
6342 GV *gv = cGVOPx_gv(rv_op);
6343 CV *maybe_cv = GvCVu(gv);
6344 if (maybe_cv && SvTYPE((SV*)maybe_cv) == SVt_PVCV)
6350 /* See if it's the indirect object for a list operator. */
6352 if (PL_oldoldbufptr &&
6353 PL_oldoldbufptr < PL_bufptr &&
6354 (PL_oldoldbufptr == PL_last_lop
6355 || PL_oldoldbufptr == PL_last_uni) &&
6356 /* NO SKIPSPACE BEFORE HERE! */
6357 (PL_expect == XREF ||
6358 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6360 bool immediate_paren = *s == '(';
6362 /* (Now we can afford to cross potential line boundary.) */
6363 s = SKIPSPACE2(s,nextPL_nextwhite);
6365 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6368 /* Two barewords in a row may indicate method call. */
6370 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6371 (tmp = intuit_method(s, gv, cv))) {
6376 /* If not a declared subroutine, it's an indirect object. */
6377 /* (But it's an indir obj regardless for sort.) */
6378 /* Also, if "_" follows a filetest operator, it's a bareword */
6381 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6383 (PL_last_lop_op != OP_MAPSTART &&
6384 PL_last_lop_op != OP_GREPSTART))))
6385 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6386 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6389 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6394 PL_expect = XOPERATOR;
6397 s = SKIPSPACE2(s,nextPL_nextwhite);
6398 PL_nextwhite = nextPL_nextwhite;
6403 /* Is this a word before a => operator? */
6404 if (*s == '=' && s[1] == '>' && !pkgname) {
6407 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6408 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6409 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6413 /* If followed by a paren, it's certainly a subroutine. */
6418 while (SPACE_OR_TAB(*d))
6420 if (*d == ')' && (sv = cv_const_sv(cv))) {
6427 PL_nextwhite = PL_thiswhite;
6430 start_force(PL_curforce);
6432 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6433 PL_expect = XOPERATOR;
6436 PL_nextwhite = nextPL_nextwhite;
6437 curmad('X', PL_thistoken);
6438 PL_thistoken = newSVpvs("");
6447 /* If followed by var or block, call it a method (unless sub) */
6449 if ((*s == '$' || *s == '{') && !cv) {
6451 PL_last_lop = PL_oldbufptr;
6452 PL_last_lop_op = OP_METHOD;
6456 /* If followed by a bareword, see if it looks like indir obj. */
6459 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6460 && (tmp = intuit_method(s, gv, cv))) {
6465 /* Not a method, so call it a subroutine (if defined) */
6468 if (lastchar == '-')
6469 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6470 "Ambiguous use of -%s resolved as -&%s()",
6471 PL_tokenbuf, PL_tokenbuf);
6472 /* Check for a constant sub */
6473 if ((sv = cv_const_sv(cv))) {
6476 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6477 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6478 pl_yylval.opval->op_private = 0;
6482 op_free(pl_yylval.opval);
6483 pl_yylval.opval = rv2cv_op;
6484 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6485 PL_last_lop = PL_oldbufptr;
6486 PL_last_lop_op = OP_ENTERSUB;
6487 /* Is there a prototype? */
6495 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
6498 while (*proto == ';')
6503 *proto == '$' || *proto == '_'
6509 *proto == '\\' && proto[1] && proto[2] == '\0'
6513 if (*proto == '\\' && proto[1] == '[') {
6514 const char *p = proto + 2;
6515 while(*p && *p != ']')
6517 if(*p == ']' && !p[1]) OPERATOR(UNIOPSUB);
6519 if (*proto == '&' && *s == '{') {
6521 sv_setpvs(PL_subname, "__ANON__");
6523 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6530 PL_nextwhite = PL_thiswhite;
6533 start_force(PL_curforce);
6534 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6537 PL_nextwhite = nextPL_nextwhite;
6538 curmad('X', PL_thistoken);
6539 PL_thistoken = newSVpvs("");
6546 /* Guess harder when madskills require "best effort". */
6547 if (PL_madskills && (!gv || !GvCVu(gv))) {
6548 int probable_sub = 0;
6549 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6551 else if (isALPHA(*s)) {
6555 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6556 if (!keyword(tmpbuf, tmplen, 0))
6559 while (d < PL_bufend && isSPACE(*d))
6561 if (*d == '=' && d[1] == '>')
6566 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
6567 op_free(pl_yylval.opval);
6568 pl_yylval.opval = rv2cv_op;
6569 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6570 PL_last_lop = PL_oldbufptr;
6571 PL_last_lop_op = OP_ENTERSUB;
6572 PL_nextwhite = PL_thiswhite;
6574 start_force(PL_curforce);
6575 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6577 PL_nextwhite = nextPL_nextwhite;
6578 curmad('X', PL_thistoken);
6579 PL_thistoken = newSVpvs("");
6584 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6591 /* Call it a bare word */
6593 if (PL_hints & HINT_STRICT_SUBS)
6594 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6597 /* after "print" and similar functions (corresponding to
6598 * "F? L" in opcode.pl), whatever wasn't already parsed as
6599 * a filehandle should be subject to "strict subs".
6600 * Likewise for the optional indirect-object argument to system
6601 * or exec, which can't be a bareword */
6602 if ((PL_last_lop_op == OP_PRINT
6603 || PL_last_lop_op == OP_PRTF
6604 || PL_last_lop_op == OP_SAY
6605 || PL_last_lop_op == OP_SYSTEM
6606 || PL_last_lop_op == OP_EXEC)
6607 && (PL_hints & HINT_STRICT_SUBS))
6608 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6609 if (lastchar != '-') {
6610 if (ckWARN(WARN_RESERVED)) {
6614 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
6615 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6623 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
6624 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6625 "Operator or semicolon missing before %c%s",
6626 lastchar, PL_tokenbuf);
6627 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6628 "Ambiguous use of %c resolved as operator %c",
6629 lastchar, lastchar);
6635 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6636 newSVpv(CopFILE(PL_curcop),0));
6640 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6641 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
6644 case KEY___PACKAGE__:
6645 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6647 ? newSVhek(HvNAME_HEK(PL_curstash))
6654 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6655 const char *pname = "main";
6656 if (PL_tokenbuf[2] == 'D')
6657 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
6658 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
6662 GvIOp(gv) = newIO();
6663 IoIFP(GvIOp(gv)) = PL_rsfp;
6664 #if defined(HAS_FCNTL) && defined(F_SETFD)
6666 const int fd = PerlIO_fileno(PL_rsfp);
6667 fcntl(fd,F_SETFD,fd >= 3);
6670 /* Mark this internal pseudo-handle as clean */
6671 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6672 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6673 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6675 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6676 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6677 /* if the script was opened in binmode, we need to revert
6678 * it to text mode for compatibility; but only iff it has CRs
6679 * XXX this is a questionable hack at best. */
6680 if (PL_bufend-PL_bufptr > 2
6681 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6684 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6685 loc = PerlIO_tell(PL_rsfp);
6686 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6689 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6691 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6692 #endif /* NETWARE */
6693 #ifdef PERLIO_IS_STDIO /* really? */
6694 # if defined(__BORLANDC__)
6695 /* XXX see note in do_binmode() */
6696 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
6700 PerlIO_seek(PL_rsfp, loc, 0);
6704 #ifdef PERLIO_LAYERS
6707 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6708 else if (PL_encoding) {
6715 XPUSHs(PL_encoding);
6717 call_method("name", G_SCALAR);
6721 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
6722 Perl_form(aTHX_ ":encoding(%"SVf")",
6731 if (PL_realtokenstart >= 0) {
6732 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6734 PL_endwhite = newSVpvs("");
6735 sv_catsv(PL_endwhite, PL_thiswhite);
6737 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
6738 PL_realtokenstart = -1;
6740 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
6756 if (PL_expect == XSTATE) {
6763 if (*s == ':' && s[1] == ':') {
6766 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6767 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
6768 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
6771 else if (tmp == KEY_require || tmp == KEY_do)
6772 /* that's a way to remember we saw "CORE::" */
6785 LOP(OP_ACCEPT,XTERM);
6791 LOP(OP_ATAN2,XTERM);
6797 LOP(OP_BINMODE,XTERM);
6800 LOP(OP_BLESS,XTERM);
6809 /* When 'use switch' is in effect, continue has a dual
6810 life as a control operator. */
6812 if (!FEATURE_IS_ENABLED("switch"))
6815 /* We have to disambiguate the two senses of
6816 "continue". If the next token is a '{' then
6817 treat it as the start of a continue block;
6818 otherwise treat it as a control operator.
6830 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
6847 if (!PL_cryptseen) {
6848 PL_cryptseen = TRUE;
6852 LOP(OP_CRYPT,XTERM);
6855 LOP(OP_CHMOD,XTERM);
6858 LOP(OP_CHOWN,XTERM);
6861 LOP(OP_CONNECT,XTERM);
6880 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6881 if (orig_keyword == KEY_do) {
6890 PL_hints |= HINT_BLOCK_SCOPE;
6900 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
6901 LOP(OP_DBMOPEN,XTERM);
6907 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6914 pl_yylval.ival = CopLINE(PL_curcop);
6930 if (*s == '{') { /* block eval */
6931 PL_expect = XTERMBLOCK;
6932 UNIBRACK(OP_ENTERTRY);
6934 else { /* string eval */
6936 UNIBRACK(OP_ENTEREVAL);
6951 case KEY_endhostent:
6957 case KEY_endservent:
6960 case KEY_endprotoent:
6971 pl_yylval.ival = CopLINE(PL_curcop);
6973 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6976 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6979 if ((PL_bufend - p) >= 3 &&
6980 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6982 else if ((PL_bufend - p) >= 4 &&
6983 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6986 if (isIDFIRST_lazy_if(p,UTF)) {
6987 p = scan_ident(p, PL_bufend,
6988 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6992 Perl_croak(aTHX_ "Missing $ on loop variable");
6994 s = SvPVX(PL_linestr) + soff;
7000 LOP(OP_FORMLINE,XTERM);
7006 LOP(OP_FCNTL,XTERM);
7012 LOP(OP_FLOCK,XTERM);
7021 LOP(OP_GREPSTART, XREF);
7024 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7039 case KEY_getpriority:
7040 LOP(OP_GETPRIORITY,XTERM);
7042 case KEY_getprotobyname:
7045 case KEY_getprotobynumber:
7046 LOP(OP_GPBYNUMBER,XTERM);
7048 case KEY_getprotoent:
7060 case KEY_getpeername:
7061 UNI(OP_GETPEERNAME);
7063 case KEY_gethostbyname:
7066 case KEY_gethostbyaddr:
7067 LOP(OP_GHBYADDR,XTERM);
7069 case KEY_gethostent:
7072 case KEY_getnetbyname:
7075 case KEY_getnetbyaddr:
7076 LOP(OP_GNBYADDR,XTERM);
7081 case KEY_getservbyname:
7082 LOP(OP_GSBYNAME,XTERM);
7084 case KEY_getservbyport:
7085 LOP(OP_GSBYPORT,XTERM);
7087 case KEY_getservent:
7090 case KEY_getsockname:
7091 UNI(OP_GETSOCKNAME);
7093 case KEY_getsockopt:
7094 LOP(OP_GSOCKOPT,XTERM);
7109 pl_yylval.ival = CopLINE(PL_curcop);
7119 pl_yylval.ival = CopLINE(PL_curcop);
7123 LOP(OP_INDEX,XTERM);
7129 LOP(OP_IOCTL,XTERM);
7141 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7173 LOP(OP_LISTEN,XTERM);
7182 s = scan_pat(s,OP_MATCH);
7183 TERM(sublex_start());
7186 LOP(OP_MAPSTART, XREF);
7189 LOP(OP_MKDIR,XTERM);
7192 LOP(OP_MSGCTL,XTERM);
7195 LOP(OP_MSGGET,XTERM);
7198 LOP(OP_MSGRCV,XTERM);
7201 LOP(OP_MSGSND,XTERM);
7206 PL_in_my = (U16)tmp;
7208 if (isIDFIRST_lazy_if(s,UTF)) {
7212 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7213 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7215 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7216 if (!PL_in_my_stash) {
7219 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7223 if (PL_madskills) { /* just add type to declarator token */
7224 sv_catsv(PL_thistoken, PL_nextwhite);
7226 sv_catpvn(PL_thistoken, start, s - start);
7234 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7241 s = tokenize_use(0, s);
7245 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7252 if (isIDFIRST_lazy_if(s,UTF)) {
7254 for (d = s; isALNUM_lazy_if(d,UTF);)
7256 for (t=d; isSPACE(*t);)
7258 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7260 && !(t[0] == '=' && t[1] == '>')
7262 int parms_len = (int)(d-s);
7263 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7264 "Precedence problem: open %.*s should be open(%.*s)",
7265 parms_len, s, parms_len, s);
7271 pl_yylval.ival = OP_OR;
7281 LOP(OP_OPEN_DIR,XTERM);
7284 checkcomma(s,PL_tokenbuf,"filehandle");
7288 checkcomma(s,PL_tokenbuf,"filehandle");
7307 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7309 s = force_strict_version(s);
7310 PL_lex_expect = XBLOCK;
7314 LOP(OP_PIPE_OP,XTERM);
7317 s = scan_str(s,!!PL_madskills,FALSE);
7320 pl_yylval.ival = OP_CONST;
7321 TERM(sublex_start());
7327 s = scan_str(s,!!PL_madskills,FALSE);
7330 PL_expect = XOPERATOR;
7332 if (SvCUR(PL_lex_stuff)) {
7335 d = SvPV_force(PL_lex_stuff, len);
7337 for (; isSPACE(*d) && len; --len, ++d)
7342 if (!warned && ckWARN(WARN_QW)) {
7343 for (; !isSPACE(*d) && len; --len, ++d) {
7345 Perl_warner(aTHX_ packWARN(WARN_QW),
7346 "Possible attempt to separate words with commas");
7349 else if (*d == '#') {
7350 Perl_warner(aTHX_ packWARN(WARN_QW),
7351 "Possible attempt to put comments in qw() list");
7357 for (; !isSPACE(*d) && len; --len, ++d)
7360 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7361 words = append_elem(OP_LIST, words,
7362 newSVOP(OP_CONST, 0, tokeq(sv)));
7366 start_force(PL_curforce);
7367 NEXTVAL_NEXTTOKE.opval = words;
7372 SvREFCNT_dec(PL_lex_stuff);
7373 PL_lex_stuff = NULL;
7379 s = scan_str(s,!!PL_madskills,FALSE);
7382 pl_yylval.ival = OP_STRINGIFY;
7383 if (SvIVX(PL_lex_stuff) == '\'')
7384 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
7385 TERM(sublex_start());
7388 s = scan_pat(s,OP_QR);
7389 TERM(sublex_start());
7392 s = scan_str(s,!!PL_madskills,FALSE);
7395 readpipe_override();
7396 TERM(sublex_start());
7404 s = force_version(s, FALSE);
7406 else if (*s != 'v' || !isDIGIT(s[1])
7407 || (s = force_version(s, TRUE), *s == 'v'))
7409 *PL_tokenbuf = '\0';
7410 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7411 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7412 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7414 yyerror("<> should be quotes");
7416 if (orig_keyword == KEY_require) {
7424 PL_last_uni = PL_oldbufptr;
7425 PL_last_lop_op = OP_REQUIRE;
7427 return REPORT( (int)REQUIRE );
7433 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7437 LOP(OP_RENAME,XTERM);
7446 LOP(OP_RINDEX,XTERM);
7455 UNIDOR(OP_READLINE);
7458 UNIDOR(OP_BACKTICK);
7467 LOP(OP_REVERSE,XTERM);
7470 UNIDOR(OP_READLINK);
7477 if (pl_yylval.opval)
7478 TERM(sublex_start());
7480 TOKEN(1); /* force error */
7483 checkcomma(s,PL_tokenbuf,"filehandle");
7493 LOP(OP_SELECT,XTERM);
7499 LOP(OP_SEMCTL,XTERM);
7502 LOP(OP_SEMGET,XTERM);
7505 LOP(OP_SEMOP,XTERM);
7511 LOP(OP_SETPGRP,XTERM);
7513 case KEY_setpriority:
7514 LOP(OP_SETPRIORITY,XTERM);
7516 case KEY_sethostent:
7522 case KEY_setservent:
7525 case KEY_setprotoent:
7535 LOP(OP_SEEKDIR,XTERM);
7537 case KEY_setsockopt:
7538 LOP(OP_SSOCKOPT,XTERM);
7544 LOP(OP_SHMCTL,XTERM);
7547 LOP(OP_SHMGET,XTERM);
7550 LOP(OP_SHMREAD,XTERM);
7553 LOP(OP_SHMWRITE,XTERM);
7556 LOP(OP_SHUTDOWN,XTERM);
7565 LOP(OP_SOCKET,XTERM);
7567 case KEY_socketpair:
7568 LOP(OP_SOCKPAIR,XTERM);
7571 checkcomma(s,PL_tokenbuf,"subroutine name");
7573 if (*s == ';' || *s == ')') /* probably a close */
7574 Perl_croak(aTHX_ "sort is now a reserved word");
7576 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7580 LOP(OP_SPLIT,XTERM);
7583 LOP(OP_SPRINTF,XTERM);
7586 LOP(OP_SPLICE,XTERM);
7601 LOP(OP_SUBSTR,XTERM);
7607 char tmpbuf[sizeof PL_tokenbuf];
7608 SSize_t tboffset = 0;
7609 expectation attrful;
7610 bool have_name, have_proto;
7611 const int key = tmp;
7616 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7617 SV *subtoken = newSVpvn(tstart, s - tstart);
7621 s = SKIPSPACE2(s,tmpwhite);
7626 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7627 (*s == ':' && s[1] == ':'))
7630 SV *nametoke = NULL;
7634 attrful = XATTRBLOCK;
7635 /* remember buffer pos'n for later force_word */
7636 tboffset = s - PL_oldbufptr;
7637 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7640 nametoke = newSVpvn(s, d - s);
7642 if (memchr(tmpbuf, ':', len))
7643 sv_setpvn(PL_subname, tmpbuf, len);
7645 sv_setsv(PL_subname,PL_curstname);
7646 sv_catpvs(PL_subname,"::");
7647 sv_catpvn(PL_subname,tmpbuf,len);
7654 CURMAD('X', nametoke);
7655 CURMAD('_', tmpwhite);
7656 (void) force_word(PL_oldbufptr + tboffset, WORD,
7659 s = SKIPSPACE2(d,tmpwhite);
7666 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7667 PL_expect = XTERMBLOCK;
7668 attrful = XATTRTERM;
7669 sv_setpvs(PL_subname,"?");
7673 if (key == KEY_format) {
7675 PL_lex_formbrack = PL_lex_brackets + 1;
7677 PL_thistoken = subtoken;
7681 (void) force_word(PL_oldbufptr + tboffset, WORD,
7687 /* Look for a prototype */
7690 bool bad_proto = FALSE;
7691 bool in_brackets = FALSE;
7692 char greedy_proto = ' ';
7693 bool proto_after_greedy_proto = FALSE;
7694 bool must_be_last = FALSE;
7695 bool underscore = FALSE;
7696 bool seen_underscore = FALSE;
7697 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
7699 s = scan_str(s,!!PL_madskills,FALSE);
7701 Perl_croak(aTHX_ "Prototype not terminated");
7702 /* strip spaces and check for bad characters */
7703 d = SvPVX(PL_lex_stuff);
7705 for (p = d; *p; ++p) {
7709 if (warnillegalproto) {
7711 proto_after_greedy_proto = TRUE;
7712 if (!strchr("$@%*;[]&\\_", *p)) {
7724 else if ( *p == ']' ) {
7725 in_brackets = FALSE;
7727 else if ( (*p == '@' || *p == '%') &&
7728 ( tmp < 2 || d[tmp-2] != '\\' ) &&
7730 must_be_last = TRUE;
7733 else if ( *p == '_' ) {
7734 underscore = seen_underscore = TRUE;
7741 if (proto_after_greedy_proto)
7742 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7743 "Prototype after '%c' for %"SVf" : %s",
7744 greedy_proto, SVfARG(PL_subname), d);
7746 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7747 "Illegal character %sin prototype for %"SVf" : %s",
7748 seen_underscore ? "after '_' " : "",
7749 SVfARG(PL_subname), d);
7750 SvCUR_set(PL_lex_stuff, tmp);
7755 CURMAD('q', PL_thisopen);
7756 CURMAD('_', tmpwhite);
7757 CURMAD('=', PL_thisstuff);
7758 CURMAD('Q', PL_thisclose);
7759 NEXTVAL_NEXTTOKE.opval =
7760 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7761 PL_lex_stuff = NULL;
7764 s = SKIPSPACE2(s,tmpwhite);
7772 if (*s == ':' && s[1] != ':')
7773 PL_expect = attrful;
7774 else if (*s != '{' && key == KEY_sub) {
7776 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
7777 else if (*s != ';' && *s != '}')
7778 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
7785 curmad('^', newSVpvs(""));
7786 CURMAD('_', tmpwhite);
7790 PL_thistoken = subtoken;
7793 NEXTVAL_NEXTTOKE.opval =
7794 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7795 PL_lex_stuff = NULL;
7801 sv_setpvs(PL_subname, "__ANON__");
7803 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7807 (void) force_word(PL_oldbufptr + tboffset, WORD,
7816 LOP(OP_SYSTEM,XREF);
7819 LOP(OP_SYMLINK,XTERM);
7822 LOP(OP_SYSCALL,XTERM);
7825 LOP(OP_SYSOPEN,XTERM);
7828 LOP(OP_SYSSEEK,XTERM);
7831 LOP(OP_SYSREAD,XTERM);
7834 LOP(OP_SYSWRITE,XTERM);
7838 TERM(sublex_start());
7859 LOP(OP_TRUNCATE,XTERM);
7871 pl_yylval.ival = CopLINE(PL_curcop);
7875 pl_yylval.ival = CopLINE(PL_curcop);
7879 LOP(OP_UNLINK,XTERM);
7885 LOP(OP_UNPACK,XTERM);
7888 LOP(OP_UTIME,XTERM);
7894 LOP(OP_UNSHIFT,XTERM);
7897 s = tokenize_use(1, s);
7907 pl_yylval.ival = CopLINE(PL_curcop);
7911 pl_yylval.ival = CopLINE(PL_curcop);
7915 PL_hints |= HINT_BLOCK_SCOPE;
7922 LOP(OP_WAITPID,XTERM);
7931 ctl_l[0] = toCTRL('L');
7933 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7936 /* Make sure $^L is defined */
7937 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7942 if (PL_expect == XOPERATOR)
7948 pl_yylval.ival = OP_XOR;
7953 TERM(sublex_start());
7958 #pragma segment Main
7962 S_pending_ident(pTHX)
7967 /* pit holds the identifier we read and pending_ident is reset */
7968 char pit = PL_pending_ident;
7969 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7970 /* All routes through this function want to know if there is a colon. */
7971 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7972 PL_pending_ident = 0;
7974 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7975 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7976 "### Pending identifier '%s'\n", PL_tokenbuf); });
7978 /* if we're in a my(), we can't allow dynamics here.
7979 $foo'bar has already been turned into $foo::bar, so
7980 just check for colons.
7982 if it's a legal name, the OP is a PADANY.
7985 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7987 yyerror(Perl_form(aTHX_ "No package name allowed for "
7988 "variable %s in \"our\"",
7990 tmp = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7994 yyerror(Perl_form(aTHX_ PL_no_myglob,
7995 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7997 pl_yylval.opval = newOP(OP_PADANY, 0);
7998 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len, 0);
8004 build the ops for accesses to a my() variable.
8006 Deny my($a) or my($b) in a sort block, *if* $a or $b is
8007 then used in a comparison. This catches most, but not
8008 all cases. For instance, it catches
8009 sort { my($a); $a <=> $b }
8011 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
8012 (although why you'd do that is anyone's guess).
8017 tmp = pad_findmy(PL_tokenbuf, tokenbuf_len, 0);
8018 if (tmp != NOT_IN_PAD) {
8019 /* might be an "our" variable" */
8020 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8021 /* build ops for a bareword */
8022 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8023 HEK * const stashname = HvNAME_HEK(stash);
8024 SV * const sym = newSVhek(stashname);
8025 sv_catpvs(sym, "::");
8026 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
8027 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8028 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8031 ? (GV_ADDMULTI | GV_ADDINEVAL)
8034 ((PL_tokenbuf[0] == '$') ? SVt_PV
8035 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8040 /* if it's a sort block and they're naming $a or $b */
8041 if (PL_last_lop_op == OP_SORT &&
8042 PL_tokenbuf[0] == '$' &&
8043 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
8046 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
8047 d < PL_bufend && *d != '\n';
8050 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
8051 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
8057 pl_yylval.opval = newOP(OP_PADANY, 0);
8058 pl_yylval.opval->op_targ = tmp;
8064 Whine if they've said @foo in a doublequoted string,
8065 and @foo isn't a variable we can find in the symbol
8068 if (ckWARN(WARN_AMBIGUOUS) &&
8069 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8070 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
8072 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8073 /* DO NOT warn for @- and @+ */
8074 && !( PL_tokenbuf[2] == '\0' &&
8075 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8078 /* Downgraded from fatal to warning 20000522 mjd */
8079 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8080 "Possible unintended interpolation of %s in string",
8085 /* build ops for a bareword */
8086 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
8088 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8089 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8090 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD,
8091 ((PL_tokenbuf[0] == '$') ? SVt_PV
8092 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8098 * The following code was generated by perl_keyword.pl.
8102 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
8106 PERL_ARGS_ASSERT_KEYWORD;
8110 case 1: /* 5 tokens of length 1 */
8142 case 2: /* 18 tokens of length 2 */
8288 case 3: /* 29 tokens of length 3 */
8292 if (name[1] == 'N' &&
8355 if (name[1] == 'i' &&
8387 if (name[1] == 'o' &&
8396 if (name[1] == 'e' &&
8405 if (name[1] == 'n' &&
8414 if (name[1] == 'o' &&
8423 if (name[1] == 'a' &&
8432 if (name[1] == 'o' &&
8494 if (name[1] == 'e' &&
8508 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
8534 if (name[1] == 'i' &&
8543 if (name[1] == 's' &&
8552 if (name[1] == 'e' &&
8561 if (name[1] == 'o' &&
8573 case 4: /* 41 tokens of length 4 */
8577 if (name[1] == 'O' &&
8587 if (name[1] == 'N' &&
8597 if (name[1] == 'i' &&
8607 if (name[1] == 'h' &&
8617 if (name[1] == 'u' &&
8630 if (name[2] == 'c' &&
8639 if (name[2] == 's' &&
8648 if (name[2] == 'a' &&
8684 if (name[1] == 'o' &&
8697 if (name[2] == 't' &&
8706 if (name[2] == 'o' &&
8715 if (name[2] == 't' &&
8724 if (name[2] == 'e' &&
8737 if (name[1] == 'o' &&
8750 if (name[2] == 'y' &&
8759 if (name[2] == 'l' &&
8775 if (name[2] == 's' &&
8784 if (name[2] == 'n' &&
8793 if (name[2] == 'c' &&
8806 if (name[1] == 'e' &&
8816 if (name[1] == 'p' &&
8829 if (name[2] == 'c' &&
8838 if (name[2] == 'p' &&
8847 if (name[2] == 's' &&
8863 if (name[2] == 'n' &&
8933 if (name[2] == 'r' &&
8942 if (name[2] == 'r' &&
8951 if (name[2] == 'a' &&
8967 if (name[2] == 'l' &&
9029 if (name[2] == 'e' &&
9032 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
9045 case 5: /* 39 tokens of length 5 */
9049 if (name[1] == 'E' &&
9060 if (name[1] == 'H' &&
9074 if (name[2] == 'a' &&
9084 if (name[2] == 'a' &&
9101 if (name[2] == 'e' &&
9111 if (name[2] == 'e' &&
9115 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
9131 if (name[3] == 'i' &&
9140 if (name[3] == 'o' &&
9176 if (name[2] == 'o' &&
9186 if (name[2] == 'y' &&
9200 if (name[1] == 'l' &&
9214 if (name[2] == 'n' &&
9224 if (name[2] == 'o' &&
9238 if (name[1] == 'i' &&
9243 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
9252 if (name[2] == 'd' &&
9262 if (name[2] == 'c' &&
9279 if (name[2] == 'c' &&
9289 if (name[2] == 't' &&
9303 if (name[1] == 'k' &&
9314 if (name[1] == 'r' &&
9328 if (name[2] == 's' &&
9338 if (name[2] == 'd' &&
9355 if (name[2] == 'm' &&
9365 if (name[2] == 'i' &&
9375 if (name[2] == 'e' &&
9385 if (name[2] == 'l' &&
9395 if (name[2] == 'a' &&
9408 if (name[3] == 't' &&
9411 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
9417 if (name[3] == 'd' &&
9434 if (name[1] == 'i' &&
9448 if (name[2] == 'a' &&
9461 if (name[3] == 'e' &&
9496 if (name[2] == 'i' &&
9513 if (name[2] == 'i' &&
9523 if (name[2] == 'i' &&
9540 case 6: /* 33 tokens of length 6 */
9544 if (name[1] == 'c' &&
9559 if (name[2] == 'l' &&
9570 if (name[2] == 'r' &&
9585 if (name[1] == 'e' &&
9600 if (name[2] == 's' &&
9605 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
9611 if (name[2] == 'i' &&
9629 if (name[2] == 'l' &&
9640 if (name[2] == 'r' &&
9655 if (name[1] == 'm' &&
9670 if (name[2] == 'n' &&
9681 if (name[2] == 's' &&
9696 if (name[1] == 's' &&
9702 if (name[4] == 't' &&
9711 if (name[4] == 'e' &&
9720 if (name[4] == 'c' &&
9729 if (name[4] == 'n' &&
9745 if (name[1] == 'r' &&
9763 if (name[3] == 'a' &&
9773 if (name[3] == 'u' &&
9787 if (name[2] == 'n' &&
9805 if (name[2] == 'a' &&
9819 if (name[3] == 'e' &&
9832 if (name[4] == 't' &&
9841 if (name[4] == 'e' &&
9863 if (name[4] == 't' &&
9872 if (name[4] == 'e' &&
9888 if (name[2] == 'c' &&
9899 if (name[2] == 'l' &&
9910 if (name[2] == 'b' &&
9921 if (name[2] == 's' &&
9944 if (name[4] == 's' &&
9953 if (name[4] == 'n' &&
9966 if (name[3] == 'a' &&
9983 if (name[1] == 'a' &&
9998 case 7: /* 29 tokens of length 7 */
10002 if (name[1] == 'E' &&
10009 return KEY_DESTROY;
10015 if (name[1] == '_' &&
10022 return KEY___END__;
10028 if (name[1] == 'i' &&
10035 return -KEY_binmode;
10041 if (name[1] == 'o' &&
10048 return -KEY_connect;
10057 if (name[2] == 'm' &&
10063 return -KEY_dbmopen;
10069 if (name[2] == 'f')
10074 if (name[4] == 'u' &&
10078 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
10084 if (name[4] == 'n' &&
10088 return KEY_defined;
10105 if (name[1] == 'o' &&
10112 return KEY_foreach;
10118 if (name[1] == 'e' &&
10125 if (name[5] == 'r' &&
10128 return -KEY_getpgrp;
10134 if (name[5] == 'i' &&
10137 return -KEY_getppid;
10150 if (name[1] == 'c' &&
10157 return -KEY_lcfirst;
10163 if (name[1] == 'p' &&
10170 return -KEY_opendir;
10176 if (name[1] == 'a' &&
10183 return KEY_package;
10189 if (name[1] == 'e')
10194 if (name[3] == 'd' &&
10199 return -KEY_readdir;
10205 if (name[3] == 'u' &&
10210 return KEY_require;
10216 if (name[3] == 'e' &&
10221 return -KEY_reverse;
10240 if (name[3] == 'k' &&
10245 return -KEY_seekdir;
10251 if (name[3] == 'p' &&
10256 return -KEY_setpgrp;
10266 if (name[2] == 'm' &&
10272 return -KEY_shmread;
10278 if (name[2] == 'r' &&
10284 return -KEY_sprintf;
10293 if (name[3] == 'l' &&
10298 return -KEY_symlink;
10307 if (name[4] == 'a' &&
10311 return -KEY_syscall;
10317 if (name[4] == 'p' &&
10321 return -KEY_sysopen;
10327 if (name[4] == 'e' &&
10331 return -KEY_sysread;
10337 if (name[4] == 'e' &&
10341 return -KEY_sysseek;
10359 if (name[1] == 'e' &&
10366 return -KEY_telldir;
10375 if (name[2] == 'f' &&
10381 return -KEY_ucfirst;
10387 if (name[2] == 's' &&
10393 return -KEY_unshift;
10403 if (name[1] == 'a' &&
10410 return -KEY_waitpid;
10419 case 8: /* 26 tokens of length 8 */
10423 if (name[1] == 'U' &&
10431 return KEY_AUTOLOAD;
10437 if (name[1] == '_')
10442 if (name[3] == 'A' &&
10448 return KEY___DATA__;
10454 if (name[3] == 'I' &&
10460 return -KEY___FILE__;
10466 if (name[3] == 'I' &&
10472 return -KEY___LINE__;
10488 if (name[2] == 'o' &&
10495 return -KEY_closedir;
10501 if (name[2] == 'n' &&
10508 return -KEY_continue;
10518 if (name[1] == 'b' &&
10526 return -KEY_dbmclose;
10532 if (name[1] == 'n' &&
10538 if (name[4] == 'r' &&
10543 return -KEY_endgrent;
10549 if (name[4] == 'w' &&
10554 return -KEY_endpwent;
10567 if (name[1] == 'o' &&
10575 return -KEY_formline;
10581 if (name[1] == 'e' &&
10587 if (name[4] == 'r')
10592 if (name[6] == 'n' &&
10595 return -KEY_getgrent;
10601 if (name[6] == 'i' &&
10604 return -KEY_getgrgid;
10610 if (name[6] == 'a' &&
10613 return -KEY_getgrnam;
10626 if (name[4] == 'o' &&
10631 return -KEY_getlogin;
10637 if (name[4] == 'w')
10642 if (name[6] == 'n' &&
10645 return -KEY_getpwent;
10651 if (name[6] == 'a' &&
10654 return -KEY_getpwnam;
10660 if (name[6] == 'i' &&
10663 return -KEY_getpwuid;
10683 if (name[1] == 'e' &&
10690 if (name[5] == 'i' &&
10697 return -KEY_readline;
10702 return -KEY_readlink;
10713 if (name[5] == 'i' &&
10717 return -KEY_readpipe;
10733 if (name[2] == 't')
10738 if (name[4] == 'r' &&
10743 return -KEY_setgrent;
10749 if (name[4] == 'w' &&
10754 return -KEY_setpwent;
10770 if (name[3] == 'w' &&
10776 return -KEY_shmwrite;
10782 if (name[3] == 't' &&
10788 return -KEY_shutdown;
10798 if (name[2] == 's' &&
10805 return -KEY_syswrite;
10815 if (name[1] == 'r' &&
10823 return -KEY_truncate;
10832 case 9: /* 9 tokens of length 9 */
10836 if (name[1] == 'N' &&
10845 return KEY_UNITCHECK;
10851 if (name[1] == 'n' &&
10860 return -KEY_endnetent;
10866 if (name[1] == 'e' &&
10875 return -KEY_getnetent;
10881 if (name[1] == 'o' &&
10890 return -KEY_localtime;
10896 if (name[1] == 'r' &&
10905 return KEY_prototype;
10911 if (name[1] == 'u' &&
10920 return -KEY_quotemeta;
10926 if (name[1] == 'e' &&
10935 return -KEY_rewinddir;
10941 if (name[1] == 'e' &&
10950 return -KEY_setnetent;
10956 if (name[1] == 'a' &&
10965 return -KEY_wantarray;
10974 case 10: /* 9 tokens of length 10 */
10978 if (name[1] == 'n' &&
10984 if (name[4] == 'o' &&
10991 return -KEY_endhostent;
10997 if (name[4] == 'e' &&
11004 return -KEY_endservent;
11017 if (name[1] == 'e' &&
11023 if (name[4] == 'o' &&
11030 return -KEY_gethostent;
11039 if (name[5] == 'r' &&
11045 return -KEY_getservent;
11051 if (name[5] == 'c' &&
11057 return -KEY_getsockopt;
11077 if (name[2] == 't')
11082 if (name[4] == 'o' &&
11089 return -KEY_sethostent;
11098 if (name[5] == 'r' &&
11104 return -KEY_setservent;
11110 if (name[5] == 'c' &&
11116 return -KEY_setsockopt;
11133 if (name[2] == 'c' &&
11142 return -KEY_socketpair;
11155 case 11: /* 8 tokens of length 11 */
11159 if (name[1] == '_' &&
11169 { /* __PACKAGE__ */
11170 return -KEY___PACKAGE__;
11176 if (name[1] == 'n' &&
11186 { /* endprotoent */
11187 return -KEY_endprotoent;
11193 if (name[1] == 'e' &&
11202 if (name[5] == 'e' &&
11208 { /* getpeername */
11209 return -KEY_getpeername;
11218 if (name[6] == 'o' &&
11223 { /* getpriority */
11224 return -KEY_getpriority;
11230 if (name[6] == 't' &&
11235 { /* getprotoent */
11236 return -KEY_getprotoent;
11250 if (name[4] == 'o' &&
11257 { /* getsockname */
11258 return -KEY_getsockname;
11271 if (name[1] == 'e' &&
11279 if (name[6] == 'o' &&
11284 { /* setpriority */
11285 return -KEY_setpriority;
11291 if (name[6] == 't' &&
11296 { /* setprotoent */
11297 return -KEY_setprotoent;
11313 case 12: /* 2 tokens of length 12 */
11314 if (name[0] == 'g' &&
11326 if (name[9] == 'd' &&
11329 { /* getnetbyaddr */
11330 return -KEY_getnetbyaddr;
11336 if (name[9] == 'a' &&
11339 { /* getnetbyname */
11340 return -KEY_getnetbyname;
11352 case 13: /* 4 tokens of length 13 */
11353 if (name[0] == 'g' &&
11360 if (name[4] == 'o' &&
11369 if (name[10] == 'd' &&
11372 { /* gethostbyaddr */
11373 return -KEY_gethostbyaddr;
11379 if (name[10] == 'a' &&
11382 { /* gethostbyname */
11383 return -KEY_gethostbyname;
11396 if (name[4] == 'e' &&
11405 if (name[10] == 'a' &&
11408 { /* getservbyname */
11409 return -KEY_getservbyname;
11415 if (name[10] == 'o' &&
11418 { /* getservbyport */
11419 return -KEY_getservbyport;
11438 case 14: /* 1 tokens of length 14 */
11439 if (name[0] == 'g' &&
11453 { /* getprotobyname */
11454 return -KEY_getprotobyname;
11459 case 16: /* 1 tokens of length 16 */
11460 if (name[0] == 'g' &&
11476 { /* getprotobynumber */
11477 return -KEY_getprotobynumber;
11491 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
11495 PERL_ARGS_ASSERT_CHECKCOMMA;
11497 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
11498 if (ckWARN(WARN_SYNTAX)) {
11501 for (w = s+2; *w && level; w++) {
11504 else if (*w == ')')
11507 while (isSPACE(*w))
11509 /* the list of chars below is for end of statements or
11510 * block / parens, boolean operators (&&, ||, //) and branch
11511 * constructs (or, and, if, until, unless, while, err, for).
11512 * Not a very solid hack... */
11513 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
11514 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11515 "%s (...) interpreted as function",name);
11518 while (s < PL_bufend && isSPACE(*s))
11522 while (s < PL_bufend && isSPACE(*s))
11524 if (isIDFIRST_lazy_if(s,UTF)) {
11525 const char * const w = s++;
11526 while (isALNUM_lazy_if(s,UTF))
11528 while (s < PL_bufend && isSPACE(*s))
11532 if (keyword(w, s - w, 0))
11535 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
11536 if (gv && GvCVu(gv))
11538 Perl_croak(aTHX_ "No comma allowed after %s", what);
11543 /* Either returns sv, or mortalizes sv and returns a new SV*.
11544 Best used as sv=new_constant(..., sv, ...).
11545 If s, pv are NULL, calls subroutine with one argument,
11546 and type is used with error messages only. */
11549 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
11550 SV *sv, SV *pv, const char *type, STRLEN typelen)
11553 HV * const table = GvHV(PL_hintgv); /* ^H */
11557 const char *why1 = "", *why2 = "", *why3 = "";
11559 PERL_ARGS_ASSERT_NEW_CONSTANT;
11561 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
11564 why2 = (const char *)
11565 (strEQ(key,"charnames")
11566 ? "(possibly a missing \"use charnames ...\")"
11568 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
11569 (type ? type: "undef"), why2);
11571 /* This is convoluted and evil ("goto considered harmful")
11572 * but I do not understand the intricacies of all the different
11573 * failure modes of %^H in here. The goal here is to make
11574 * the most probable error message user-friendly. --jhi */
11579 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
11580 (type ? type: "undef"), why1, why2, why3);
11582 yyerror(SvPVX_const(msg));
11587 /* charnames doesn't work well if there have been errors found */
11588 if (PL_error_count > 0 && strEQ(key,"charnames"))
11589 return &PL_sv_undef;
11591 cvp = hv_fetch(table, key, keylen, FALSE);
11592 if (!cvp || !SvOK(*cvp)) {
11595 why3 = "} is not defined";
11598 sv_2mortal(sv); /* Parent created it permanently */
11601 pv = newSVpvn_flags(s, len, SVs_TEMP);
11603 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
11605 typesv = &PL_sv_undef;
11607 PUSHSTACKi(PERLSI_OVERLOAD);
11619 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
11623 /* Check the eval first */
11624 if (!PL_in_eval && SvTRUE(ERRSV)) {
11625 sv_catpvs(ERRSV, "Propagated");
11626 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
11628 res = SvREFCNT_inc_simple(sv);
11632 SvREFCNT_inc_simple_void(res);
11641 why1 = "Call to &{$^H{";
11643 why3 = "}} did not return a defined value";
11651 /* Returns a NUL terminated string, with the length of the string written to
11655 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
11658 register char *d = dest;
11659 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11661 PERL_ARGS_ASSERT_SCAN_WORD;
11665 Perl_croak(aTHX_ ident_too_long);
11666 if (isALNUM(*s)) /* UTF handled below */
11668 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
11673 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
11677 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11678 char *t = s + UTF8SKIP(s);
11680 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11684 Perl_croak(aTHX_ ident_too_long);
11685 Copy(s, d, len, char);
11698 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
11701 char *bracket = NULL;
11703 register char *d = dest;
11704 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11706 PERL_ARGS_ASSERT_SCAN_IDENT;
11711 while (isDIGIT(*s)) {
11713 Perl_croak(aTHX_ ident_too_long);
11720 Perl_croak(aTHX_ ident_too_long);
11721 if (isALNUM(*s)) /* UTF handled below */
11723 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
11728 else if (*s == ':' && s[1] == ':') {
11732 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11733 char *t = s + UTF8SKIP(s);
11734 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11736 if (d + (t - s) > e)
11737 Perl_croak(aTHX_ ident_too_long);
11738 Copy(s, d, t - s, char);
11749 if (PL_lex_state != LEX_NORMAL)
11750 PL_lex_state = LEX_INTERPENDMAYBE;
11753 if (*s == '$' && s[1] &&
11754 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
11767 if (*d == '^' && *s && isCONTROLVAR(*s)) {
11772 if (isSPACE(s[-1])) {
11774 const char ch = *s++;
11775 if (!SPACE_OR_TAB(ch)) {
11781 if (isIDFIRST_lazy_if(d,UTF)) {
11785 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
11786 end += UTF8SKIP(end);
11787 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
11788 end += UTF8SKIP(end);
11790 Copy(s, d, end - s, char);
11795 while ((isALNUM(*s) || *s == ':') && d < e)
11798 Perl_croak(aTHX_ ident_too_long);
11801 while (s < send && SPACE_OR_TAB(*s))
11803 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
11804 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
11805 const char * const brack =
11807 ((*s == '[') ? "[...]" : "{...}");
11808 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11809 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
11810 funny, dest, brack, funny, dest, brack);
11813 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
11817 /* Handle extended ${^Foo} variables
11818 * 1999-02-27 mjd-perl-patch@plover.com */
11819 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
11823 while (isALNUM(*s) && d < e) {
11827 Perl_croak(aTHX_ ident_too_long);
11832 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
11833 PL_lex_state = LEX_INTERPEND;
11836 if (PL_lex_state == LEX_NORMAL) {
11837 if (ckWARN(WARN_AMBIGUOUS) &&
11838 (keyword(dest, d - dest, 0)
11839 || get_cvn_flags(dest, d - dest, 0)))
11843 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11844 "Ambiguous use of %c{%s} resolved to %c%s",
11845 funny, dest, funny, dest);
11850 s = bracket; /* let the parser handle it */
11854 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
11855 PL_lex_state = LEX_INTERPEND;
11860 S_pmflag(U32 pmfl, const char ch) {
11862 CASE_STD_PMMOD_FLAGS_PARSE_SET(&pmfl);
11863 case GLOBAL_PAT_MOD: pmfl |= PMf_GLOBAL; break;
11864 case CONTINUE_PAT_MOD: pmfl |= PMf_CONTINUE; break;
11865 case ONCE_PAT_MOD: pmfl |= PMf_KEEP; break;
11866 case KEEPCOPY_PAT_MOD: pmfl |= PMf_KEEPCOPY; break;
11867 case NONDESTRUCT_PAT_MOD: pmfl |= PMf_NONDESTRUCT; break;
11873 S_scan_pat(pTHX_ char *start, I32 type)
11877 char *s = scan_str(start,!!PL_madskills,FALSE);
11878 const char * const valid_flags =
11879 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
11884 PERL_ARGS_ASSERT_SCAN_PAT;
11887 const char * const delimiter = skipspace(start);
11891 ? "Search pattern not terminated or ternary operator parsed as search pattern"
11892 : "Search pattern not terminated" ));
11895 pm = (PMOP*)newPMOP(type, 0);
11896 if (PL_multi_open == '?') {
11897 /* This is the only point in the code that sets PMf_ONCE: */
11898 pm->op_pmflags |= PMf_ONCE;
11900 /* Hence it's safe to do this bit of PMOP book-keeping here, which
11901 allows us to restrict the list needed by reset to just the ??
11903 assert(type != OP_TRANS);
11905 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11908 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11911 elements = mg->mg_len / sizeof(PMOP**);
11912 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11913 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11914 mg->mg_len = elements * sizeof(PMOP**);
11915 PmopSTASH_set(pm,PL_curstash);
11921 while (*s && strchr(valid_flags, *s))
11922 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11925 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX),
11926 "Having no space between pattern and following word is deprecated");
11930 if (PL_madskills && modstart != s) {
11931 SV* tmptoken = newSVpvn(modstart, s - modstart);
11932 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11935 /* issue a warning if /c is specified,but /g is not */
11936 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11938 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11939 "Use of /c modifier is meaningless without /g" );
11942 PL_lex_op = (OP*)pm;
11943 pl_yylval.ival = OP_MATCH;
11948 S_scan_subst(pTHX_ char *start)
11959 PERL_ARGS_ASSERT_SCAN_SUBST;
11961 pl_yylval.ival = OP_NULL;
11963 s = scan_str(start,!!PL_madskills,FALSE);
11966 Perl_croak(aTHX_ "Substitution pattern not terminated");
11968 if (s[-1] == PL_multi_open)
11971 if (PL_madskills) {
11972 CURMAD('q', PL_thisopen);
11973 CURMAD('_', PL_thiswhite);
11974 CURMAD('E', PL_thisstuff);
11975 CURMAD('Q', PL_thisclose);
11976 PL_realtokenstart = s - SvPVX(PL_linestr);
11980 first_start = PL_multi_start;
11981 s = scan_str(s,!!PL_madskills,FALSE);
11983 if (PL_lex_stuff) {
11984 SvREFCNT_dec(PL_lex_stuff);
11985 PL_lex_stuff = NULL;
11987 Perl_croak(aTHX_ "Substitution replacement not terminated");
11989 PL_multi_start = first_start; /* so whole substitution is taken together */
11991 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11994 if (PL_madskills) {
11995 CURMAD('z', PL_thisopen);
11996 CURMAD('R', PL_thisstuff);
11997 CURMAD('Z', PL_thisclose);
12003 if (*s == EXEC_PAT_MOD) {
12007 else if (strchr(S_PAT_MODS, *s))
12008 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
12011 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX),
12012 "Having no space between pattern and following word is deprecated");
12020 if (PL_madskills) {
12022 curmad('m', newSVpvn(modstart, s - modstart));
12023 append_madprops(PL_thismad, (OP*)pm, 0);
12027 if ((pm->op_pmflags & PMf_CONTINUE)) {
12028 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
12032 SV * const repl = newSVpvs("");
12034 PL_sublex_info.super_bufptr = s;
12035 PL_sublex_info.super_bufend = PL_bufend;
12037 pm->op_pmflags |= PMf_EVAL;
12040 sv_catpvs(repl, "eval ");
12042 sv_catpvs(repl, "do ");
12044 sv_catpvs(repl, "{");
12045 sv_catsv(repl, PL_lex_repl);
12046 if (strchr(SvPVX(PL_lex_repl), '#'))
12047 sv_catpvs(repl, "\n");
12048 sv_catpvs(repl, "}");
12050 SvREFCNT_dec(PL_lex_repl);
12051 PL_lex_repl = repl;
12054 PL_lex_op = (OP*)pm;
12055 pl_yylval.ival = OP_SUBST;
12060 S_scan_trans(pTHX_ char *start)
12073 PERL_ARGS_ASSERT_SCAN_TRANS;
12075 pl_yylval.ival = OP_NULL;
12077 s = scan_str(start,!!PL_madskills,FALSE);
12079 Perl_croak(aTHX_ "Transliteration pattern not terminated");
12081 if (s[-1] == PL_multi_open)
12084 if (PL_madskills) {
12085 CURMAD('q', PL_thisopen);
12086 CURMAD('_', PL_thiswhite);
12087 CURMAD('E', PL_thisstuff);
12088 CURMAD('Q', PL_thisclose);
12089 PL_realtokenstart = s - SvPVX(PL_linestr);
12093 s = scan_str(s,!!PL_madskills,FALSE);
12095 if (PL_lex_stuff) {
12096 SvREFCNT_dec(PL_lex_stuff);
12097 PL_lex_stuff = NULL;
12099 Perl_croak(aTHX_ "Transliteration replacement not terminated");
12101 if (PL_madskills) {
12102 CURMAD('z', PL_thisopen);
12103 CURMAD('R', PL_thisstuff);
12104 CURMAD('Z', PL_thisclose);
12107 complement = del = squash = 0;
12114 complement = OPpTRANS_COMPLEMENT;
12117 del = OPpTRANS_DELETE;
12120 squash = OPpTRANS_SQUASH;
12129 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
12130 o = newPVOP(OP_TRANS, 0, (char*)tbl);
12131 o->op_private &= ~OPpTRANS_ALL;
12132 o->op_private |= del|squash|complement|
12133 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
12134 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
12137 pl_yylval.ival = OP_TRANS;
12140 if (PL_madskills) {
12142 curmad('m', newSVpvn(modstart, s - modstart));
12143 append_madprops(PL_thismad, o, 0);
12152 S_scan_heredoc(pTHX_ register char *s)
12156 I32 op_type = OP_SCALAR;
12160 const char *found_newline;
12164 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
12166 I32 stuffstart = s - SvPVX(PL_linestr);
12169 PL_realtokenstart = -1;
12172 PERL_ARGS_ASSERT_SCAN_HEREDOC;
12176 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
12180 while (SPACE_OR_TAB(*peek))
12182 if (*peek == '`' || *peek == '\'' || *peek =='"') {
12185 s = delimcpy(d, e, s, PL_bufend, term, &len);
12195 if (!isALNUM_lazy_if(s,UTF))
12196 deprecate("bare << to mean <<\"\"");
12197 for (; isALNUM_lazy_if(s,UTF); s++) {
12202 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
12203 Perl_croak(aTHX_ "Delimiter for here document is too long");
12206 len = d - PL_tokenbuf;
12209 if (PL_madskills) {
12210 tstart = PL_tokenbuf + !outer;
12211 PL_thisclose = newSVpvn(tstart, len - !outer);
12212 tstart = SvPVX(PL_linestr) + stuffstart;
12213 PL_thisopen = newSVpvn(tstart, s - tstart);
12214 stuffstart = s - SvPVX(PL_linestr);
12217 #ifndef PERL_STRICT_CR
12218 d = strchr(s, '\r');
12220 char * const olds = s;
12222 while (s < PL_bufend) {
12228 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
12237 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12244 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
12245 herewas = newSVpvn(s,PL_bufend-s);
12249 herewas = newSVpvn(s-1,found_newline-s+1);
12252 herewas = newSVpvn(s,found_newline-s);
12256 if (PL_madskills) {
12257 tstart = SvPVX(PL_linestr) + stuffstart;
12259 sv_catpvn(PL_thisstuff, tstart, s - tstart);
12261 PL_thisstuff = newSVpvn(tstart, s - tstart);
12264 s += SvCUR(herewas);
12267 stuffstart = s - SvPVX(PL_linestr);
12273 tmpstr = newSV_type(SVt_PVIV);
12274 SvGROW(tmpstr, 80);
12275 if (term == '\'') {
12276 op_type = OP_CONST;
12277 SvIV_set(tmpstr, -1);
12279 else if (term == '`') {
12280 op_type = OP_BACKTICK;
12281 SvIV_set(tmpstr, '\\');
12285 PL_multi_start = CopLINE(PL_curcop);
12286 PL_multi_open = PL_multi_close = '<';
12287 term = *PL_tokenbuf;
12288 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
12289 char * const bufptr = PL_sublex_info.super_bufptr;
12290 char * const bufend = PL_sublex_info.super_bufend;
12291 char * const olds = s - SvCUR(herewas);
12292 s = strchr(bufptr, '\n');
12296 while (s < bufend &&
12297 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12299 CopLINE_inc(PL_curcop);
12302 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12303 missingterm(PL_tokenbuf);
12305 sv_setpvn(herewas,bufptr,d-bufptr+1);
12306 sv_setpvn(tmpstr,d+1,s-d);
12308 sv_catpvn(herewas,s,bufend-s);
12309 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
12316 while (s < PL_bufend &&
12317 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12319 CopLINE_inc(PL_curcop);
12321 if (s >= PL_bufend) {
12322 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12323 missingterm(PL_tokenbuf);
12325 sv_setpvn(tmpstr,d+1,s-d);
12327 if (PL_madskills) {
12329 sv_catpvn(PL_thisstuff, d + 1, s - d);
12331 PL_thisstuff = newSVpvn(d + 1, s - d);
12332 stuffstart = s - SvPVX(PL_linestr);
12336 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
12338 sv_catpvn(herewas,s,PL_bufend-s);
12339 sv_setsv(PL_linestr,herewas);
12340 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
12341 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12342 PL_last_lop = PL_last_uni = NULL;
12345 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
12346 while (s >= PL_bufend) { /* multiple line string? */
12348 if (PL_madskills) {
12349 tstart = SvPVX(PL_linestr) + stuffstart;
12351 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12353 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12357 CopLINE_inc(PL_curcop);
12358 if (!outer || !lex_next_chunk(0)) {
12359 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12360 missingterm(PL_tokenbuf);
12362 CopLINE_dec(PL_curcop);
12365 stuffstart = s - SvPVX(PL_linestr);
12367 CopLINE_inc(PL_curcop);
12368 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12369 PL_last_lop = PL_last_uni = NULL;
12370 #ifndef PERL_STRICT_CR
12371 if (PL_bufend - PL_linestart >= 2) {
12372 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
12373 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
12375 PL_bufend[-2] = '\n';
12377 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12379 else if (PL_bufend[-1] == '\r')
12380 PL_bufend[-1] = '\n';
12382 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
12383 PL_bufend[-1] = '\n';
12385 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
12386 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
12387 *(SvPVX(PL_linestr) + off ) = ' ';
12388 sv_catsv(PL_linestr,herewas);
12389 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12390 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
12394 sv_catsv(tmpstr,PL_linestr);
12399 PL_multi_end = CopLINE(PL_curcop);
12400 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
12401 SvPV_shrink_to_cur(tmpstr);
12403 SvREFCNT_dec(herewas);
12405 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
12407 else if (PL_encoding)
12408 sv_recode_to_utf8(tmpstr, PL_encoding);
12410 PL_lex_stuff = tmpstr;
12411 pl_yylval.ival = op_type;
12415 /* scan_inputsymbol
12416 takes: current position in input buffer
12417 returns: new position in input buffer
12418 side-effects: pl_yylval and lex_op are set.
12423 <FH> read from filehandle
12424 <pkg::FH> read from package qualified filehandle
12425 <pkg'FH> read from package qualified filehandle
12426 <$fh> read from filehandle in $fh
12427 <*.h> filename glob
12432 S_scan_inputsymbol(pTHX_ char *start)
12435 register char *s = start; /* current position in buffer */
12438 char *d = PL_tokenbuf; /* start of temp holding space */
12439 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
12441 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
12443 end = strchr(s, '\n');
12446 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
12448 /* die if we didn't have space for the contents of the <>,
12449 or if it didn't end, or if we see a newline
12452 if (len >= (I32)sizeof PL_tokenbuf)
12453 Perl_croak(aTHX_ "Excessively long <> operator");
12455 Perl_croak(aTHX_ "Unterminated <> operator");
12460 Remember, only scalar variables are interpreted as filehandles by
12461 this code. Anything more complex (e.g., <$fh{$num}>) will be
12462 treated as a glob() call.
12463 This code makes use of the fact that except for the $ at the front,
12464 a scalar variable and a filehandle look the same.
12466 if (*d == '$' && d[1]) d++;
12468 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
12469 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
12472 /* If we've tried to read what we allow filehandles to look like, and
12473 there's still text left, then it must be a glob() and not a getline.
12474 Use scan_str to pull out the stuff between the <> and treat it
12475 as nothing more than a string.
12478 if (d - PL_tokenbuf != len) {
12479 pl_yylval.ival = OP_GLOB;
12480 s = scan_str(start,!!PL_madskills,FALSE);
12482 Perl_croak(aTHX_ "Glob not terminated");
12486 bool readline_overriden = FALSE;
12489 /* we're in a filehandle read situation */
12492 /* turn <> into <ARGV> */
12494 Copy("ARGV",d,5,char);
12496 /* Check whether readline() is overriden */
12497 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
12499 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
12501 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
12502 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
12503 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
12504 readline_overriden = TRUE;
12506 /* if <$fh>, create the ops to turn the variable into a
12510 /* try to find it in the pad for this block, otherwise find
12511 add symbol table ops
12513 const PADOFFSET tmp = pad_findmy(d, len, 0);
12514 if (tmp != NOT_IN_PAD) {
12515 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
12516 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
12517 HEK * const stashname = HvNAME_HEK(stash);
12518 SV * const sym = sv_2mortal(newSVhek(stashname));
12519 sv_catpvs(sym, "::");
12520 sv_catpv(sym, d+1);
12525 OP * const o = newOP(OP_PADSV, 0);
12527 PL_lex_op = readline_overriden
12528 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12529 append_elem(OP_LIST, o,
12530 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
12531 : (OP*)newUNOP(OP_READLINE, 0, o);
12540 ? (GV_ADDMULTI | GV_ADDINEVAL)
12543 PL_lex_op = readline_overriden
12544 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12545 append_elem(OP_LIST,
12546 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
12547 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12548 : (OP*)newUNOP(OP_READLINE, 0,
12549 newUNOP(OP_RV2SV, 0,
12550 newGVOP(OP_GV, 0, gv)));
12552 if (!readline_overriden)
12553 PL_lex_op->op_flags |= OPf_SPECIAL;
12554 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
12555 pl_yylval.ival = OP_NULL;
12558 /* If it's none of the above, it must be a literal filehandle
12559 (<Foo::BAR> or <FOO>) so build a simple readline OP */
12561 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
12562 PL_lex_op = readline_overriden
12563 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12564 append_elem(OP_LIST,
12565 newGVOP(OP_GV, 0, gv),
12566 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12567 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
12568 pl_yylval.ival = OP_NULL;
12577 takes: start position in buffer
12578 keep_quoted preserve \ on the embedded delimiter(s)
12579 keep_delims preserve the delimiters around the string
12580 returns: position to continue reading from buffer
12581 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
12582 updates the read buffer.
12584 This subroutine pulls a string out of the input. It is called for:
12585 q single quotes q(literal text)
12586 ' single quotes 'literal text'
12587 qq double quotes qq(interpolate $here please)
12588 " double quotes "interpolate $here please"
12589 qx backticks qx(/bin/ls -l)
12590 ` backticks `/bin/ls -l`
12591 qw quote words @EXPORT_OK = qw( func() $spam )
12592 m// regexp match m/this/
12593 s/// regexp substitute s/this/that/
12594 tr/// string transliterate tr/this/that/
12595 y/// string transliterate y/this/that/
12596 ($*@) sub prototypes sub foo ($)
12597 (stuff) sub attr parameters sub foo : attr(stuff)
12598 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
12600 In most of these cases (all but <>, patterns and transliterate)
12601 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
12602 calls scan_str(). s/// makes yylex() call scan_subst() which calls
12603 scan_str(). tr/// and y/// make yylex() call scan_trans() which
12606 It skips whitespace before the string starts, and treats the first
12607 character as the delimiter. If the delimiter is one of ([{< then
12608 the corresponding "close" character )]}> is used as the closing
12609 delimiter. It allows quoting of delimiters, and if the string has
12610 balanced delimiters ([{<>}]) it allows nesting.
12612 On success, the SV with the resulting string is put into lex_stuff or,
12613 if that is already non-NULL, into lex_repl. The second case occurs only
12614 when parsing the RHS of the special constructs s/// and tr/// (y///).
12615 For convenience, the terminating delimiter character is stuffed into
12620 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
12623 SV *sv; /* scalar value: string */
12624 const char *tmps; /* temp string, used for delimiter matching */
12625 register char *s = start; /* current position in the buffer */
12626 register char term; /* terminating character */
12627 register char *to; /* current position in the sv's data */
12628 I32 brackets = 1; /* bracket nesting level */
12629 bool has_utf8 = FALSE; /* is there any utf8 content? */
12630 I32 termcode; /* terminating char. code */
12631 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
12632 STRLEN termlen; /* length of terminating string */
12633 int last_off = 0; /* last position for nesting bracket */
12639 PERL_ARGS_ASSERT_SCAN_STR;
12641 /* skip space before the delimiter */
12647 if (PL_realtokenstart >= 0) {
12648 stuffstart = PL_realtokenstart;
12649 PL_realtokenstart = -1;
12652 stuffstart = start - SvPVX(PL_linestr);
12654 /* mark where we are, in case we need to report errors */
12657 /* after skipping whitespace, the next character is the terminator */
12660 termcode = termstr[0] = term;
12664 termcode = utf8_to_uvchr((U8*)s, &termlen);
12665 Copy(s, termstr, termlen, U8);
12666 if (!UTF8_IS_INVARIANT(term))
12670 /* mark where we are */
12671 PL_multi_start = CopLINE(PL_curcop);
12672 PL_multi_open = term;
12674 /* find corresponding closing delimiter */
12675 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
12676 termcode = termstr[0] = term = tmps[5];
12678 PL_multi_close = term;
12680 /* create a new SV to hold the contents. 79 is the SV's initial length.
12681 What a random number. */
12682 sv = newSV_type(SVt_PVIV);
12684 SvIV_set(sv, termcode);
12685 (void)SvPOK_only(sv); /* validate pointer */
12687 /* move past delimiter and try to read a complete string */
12689 sv_catpvn(sv, s, termlen);
12692 tstart = SvPVX(PL_linestr) + stuffstart;
12693 if (!PL_thisopen && !keep_delims) {
12694 PL_thisopen = newSVpvn(tstart, s - tstart);
12695 stuffstart = s - SvPVX(PL_linestr);
12699 if (PL_encoding && !UTF) {
12703 int offset = s - SvPVX_const(PL_linestr);
12704 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
12705 &offset, (char*)termstr, termlen);
12706 const char * const ns = SvPVX_const(PL_linestr) + offset;
12707 char * const svlast = SvEND(sv) - 1;
12709 for (; s < ns; s++) {
12710 if (*s == '\n' && !PL_rsfp)
12711 CopLINE_inc(PL_curcop);
12714 goto read_more_line;
12716 /* handle quoted delimiters */
12717 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
12719 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
12721 if ((svlast-1 - t) % 2) {
12722 if (!keep_quoted) {
12723 *(svlast-1) = term;
12725 SvCUR_set(sv, SvCUR(sv) - 1);
12730 if (PL_multi_open == PL_multi_close) {
12736 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
12737 /* At here, all closes are "was quoted" one,
12738 so we don't check PL_multi_close. */
12740 if (!keep_quoted && *(t+1) == PL_multi_open)
12745 else if (*t == PL_multi_open)
12753 SvCUR_set(sv, w - SvPVX_const(sv));
12755 last_off = w - SvPVX(sv);
12756 if (--brackets <= 0)
12761 if (!keep_delims) {
12762 SvCUR_set(sv, SvCUR(sv) - 1);
12768 /* extend sv if need be */
12769 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
12770 /* set 'to' to the next character in the sv's string */
12771 to = SvPVX(sv)+SvCUR(sv);
12773 /* if open delimiter is the close delimiter read unbridle */
12774 if (PL_multi_open == PL_multi_close) {
12775 for (; s < PL_bufend; s++,to++) {
12776 /* embedded newlines increment the current line number */
12777 if (*s == '\n' && !PL_rsfp)
12778 CopLINE_inc(PL_curcop);
12779 /* handle quoted delimiters */
12780 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
12781 if (!keep_quoted && s[1] == term)
12783 /* any other quotes are simply copied straight through */
12787 /* terminate when run out of buffer (the for() condition), or
12788 have found the terminator */
12789 else if (*s == term) {
12792 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
12795 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12801 /* if the terminator isn't the same as the start character (e.g.,
12802 matched brackets), we have to allow more in the quoting, and
12803 be prepared for nested brackets.
12806 /* read until we run out of string, or we find the terminator */
12807 for (; s < PL_bufend; s++,to++) {
12808 /* embedded newlines increment the line count */
12809 if (*s == '\n' && !PL_rsfp)
12810 CopLINE_inc(PL_curcop);
12811 /* backslashes can escape the open or closing characters */
12812 if (*s == '\\' && s+1 < PL_bufend) {
12813 if (!keep_quoted &&
12814 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
12819 /* allow nested opens and closes */
12820 else if (*s == PL_multi_close && --brackets <= 0)
12822 else if (*s == PL_multi_open)
12824 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12829 /* terminate the copied string and update the sv's end-of-string */
12831 SvCUR_set(sv, to - SvPVX_const(sv));
12834 * this next chunk reads more into the buffer if we're not done yet
12838 break; /* handle case where we are done yet :-) */
12840 #ifndef PERL_STRICT_CR
12841 if (to - SvPVX_const(sv) >= 2) {
12842 if ((to[-2] == '\r' && to[-1] == '\n') ||
12843 (to[-2] == '\n' && to[-1] == '\r'))
12847 SvCUR_set(sv, to - SvPVX_const(sv));
12849 else if (to[-1] == '\r')
12852 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
12857 /* if we're out of file, or a read fails, bail and reset the current
12858 line marker so we can report where the unterminated string began
12861 if (PL_madskills) {
12862 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12864 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12866 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12869 CopLINE_inc(PL_curcop);
12870 PL_bufptr = PL_bufend;
12871 if (!lex_next_chunk(0)) {
12873 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12882 /* at this point, we have successfully read the delimited string */
12884 if (!PL_encoding || UTF) {
12886 if (PL_madskills) {
12887 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12888 const int len = s - tstart;
12890 sv_catpvn(PL_thisstuff, tstart, len);
12892 PL_thisstuff = newSVpvn(tstart, len);
12893 if (!PL_thisclose && !keep_delims)
12894 PL_thisclose = newSVpvn(s,termlen);
12899 sv_catpvn(sv, s, termlen);
12904 if (PL_madskills) {
12905 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12906 const int len = s - tstart - termlen;
12908 sv_catpvn(PL_thisstuff, tstart, len);
12910 PL_thisstuff = newSVpvn(tstart, len);
12911 if (!PL_thisclose && !keep_delims)
12912 PL_thisclose = newSVpvn(s - termlen,termlen);
12916 if (has_utf8 || PL_encoding)
12919 PL_multi_end = CopLINE(PL_curcop);
12921 /* if we allocated too much space, give some back */
12922 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12923 SvLEN_set(sv, SvCUR(sv) + 1);
12924 SvPV_renew(sv, SvLEN(sv));
12927 /* decide whether this is the first or second quoted string we've read
12940 takes: pointer to position in buffer
12941 returns: pointer to new position in buffer
12942 side-effects: builds ops for the constant in pl_yylval.op
12944 Read a number in any of the formats that Perl accepts:
12946 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12947 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12950 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12952 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12955 If it reads a number without a decimal point or an exponent, it will
12956 try converting the number to an integer and see if it can do so
12957 without loss of precision.
12961 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12964 register const char *s = start; /* current position in buffer */
12965 register char *d; /* destination in temp buffer */
12966 register char *e; /* end of temp buffer */
12967 NV nv; /* number read, as a double */
12968 SV *sv = NULL; /* place to put the converted number */
12969 bool floatit; /* boolean: int or float? */
12970 const char *lastub = NULL; /* position of last underbar */
12971 static char const number_too_long[] = "Number too long";
12973 PERL_ARGS_ASSERT_SCAN_NUM;
12975 /* We use the first character to decide what type of number this is */
12979 Perl_croak(aTHX_ "panic: scan_num");
12981 /* if it starts with a 0, it could be an octal number, a decimal in
12982 0.13 disguise, or a hexadecimal number, or a binary number. */
12986 u holds the "number so far"
12987 shift the power of 2 of the base
12988 (hex == 4, octal == 3, binary == 1)
12989 overflowed was the number more than we can hold?
12991 Shift is used when we add a digit. It also serves as an "are
12992 we in octal/hex/binary?" indicator to disallow hex characters
12993 when in octal mode.
12998 bool overflowed = FALSE;
12999 bool just_zero = TRUE; /* just plain 0 or binary number? */
13000 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
13001 static const char* const bases[5] =
13002 { "", "binary", "", "octal", "hexadecimal" };
13003 static const char* const Bases[5] =
13004 { "", "Binary", "", "Octal", "Hexadecimal" };
13005 static const char* const maxima[5] =
13007 "0b11111111111111111111111111111111",
13011 const char *base, *Base, *max;
13013 /* check for hex */
13014 if (s[1] == 'x' || s[1] == 'X') {
13018 } else if (s[1] == 'b' || s[1] == 'B') {
13023 /* check for a decimal in disguise */
13024 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
13026 /* so it must be octal */
13033 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13034 "Misplaced _ in number");
13038 base = bases[shift];
13039 Base = Bases[shift];
13040 max = maxima[shift];
13042 /* read the rest of the number */
13044 /* x is used in the overflow test,
13045 b is the digit we're adding on. */
13050 /* if we don't mention it, we're done */
13054 /* _ are ignored -- but warned about if consecutive */
13056 if (lastub && s == lastub + 1)
13057 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13058 "Misplaced _ in number");
13062 /* 8 and 9 are not octal */
13063 case '8': case '9':
13065 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
13069 case '2': case '3': case '4':
13070 case '5': case '6': case '7':
13072 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
13075 case '0': case '1':
13076 b = *s++ & 15; /* ASCII digit -> value of digit */
13080 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
13081 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
13082 /* make sure they said 0x */
13085 b = (*s++ & 7) + 9;
13087 /* Prepare to put the digit we have onto the end
13088 of the number so far. We check for overflows.
13094 x = u << shift; /* make room for the digit */
13096 if ((x >> shift) != u
13097 && !(PL_hints & HINT_NEW_BINARY)) {
13100 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13101 "Integer overflow in %s number",
13104 u = x | b; /* add the digit to the end */
13107 n *= nvshift[shift];
13108 /* If an NV has not enough bits in its
13109 * mantissa to represent an UV this summing of
13110 * small low-order numbers is a waste of time
13111 * (because the NV cannot preserve the
13112 * low-order bits anyway): we could just
13113 * remember when did we overflow and in the
13114 * end just multiply n by the right
13122 /* if we get here, we had success: make a scalar value from
13127 /* final misplaced underbar check */
13128 if (s[-1] == '_') {
13129 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13133 if (n > 4294967295.0)
13134 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13135 "%s number > %s non-portable",
13141 if (u > 0xffffffff)
13142 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13143 "%s number > %s non-portable",
13148 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
13149 sv = new_constant(start, s - start, "integer",
13150 sv, NULL, NULL, 0);
13151 else if (PL_hints & HINT_NEW_BINARY)
13152 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
13157 handle decimal numbers.
13158 we're also sent here when we read a 0 as the first digit
13160 case '1': case '2': case '3': case '4': case '5':
13161 case '6': case '7': case '8': case '9': case '.':
13164 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
13167 /* read next group of digits and _ and copy into d */
13168 while (isDIGIT(*s) || *s == '_') {
13169 /* skip underscores, checking for misplaced ones
13173 if (lastub && s == lastub + 1)
13174 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13175 "Misplaced _ in number");
13179 /* check for end of fixed-length buffer */
13181 Perl_croak(aTHX_ number_too_long);
13182 /* if we're ok, copy the character */
13187 /* final misplaced underbar check */
13188 if (lastub && s == lastub + 1) {
13189 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13192 /* read a decimal portion if there is one. avoid
13193 3..5 being interpreted as the number 3. followed
13196 if (*s == '.' && s[1] != '.') {
13201 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13202 "Misplaced _ in number");
13206 /* copy, ignoring underbars, until we run out of digits.
13208 for (; isDIGIT(*s) || *s == '_'; s++) {
13209 /* fixed length buffer check */
13211 Perl_croak(aTHX_ number_too_long);
13213 if (lastub && s == lastub + 1)
13214 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13215 "Misplaced _ in number");
13221 /* fractional part ending in underbar? */
13222 if (s[-1] == '_') {
13223 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13224 "Misplaced _ in number");
13226 if (*s == '.' && isDIGIT(s[1])) {
13227 /* oops, it's really a v-string, but without the "v" */
13233 /* read exponent part, if present */
13234 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
13238 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
13239 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
13241 /* stray preinitial _ */
13243 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13244 "Misplaced _ in number");
13248 /* allow positive or negative exponent */
13249 if (*s == '+' || *s == '-')
13252 /* stray initial _ */
13254 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13255 "Misplaced _ in number");
13259 /* read digits of exponent */
13260 while (isDIGIT(*s) || *s == '_') {
13263 Perl_croak(aTHX_ number_too_long);
13267 if (((lastub && s == lastub + 1) ||
13268 (!isDIGIT(s[1]) && s[1] != '_')))
13269 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13270 "Misplaced _ in number");
13278 We try to do an integer conversion first if no characters
13279 indicating "float" have been found.
13284 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
13286 if (flags == IS_NUMBER_IN_UV) {
13288 sv = newSViv(uv); /* Prefer IVs over UVs. */
13291 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
13292 if (uv <= (UV) IV_MIN)
13293 sv = newSViv(-(IV)uv);
13300 /* terminate the string */
13302 nv = Atof(PL_tokenbuf);
13307 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
13308 const char *const key = floatit ? "float" : "integer";
13309 const STRLEN keylen = floatit ? 5 : 7;
13310 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
13311 key, keylen, sv, NULL, NULL, 0);
13315 /* if it starts with a v, it could be a v-string */
13318 sv = newSV(5); /* preallocate storage space */
13319 s = scan_vstring(s, PL_bufend, sv);
13323 /* make the op for the constant and return */
13326 lvalp->opval = newSVOP(OP_CONST, 0, sv);
13328 lvalp->opval = NULL;
13334 S_scan_formline(pTHX_ register char *s)
13337 register char *eol;
13339 SV * const stuff = newSVpvs("");
13340 bool needargs = FALSE;
13341 bool eofmt = FALSE;
13343 char *tokenstart = s;
13344 SV* savewhite = NULL;
13346 if (PL_madskills) {
13347 savewhite = PL_thiswhite;
13352 PERL_ARGS_ASSERT_SCAN_FORMLINE;
13354 while (!needargs) {
13357 #ifdef PERL_STRICT_CR
13358 while (SPACE_OR_TAB(*t))
13361 while (SPACE_OR_TAB(*t) || *t == '\r')
13364 if (*t == '\n' || t == PL_bufend) {
13369 if (PL_in_eval && !PL_rsfp) {
13370 eol = (char *) memchr(s,'\n',PL_bufend-s);
13375 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
13377 for (t = s; t < eol; t++) {
13378 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
13380 goto enough; /* ~~ must be first line in formline */
13382 if (*t == '@' || *t == '^')
13386 sv_catpvn(stuff, s, eol-s);
13387 #ifndef PERL_STRICT_CR
13388 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
13389 char *end = SvPVX(stuff) + SvCUR(stuff);
13392 SvCUR_set(stuff, SvCUR(stuff) - 1);
13403 if (PL_madskills) {
13405 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
13407 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
13410 PL_bufptr = PL_bufend;
13411 CopLINE_inc(PL_curcop);
13412 got_some = lex_next_chunk(0);
13413 CopLINE_dec(PL_curcop);
13416 tokenstart = PL_bufptr;
13424 if (SvCUR(stuff)) {
13427 PL_lex_state = LEX_NORMAL;
13428 start_force(PL_curforce);
13429 NEXTVAL_NEXTTOKE.ival = 0;
13433 PL_lex_state = LEX_FORMLINE;
13435 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
13437 else if (PL_encoding)
13438 sv_recode_to_utf8(stuff, PL_encoding);
13440 start_force(PL_curforce);
13441 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
13443 start_force(PL_curforce);
13444 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
13448 SvREFCNT_dec(stuff);
13450 PL_lex_formbrack = 0;
13454 if (PL_madskills) {
13456 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
13458 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
13459 PL_thiswhite = savewhite;
13466 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
13469 const I32 oldsavestack_ix = PL_savestack_ix;
13470 CV* const outsidecv = PL_compcv;
13473 assert(SvTYPE(PL_compcv) == SVt_PVCV);
13475 SAVEI32(PL_subline);
13476 save_item(PL_subname);
13477 SAVESPTR(PL_compcv);
13479 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
13480 CvFLAGS(PL_compcv) |= flags;
13482 PL_subline = CopLINE(PL_curcop);
13483 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
13484 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
13485 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
13487 return oldsavestack_ix;
13491 #pragma segment Perl_yylex
13494 S_yywarn(pTHX_ const char *const s)
13498 PERL_ARGS_ASSERT_YYWARN;
13500 PL_in_eval |= EVAL_WARNONLY;
13502 PL_in_eval &= ~EVAL_WARNONLY;
13507 Perl_yyerror(pTHX_ const char *const s)
13510 const char *where = NULL;
13511 const char *context = NULL;
13514 int yychar = PL_parser->yychar;
13516 PERL_ARGS_ASSERT_YYERROR;
13518 if (!yychar || (yychar == ';' && !PL_rsfp))
13520 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
13521 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
13522 PL_oldbufptr != PL_bufptr) {
13525 The code below is removed for NetWare because it abends/crashes on NetWare
13526 when the script has error such as not having the closing quotes like:
13527 if ($var eq "value)
13528 Checking of white spaces is anyway done in NetWare code.
13531 while (isSPACE(*PL_oldoldbufptr))
13534 context = PL_oldoldbufptr;
13535 contlen = PL_bufptr - PL_oldoldbufptr;
13537 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
13538 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
13541 The code below is removed for NetWare because it abends/crashes on NetWare
13542 when the script has error such as not having the closing quotes like:
13543 if ($var eq "value)
13544 Checking of white spaces is anyway done in NetWare code.
13547 while (isSPACE(*PL_oldbufptr))
13550 context = PL_oldbufptr;
13551 contlen = PL_bufptr - PL_oldbufptr;
13553 else if (yychar > 255)
13554 where = "next token ???";
13555 else if (yychar == -2) { /* YYEMPTY */
13556 if (PL_lex_state == LEX_NORMAL ||
13557 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
13558 where = "at end of line";
13559 else if (PL_lex_inpat)
13560 where = "within pattern";
13562 where = "within string";
13565 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
13567 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
13568 else if (isPRINT_LC(yychar)) {
13569 const char string = yychar;
13570 sv_catpvn(where_sv, &string, 1);
13573 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
13574 where = SvPVX_const(where_sv);
13576 msg = sv_2mortal(newSVpv(s, 0));
13577 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
13578 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
13580 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
13582 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
13583 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
13584 Perl_sv_catpvf(aTHX_ msg,
13585 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
13586 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
13589 if (PL_in_eval & EVAL_WARNONLY) {
13590 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
13594 if (PL_error_count >= 10) {
13595 if (PL_in_eval && SvCUR(ERRSV))
13596 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
13597 SVfARG(ERRSV), OutCopFILE(PL_curcop));
13599 Perl_croak(aTHX_ "%s has too many errors.\n",
13600 OutCopFILE(PL_curcop));
13603 PL_in_my_stash = NULL;
13607 #pragma segment Main
13611 S_swallow_bom(pTHX_ U8 *s)
13614 const STRLEN slen = SvCUR(PL_linestr);
13616 PERL_ARGS_ASSERT_SWALLOW_BOM;
13620 if (s[1] == 0xFE) {
13621 /* UTF-16 little-endian? (or UTF-32LE?) */
13622 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
13623 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
13624 #ifndef PERL_NO_UTF16_FILTER
13625 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
13627 if (PL_bufend > (char*)s) {
13628 s = add_utf16_textfilter(s, TRUE);
13631 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13636 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
13637 #ifndef PERL_NO_UTF16_FILTER
13638 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
13640 if (PL_bufend > (char *)s) {
13641 s = add_utf16_textfilter(s, FALSE);
13644 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13649 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
13650 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13651 s += 3; /* UTF-8 */
13657 if (s[2] == 0xFE && s[3] == 0xFF) {
13658 /* UTF-32 big-endian */
13659 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
13662 else if (s[2] == 0 && s[3] != 0) {
13665 * are a good indicator of UTF-16BE. */
13666 #ifndef PERL_NO_UTF16_FILTER
13667 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
13668 s = add_utf16_textfilter(s, FALSE);
13670 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13676 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
13677 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13678 s += 4; /* UTF-8 */
13684 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
13687 * are a good indicator of UTF-16LE. */
13688 #ifndef PERL_NO_UTF16_FILTER
13689 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
13690 s = add_utf16_textfilter(s, TRUE);
13692 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13700 #ifndef PERL_NO_UTF16_FILTER
13702 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
13705 SV *const filter = FILTER_DATA(idx);
13706 /* We re-use this each time round, throwing the contents away before we
13708 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
13709 SV *const utf8_buffer = filter;
13710 IV status = IoPAGE(filter);
13711 const bool reverse = cBOOL(IoLINES(filter));
13714 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
13716 /* As we're automatically added, at the lowest level, and hence only called
13717 from this file, we can be sure that we're not called in block mode. Hence
13718 don't bother writing code to deal with block mode. */
13720 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
13723 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
13725 DEBUG_P(PerlIO_printf(Perl_debug_log,
13726 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13727 FPTR2DPTR(void *, S_utf16_textfilter),
13728 reverse ? 'l' : 'b', idx, maxlen, status,
13729 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13736 /* First, look in our buffer of existing UTF-8 data: */
13737 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
13741 } else if (status == 0) {
13743 IoPAGE(filter) = 0;
13744 nl = SvEND(utf8_buffer);
13747 STRLEN got = nl - SvPVX(utf8_buffer);
13748 /* Did we have anything to append? */
13750 sv_catpvn(sv, SvPVX(utf8_buffer), got);
13751 /* Everything else in this code works just fine if SVp_POK isn't
13752 set. This, however, needs it, and we need it to work, else
13753 we loop infinitely because the buffer is never consumed. */
13754 sv_chop(utf8_buffer, nl);
13758 /* OK, not a complete line there, so need to read some more UTF-16.
13759 Read an extra octect if the buffer currently has an odd number. */
13763 if (SvCUR(utf16_buffer) >= 2) {
13764 /* Location of the high octet of the last complete code point.
13765 Gosh, UTF-16 is a pain. All the benefits of variable length,
13766 *coupled* with all the benefits of partial reads and
13768 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
13769 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
13771 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
13775 /* We have the first half of a surrogate. Read more. */
13776 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
13779 status = FILTER_READ(idx + 1, utf16_buffer,
13780 160 + (SvCUR(utf16_buffer) & 1));
13781 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
13782 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
13785 IoPAGE(filter) = status;
13790 chars = SvCUR(utf16_buffer) >> 1;
13791 have = SvCUR(utf8_buffer);
13792 SvGROW(utf8_buffer, have + chars * 3 + 1);
13795 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
13796 (U8*)SvPVX_const(utf8_buffer) + have,
13797 chars * 2, &newlen);
13799 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
13800 (U8*)SvPVX_const(utf8_buffer) + have,
13801 chars * 2, &newlen);
13803 SvCUR_set(utf8_buffer, have + newlen);
13806 /* No need to keep this SV "well-formed" with a '\0' after the end, as
13807 it's private to us, and utf16_to_utf8{,reversed} take a
13808 (pointer,length) pair, rather than a NUL-terminated string. */
13809 if(SvCUR(utf16_buffer) & 1) {
13810 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
13811 SvCUR_set(utf16_buffer, 1);
13813 SvCUR_set(utf16_buffer, 0);
13816 DEBUG_P(PerlIO_printf(Perl_debug_log,
13817 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13819 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13820 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
13825 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
13827 SV *filter = filter_add(S_utf16_textfilter, NULL);
13829 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
13831 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
13832 sv_setpvs(filter, "");
13833 IoLINES(filter) = reversed;
13834 IoPAGE(filter) = 1; /* Not EOF */
13836 /* Sadly, we have to return a valid pointer, come what may, so we have to
13837 ignore any error return from this. */
13838 SvCUR_set(PL_linestr, 0);
13839 if (FILTER_READ(0, PL_linestr, 0)) {
13840 SvUTF8_on(PL_linestr);
13842 SvUTF8_on(PL_linestr);
13844 PL_bufend = SvEND(PL_linestr);
13845 return (U8*)SvPVX(PL_linestr);
13850 Returns a pointer to the next character after the parsed
13851 vstring, as well as updating the passed in sv.
13853 Function must be called like
13856 s = scan_vstring(s,e,sv);
13858 where s and e are the start and end of the string.
13859 The sv should already be large enough to store the vstring
13860 passed in, for performance reasons.
13865 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
13868 const char *pos = s;
13869 const char *start = s;
13871 PERL_ARGS_ASSERT_SCAN_VSTRING;
13873 if (*pos == 'v') pos++; /* get past 'v' */
13874 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13876 if ( *pos != '.') {
13877 /* this may not be a v-string if followed by => */
13878 const char *next = pos;
13879 while (next < e && isSPACE(*next))
13881 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
13882 /* return string not v-string */
13883 sv_setpvn(sv,(char *)s,pos-s);
13884 return (char *)pos;
13888 if (!isALPHA(*pos)) {
13889 U8 tmpbuf[UTF8_MAXBYTES+1];
13892 s++; /* get past 'v' */
13897 /* this is atoi() that tolerates underscores */
13900 const char *end = pos;
13902 while (--end >= s) {
13904 const UV orev = rev;
13905 rev += (*end - '0') * mult;
13908 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13909 "Integer overflow in decimal number");
13913 if (rev > 0x7FFFFFFF)
13914 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
13916 /* Append native character for the rev point */
13917 tmpend = uvchr_to_utf8(tmpbuf, rev);
13918 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
13919 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
13921 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
13927 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13931 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
13938 Perl_keyword_plugin_standard(pTHX_
13939 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
13941 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
13942 PERL_UNUSED_CONTEXT;
13943 PERL_UNUSED_ARG(keyword_ptr);
13944 PERL_UNUSED_ARG(keyword_len);
13945 PERL_UNUSED_ARG(op_ptr);
13946 return KEYWORD_PLUGIN_DECLINE;
13951 * c-indentation-style: bsd
13952 * c-basic-offset: 4
13953 * indent-tabs-mode: t
13956 * ex: set ts=8 sts=4 sw=4 noet: