3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * 'It all comes from here, the stench and the peril.' --Frodo
14 * [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
18 * This file is the lexer for Perl. It's closely linked to the
21 * The main routine is yylex(), which returns the next token.
25 =head1 Lexer interface
27 This is the lower layer of the Perl parser, managing characters and tokens.
29 =for apidoc AmU|yy_parser *|PL_parser
31 Pointer to a structure encapsulating the state of the parsing operation
32 currently in progress. The pointer can be locally changed to perform
33 a nested parse without interfering with the state of an outer parse.
34 Individual members of C<PL_parser> have their own documentation.
40 #define PERL_IN_TOKE_C
43 #define new_constant(a,b,c,d,e,f,g) \
44 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
46 #define pl_yylval (PL_parser->yylval)
48 /* YYINITDEPTH -- initial size of the parser's stacks. */
49 #define YYINITDEPTH 200
51 /* XXX temporary backwards compatibility */
52 #define PL_lex_brackets (PL_parser->lex_brackets)
53 #define PL_lex_brackstack (PL_parser->lex_brackstack)
54 #define PL_lex_casemods (PL_parser->lex_casemods)
55 #define PL_lex_casestack (PL_parser->lex_casestack)
56 #define PL_lex_defer (PL_parser->lex_defer)
57 #define PL_lex_dojoin (PL_parser->lex_dojoin)
58 #define PL_lex_expect (PL_parser->lex_expect)
59 #define PL_lex_formbrack (PL_parser->lex_formbrack)
60 #define PL_lex_inpat (PL_parser->lex_inpat)
61 #define PL_lex_inwhat (PL_parser->lex_inwhat)
62 #define PL_lex_op (PL_parser->lex_op)
63 #define PL_lex_repl (PL_parser->lex_repl)
64 #define PL_lex_starts (PL_parser->lex_starts)
65 #define PL_lex_stuff (PL_parser->lex_stuff)
66 #define PL_multi_start (PL_parser->multi_start)
67 #define PL_multi_open (PL_parser->multi_open)
68 #define PL_multi_close (PL_parser->multi_close)
69 #define PL_pending_ident (PL_parser->pending_ident)
70 #define PL_preambled (PL_parser->preambled)
71 #define PL_sublex_info (PL_parser->sublex_info)
72 #define PL_linestr (PL_parser->linestr)
73 #define PL_expect (PL_parser->expect)
74 #define PL_copline (PL_parser->copline)
75 #define PL_bufptr (PL_parser->bufptr)
76 #define PL_oldbufptr (PL_parser->oldbufptr)
77 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
78 #define PL_linestart (PL_parser->linestart)
79 #define PL_bufend (PL_parser->bufend)
80 #define PL_last_uni (PL_parser->last_uni)
81 #define PL_last_lop (PL_parser->last_lop)
82 #define PL_last_lop_op (PL_parser->last_lop_op)
83 #define PL_lex_state (PL_parser->lex_state)
84 #define PL_rsfp (PL_parser->rsfp)
85 #define PL_rsfp_filters (PL_parser->rsfp_filters)
86 #define PL_in_my (PL_parser->in_my)
87 #define PL_in_my_stash (PL_parser->in_my_stash)
88 #define PL_tokenbuf (PL_parser->tokenbuf)
89 #define PL_multi_end (PL_parser->multi_end)
90 #define PL_error_count (PL_parser->error_count)
93 # define PL_endwhite (PL_parser->endwhite)
94 # define PL_faketokens (PL_parser->faketokens)
95 # define PL_lasttoke (PL_parser->lasttoke)
96 # define PL_nextwhite (PL_parser->nextwhite)
97 # define PL_realtokenstart (PL_parser->realtokenstart)
98 # define PL_skipwhite (PL_parser->skipwhite)
99 # define PL_thisclose (PL_parser->thisclose)
100 # define PL_thismad (PL_parser->thismad)
101 # define PL_thisopen (PL_parser->thisopen)
102 # define PL_thisstuff (PL_parser->thisstuff)
103 # define PL_thistoken (PL_parser->thistoken)
104 # define PL_thiswhite (PL_parser->thiswhite)
105 # define PL_thiswhite (PL_parser->thiswhite)
106 # define PL_nexttoke (PL_parser->nexttoke)
107 # define PL_curforce (PL_parser->curforce)
109 # define PL_nexttoke (PL_parser->nexttoke)
110 # define PL_nexttype (PL_parser->nexttype)
111 # define PL_nextval (PL_parser->nextval)
114 /* This can't be done with embed.fnc, because struct yy_parser contains a
115 member named pending_ident, which clashes with the generated #define */
117 S_pending_ident(pTHX);
119 static const char ident_too_long[] = "Identifier too long";
122 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
123 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
125 # define CURMAD(slot,sv)
126 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
129 #define XFAKEBRACK 128
130 #define XENUMMASK 127
132 #ifdef USE_UTF8_SCRIPTS
133 # define UTF (!IN_BYTES)
135 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
138 /* The maximum number of characters preceding the unrecognized one to display */
139 #define UNRECOGNIZED_PRECEDE_COUNT 10
141 /* In variables named $^X, these are the legal values for X.
142 * 1999-02-27 mjd-perl-patch@plover.com */
143 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
145 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
147 /* LEX_* are values for PL_lex_state, the state of the lexer.
148 * They are arranged oddly so that the guard on the switch statement
149 * can get by with a single comparison (if the compiler is smart enough).
152 /* #define LEX_NOTPARSING 11 is done in perl.h. */
154 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
155 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
156 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
157 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
158 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
160 /* at end of code, eg "$x" followed by: */
161 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
162 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
164 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
165 string or after \E, $foo, etc */
166 #define LEX_INTERPCONST 2 /* NOT USED */
167 #define LEX_FORMLINE 1 /* expecting a format line */
168 #define LEX_KNOWNEXT 0 /* next token known; just return it */
172 static const char* const lex_state_names[] = {
191 #include "keywords.h"
193 /* CLINE is a macro that ensures PL_copline has a sane value */
198 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
201 # define SKIPSPACE0(s) skipspace0(s)
202 # define SKIPSPACE1(s) skipspace1(s)
203 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
204 # define PEEKSPACE(s) skipspace2(s,0)
206 # define SKIPSPACE0(s) skipspace(s)
207 # define SKIPSPACE1(s) skipspace(s)
208 # define SKIPSPACE2(s,tsv) skipspace(s)
209 # define PEEKSPACE(s) skipspace(s)
213 * Convenience functions to return different tokens and prime the
214 * lexer for the next token. They all take an argument.
216 * TOKEN : generic token (used for '(', DOLSHARP, etc)
217 * OPERATOR : generic operator
218 * AOPERATOR : assignment operator
219 * PREBLOCK : beginning the block after an if, while, foreach, ...
220 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
221 * PREREF : *EXPR where EXPR is not a simple identifier
222 * TERM : expression term
223 * LOOPX : loop exiting command (goto, last, dump, etc)
224 * FTST : file test operator
225 * FUN0 : zero-argument function
226 * FUN1 : not used, except for not, which isn't a UNIOP
227 * BOop : bitwise or or xor
229 * SHop : shift operator
230 * PWop : power operator
231 * PMop : pattern-matching operator
232 * Aop : addition-level operator
233 * Mop : multiplication-level operator
234 * Eop : equality-testing operator
235 * Rop : relational operator <= != gt
237 * Also see LOP and lop() below.
240 #ifdef DEBUGGING /* Serve -DT. */
241 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
243 # define REPORT(retval) (retval)
246 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
247 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
248 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
249 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
250 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
251 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
252 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
253 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
254 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
255 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
256 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
257 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
258 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
259 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
260 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
261 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
262 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
263 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
264 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
265 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
267 /* This bit of chicanery makes a unary function followed by
268 * a parenthesis into a function with one argument, highest precedence.
269 * The UNIDOR macro is for unary functions that can be followed by the //
270 * operator (such as C<shift // 0>).
272 #define UNI2(f,x) { \
273 pl_yylval.ival = f; \
276 PL_last_uni = PL_oldbufptr; \
277 PL_last_lop_op = f; \
279 return REPORT( (int)FUNC1 ); \
281 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
283 #define UNI(f) UNI2(f,XTERM)
284 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
286 #define UNIBRACK(f) { \
287 pl_yylval.ival = f; \
289 PL_last_uni = PL_oldbufptr; \
291 return REPORT( (int)FUNC1 ); \
293 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
296 /* grandfather return to old style */
297 #define OLDLOP(f) return(pl_yylval.ival=f,PL_expect = XTERM,PL_bufptr = s,(int)LSTOP)
301 /* how to interpret the pl_yylval associated with the token */
305 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
311 static struct debug_tokens {
313 enum token_type type;
315 } const debug_tokens[] =
317 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
318 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
319 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
320 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
321 { ARROW, TOKENTYPE_NONE, "ARROW" },
322 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
323 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
324 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
325 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
326 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
327 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
328 { DO, TOKENTYPE_NONE, "DO" },
329 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
330 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
331 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
332 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
333 { ELSE, TOKENTYPE_NONE, "ELSE" },
334 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
335 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
336 { FOR, TOKENTYPE_IVAL, "FOR" },
337 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
338 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
339 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
340 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
341 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
342 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
343 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
344 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
345 { IF, TOKENTYPE_IVAL, "IF" },
346 { LABEL, TOKENTYPE_PVAL, "LABEL" },
347 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
348 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
349 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
350 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
351 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
352 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
353 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
354 { MY, TOKENTYPE_IVAL, "MY" },
355 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
356 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
357 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
358 { OROP, TOKENTYPE_IVAL, "OROP" },
359 { OROR, TOKENTYPE_NONE, "OROR" },
360 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
361 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
362 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
363 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
364 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
365 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
366 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
367 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
368 { PREINC, TOKENTYPE_NONE, "PREINC" },
369 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
370 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
371 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
372 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
373 { SUB, TOKENTYPE_NONE, "SUB" },
374 { THING, TOKENTYPE_OPVAL, "THING" },
375 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
376 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
377 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
378 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
379 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
380 { USE, TOKENTYPE_IVAL, "USE" },
381 { WHEN, TOKENTYPE_IVAL, "WHEN" },
382 { WHILE, TOKENTYPE_IVAL, "WHILE" },
383 { WORD, TOKENTYPE_OPVAL, "WORD" },
384 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
385 { 0, TOKENTYPE_NONE, NULL }
388 /* dump the returned token in rv, plus any optional arg in pl_yylval */
391 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
395 PERL_ARGS_ASSERT_TOKEREPORT;
398 const char *name = NULL;
399 enum token_type type = TOKENTYPE_NONE;
400 const struct debug_tokens *p;
401 SV* const report = newSVpvs("<== ");
403 for (p = debug_tokens; p->token; p++) {
404 if (p->token == (int)rv) {
411 Perl_sv_catpv(aTHX_ report, name);
412 else if ((char)rv > ' ' && (char)rv < '~')
413 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
415 sv_catpvs(report, "EOF");
417 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
420 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
423 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
425 case TOKENTYPE_OPNUM:
426 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
427 PL_op_name[lvalp->ival]);
430 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
432 case TOKENTYPE_OPVAL:
434 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
435 PL_op_name[lvalp->opval->op_type]);
436 if (lvalp->opval->op_type == OP_CONST) {
437 Perl_sv_catpvf(aTHX_ report, " %s",
438 SvPEEK(cSVOPx_sv(lvalp->opval)));
443 sv_catpvs(report, "(opval=null)");
446 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
452 /* print the buffer with suitable escapes */
455 S_printbuf(pTHX_ const char *const fmt, const char *const s)
457 SV* const tmp = newSVpvs("");
459 PERL_ARGS_ASSERT_PRINTBUF;
461 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
468 S_deprecate_commaless_var_list(pTHX) {
470 deprecate("comma-less variable list");
471 return REPORT(','); /* grandfather non-comma-format format */
477 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
478 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
482 S_ao(pTHX_ int toketype)
485 if (*PL_bufptr == '=') {
487 if (toketype == ANDAND)
488 pl_yylval.ival = OP_ANDASSIGN;
489 else if (toketype == OROR)
490 pl_yylval.ival = OP_ORASSIGN;
491 else if (toketype == DORDOR)
492 pl_yylval.ival = OP_DORASSIGN;
500 * When Perl expects an operator and finds something else, no_op
501 * prints the warning. It always prints "<something> found where
502 * operator expected. It prints "Missing semicolon on previous line?"
503 * if the surprise occurs at the start of the line. "do you need to
504 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
505 * where the compiler doesn't know if foo is a method call or a function.
506 * It prints "Missing operator before end of line" if there's nothing
507 * after the missing operator, or "... before <...>" if there is something
508 * after the missing operator.
512 S_no_op(pTHX_ const char *const what, char *s)
515 char * const oldbp = PL_bufptr;
516 const bool is_first = (PL_oldbufptr == PL_linestart);
518 PERL_ARGS_ASSERT_NO_OP;
524 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
525 if (ckWARN_d(WARN_SYNTAX)) {
527 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
528 "\t(Missing semicolon on previous line?)\n");
529 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
531 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
533 if (t < PL_bufptr && isSPACE(*t))
534 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
535 "\t(Do you need to predeclare %.*s?)\n",
536 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
540 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
541 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
549 * Complain about missing quote/regexp/heredoc terminator.
550 * If it's called with NULL then it cauterizes the line buffer.
551 * If we're in a delimited string and the delimiter is a control
552 * character, it's reformatted into a two-char sequence like ^C.
557 S_missingterm(pTHX_ char *s)
563 char * const nl = strrchr(s,'\n');
567 else if (isCNTRL(PL_multi_close)) {
569 tmpbuf[1] = (char)toCTRL(PL_multi_close);
574 *tmpbuf = (char)PL_multi_close;
578 q = strchr(s,'"') ? '\'' : '"';
579 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
582 #define FEATURE_IS_ENABLED(name) \
583 ((0 != (PL_hints & HINT_LOCALIZE_HH)) \
584 && S_feature_is_enabled(aTHX_ STR_WITH_LEN(name)))
585 /* The longest string we pass in. */
586 #define MAX_FEATURE_LEN (sizeof("unicode_strings")-1)
589 * S_feature_is_enabled
590 * Check whether the named feature is enabled.
593 S_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
596 HV * const hinthv = GvHV(PL_hintgv);
597 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
599 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
601 assert(namelen <= MAX_FEATURE_LEN);
602 memcpy(&he_name[8], name, namelen);
604 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
608 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
609 * utf16-to-utf8-reversed.
612 #ifdef PERL_CR_FILTER
616 register const char *s = SvPVX_const(sv);
617 register const char * const e = s + SvCUR(sv);
619 PERL_ARGS_ASSERT_STRIP_RETURN;
621 /* outer loop optimized to do nothing if there are no CR-LFs */
623 if (*s++ == '\r' && *s == '\n') {
624 /* hit a CR-LF, need to copy the rest */
625 register char *d = s - 1;
628 if (*s == '\r' && s[1] == '\n')
639 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
641 const I32 count = FILTER_READ(idx+1, sv, maxlen);
642 if (count > 0 && !maxlen)
653 * Create a parser object and initialise its parser and lexer fields
655 * rsfp is the opened file handle to read from (if any),
657 * line holds any initial content already read from the file (or in
658 * the case of no file, such as an eval, the whole contents);
660 * new_filter indicates that this is a new file and it shouldn't inherit
661 * the filters from the current parser (ie require).
665 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, bool new_filter)
668 const char *s = NULL;
670 yy_parser *parser, *oparser;
672 /* create and initialise a parser */
674 Newxz(parser, 1, yy_parser);
675 parser->old_parser = oparser = PL_parser;
678 Newx(parser->stack, YYINITDEPTH, yy_stack_frame);
679 parser->ps = parser->stack;
680 parser->stack_size = YYINITDEPTH;
682 parser->stack->state = 0;
683 parser->yyerrstatus = 0;
684 parser->yychar = YYEMPTY; /* Cause a token to be read. */
686 /* on scope exit, free this parser and restore any outer one */
688 parser->saved_curcop = PL_curcop;
690 /* initialise lexer state */
693 parser->curforce = -1;
695 parser->nexttoke = 0;
697 parser->error_count = oparser ? oparser->error_count : 0;
698 parser->copline = NOLINE;
699 parser->lex_state = LEX_NORMAL;
700 parser->expect = XSTATE;
702 parser->rsfp_filters = (new_filter || !oparser) ? newAV()
703 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
705 Newx(parser->lex_brackstack, 120, char);
706 Newx(parser->lex_casestack, 12, char);
707 *parser->lex_casestack = '\0';
710 s = SvPV_const(line, len);
716 parser->linestr = newSVpvs("\n;");
717 } else if (SvREADONLY(line) || s[len-1] != ';') {
718 parser->linestr = newSVsv(line);
720 sv_catpvs(parser->linestr, "\n;");
723 SvREFCNT_inc_simple_void_NN(line);
724 parser->linestr = line;
726 parser->oldoldbufptr =
729 parser->linestart = SvPVX(parser->linestr);
730 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
731 parser->last_lop = parser->last_uni = NULL;
735 /* delete a parser object */
738 Perl_parser_free(pTHX_ const yy_parser *parser)
740 PERL_ARGS_ASSERT_PARSER_FREE;
742 PL_curcop = parser->saved_curcop;
743 SvREFCNT_dec(parser->linestr);
745 if (parser->rsfp == PerlIO_stdin())
746 PerlIO_clearerr(parser->rsfp);
747 else if (parser->rsfp && (!parser->old_parser ||
748 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
749 PerlIO_close(parser->rsfp);
750 SvREFCNT_dec(parser->rsfp_filters);
752 Safefree(parser->stack);
753 Safefree(parser->lex_brackstack);
754 Safefree(parser->lex_casestack);
755 PL_parser = parser->old_parser;
762 * Finalizer for lexing operations. Must be called when the parser is
763 * done with the lexer.
770 PL_doextract = FALSE;
774 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
776 Buffer scalar containing the chunk currently under consideration of the
777 text currently being lexed. This is always a plain string scalar (for
778 which C<SvPOK> is true). It is not intended to be used as a scalar by
779 normal scalar means; instead refer to the buffer directly by the pointer
780 variables described below.
782 The lexer maintains various C<char*> pointers to things in the
783 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
784 reallocated, all of these pointers must be updated. Don't attempt to
785 do this manually, but rather use L</lex_grow_linestr> if you need to
786 reallocate the buffer.
788 The content of the text chunk in the buffer is commonly exactly one
789 complete line of input, up to and including a newline terminator,
790 but there are situations where it is otherwise. The octets of the
791 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
792 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
793 flag on this scalar, which may disagree with it.
795 For direct examination of the buffer, the variable
796 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
797 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
798 of these pointers is usually preferable to examination of the scalar
799 through normal scalar means.
801 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
803 Direct pointer to the end of the chunk of text currently being lexed, the
804 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
805 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
806 always located at the end of the buffer, and does not count as part of
807 the buffer's contents.
809 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
811 Points to the current position of lexing inside the lexer buffer.
812 Characters around this point may be freely examined, within
813 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
814 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
815 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
817 Lexing code (whether in the Perl core or not) moves this pointer past
818 the characters that it consumes. It is also expected to perform some
819 bookkeeping whenever a newline character is consumed. This movement
820 can be more conveniently performed by the function L</lex_read_to>,
821 which handles newlines appropriately.
823 Interpretation of the buffer's octets can be abstracted out by
824 using the slightly higher-level functions L</lex_peek_unichar> and
825 L</lex_read_unichar>.
827 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
829 Points to the start of the current line inside the lexer buffer.
830 This is useful for indicating at which column an error occurred, and
831 not much else. This must be updated by any lexing code that consumes
832 a newline; the function L</lex_read_to> handles this detail.
838 =for apidoc Amx|bool|lex_bufutf8
840 Indicates whether the octets in the lexer buffer
841 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
842 of Unicode characters. If not, they should be interpreted as Latin-1
843 characters. This is analogous to the C<SvUTF8> flag for scalars.
845 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
846 contains valid UTF-8. Lexing code must be robust in the face of invalid
849 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
850 is significant, but not the whole story regarding the input character
851 encoding. Normally, when a file is being read, the scalar contains octets
852 and its C<SvUTF8> flag is off, but the octets should be interpreted as
853 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
854 however, the scalar may have the C<SvUTF8> flag on, and in this case its
855 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
856 is in effect. This logic may change in the future; use this function
857 instead of implementing the logic yourself.
863 Perl_lex_bufutf8(pTHX)
869 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
871 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
872 at least I<len> octets (including terminating NUL). Returns a
873 pointer to the reallocated buffer. This is necessary before making
874 any direct modification of the buffer that would increase its length.
875 L</lex_stuff_pvn> provides a more convenient way to insert text into
878 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
879 this function updates all of the lexer's variables that point directly
886 Perl_lex_grow_linestr(pTHX_ STRLEN len)
890 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
891 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
892 linestr = PL_parser->linestr;
893 buf = SvPVX(linestr);
894 if (len <= SvLEN(linestr))
896 bufend_pos = PL_parser->bufend - buf;
897 bufptr_pos = PL_parser->bufptr - buf;
898 oldbufptr_pos = PL_parser->oldbufptr - buf;
899 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
900 linestart_pos = PL_parser->linestart - buf;
901 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
902 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
903 buf = sv_grow(linestr, len);
904 PL_parser->bufend = buf + bufend_pos;
905 PL_parser->bufptr = buf + bufptr_pos;
906 PL_parser->oldbufptr = buf + oldbufptr_pos;
907 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
908 PL_parser->linestart = buf + linestart_pos;
909 if (PL_parser->last_uni)
910 PL_parser->last_uni = buf + last_uni_pos;
911 if (PL_parser->last_lop)
912 PL_parser->last_lop = buf + last_lop_pos;
917 =for apidoc Amx|void|lex_stuff_pvn|char *pv|STRLEN len|U32 flags
919 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
920 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
921 reallocating the buffer if necessary. This means that lexing code that
922 runs later will see the characters as if they had appeared in the input.
923 It is not recommended to do this as part of normal parsing, and most
924 uses of this facility run the risk of the inserted characters being
925 interpreted in an unintended manner.
927 The string to be inserted is represented by I<len> octets starting
928 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
929 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
930 The characters are recoded for the lexer buffer, according to how the
931 buffer is currently being interpreted (L</lex_bufutf8>). If a string
932 to be interpreted is available as a Perl scalar, the L</lex_stuff_sv>
933 function is more convenient.
939 Perl_lex_stuff_pvn(pTHX_ char *pv, STRLEN len, U32 flags)
942 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
943 if (flags & ~(LEX_STUFF_UTF8))
944 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
946 if (flags & LEX_STUFF_UTF8) {
950 char *p, *e = pv+len;
951 for (p = pv; p != e; p++)
952 highhalf += !!(((U8)*p) & 0x80);
955 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
956 bufptr = PL_parser->bufptr;
957 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
958 PL_parser->bufend += len+highhalf;
959 for (p = pv; p != e; p++) {
962 *bufptr++ = (char)(0xc0 | (c >> 6));
963 *bufptr++ = (char)(0x80 | (c & 0x3f));
970 if (flags & LEX_STUFF_UTF8) {
972 char *p, *e = pv+len;
973 for (p = pv; p != e; p++) {
976 Perl_croak(aTHX_ "Lexing code attempted to stuff "
977 "non-Latin-1 character into Latin-1 input");
978 } else if (c >= 0xc2 && p+1 != e &&
979 (((U8)p[1]) & 0xc0) == 0x80) {
982 } else if (c >= 0x80) {
983 /* malformed UTF-8 */
985 SAVESPTR(PL_warnhook);
986 PL_warnhook = PERL_WARNHOOK_FATAL;
987 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
993 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
994 bufptr = PL_parser->bufptr;
995 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
996 PL_parser->bufend += len-highhalf;
997 for (p = pv; p != e; p++) {
1000 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1003 *bufptr++ = (char)c;
1008 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1009 bufptr = PL_parser->bufptr;
1010 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1011 PL_parser->bufend += len;
1012 Copy(pv, bufptr, len, char);
1018 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1020 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1021 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1022 reallocating the buffer if necessary. This means that lexing code that
1023 runs later will see the characters as if they had appeared in the input.
1024 It is not recommended to do this as part of normal parsing, and most
1025 uses of this facility run the risk of the inserted characters being
1026 interpreted in an unintended manner.
1028 The string to be inserted is the string value of I<sv>. The characters
1029 are recoded for the lexer buffer, according to how the buffer is currently
1030 being interpreted (L</lex_bufutf8>). If a string to be interpreted is
1031 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1032 need to construct a scalar.
1038 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1042 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1044 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1046 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1050 =for apidoc Amx|void|lex_unstuff|char *ptr
1052 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1053 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1054 This hides the discarded text from any lexing code that runs later,
1055 as if the text had never appeared.
1057 This is not the normal way to consume lexed text. For that, use
1064 Perl_lex_unstuff(pTHX_ char *ptr)
1068 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1069 buf = PL_parser->bufptr;
1071 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1074 bufend = PL_parser->bufend;
1076 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1077 unstuff_len = ptr - buf;
1078 Move(ptr, buf, bufend+1-ptr, char);
1079 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1080 PL_parser->bufend = bufend - unstuff_len;
1084 =for apidoc Amx|void|lex_read_to|char *ptr
1086 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1087 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1088 performing the correct bookkeeping whenever a newline character is passed.
1089 This is the normal way to consume lexed text.
1091 Interpretation of the buffer's octets can be abstracted out by
1092 using the slightly higher-level functions L</lex_peek_unichar> and
1093 L</lex_read_unichar>.
1099 Perl_lex_read_to(pTHX_ char *ptr)
1102 PERL_ARGS_ASSERT_LEX_READ_TO;
1103 s = PL_parser->bufptr;
1104 if (ptr < s || ptr > PL_parser->bufend)
1105 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1106 for (; s != ptr; s++)
1108 CopLINE_inc(PL_curcop);
1109 PL_parser->linestart = s+1;
1111 PL_parser->bufptr = ptr;
1115 =for apidoc Amx|void|lex_discard_to|char *ptr
1117 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1118 up to I<ptr>. The remaining content of the buffer will be moved, and
1119 all pointers into the buffer updated appropriately. I<ptr> must not
1120 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1121 it is not permitted to discard text that has yet to be lexed.
1123 Normally it is not necessarily to do this directly, because it suffices to
1124 use the implicit discarding behaviour of L</lex_next_chunk> and things
1125 based on it. However, if a token stretches across multiple lines,
1126 and the lexing code has kept multiple lines of text in the buffer fof
1127 that purpose, then after completion of the token it would be wise to
1128 explicitly discard the now-unneeded earlier lines, to avoid future
1129 multi-line tokens growing the buffer without bound.
1135 Perl_lex_discard_to(pTHX_ char *ptr)
1139 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1140 buf = SvPVX(PL_parser->linestr);
1142 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1145 if (ptr > PL_parser->bufptr)
1146 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1147 discard_len = ptr - buf;
1148 if (PL_parser->oldbufptr < ptr)
1149 PL_parser->oldbufptr = ptr;
1150 if (PL_parser->oldoldbufptr < ptr)
1151 PL_parser->oldoldbufptr = ptr;
1152 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1153 PL_parser->last_uni = NULL;
1154 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1155 PL_parser->last_lop = NULL;
1156 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1157 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1158 PL_parser->bufend -= discard_len;
1159 PL_parser->bufptr -= discard_len;
1160 PL_parser->oldbufptr -= discard_len;
1161 PL_parser->oldoldbufptr -= discard_len;
1162 if (PL_parser->last_uni)
1163 PL_parser->last_uni -= discard_len;
1164 if (PL_parser->last_lop)
1165 PL_parser->last_lop -= discard_len;
1169 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1171 Reads in the next chunk of text to be lexed, appending it to
1172 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1173 looked to the end of the current chunk and wants to know more. It is
1174 usual, but not necessary, for lexing to have consumed the entirety of
1175 the current chunk at this time.
1177 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1178 chunk (i.e., the current chunk has been entirely consumed), normally the
1179 current chunk will be discarded at the same time that the new chunk is
1180 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1181 will not be discarded. If the current chunk has not been entirely
1182 consumed, then it will not be discarded regardless of the flag.
1184 Returns true if some new text was added to the buffer, or false if the
1185 buffer has reached the end of the input text.
1190 #define LEX_FAKE_EOF 0x80000000
1193 Perl_lex_next_chunk(pTHX_ U32 flags)
1197 STRLEN old_bufend_pos, new_bufend_pos;
1198 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1199 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1200 bool got_some_for_debugger = 0;
1202 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1203 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1204 linestr = PL_parser->linestr;
1205 buf = SvPVX(linestr);
1206 if (!(flags & LEX_KEEP_PREVIOUS) &&
1207 PL_parser->bufptr == PL_parser->bufend) {
1208 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1210 if (PL_parser->last_uni != PL_parser->bufend)
1211 PL_parser->last_uni = NULL;
1212 if (PL_parser->last_lop != PL_parser->bufend)
1213 PL_parser->last_lop = NULL;
1214 last_uni_pos = last_lop_pos = 0;
1218 old_bufend_pos = PL_parser->bufend - buf;
1219 bufptr_pos = PL_parser->bufptr - buf;
1220 oldbufptr_pos = PL_parser->oldbufptr - buf;
1221 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1222 linestart_pos = PL_parser->linestart - buf;
1223 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1224 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1226 if (flags & LEX_FAKE_EOF) {
1228 } else if (!PL_parser->rsfp) {
1230 } else if (filter_gets(linestr, old_bufend_pos)) {
1232 got_some_for_debugger = 1;
1234 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1235 sv_setpvs(linestr, "");
1237 /* End of real input. Close filehandle (unless it was STDIN),
1238 * then add implicit termination.
1240 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1241 PerlIO_clearerr(PL_parser->rsfp);
1242 else if (PL_parser->rsfp)
1243 (void)PerlIO_close(PL_parser->rsfp);
1244 PL_parser->rsfp = NULL;
1245 PL_doextract = FALSE;
1247 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1250 if (!PL_in_eval && PL_minus_p) {
1252 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1253 PL_minus_n = PL_minus_p = 0;
1254 } else if (!PL_in_eval && PL_minus_n) {
1255 sv_catpvs(linestr, /*{*/";}");
1258 sv_catpvs(linestr, ";");
1261 buf = SvPVX(linestr);
1262 new_bufend_pos = SvCUR(linestr);
1263 PL_parser->bufend = buf + new_bufend_pos;
1264 PL_parser->bufptr = buf + bufptr_pos;
1265 PL_parser->oldbufptr = buf + oldbufptr_pos;
1266 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1267 PL_parser->linestart = buf + linestart_pos;
1268 if (PL_parser->last_uni)
1269 PL_parser->last_uni = buf + last_uni_pos;
1270 if (PL_parser->last_lop)
1271 PL_parser->last_lop = buf + last_lop_pos;
1272 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1273 PL_curstash != PL_debstash) {
1274 /* debugger active and we're not compiling the debugger code,
1275 * so store the line into the debugger's array of lines
1277 update_debugger_info(NULL, buf+old_bufend_pos,
1278 new_bufend_pos-old_bufend_pos);
1284 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1286 Looks ahead one (Unicode) character in the text currently being lexed.
1287 Returns the codepoint (unsigned integer value) of the next character,
1288 or -1 if lexing has reached the end of the input text. To consume the
1289 peeked character, use L</lex_read_unichar>.
1291 If the next character is in (or extends into) the next chunk of input
1292 text, the next chunk will be read in. Normally the current chunk will be
1293 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1294 then the current chunk will not be discarded.
1296 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1297 is encountered, an exception is generated.
1303 Perl_lex_peek_unichar(pTHX_ U32 flags)
1306 if (flags & ~(LEX_KEEP_PREVIOUS))
1307 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1308 s = PL_parser->bufptr;
1309 bufend = PL_parser->bufend;
1315 if (!lex_next_chunk(flags))
1317 s = PL_parser->bufptr;
1318 bufend = PL_parser->bufend;
1324 len = PL_utf8skip[head];
1325 while ((STRLEN)(bufend-s) < len) {
1326 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1328 s = PL_parser->bufptr;
1329 bufend = PL_parser->bufend;
1332 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1333 if (retlen == (STRLEN)-1) {
1334 /* malformed UTF-8 */
1336 SAVESPTR(PL_warnhook);
1337 PL_warnhook = PERL_WARNHOOK_FATAL;
1338 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1344 if (!lex_next_chunk(flags))
1346 s = PL_parser->bufptr;
1353 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1355 Reads the next (Unicode) character in the text currently being lexed.
1356 Returns the codepoint (unsigned integer value) of the character read,
1357 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1358 if lexing has reached the end of the input text. To non-destructively
1359 examine the next character, use L</lex_peek_unichar> instead.
1361 If the next character is in (or extends into) the next chunk of input
1362 text, the next chunk will be read in. Normally the current chunk will be
1363 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1364 then the current chunk will not be discarded.
1366 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1367 is encountered, an exception is generated.
1373 Perl_lex_read_unichar(pTHX_ U32 flags)
1376 if (flags & ~(LEX_KEEP_PREVIOUS))
1377 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1378 c = lex_peek_unichar(flags);
1381 CopLINE_inc(PL_curcop);
1382 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1388 =for apidoc Amx|void|lex_read_space|U32 flags
1390 Reads optional spaces, in Perl style, in the text currently being
1391 lexed. The spaces may include ordinary whitespace characters and
1392 Perl-style comments. C<#line> directives are processed if encountered.
1393 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1394 at a non-space character (or the end of the input text).
1396 If spaces extend into the next chunk of input text, the next chunk will
1397 be read in. Normally the current chunk will be discarded at the same
1398 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1399 chunk will not be discarded.
1404 #define LEX_NO_NEXT_CHUNK 0x80000000
1407 Perl_lex_read_space(pTHX_ U32 flags)
1410 bool need_incline = 0;
1411 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1412 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1415 sv_free(PL_skipwhite);
1416 PL_skipwhite = NULL;
1419 PL_skipwhite = newSVpvs("");
1420 #endif /* PERL_MAD */
1421 s = PL_parser->bufptr;
1422 bufend = PL_parser->bufend;
1428 } while (!(c == '\n' || (c == 0 && s == bufend)));
1429 } else if (c == '\n') {
1431 PL_parser->linestart = s;
1436 } else if (isSPACE(c)) {
1438 } else if (c == 0 && s == bufend) {
1442 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1443 #endif /* PERL_MAD */
1444 if (flags & LEX_NO_NEXT_CHUNK)
1446 PL_parser->bufptr = s;
1447 CopLINE_inc(PL_curcop);
1448 got_more = lex_next_chunk(flags);
1449 CopLINE_dec(PL_curcop);
1450 s = PL_parser->bufptr;
1451 bufend = PL_parser->bufend;
1454 if (need_incline && PL_parser->rsfp) {
1464 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1465 #endif /* PERL_MAD */
1466 PL_parser->bufptr = s;
1471 * This subroutine has nothing to do with tilting, whether at windmills
1472 * or pinball tables. Its name is short for "increment line". It
1473 * increments the current line number in CopLINE(PL_curcop) and checks
1474 * to see whether the line starts with a comment of the form
1475 * # line 500 "foo.pm"
1476 * If so, it sets the current line number and file to the values in the comment.
1480 S_incline(pTHX_ const char *s)
1487 PERL_ARGS_ASSERT_INCLINE;
1489 CopLINE_inc(PL_curcop);
1492 while (SPACE_OR_TAB(*s))
1494 if (strnEQ(s, "line", 4))
1498 if (SPACE_OR_TAB(*s))
1502 while (SPACE_OR_TAB(*s))
1510 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1512 while (SPACE_OR_TAB(*s))
1514 if (*s == '"' && (t = strchr(s+1, '"'))) {
1520 while (!isSPACE(*t))
1524 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1526 if (*e != '\n' && *e != '\0')
1527 return; /* false alarm */
1530 const STRLEN len = t - s;
1531 #ifndef USE_ITHREADS
1532 SV *const temp_sv = CopFILESV(PL_curcop);
1537 cf = SvPVX(temp_sv);
1538 tmplen = SvCUR(temp_sv);
1544 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
1545 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1546 * to *{"::_<newfilename"} */
1547 /* However, the long form of evals is only turned on by the
1548 debugger - usually they're "(eval %lu)" */
1552 STRLEN tmplen2 = len;
1553 if (tmplen + 2 <= sizeof smallbuf)
1556 Newx(tmpbuf, tmplen + 2, char);
1559 memcpy(tmpbuf + 2, cf, tmplen);
1561 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1566 if (tmplen2 + 2 <= sizeof smallbuf)
1569 Newx(tmpbuf2, tmplen2 + 2, char);
1571 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1572 /* Either they malloc'd it, or we malloc'd it,
1573 so no prefix is present in ours. */
1578 memcpy(tmpbuf2 + 2, s, tmplen2);
1581 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1583 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1584 /* adjust ${"::_<newfilename"} to store the new file name */
1585 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1586 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1587 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1590 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1592 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1595 CopFILE_free(PL_curcop);
1596 CopFILE_setn(PL_curcop, s, len);
1598 CopLINE_set(PL_curcop, atoi(n)-1);
1602 /* skip space before PL_thistoken */
1605 S_skipspace0(pTHX_ register char *s)
1607 PERL_ARGS_ASSERT_SKIPSPACE0;
1614 PL_thiswhite = newSVpvs("");
1615 sv_catsv(PL_thiswhite, PL_skipwhite);
1616 sv_free(PL_skipwhite);
1619 PL_realtokenstart = s - SvPVX(PL_linestr);
1623 /* skip space after PL_thistoken */
1626 S_skipspace1(pTHX_ register char *s)
1628 const char *start = s;
1629 I32 startoff = start - SvPVX(PL_linestr);
1631 PERL_ARGS_ASSERT_SKIPSPACE1;
1636 start = SvPVX(PL_linestr) + startoff;
1637 if (!PL_thistoken && PL_realtokenstart >= 0) {
1638 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1639 PL_thistoken = newSVpvn(tstart, start - tstart);
1641 PL_realtokenstart = -1;
1644 PL_nextwhite = newSVpvs("");
1645 sv_catsv(PL_nextwhite, PL_skipwhite);
1646 sv_free(PL_skipwhite);
1653 S_skipspace2(pTHX_ register char *s, SV **svp)
1656 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1657 const I32 startoff = s - SvPVX(PL_linestr);
1659 PERL_ARGS_ASSERT_SKIPSPACE2;
1662 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1663 if (!PL_madskills || !svp)
1665 start = SvPVX(PL_linestr) + startoff;
1666 if (!PL_thistoken && PL_realtokenstart >= 0) {
1667 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1668 PL_thistoken = newSVpvn(tstart, start - tstart);
1669 PL_realtokenstart = -1;
1673 *svp = newSVpvs("");
1674 sv_setsv(*svp, PL_skipwhite);
1675 sv_free(PL_skipwhite);
1684 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1686 AV *av = CopFILEAVx(PL_curcop);
1688 SV * const sv = newSV_type(SVt_PVMG);
1690 sv_setsv(sv, orig_sv);
1692 sv_setpvn(sv, buf, len);
1695 av_store(av, (I32)CopLINE(PL_curcop), sv);
1701 * Called to gobble the appropriate amount and type of whitespace.
1702 * Skips comments as well.
1706 S_skipspace(pTHX_ register char *s)
1710 #endif /* PERL_MAD */
1711 PERL_ARGS_ASSERT_SKIPSPACE;
1714 sv_free(PL_skipwhite);
1715 PL_skipwhite = NULL;
1717 #endif /* PERL_MAD */
1718 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1719 while (s < PL_bufend && SPACE_OR_TAB(*s))
1722 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1724 lex_read_space(LEX_KEEP_PREVIOUS |
1725 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1726 LEX_NO_NEXT_CHUNK : 0));
1728 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1729 if (PL_linestart > PL_bufptr)
1730 PL_bufptr = PL_linestart;
1735 PL_skipwhite = newSVpvn(start, s-start);
1736 #endif /* PERL_MAD */
1742 * Check the unary operators to ensure there's no ambiguity in how they're
1743 * used. An ambiguous piece of code would be:
1745 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1746 * the +5 is its argument.
1756 if (PL_oldoldbufptr != PL_last_uni)
1758 while (isSPACE(*PL_last_uni))
1761 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1763 if ((t = strchr(s, '(')) && t < PL_bufptr)
1766 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1767 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1768 (int)(s - PL_last_uni), PL_last_uni);
1772 * LOP : macro to build a list operator. Its behaviour has been replaced
1773 * with a subroutine, S_lop() for which LOP is just another name.
1776 #define LOP(f,x) return lop(f,x,s)
1780 * Build a list operator (or something that might be one). The rules:
1781 * - if we have a next token, then it's a list operator [why?]
1782 * - if the next thing is an opening paren, then it's a function
1783 * - else it's a list operator
1787 S_lop(pTHX_ I32 f, int x, char *s)
1791 PERL_ARGS_ASSERT_LOP;
1797 PL_last_lop = PL_oldbufptr;
1798 PL_last_lop_op = (OPCODE)f;
1801 return REPORT(LSTOP);
1804 return REPORT(LSTOP);
1807 return REPORT(FUNC);
1810 return REPORT(FUNC);
1812 return REPORT(LSTOP);
1818 * Sets up for an eventual force_next(). start_force(0) basically does
1819 * an unshift, while start_force(-1) does a push. yylex removes items
1824 S_start_force(pTHX_ int where)
1828 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1829 where = PL_lasttoke;
1830 assert(PL_curforce < 0 || PL_curforce == where);
1831 if (PL_curforce != where) {
1832 for (i = PL_lasttoke; i > where; --i) {
1833 PL_nexttoke[i] = PL_nexttoke[i-1];
1837 if (PL_curforce < 0) /* in case of duplicate start_force() */
1838 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1839 PL_curforce = where;
1842 curmad('^', newSVpvs(""));
1843 CURMAD('_', PL_nextwhite);
1848 S_curmad(pTHX_ char slot, SV *sv)
1854 if (PL_curforce < 0)
1855 where = &PL_thismad;
1857 where = &PL_nexttoke[PL_curforce].next_mad;
1863 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1865 else if (PL_encoding) {
1866 sv_recode_to_utf8(sv, PL_encoding);
1871 /* keep a slot open for the head of the list? */
1872 if (slot != '_' && *where && (*where)->mad_key == '^') {
1873 (*where)->mad_key = slot;
1874 sv_free(MUTABLE_SV(((*where)->mad_val)));
1875 (*where)->mad_val = (void*)sv;
1878 addmad(newMADsv(slot, sv), where, 0);
1881 # define start_force(where) NOOP
1882 # define curmad(slot, sv) NOOP
1887 * When the lexer realizes it knows the next token (for instance,
1888 * it is reordering tokens for the parser) then it can call S_force_next
1889 * to know what token to return the next time the lexer is called. Caller
1890 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1891 * and possibly PL_expect to ensure the lexer handles the token correctly.
1895 S_force_next(pTHX_ I32 type)
1900 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1901 tokereport(type, &NEXTVAL_NEXTTOKE);
1905 if (PL_curforce < 0)
1906 start_force(PL_lasttoke);
1907 PL_nexttoke[PL_curforce].next_type = type;
1908 if (PL_lex_state != LEX_KNOWNEXT)
1909 PL_lex_defer = PL_lex_state;
1910 PL_lex_state = LEX_KNOWNEXT;
1911 PL_lex_expect = PL_expect;
1914 PL_nexttype[PL_nexttoke] = type;
1916 if (PL_lex_state != LEX_KNOWNEXT) {
1917 PL_lex_defer = PL_lex_state;
1918 PL_lex_expect = PL_expect;
1919 PL_lex_state = LEX_KNOWNEXT;
1925 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1928 SV * const sv = newSVpvn_utf8(start, len,
1931 && !is_ascii_string((const U8*)start, len)
1932 && is_utf8_string((const U8*)start, len));
1938 * When the lexer knows the next thing is a word (for instance, it has
1939 * just seen -> and it knows that the next char is a word char, then
1940 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1944 * char *start : buffer position (must be within PL_linestr)
1945 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1946 * int check_keyword : if true, Perl checks to make sure the word isn't
1947 * a keyword (do this if the word is a label, e.g. goto FOO)
1948 * int allow_pack : if true, : characters will also be allowed (require,
1949 * use, etc. do this)
1950 * int allow_initial_tick : used by the "sub" lexer only.
1954 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1960 PERL_ARGS_ASSERT_FORCE_WORD;
1962 start = SKIPSPACE1(start);
1964 if (isIDFIRST_lazy_if(s,UTF) ||
1965 (allow_pack && *s == ':') ||
1966 (allow_initial_tick && *s == '\'') )
1968 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1969 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1971 start_force(PL_curforce);
1973 curmad('X', newSVpvn(start,s-start));
1974 if (token == METHOD) {
1979 PL_expect = XOPERATOR;
1983 curmad('g', newSVpvs( "forced" ));
1984 NEXTVAL_NEXTTOKE.opval
1985 = (OP*)newSVOP(OP_CONST,0,
1986 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1987 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1995 * Called when the lexer wants $foo *foo &foo etc, but the program
1996 * text only contains the "foo" portion. The first argument is a pointer
1997 * to the "foo", and the second argument is the type symbol to prefix.
1998 * Forces the next token to be a "WORD".
1999 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2003 S_force_ident(pTHX_ register const char *s, int kind)
2007 PERL_ARGS_ASSERT_FORCE_IDENT;
2010 const STRLEN len = strlen(s);
2011 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
2012 start_force(PL_curforce);
2013 NEXTVAL_NEXTTOKE.opval = o;
2016 o->op_private = OPpCONST_ENTERED;
2017 /* XXX see note in pp_entereval() for why we forgo typo
2018 warnings if the symbol must be introduced in an eval.
2020 gv_fetchpvn_flags(s, len,
2021 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2023 kind == '$' ? SVt_PV :
2024 kind == '@' ? SVt_PVAV :
2025 kind == '%' ? SVt_PVHV :
2033 Perl_str_to_version(pTHX_ SV *sv)
2038 const char *start = SvPV_const(sv,len);
2039 const char * const end = start + len;
2040 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2042 PERL_ARGS_ASSERT_STR_TO_VERSION;
2044 while (start < end) {
2048 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2053 retval += ((NV)n)/nshift;
2062 * Forces the next token to be a version number.
2063 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2064 * and if "guessing" is TRUE, then no new token is created (and the caller
2065 * must use an alternative parsing method).
2069 S_force_version(pTHX_ char *s, int guessing)
2075 I32 startoff = s - SvPVX(PL_linestr);
2078 PERL_ARGS_ASSERT_FORCE_VERSION;
2086 while (isDIGIT(*d) || *d == '_' || *d == '.')
2090 start_force(PL_curforce);
2091 curmad('X', newSVpvn(s,d-s));
2094 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
2096 #ifdef USE_LOCALE_NUMERIC
2097 char *loc = setlocale(LC_NUMERIC, "C");
2099 s = scan_num(s, &pl_yylval);
2100 #ifdef USE_LOCALE_NUMERIC
2101 setlocale(LC_NUMERIC, loc);
2103 version = pl_yylval.opval;
2104 ver = cSVOPx(version)->op_sv;
2105 if (SvPOK(ver) && !SvNIOK(ver)) {
2106 SvUPGRADE(ver, SVt_PVNV);
2107 SvNV_set(ver, str_to_version(ver));
2108 SvNOK_on(ver); /* hint that it is a version */
2111 else if (guessing) {
2114 sv_free(PL_nextwhite); /* let next token collect whitespace */
2116 s = SvPVX(PL_linestr) + startoff;
2124 if (PL_madskills && !version) {
2125 sv_free(PL_nextwhite); /* let next token collect whitespace */
2127 s = SvPVX(PL_linestr) + startoff;
2130 /* NOTE: The parser sees the package name and the VERSION swapped */
2131 start_force(PL_curforce);
2132 NEXTVAL_NEXTTOKE.opval = version;
2139 * S_force_strict_version
2140 * Forces the next token to be a version number using strict syntax rules.
2144 S_force_strict_version(pTHX_ char *s)
2149 I32 startoff = s - SvPVX(PL_linestr);
2151 const char *errstr = NULL;
2153 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2155 while (isSPACE(*s)) /* leading whitespace */
2158 if (is_STRICT_VERSION(s,&errstr)) {
2160 s = (char *)scan_version(s, ver, 0);
2161 version = newSVOP(OP_CONST, 0, ver);
2163 else if ( (*s != ';' && *s != '}' ) && (s = SKIPSPACE1(s), (*s != ';' && *s !='}' ))) {
2166 yyerror(errstr); /* version required */
2171 if (PL_madskills && !version) {
2172 sv_free(PL_nextwhite); /* let next token collect whitespace */
2174 s = SvPVX(PL_linestr) + startoff;
2177 /* NOTE: The parser sees the package name and the VERSION swapped */
2178 start_force(PL_curforce);
2179 NEXTVAL_NEXTTOKE.opval = version;
2187 * Tokenize a quoted string passed in as an SV. It finds the next
2188 * chunk, up to end of string or a backslash. It may make a new
2189 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2194 S_tokeq(pTHX_ SV *sv)
2198 register char *send;
2203 PERL_ARGS_ASSERT_TOKEQ;
2208 s = SvPV_force(sv, len);
2209 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2212 while (s < send && *s != '\\')
2217 if ( PL_hints & HINT_NEW_STRING ) {
2218 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2222 if (s + 1 < send && (s[1] == '\\'))
2223 s++; /* all that, just for this */
2228 SvCUR_set(sv, d - SvPVX_const(sv));
2230 if ( PL_hints & HINT_NEW_STRING )
2231 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2236 * Now come three functions related to double-quote context,
2237 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2238 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2239 * interact with PL_lex_state, and create fake ( ... ) argument lists
2240 * to handle functions and concatenation.
2241 * They assume that whoever calls them will be setting up a fake
2242 * join call, because each subthing puts a ',' after it. This lets
2245 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2247 * (I'm not sure whether the spurious commas at the end of lcfirst's
2248 * arguments and join's arguments are created or not).
2253 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2255 * Pattern matching will set PL_lex_op to the pattern-matching op to
2256 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2258 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2260 * Everything else becomes a FUNC.
2262 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2263 * had an OP_CONST or OP_READLINE). This just sets us up for a
2264 * call to S_sublex_push().
2268 S_sublex_start(pTHX)
2271 register const I32 op_type = pl_yylval.ival;
2273 if (op_type == OP_NULL) {
2274 pl_yylval.opval = PL_lex_op;
2278 if (op_type == OP_CONST || op_type == OP_READLINE) {
2279 SV *sv = tokeq(PL_lex_stuff);
2281 if (SvTYPE(sv) == SVt_PVIV) {
2282 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2284 const char * const p = SvPV_const(sv, len);
2285 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2289 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2290 PL_lex_stuff = NULL;
2291 /* Allow <FH> // "foo" */
2292 if (op_type == OP_READLINE)
2293 PL_expect = XTERMORDORDOR;
2296 else if (op_type == OP_BACKTICK && PL_lex_op) {
2297 /* readpipe() vas overriden */
2298 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2299 pl_yylval.opval = PL_lex_op;
2301 PL_lex_stuff = NULL;
2305 PL_sublex_info.super_state = PL_lex_state;
2306 PL_sublex_info.sub_inwhat = (U16)op_type;
2307 PL_sublex_info.sub_op = PL_lex_op;
2308 PL_lex_state = LEX_INTERPPUSH;
2312 pl_yylval.opval = PL_lex_op;
2322 * Create a new scope to save the lexing state. The scope will be
2323 * ended in S_sublex_done. Returns a '(', starting the function arguments
2324 * to the uc, lc, etc. found before.
2325 * Sets PL_lex_state to LEX_INTERPCONCAT.
2334 PL_lex_state = PL_sublex_info.super_state;
2335 SAVEBOOL(PL_lex_dojoin);
2336 SAVEI32(PL_lex_brackets);
2337 SAVEI32(PL_lex_casemods);
2338 SAVEI32(PL_lex_starts);
2339 SAVEI8(PL_lex_state);
2340 SAVEVPTR(PL_lex_inpat);
2341 SAVEI16(PL_lex_inwhat);
2342 SAVECOPLINE(PL_curcop);
2343 SAVEPPTR(PL_bufptr);
2344 SAVEPPTR(PL_bufend);
2345 SAVEPPTR(PL_oldbufptr);
2346 SAVEPPTR(PL_oldoldbufptr);
2347 SAVEPPTR(PL_last_lop);
2348 SAVEPPTR(PL_last_uni);
2349 SAVEPPTR(PL_linestart);
2350 SAVESPTR(PL_linestr);
2351 SAVEGENERICPV(PL_lex_brackstack);
2352 SAVEGENERICPV(PL_lex_casestack);
2354 PL_linestr = PL_lex_stuff;
2355 PL_lex_stuff = NULL;
2357 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2358 = SvPVX(PL_linestr);
2359 PL_bufend += SvCUR(PL_linestr);
2360 PL_last_lop = PL_last_uni = NULL;
2361 SAVEFREESV(PL_linestr);
2363 PL_lex_dojoin = FALSE;
2364 PL_lex_brackets = 0;
2365 Newx(PL_lex_brackstack, 120, char);
2366 Newx(PL_lex_casestack, 12, char);
2367 PL_lex_casemods = 0;
2368 *PL_lex_casestack = '\0';
2370 PL_lex_state = LEX_INTERPCONCAT;
2371 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2373 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2374 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2375 PL_lex_inpat = PL_sublex_info.sub_op;
2377 PL_lex_inpat = NULL;
2384 * Restores lexer state after a S_sublex_push.
2391 if (!PL_lex_starts++) {
2392 SV * const sv = newSVpvs("");
2393 if (SvUTF8(PL_linestr))
2395 PL_expect = XOPERATOR;
2396 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2400 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2401 PL_lex_state = LEX_INTERPCASEMOD;
2405 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2406 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2407 PL_linestr = PL_lex_repl;
2409 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2410 PL_bufend += SvCUR(PL_linestr);
2411 PL_last_lop = PL_last_uni = NULL;
2412 SAVEFREESV(PL_linestr);
2413 PL_lex_dojoin = FALSE;
2414 PL_lex_brackets = 0;
2415 PL_lex_casemods = 0;
2416 *PL_lex_casestack = '\0';
2418 if (SvEVALED(PL_lex_repl)) {
2419 PL_lex_state = LEX_INTERPNORMAL;
2421 /* we don't clear PL_lex_repl here, so that we can check later
2422 whether this is an evalled subst; that means we rely on the
2423 logic to ensure sublex_done() is called again only via the
2424 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2427 PL_lex_state = LEX_INTERPCONCAT;
2437 PL_endwhite = newSVpvs("");
2438 sv_catsv(PL_endwhite, PL_thiswhite);
2442 sv_setpvs(PL_thistoken,"");
2444 PL_realtokenstart = -1;
2448 PL_bufend = SvPVX(PL_linestr);
2449 PL_bufend += SvCUR(PL_linestr);
2450 PL_expect = XOPERATOR;
2451 PL_sublex_info.sub_inwhat = 0;
2459 Extracts a pattern, double-quoted string, or transliteration. This
2462 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2463 processing a pattern (PL_lex_inpat is true), a transliteration
2464 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2466 Returns a pointer to the character scanned up to. If this is
2467 advanced from the start pointer supplied (i.e. if anything was
2468 successfully parsed), will leave an OP for the substring scanned
2469 in pl_yylval. Caller must intuit reason for not parsing further
2470 by looking at the next characters herself.
2474 constants: \N{NAME} only
2475 case and quoting: \U \Q \E
2476 stops on @ and $, but not for $ as tail anchor
2478 In transliterations:
2479 characters are VERY literal, except for - not at the start or end
2480 of the string, which indicates a range. If the range is in bytes,
2481 scan_const expands the range to the full set of intermediate
2482 characters. If the range is in utf8, the hyphen is replaced with
2483 a certain range mark which will be handled by pmtrans() in op.c.
2485 In double-quoted strings:
2487 double-quoted style: \r and \n
2488 constants: \x31, etc.
2489 deprecated backrefs: \1 (in substitution replacements)
2490 case and quoting: \U \Q \E
2493 scan_const does *not* construct ops to handle interpolated strings.
2494 It stops processing as soon as it finds an embedded $ or @ variable
2495 and leaves it to the caller to work out what's going on.
2497 embedded arrays (whether in pattern or not) could be:
2498 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2500 $ in double-quoted strings must be the symbol of an embedded scalar.
2502 $ in pattern could be $foo or could be tail anchor. Assumption:
2503 it's a tail anchor if $ is the last thing in the string, or if it's
2504 followed by one of "()| \r\n\t"
2506 \1 (backreferences) are turned into $1
2508 The structure of the code is
2509 while (there's a character to process) {
2510 handle transliteration ranges
2511 skip regexp comments /(?#comment)/ and codes /(?{code})/
2512 skip #-initiated comments in //x patterns
2513 check for embedded arrays
2514 check for embedded scalars
2516 deprecate \1 in substitution replacements
2517 handle string-changing backslashes \l \U \Q \E, etc.
2518 switch (what was escaped) {
2519 handle \- in a transliteration (becomes a literal -)
2520 if a pattern and not \N{, go treat as regular character
2521 handle \132 (octal characters)
2522 handle \x15 and \x{1234} (hex characters)
2523 handle \N{name} (named characters, also \N{3,5} in a pattern)
2524 handle \cV (control characters)
2525 handle printf-style backslashes (\f, \r, \n, etc)
2528 } (end if backslash)
2529 handle regular character
2530 } (end while character to read)
2535 S_scan_const(pTHX_ char *start)
2538 register char *send = PL_bufend; /* end of the constant */
2539 SV *sv = newSV(send - start); /* sv for the constant. See
2540 note below on sizing. */
2541 register char *s = start; /* start of the constant */
2542 register char *d = SvPVX(sv); /* destination for copies */
2543 bool dorange = FALSE; /* are we in a translit range? */
2544 bool didrange = FALSE; /* did we just finish a range? */
2545 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
2546 I32 this_utf8 = UTF; /* Is the source string assumed
2547 to be UTF8? But, this can
2548 show as true when the source
2549 isn't utf8, as for example
2550 when it is entirely composed
2553 /* Note on sizing: The scanned constant is placed into sv, which is
2554 * initialized by newSV() assuming one byte of output for every byte of
2555 * input. This routine expects newSV() to allocate an extra byte for a
2556 * trailing NUL, which this routine will append if it gets to the end of
2557 * the input. There may be more bytes of input than output (eg., \N{LATIN
2558 * CAPITAL LETTER A}), or more output than input if the constant ends up
2559 * recoded to utf8, but each time a construct is found that might increase
2560 * the needed size, SvGROW() is called. Its size parameter each time is
2561 * based on the best guess estimate at the time, namely the length used so
2562 * far, plus the length the current construct will occupy, plus room for
2563 * the trailing NUL, plus one byte for every input byte still unscanned */
2567 UV literal_endpoint = 0;
2568 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2571 PERL_ARGS_ASSERT_SCAN_CONST;
2573 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2574 /* If we are doing a trans and we know we want UTF8 set expectation */
2575 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2576 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2580 while (s < send || dorange) {
2582 /* get transliterations out of the way (they're most literal) */
2583 if (PL_lex_inwhat == OP_TRANS) {
2584 /* expand a range A-Z to the full set of characters. AIE! */
2586 I32 i; /* current expanded character */
2587 I32 min; /* first character in range */
2588 I32 max; /* last character in range */
2599 char * const c = (char*)utf8_hop((U8*)d, -1);
2603 *c = (char)UTF_TO_NATIVE(0xff);
2604 /* mark the range as done, and continue */
2610 i = d - SvPVX_const(sv); /* remember current offset */
2613 SvLEN(sv) + (has_utf8 ?
2614 (512 - UTF_CONTINUATION_MARK +
2617 /* How many two-byte within 0..255: 128 in UTF-8,
2618 * 96 in UTF-8-mod. */
2620 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2622 d = SvPVX(sv) + i; /* refresh d after realloc */
2626 for (j = 0; j <= 1; j++) {
2627 char * const c = (char*)utf8_hop((U8*)d, -1);
2628 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2634 max = (U8)0xff; /* only to \xff */
2635 uvmax = uv; /* \x{100} to uvmax */
2637 d = c; /* eat endpoint chars */
2642 d -= 2; /* eat the first char and the - */
2643 min = (U8)*d; /* first char in range */
2644 max = (U8)d[1]; /* last char in range */
2651 "Invalid range \"%c-%c\" in transliteration operator",
2652 (char)min, (char)max);
2656 if (literal_endpoint == 2 &&
2657 ((isLOWER(min) && isLOWER(max)) ||
2658 (isUPPER(min) && isUPPER(max)))) {
2660 for (i = min; i <= max; i++)
2662 *d++ = NATIVE_TO_NEED(has_utf8,i);
2664 for (i = min; i <= max; i++)
2666 *d++ = NATIVE_TO_NEED(has_utf8,i);
2671 for (i = min; i <= max; i++)
2674 const U8 ch = (U8)NATIVE_TO_UTF(i);
2675 if (UNI_IS_INVARIANT(ch))
2678 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2679 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2688 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2690 *d++ = (char)UTF_TO_NATIVE(0xff);
2692 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2696 /* mark the range as done, and continue */
2700 literal_endpoint = 0;
2705 /* range begins (ignore - as first or last char) */
2706 else if (*s == '-' && s+1 < send && s != start) {
2708 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2715 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2725 literal_endpoint = 0;
2726 native_range = TRUE;
2731 /* if we get here, we're not doing a transliteration */
2733 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2734 except for the last char, which will be done separately. */
2735 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2737 while (s+1 < send && *s != ')')
2738 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2740 else if (s[2] == '{' /* This should match regcomp.c */
2741 || (s[2] == '?' && s[3] == '{'))
2744 char *regparse = s + (s[2] == '{' ? 3 : 4);
2747 while (count && (c = *regparse)) {
2748 if (c == '\\' && regparse[1])
2756 if (*regparse != ')')
2757 regparse--; /* Leave one char for continuation. */
2758 while (s < regparse)
2759 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2763 /* likewise skip #-initiated comments in //x patterns */
2764 else if (*s == '#' && PL_lex_inpat &&
2765 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2766 while (s+1 < send && *s != '\n')
2767 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2770 /* check for embedded arrays
2771 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2773 else if (*s == '@' && s[1]) {
2774 if (isALNUM_lazy_if(s+1,UTF))
2776 if (strchr(":'{$", s[1]))
2778 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2779 break; /* in regexp, neither @+ nor @- are interpolated */
2782 /* check for embedded scalars. only stop if we're sure it's a
2785 else if (*s == '$') {
2786 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2788 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2790 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2791 "Possible unintended interpolation of $\\ in regex");
2793 break; /* in regexp, $ might be tail anchor */
2797 /* End of else if chain - OP_TRANS rejoin rest */
2800 if (*s == '\\' && s+1 < send) {
2801 char* e; /* Can be used for ending '}', etc. */
2805 /* deprecate \1 in strings and substitution replacements */
2806 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2807 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2809 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2814 /* string-change backslash escapes */
2815 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2819 /* In a pattern, process \N, but skip any other backslash escapes.
2820 * This is because we don't want to translate an escape sequence
2821 * into a meta symbol and have the regex compiler use the meta
2822 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2823 * in spite of this, we do have to process \N here while the proper
2824 * charnames handler is in scope. See bugs #56444 and #62056.
2825 * There is a complication because \N in a pattern may also stand
2826 * for 'match a non-nl', and not mean a charname, in which case its
2827 * processing should be deferred to the regex compiler. To be a
2828 * charname it must be followed immediately by a '{', and not look
2829 * like \N followed by a curly quantifier, i.e., not something like
2830 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2832 else if (PL_lex_inpat
2835 || regcurly(s + 1)))
2837 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2838 goto default_action;
2843 /* quoted - in transliterations */
2845 if (PL_lex_inwhat == OP_TRANS) {
2852 if ((isALPHA(*s) || isDIGIT(*s)))
2853 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2854 "Unrecognized escape \\%c passed through",
2856 /* default action is to copy the quoted character */
2857 goto default_action;
2860 /* eg. \132 indicates the octal constant 0x132 */
2861 case '0': case '1': case '2': case '3':
2862 case '4': case '5': case '6': case '7':
2866 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2869 goto NUM_ESCAPE_INSERT;
2871 /* eg. \x24 indicates the hex constant 0x24 */
2875 char* const e = strchr(s, '}');
2876 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2877 PERL_SCAN_DISALLOW_PREFIX;
2882 yyerror("Missing right brace on \\x{}");
2886 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2892 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2893 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2899 /* Insert oct or hex escaped character. There will always be
2900 * enough room in sv since such escapes will be longer than any
2901 * UTF-8 sequence they can end up as, except if they force us
2902 * to recode the rest of the string into utf8 */
2904 /* Here uv is the ordinal of the next character being added in
2905 * unicode (converted from native). */
2906 if (!UNI_IS_INVARIANT(uv)) {
2907 if (!has_utf8 && uv > 255) {
2908 /* Might need to recode whatever we have accumulated so
2909 * far if it contains any chars variant in utf8 or
2912 SvCUR_set(sv, d - SvPVX_const(sv));
2915 /* See Note on sizing above. */
2916 sv_utf8_upgrade_flags_grow(sv,
2917 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2918 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2919 d = SvPVX(sv) + SvCUR(sv);
2924 d = (char*)uvuni_to_utf8((U8*)d, uv);
2925 if (PL_lex_inwhat == OP_TRANS &&
2926 PL_sublex_info.sub_op) {
2927 PL_sublex_info.sub_op->op_private |=
2928 (PL_lex_repl ? OPpTRANS_FROM_UTF
2932 if (uv > 255 && !dorange)
2933 native_range = FALSE;
2946 /* In a non-pattern \N must be a named character, like \N{LATIN
2947 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
2948 * mean to match a non-newline. For non-patterns, named
2949 * characters are converted to their string equivalents. In
2950 * patterns, named characters are not converted to their
2951 * ultimate forms for the same reasons that other escapes
2952 * aren't. Instead, they are converted to the \N{U+...} form
2953 * to get the value from the charnames that is in effect right
2954 * now, while preserving the fact that it was a named character
2955 * so that the regex compiler knows this */
2957 /* This section of code doesn't generally use the
2958 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
2959 * a close examination of this macro and determined it is a
2960 * no-op except on utfebcdic variant characters. Every
2961 * character generated by this that would normally need to be
2962 * enclosed by this macro is invariant, so the macro is not
2963 * needed, and would complicate use of copy(). There are other
2964 * parts of this file where the macro is used inconsistently,
2965 * but are saved by it being a no-op */
2967 /* The structure of this section of code (besides checking for
2968 * errors and upgrading to utf8) is:
2969 * Further disambiguate between the two meanings of \N, and if
2970 * not a charname, go process it elsewhere
2971 * If of form \N{U+...}, pass it through if a pattern; otherwise
2973 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a pattern;
2974 * otherwise convert to utf8 */
2976 /* Here, s points to the 'N'; the test below is guaranteed to
2977 * succeed if we are being called on a pattern as we already
2978 * know from a test above that the next character is a '{'.
2979 * On a non-pattern \N must mean 'named sequence, which
2980 * requires braces */
2983 yyerror("Missing braces on \\N{}");
2988 /* If there is no matching '}', it is an error outside of a
2989 * pattern, or ambiguous inside. */
2990 if (! (e = strchr(s, '}'))) {
2991 if (! PL_lex_inpat) {
2992 yyerror("Missing right brace on \\N{}");
2997 /* A missing brace means it can't be a legal character
2998 * name, and it could be a legal "match non-newline".
2999 * But it's kind of weird without an unescaped left
3000 * brace, so warn. */
3001 if (ckWARN(WARN_SYNTAX)) {
3002 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
3003 "Missing right brace on \\N{} or unescaped left brace after \\N. Assuming the latter");
3005 s -= 3; /* Backup over cur char, {, N, to the '\' */
3006 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
3007 goto default_action;
3011 /* Here it looks like a named character */
3015 /* XXX This block is temporary code. \N{} implies that the
3016 * pattern is to have Unicode semantics, and therefore
3017 * currently has to be encoded in utf8. By putting it in
3018 * utf8 now, we save a whole pass in the regular expression
3019 * compiler. Once that code is changed so Unicode
3020 * semantics doesn't necessarily have to be in utf8, this
3021 * block should be removed */
3023 SvCUR_set(sv, d - SvPVX_const(sv));
3026 /* See Note on sizing above. */
3027 sv_utf8_upgrade_flags_grow(sv,
3028 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3029 /* 5 = '\N{' + cur char + NUL */
3030 (STRLEN)(send - s) + 5);
3031 d = SvPVX(sv) + SvCUR(sv);
3036 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3037 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3038 | PERL_SCAN_DISALLOW_PREFIX;
3041 /* For \N{U+...}, the '...' is a unicode value even on
3042 * EBCDIC machines */
3043 s += 2; /* Skip to next char after the 'U+' */
3045 uv = grok_hex(s, &len, &flags, NULL);
3046 if (len == 0 || len != (STRLEN)(e - s)) {
3047 yyerror("Invalid hexadecimal number in \\N{U+...}");
3054 /* Pass through to the regex compiler unchanged. The
3055 * reason we evaluated the number above is to make sure
3056 * there wasn't a syntax error. It also makes sure
3057 * that the syntax created below, \N{Uc1.c2...}, is
3059 s -= 5; /* Include the '\N{U+' */
3060 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3063 else { /* Not a pattern: convert the hex to string */
3065 /* If destination is not in utf8, unconditionally
3066 * recode it to be so. This is because \N{} implies
3067 * Unicode semantics, and scalars have to be in utf8
3068 * to guarantee those semantics */
3070 SvCUR_set(sv, d - SvPVX_const(sv));
3073 /* See Note on sizing above. */
3074 sv_utf8_upgrade_flags_grow(
3076 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3077 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3078 d = SvPVX(sv) + SvCUR(sv);
3082 /* Add the string to the output */
3083 if (UNI_IS_INVARIANT(uv)) {
3086 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3089 else { /* Here is \N{NAME} but not \N{U+...}. */
3091 SV *res; /* result from charnames */
3092 const char *str; /* the string in 'res' */
3093 STRLEN len; /* its length */
3095 /* Get the value for NAME */
3096 res = newSVpvn(s, e - s);
3097 res = new_constant( NULL, 0, "charnames",
3098 /* includes all of: \N{...} */
3099 res, NULL, s - 3, e - s + 4 );
3101 /* Most likely res will be in utf8 already since the
3102 * standard charnames uses pack U, but a custom translator
3103 * can leave it otherwise, so make sure. XXX This can be
3104 * revisited to not have charnames use utf8 for characters
3105 * that don't need it when regexes don't have to be in utf8
3106 * for Unicode semantics. If doing so, remember EBCDIC */
3107 sv_utf8_upgrade(res);
3108 str = SvPV_const(res, len);
3110 /* Don't accept malformed input */
3111 if (! is_utf8_string((U8 *) str, len)) {
3112 yyerror("Malformed UTF-8 returned by \\N");
3114 else if (PL_lex_inpat) {
3116 if (! len) { /* The name resolved to an empty string */
3117 Copy("\\N{}", d, 4, char);
3121 /* In order to not lose information for the regex
3122 * compiler, pass the result in the specially made
3123 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3124 * the code points in hex of each character
3125 * returned by charnames */
3127 const char *str_end = str + len;
3128 STRLEN char_length; /* cur char's byte length */
3129 STRLEN output_length; /* and the number of bytes
3130 after this is translated
3132 const STRLEN off = d - SvPVX_const(sv);
3134 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3135 * max('U+', '.'); and 1 for NUL */
3136 char hex_string[2 * UTF8_MAXBYTES + 5];
3138 /* Get the first character of the result. */
3139 U32 uv = utf8n_to_uvuni((U8 *) str,
3144 /* The call to is_utf8_string() above hopefully
3145 * guarantees that there won't be an error. But
3146 * it's easy here to make sure. The function just
3147 * above warns and returns 0 if invalid utf8, but
3148 * it can also return 0 if the input is validly a
3149 * NUL. Disambiguate */
3150 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3151 uv = UNICODE_REPLACEMENT;
3154 /* Convert first code point to hex, including the
3155 * boiler plate before it */
3156 sprintf(hex_string, "\\N{U+%X", (unsigned int) uv);
3157 output_length = strlen(hex_string);
3159 /* Make sure there is enough space to hold it */
3160 d = off + SvGROW(sv, off
3162 + (STRLEN)(send - e)
3163 + 2); /* '}' + NUL */
3165 Copy(hex_string, d, output_length, char);
3168 /* For each subsequent character, append dot and
3169 * its ordinal in hex */
3170 while ((str += char_length) < str_end) {
3171 const STRLEN off = d - SvPVX_const(sv);
3172 U32 uv = utf8n_to_uvuni((U8 *) str,
3176 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3177 uv = UNICODE_REPLACEMENT;
3180 sprintf(hex_string, ".%X", (unsigned int) uv);
3181 output_length = strlen(hex_string);
3183 d = off + SvGROW(sv, off
3185 + (STRLEN)(send - e)
3186 + 2); /* '}' + NUL */
3187 Copy(hex_string, d, output_length, char);
3191 *d++ = '}'; /* Done. Add the trailing brace */
3194 else { /* Here, not in a pattern. Convert the name to a
3197 /* If destination is not in utf8, unconditionally
3198 * recode it to be so. This is because \N{} implies
3199 * Unicode semantics, and scalars have to be in utf8
3200 * to guarantee those semantics */
3202 SvCUR_set(sv, d - SvPVX_const(sv));
3205 /* See Note on sizing above. */
3206 sv_utf8_upgrade_flags_grow(sv,
3207 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3208 len + (STRLEN)(send - s) + 1);
3209 d = SvPVX(sv) + SvCUR(sv);
3211 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3213 /* See Note on sizing above. (NOTE: SvCUR() is not
3214 * set correctly here). */
3215 const STRLEN off = d - SvPVX_const(sv);
3216 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3218 Copy(str, d, len, char);
3225 native_range = FALSE; /* \N{} is defined to be Unicode */
3227 s = e + 1; /* Point to just after the '}' */
3230 /* \c is a control character */
3239 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
3242 yyerror("Missing control char name in \\c");
3246 /* printf-style backslashes, formfeeds, newlines, etc */
3248 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3251 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3254 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3257 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3260 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3263 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3266 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3272 } /* end if (backslash) */
3279 /* If we started with encoded form, or already know we want it,
3280 then encode the next character */
3281 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3285 /* One might think that it is wasted effort in the case of the
3286 * source being utf8 (this_utf8 == TRUE) to take the next character
3287 * in the source, convert it to an unsigned value, and then convert
3288 * it back again. But the source has not been validated here. The
3289 * routine that does the conversion checks for errors like
3292 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3293 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3295 SvCUR_set(sv, d - SvPVX_const(sv));
3298 /* See Note on sizing above. */
3299 sv_utf8_upgrade_flags_grow(sv,
3300 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3301 need + (STRLEN)(send - s) + 1);
3302 d = SvPVX(sv) + SvCUR(sv);
3304 } else if (need > len) {
3305 /* encoded value larger than old, may need extra space (NOTE:
3306 * SvCUR() is not set correctly here). See Note on sizing
3308 const STRLEN off = d - SvPVX_const(sv);
3309 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3313 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3315 if (uv > 255 && !dorange)
3316 native_range = FALSE;
3320 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3322 } /* while loop to process each character */
3324 /* terminate the string and set up the sv */
3326 SvCUR_set(sv, d - SvPVX_const(sv));
3327 if (SvCUR(sv) >= SvLEN(sv))
3328 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3331 if (PL_encoding && !has_utf8) {
3332 sv_recode_to_utf8(sv, PL_encoding);
3338 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3339 PL_sublex_info.sub_op->op_private |=
3340 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3344 /* shrink the sv if we allocated more than we used */
3345 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3346 SvPV_shrink_to_cur(sv);
3349 /* return the substring (via pl_yylval) only if we parsed anything */
3350 if (s > PL_bufptr) {
3351 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3352 const char *const key = PL_lex_inpat ? "qr" : "q";
3353 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3357 if (PL_lex_inwhat == OP_TRANS) {
3360 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3368 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3371 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3378 * Returns TRUE if there's more to the expression (e.g., a subscript),
3381 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3383 * ->[ and ->{ return TRUE
3384 * { and [ outside a pattern are always subscripts, so return TRUE
3385 * if we're outside a pattern and it's not { or [, then return FALSE
3386 * if we're in a pattern and the first char is a {
3387 * {4,5} (any digits around the comma) returns FALSE
3388 * if we're in a pattern and the first char is a [
3390 * [SOMETHING] has a funky algorithm to decide whether it's a
3391 * character class or not. It has to deal with things like
3392 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3393 * anything else returns TRUE
3396 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3399 S_intuit_more(pTHX_ register char *s)
3403 PERL_ARGS_ASSERT_INTUIT_MORE;
3405 if (PL_lex_brackets)
3407 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3409 if (*s != '{' && *s != '[')
3414 /* In a pattern, so maybe we have {n,m}. */
3431 /* On the other hand, maybe we have a character class */
3434 if (*s == ']' || *s == '^')
3437 /* this is terrifying, and it works */
3438 int weight = 2; /* let's weigh the evidence */
3440 unsigned char un_char = 255, last_un_char;
3441 const char * const send = strchr(s,']');
3442 char tmpbuf[sizeof PL_tokenbuf * 4];
3444 if (!send) /* has to be an expression */
3447 Zero(seen,256,char);
3450 else if (isDIGIT(*s)) {
3452 if (isDIGIT(s[1]) && s[2] == ']')
3458 for (; s < send; s++) {
3459 last_un_char = un_char;
3460 un_char = (unsigned char)*s;
3465 weight -= seen[un_char] * 10;
3466 if (isALNUM_lazy_if(s+1,UTF)) {
3468 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3469 len = (int)strlen(tmpbuf);
3470 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
3475 else if (*s == '$' && s[1] &&
3476 strchr("[#!%*<>()-=",s[1])) {
3477 if (/*{*/ strchr("])} =",s[2]))
3486 if (strchr("wds]",s[1]))
3488 else if (seen[(U8)'\''] || seen[(U8)'"'])
3490 else if (strchr("rnftbxcav",s[1]))
3492 else if (isDIGIT(s[1])) {
3494 while (s[1] && isDIGIT(s[1]))
3504 if (strchr("aA01! ",last_un_char))
3506 if (strchr("zZ79~",s[1]))
3508 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3509 weight -= 5; /* cope with negative subscript */
3512 if (!isALNUM(last_un_char)
3513 && !(last_un_char == '$' || last_un_char == '@'
3514 || last_un_char == '&')
3515 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3520 if (keyword(tmpbuf, d - tmpbuf, 0))
3523 if (un_char == last_un_char + 1)
3525 weight -= seen[un_char];
3530 if (weight >= 0) /* probably a character class */
3540 * Does all the checking to disambiguate
3542 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3543 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3545 * First argument is the stuff after the first token, e.g. "bar".
3547 * Not a method if bar is a filehandle.
3548 * Not a method if foo is a subroutine prototyped to take a filehandle.
3549 * Not a method if it's really "Foo $bar"
3550 * Method if it's "foo $bar"
3551 * Not a method if it's really "print foo $bar"
3552 * Method if it's really "foo package::" (interpreted as package->foo)
3553 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3554 * Not a method if bar is a filehandle or package, but is quoted with
3559 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3562 char *s = start + (*start == '$');
3563 char tmpbuf[sizeof PL_tokenbuf];
3570 PERL_ARGS_ASSERT_INTUIT_METHOD;
3573 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3577 const char *proto = SvPVX_const(cv);
3588 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3589 /* start is the beginning of the possible filehandle/object,
3590 * and s is the end of it
3591 * tmpbuf is a copy of it
3594 if (*start == '$') {
3595 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3596 isUPPER(*PL_tokenbuf))
3599 len = start - SvPVX(PL_linestr);
3603 start = SvPVX(PL_linestr) + len;
3607 return *s == '(' ? FUNCMETH : METHOD;
3609 if (!keyword(tmpbuf, len, 0)) {
3610 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3614 soff = s - SvPVX(PL_linestr);
3618 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
3619 if (indirgv && GvCVu(indirgv))
3621 /* filehandle or package name makes it a method */
3622 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
3624 soff = s - SvPVX(PL_linestr);
3627 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3628 return 0; /* no assumptions -- "=>" quotes bearword */
3630 start_force(PL_curforce);
3631 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3632 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3633 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3635 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3640 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3642 return *s == '(' ? FUNCMETH : METHOD;
3648 /* Encoded script support. filter_add() effectively inserts a
3649 * 'pre-processing' function into the current source input stream.
3650 * Note that the filter function only applies to the current source file
3651 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3653 * The datasv parameter (which may be NULL) can be used to pass
3654 * private data to this instance of the filter. The filter function
3655 * can recover the SV using the FILTER_DATA macro and use it to
3656 * store private buffers and state information.
3658 * The supplied datasv parameter is upgraded to a PVIO type
3659 * and the IoDIRP/IoANY field is used to store the function pointer,
3660 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3661 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3662 * private use must be set using malloc'd pointers.
3666 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3675 if (!PL_rsfp_filters)
3676 PL_rsfp_filters = newAV();
3679 SvUPGRADE(datasv, SVt_PVIO);
3680 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3681 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3682 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3683 FPTR2DPTR(void *, IoANY(datasv)),
3684 SvPV_nolen(datasv)));
3685 av_unshift(PL_rsfp_filters, 1);
3686 av_store(PL_rsfp_filters, 0, datasv) ;
3691 /* Delete most recently added instance of this filter function. */
3693 Perl_filter_del(pTHX_ filter_t funcp)
3698 PERL_ARGS_ASSERT_FILTER_DEL;
3701 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3702 FPTR2DPTR(void*, funcp)));
3704 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3706 /* if filter is on top of stack (usual case) just pop it off */
3707 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3708 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3709 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
3710 IoANY(datasv) = (void *)NULL;
3711 sv_free(av_pop(PL_rsfp_filters));
3715 /* we need to search for the correct entry and clear it */
3716 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3720 /* Invoke the idxth filter function for the current rsfp. */
3721 /* maxlen 0 = read one text line */
3723 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3728 /* This API is bad. It should have been using unsigned int for maxlen.
3729 Not sure if we want to change the API, but if not we should sanity
3730 check the value here. */
3731 const unsigned int correct_length
3740 PERL_ARGS_ASSERT_FILTER_READ;
3742 if (!PL_parser || !PL_rsfp_filters)
3744 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3745 /* Provide a default input filter to make life easy. */
3746 /* Note that we append to the line. This is handy. */
3747 DEBUG_P(PerlIO_printf(Perl_debug_log,
3748 "filter_read %d: from rsfp\n", idx));
3749 if (correct_length) {
3752 const int old_len = SvCUR(buf_sv);
3754 /* ensure buf_sv is large enough */
3755 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3756 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3757 correct_length)) <= 0) {
3758 if (PerlIO_error(PL_rsfp))
3759 return -1; /* error */
3761 return 0 ; /* end of file */
3763 SvCUR_set(buf_sv, old_len + len) ;
3764 SvPVX(buf_sv)[old_len + len] = '\0';
3767 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3768 if (PerlIO_error(PL_rsfp))
3769 return -1; /* error */
3771 return 0 ; /* end of file */
3774 return SvCUR(buf_sv);
3776 /* Skip this filter slot if filter has been deleted */
3777 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3778 DEBUG_P(PerlIO_printf(Perl_debug_log,
3779 "filter_read %d: skipped (filter deleted)\n",
3781 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
3783 /* Get function pointer hidden within datasv */
3784 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
3785 DEBUG_P(PerlIO_printf(Perl_debug_log,
3786 "filter_read %d: via function %p (%s)\n",
3787 idx, (void*)datasv, SvPV_nolen_const(datasv)));
3788 /* Call function. The function is expected to */
3789 /* call "FILTER_READ(idx+1, buf_sv)" first. */
3790 /* Return: <0:error, =0:eof, >0:not eof */
3791 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
3795 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
3799 PERL_ARGS_ASSERT_FILTER_GETS;
3801 #ifdef PERL_CR_FILTER
3802 if (!PL_rsfp_filters) {
3803 filter_add(S_cr_textfilter,NULL);
3806 if (PL_rsfp_filters) {
3808 SvCUR_set(sv, 0); /* start with empty line */
3809 if (FILTER_READ(0, sv, 0) > 0)
3810 return ( SvPVX(sv) ) ;
3815 return (sv_gets(sv, PL_rsfp, append));
3819 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3824 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3826 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3830 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3831 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3833 return GvHV(gv); /* Foo:: */
3836 /* use constant CLASS => 'MyClass' */
3837 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3838 if (gv && GvCV(gv)) {
3839 SV * const sv = cv_const_sv(GvCV(gv));
3841 pkgname = SvPV_const(sv, len);
3844 return gv_stashpvn(pkgname, len, 0);
3848 * S_readpipe_override
3849 * Check whether readpipe() is overriden, and generates the appropriate
3850 * optree, provided sublex_start() is called afterwards.
3853 S_readpipe_override(pTHX)
3856 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3857 pl_yylval.ival = OP_BACKTICK;
3859 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3861 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3862 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3863 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3865 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3866 append_elem(OP_LIST,
3867 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3868 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3875 * The intent of this yylex wrapper is to minimize the changes to the
3876 * tokener when we aren't interested in collecting madprops. It remains
3877 * to be seen how successful this strategy will be...
3884 char *s = PL_bufptr;
3886 /* make sure PL_thiswhite is initialized */
3890 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3891 if (PL_pending_ident)
3892 return S_pending_ident(aTHX);
3894 /* previous token ate up our whitespace? */
3895 if (!PL_lasttoke && PL_nextwhite) {
3896 PL_thiswhite = PL_nextwhite;
3900 /* isolate the token, and figure out where it is without whitespace */
3901 PL_realtokenstart = -1;
3905 assert(PL_curforce < 0);
3907 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3908 if (!PL_thistoken) {
3909 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3910 PL_thistoken = newSVpvs("");
3912 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3913 PL_thistoken = newSVpvn(tstart, s - tstart);
3916 if (PL_thismad) /* install head */
3917 CURMAD('X', PL_thistoken);
3920 /* last whitespace of a sublex? */
3921 if (optype == ')' && PL_endwhite) {
3922 CURMAD('X', PL_endwhite);
3927 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3928 if (!PL_thiswhite && !PL_endwhite && !optype) {
3929 sv_free(PL_thistoken);
3934 /* put off final whitespace till peg */
3935 if (optype == ';' && !PL_rsfp) {
3936 PL_nextwhite = PL_thiswhite;
3939 else if (PL_thisopen) {
3940 CURMAD('q', PL_thisopen);
3942 sv_free(PL_thistoken);
3946 /* Store actual token text as madprop X */
3947 CURMAD('X', PL_thistoken);
3951 /* add preceding whitespace as madprop _ */
3952 CURMAD('_', PL_thiswhite);
3956 /* add quoted material as madprop = */
3957 CURMAD('=', PL_thisstuff);
3961 /* add terminating quote as madprop Q */
3962 CURMAD('Q', PL_thisclose);
3966 /* special processing based on optype */
3970 /* opval doesn't need a TOKEN since it can already store mp */
3980 if (pl_yylval.opval)
3981 append_madprops(PL_thismad, pl_yylval.opval, 0);
3989 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
3998 /* remember any fake bracket that lexer is about to discard */
3999 if (PL_lex_brackets == 1 &&
4000 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4003 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4006 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4007 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4010 break; /* don't bother looking for trailing comment */
4019 /* attach a trailing comment to its statement instead of next token */
4023 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4025 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4027 if (*s == '\n' || *s == '#') {
4028 while (s < PL_bufend && *s != '\n')
4032 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4033 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4050 /* Create new token struct. Note: opvals return early above. */
4051 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4058 S_tokenize_use(pTHX_ int is_use, char *s) {
4061 PERL_ARGS_ASSERT_TOKENIZE_USE;
4063 if (PL_expect != XSTATE)
4064 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4065 is_use ? "use" : "no"));
4067 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4068 s = force_version(s, TRUE);
4069 if (*s == ';' || *s == '}'
4070 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4071 start_force(PL_curforce);
4072 NEXTVAL_NEXTTOKE.opval = NULL;
4075 else if (*s == 'v') {
4076 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4077 s = force_version(s, FALSE);
4081 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4082 s = force_version(s, FALSE);
4084 pl_yylval.ival = is_use;
4088 static const char* const exp_name[] =
4089 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4090 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4097 Works out what to call the token just pulled out of the input
4098 stream. The yacc parser takes care of taking the ops we return and
4099 stitching them into a tree.
4105 if read an identifier
4106 if we're in a my declaration
4107 croak if they tried to say my($foo::bar)
4108 build the ops for a my() declaration
4109 if it's an access to a my() variable
4110 are we in a sort block?
4111 croak if my($a); $a <=> $b
4112 build ops for access to a my() variable
4113 if in a dq string, and they've said @foo and we can't find @foo
4115 build ops for a bareword
4116 if we already built the token before, use it.
4121 #pragma segment Perl_yylex
4127 register char *s = PL_bufptr;
4133 /* orig_keyword, gvp, and gv are initialized here because
4134 * jump to the label just_a_word_zero can bypass their
4135 * initialization later. */
4136 I32 orig_keyword = 0;
4141 SV* tmp = newSVpvs("");
4142 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4143 (IV)CopLINE(PL_curcop),
4144 lex_state_names[PL_lex_state],
4145 exp_name[PL_expect],
4146 pv_display(tmp, s, strlen(s), 0, 60));
4149 /* check if there's an identifier for us to look at */
4150 if (PL_pending_ident)
4151 return REPORT(S_pending_ident(aTHX));
4153 /* no identifier pending identification */
4155 switch (PL_lex_state) {
4157 case LEX_NORMAL: /* Some compilers will produce faster */
4158 case LEX_INTERPNORMAL: /* code if we comment these out. */
4162 /* when we've already built the next token, just pull it out of the queue */
4166 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4168 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4169 PL_nexttoke[PL_lasttoke].next_mad = 0;
4170 if (PL_thismad && PL_thismad->mad_key == '_') {
4171 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4172 PL_thismad->mad_val = 0;
4173 mad_free(PL_thismad);
4178 PL_lex_state = PL_lex_defer;
4179 PL_expect = PL_lex_expect;
4180 PL_lex_defer = LEX_NORMAL;
4181 if (!PL_nexttoke[PL_lasttoke].next_type)
4186 pl_yylval = PL_nextval[PL_nexttoke];
4188 PL_lex_state = PL_lex_defer;
4189 PL_expect = PL_lex_expect;
4190 PL_lex_defer = LEX_NORMAL;
4194 /* FIXME - can these be merged? */
4195 return(PL_nexttoke[PL_lasttoke].next_type);
4197 return REPORT(PL_nexttype[PL_nexttoke]);
4200 /* interpolated case modifiers like \L \U, including \Q and \E.
4201 when we get here, PL_bufptr is at the \
4203 case LEX_INTERPCASEMOD:
4205 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4206 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
4208 /* handle \E or end of string */
4209 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4211 if (PL_lex_casemods) {
4212 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4213 PL_lex_casestack[PL_lex_casemods] = '\0';
4215 if (PL_bufptr != PL_bufend
4216 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
4218 PL_lex_state = LEX_INTERPCONCAT;
4221 PL_thistoken = newSVpvs("\\E");
4227 while (PL_bufptr != PL_bufend &&
4228 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4230 PL_thiswhite = newSVpvs("");
4231 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4235 if (PL_bufptr != PL_bufend)
4238 PL_lex_state = LEX_INTERPCONCAT;
4242 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4243 "### Saw case modifier\n"); });
4245 if (s[1] == '\\' && s[2] == 'E') {
4248 PL_thiswhite = newSVpvs("");
4249 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4252 PL_lex_state = LEX_INTERPCONCAT;
4257 if (!PL_madskills) /* when just compiling don't need correct */
4258 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4259 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4260 if ((*s == 'L' || *s == 'U') &&
4261 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4262 PL_lex_casestack[--PL_lex_casemods] = '\0';
4265 if (PL_lex_casemods > 10)
4266 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4267 PL_lex_casestack[PL_lex_casemods++] = *s;
4268 PL_lex_casestack[PL_lex_casemods] = '\0';
4269 PL_lex_state = LEX_INTERPCONCAT;
4270 start_force(PL_curforce);
4271 NEXTVAL_NEXTTOKE.ival = 0;
4273 start_force(PL_curforce);
4275 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4277 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4279 NEXTVAL_NEXTTOKE.ival = OP_LC;
4281 NEXTVAL_NEXTTOKE.ival = OP_UC;
4283 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4285 Perl_croak(aTHX_ "panic: yylex");
4287 SV* const tmpsv = newSVpvs("\\ ");
4288 /* replace the space with the character we want to escape
4290 SvPVX(tmpsv)[1] = *s;
4296 if (PL_lex_starts) {
4302 sv_free(PL_thistoken);
4303 PL_thistoken = newSVpvs("");
4306 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4307 if (PL_lex_casemods == 1 && PL_lex_inpat)
4316 case LEX_INTERPPUSH:
4317 return REPORT(sublex_push());
4319 case LEX_INTERPSTART:
4320 if (PL_bufptr == PL_bufend)
4321 return REPORT(sublex_done());
4322 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4323 "### Interpolated variable\n"); });
4325 PL_lex_dojoin = (*PL_bufptr == '@');
4326 PL_lex_state = LEX_INTERPNORMAL;
4327 if (PL_lex_dojoin) {
4328 start_force(PL_curforce);
4329 NEXTVAL_NEXTTOKE.ival = 0;
4331 start_force(PL_curforce);
4332 force_ident("\"", '$');
4333 start_force(PL_curforce);
4334 NEXTVAL_NEXTTOKE.ival = 0;
4336 start_force(PL_curforce);
4337 NEXTVAL_NEXTTOKE.ival = 0;
4339 start_force(PL_curforce);
4340 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4343 if (PL_lex_starts++) {
4348 sv_free(PL_thistoken);
4349 PL_thistoken = newSVpvs("");
4352 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4353 if (!PL_lex_casemods && PL_lex_inpat)
4360 case LEX_INTERPENDMAYBE:
4361 if (intuit_more(PL_bufptr)) {
4362 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4368 if (PL_lex_dojoin) {
4369 PL_lex_dojoin = FALSE;
4370 PL_lex_state = LEX_INTERPCONCAT;
4374 sv_free(PL_thistoken);
4375 PL_thistoken = newSVpvs("");
4380 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4381 && SvEVALED(PL_lex_repl))
4383 if (PL_bufptr != PL_bufend)
4384 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4388 case LEX_INTERPCONCAT:
4390 if (PL_lex_brackets)
4391 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4393 if (PL_bufptr == PL_bufend)
4394 return REPORT(sublex_done());
4396 if (SvIVX(PL_linestr) == '\'') {
4397 SV *sv = newSVsv(PL_linestr);
4400 else if ( PL_hints & HINT_NEW_RE )
4401 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4402 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4406 s = scan_const(PL_bufptr);
4408 PL_lex_state = LEX_INTERPCASEMOD;
4410 PL_lex_state = LEX_INTERPSTART;
4413 if (s != PL_bufptr) {
4414 start_force(PL_curforce);
4416 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4418 NEXTVAL_NEXTTOKE = pl_yylval;
4421 if (PL_lex_starts++) {
4425 sv_free(PL_thistoken);
4426 PL_thistoken = newSVpvs("");
4429 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4430 if (!PL_lex_casemods && PL_lex_inpat)
4443 PL_lex_state = LEX_NORMAL;
4444 s = scan_formline(PL_bufptr);
4445 if (!PL_lex_formbrack)
4451 PL_oldoldbufptr = PL_oldbufptr;
4457 sv_free(PL_thistoken);
4460 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4464 if (isIDFIRST_lazy_if(s,UTF))
4467 unsigned char c = *s;
4468 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4469 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4470 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4475 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4479 goto fake_eof; /* emulate EOF on ^D or ^Z */
4488 if (PL_lex_brackets) {
4489 yyerror((const char *)
4491 ? "Format not terminated"
4492 : "Missing right curly or square bracket"));
4494 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4495 "### Tokener got EOF\n");
4499 if (s++ < PL_bufend)
4500 goto retry; /* ignore stray nulls */
4503 if (!PL_in_eval && !PL_preambled) {
4504 PL_preambled = TRUE;
4510 /* Generate a string of Perl code to load the debugger.
4511 * If PERL5DB is set, it will return the contents of that,
4512 * otherwise a compile-time require of perl5db.pl. */
4514 const char * const pdb = PerlEnv_getenv("PERL5DB");
4517 sv_setpv(PL_linestr, pdb);
4518 sv_catpvs(PL_linestr,";");
4520 SETERRNO(0,SS_NORMAL);
4521 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4524 sv_setpvs(PL_linestr,"");
4525 if (PL_preambleav) {
4526 SV **svp = AvARRAY(PL_preambleav);
4527 SV **const end = svp + AvFILLp(PL_preambleav);
4529 sv_catsv(PL_linestr, *svp);
4531 sv_catpvs(PL_linestr, ";");
4533 sv_free(MUTABLE_SV(PL_preambleav));
4534 PL_preambleav = NULL;
4537 sv_catpvs(PL_linestr,
4538 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4539 if (PL_minus_n || PL_minus_p) {
4540 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4542 sv_catpvs(PL_linestr,"chomp;");
4545 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4546 || *PL_splitstr == '"')
4547 && strchr(PL_splitstr + 1, *PL_splitstr))
4548 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4550 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4551 bytes can be used as quoting characters. :-) */
4552 const char *splits = PL_splitstr;
4553 sv_catpvs(PL_linestr, "our @F=split(q\0");
4556 if (*splits == '\\')
4557 sv_catpvn(PL_linestr, splits, 1);
4558 sv_catpvn(PL_linestr, splits, 1);
4559 } while (*splits++);
4560 /* This loop will embed the trailing NUL of
4561 PL_linestr as the last thing it does before
4563 sv_catpvs(PL_linestr, ");");
4567 sv_catpvs(PL_linestr,"our @F=split(' ');");
4570 sv_catpvs(PL_linestr, "\n");
4571 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4572 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4573 PL_last_lop = PL_last_uni = NULL;
4574 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4575 update_debugger_info(PL_linestr, NULL, 0);
4580 bof = PL_rsfp ? TRUE : FALSE;
4583 fake_eof = LEX_FAKE_EOF;
4585 PL_bufptr = PL_bufend;
4586 CopLINE_inc(PL_curcop);
4587 if (!lex_next_chunk(fake_eof)) {
4588 CopLINE_dec(PL_curcop);
4590 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4592 CopLINE_dec(PL_curcop);
4595 PL_realtokenstart = -1;
4598 /* If it looks like the start of a BOM or raw UTF-16,
4599 * check if it in fact is. */
4600 if (bof && PL_rsfp &&
4605 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
4607 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4608 s = swallow_bom((U8*)s);
4612 /* Incest with pod. */
4615 sv_catsv(PL_thiswhite, PL_linestr);
4617 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4618 sv_setpvs(PL_linestr, "");
4619 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4620 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4621 PL_last_lop = PL_last_uni = NULL;
4622 PL_doextract = FALSE;
4627 } while (PL_doextract);
4628 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4629 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4630 PL_last_lop = PL_last_uni = NULL;
4631 if (CopLINE(PL_curcop) == 1) {
4632 while (s < PL_bufend && isSPACE(*s))
4634 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4638 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4642 if (*s == '#' && *(s+1) == '!')
4644 #ifdef ALTERNATE_SHEBANG
4646 static char const as[] = ALTERNATE_SHEBANG;
4647 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4648 d = s + (sizeof(as) - 1);
4650 #endif /* ALTERNATE_SHEBANG */
4659 while (*d && !isSPACE(*d))
4663 #ifdef ARG_ZERO_IS_SCRIPT
4664 if (ipathend > ipath) {
4666 * HP-UX (at least) sets argv[0] to the script name,
4667 * which makes $^X incorrect. And Digital UNIX and Linux,
4668 * at least, set argv[0] to the basename of the Perl
4669 * interpreter. So, having found "#!", we'll set it right.
4671 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4673 assert(SvPOK(x) || SvGMAGICAL(x));
4674 if (sv_eq(x, CopFILESV(PL_curcop))) {
4675 sv_setpvn(x, ipath, ipathend - ipath);
4681 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4682 const char * const lstart = SvPV_const(x,llen);
4684 bstart += blen - llen;
4685 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4686 sv_setpvn(x, ipath, ipathend - ipath);
4691 TAINT_NOT; /* $^X is always tainted, but that's OK */
4693 #endif /* ARG_ZERO_IS_SCRIPT */
4698 d = instr(s,"perl -");
4700 d = instr(s,"perl");
4702 /* avoid getting into infinite loops when shebang
4703 * line contains "Perl" rather than "perl" */
4705 for (d = ipathend-4; d >= ipath; --d) {
4706 if ((*d == 'p' || *d == 'P')
4707 && !ibcmp(d, "perl", 4))
4717 #ifdef ALTERNATE_SHEBANG
4719 * If the ALTERNATE_SHEBANG on this system starts with a
4720 * character that can be part of a Perl expression, then if
4721 * we see it but not "perl", we're probably looking at the
4722 * start of Perl code, not a request to hand off to some
4723 * other interpreter. Similarly, if "perl" is there, but
4724 * not in the first 'word' of the line, we assume the line
4725 * contains the start of the Perl program.
4727 if (d && *s != '#') {
4728 const char *c = ipath;
4729 while (*c && !strchr("; \t\r\n\f\v#", *c))
4732 d = NULL; /* "perl" not in first word; ignore */
4734 *s = '#'; /* Don't try to parse shebang line */
4736 #endif /* ALTERNATE_SHEBANG */
4741 !instr(s,"indir") &&
4742 instr(PL_origargv[0],"perl"))
4749 while (s < PL_bufend && isSPACE(*s))
4751 if (s < PL_bufend) {
4752 Newx(newargv,PL_origargc+3,char*);
4754 while (s < PL_bufend && !isSPACE(*s))
4757 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4760 newargv = PL_origargv;
4763 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4765 Perl_croak(aTHX_ "Can't exec %s", ipath);
4768 while (*d && !isSPACE(*d))
4770 while (SPACE_OR_TAB(*d))
4774 const bool switches_done = PL_doswitches;
4775 const U32 oldpdb = PL_perldb;
4776 const bool oldn = PL_minus_n;
4777 const bool oldp = PL_minus_p;
4781 bool baduni = FALSE;
4783 const char *d2 = d1 + 1;
4784 if (parse_unicode_opts((const char **)&d2)
4788 if (baduni || *d1 == 'M' || *d1 == 'm') {
4789 const char * const m = d1;
4790 while (*d1 && !isSPACE(*d1))
4792 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4795 d1 = moreswitches(d1);
4797 if (PL_doswitches && !switches_done) {
4798 int argc = PL_origargc;
4799 char **argv = PL_origargv;
4802 } while (argc && argv[0][0] == '-' && argv[0][1]);
4803 init_argv_symbols(argc,argv);
4805 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4806 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4807 /* if we have already added "LINE: while (<>) {",
4808 we must not do it again */
4810 sv_setpvs(PL_linestr, "");
4811 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4812 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4813 PL_last_lop = PL_last_uni = NULL;
4814 PL_preambled = FALSE;
4815 if (PERLDB_LINE || PERLDB_SAVESRC)
4816 (void)gv_fetchfile(PL_origfilename);
4823 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4825 PL_lex_state = LEX_FORMLINE;
4830 #ifdef PERL_STRICT_CR
4831 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4833 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4835 case ' ': case '\t': case '\f': case 013:
4837 PL_realtokenstart = -1;
4839 PL_thiswhite = newSVpvs("");
4840 sv_catpvn(PL_thiswhite, s, 1);
4847 PL_realtokenstart = -1;
4851 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4852 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4853 /* handle eval qq[#line 1 "foo"\n ...] */
4854 CopLINE_dec(PL_curcop);
4857 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4859 if (!PL_in_eval || PL_rsfp)
4864 while (d < PL_bufend && *d != '\n')
4868 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4869 Perl_croak(aTHX_ "panic: input overflow");
4872 PL_thiswhite = newSVpvn(s, d - s);
4877 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4879 PL_lex_state = LEX_FORMLINE;
4885 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4886 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4889 TOKEN(PEG); /* make sure any #! line is accessible */
4894 /* if (PL_madskills && PL_lex_formbrack) { */
4896 while (d < PL_bufend && *d != '\n')
4900 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4901 Perl_croak(aTHX_ "panic: input overflow");
4902 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4904 PL_thiswhite = newSVpvs("");
4905 if (CopLINE(PL_curcop) == 1) {
4906 sv_setpvs(PL_thiswhite, "");
4909 sv_catpvn(PL_thiswhite, s, d - s);
4923 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4931 while (s < PL_bufend && SPACE_OR_TAB(*s))
4934 if (strnEQ(s,"=>",2)) {
4935 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4936 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4937 OPERATOR('-'); /* unary minus */
4939 PL_last_uni = PL_oldbufptr;
4941 case 'r': ftst = OP_FTEREAD; break;
4942 case 'w': ftst = OP_FTEWRITE; break;
4943 case 'x': ftst = OP_FTEEXEC; break;
4944 case 'o': ftst = OP_FTEOWNED; break;
4945 case 'R': ftst = OP_FTRREAD; break;
4946 case 'W': ftst = OP_FTRWRITE; break;
4947 case 'X': ftst = OP_FTREXEC; break;
4948 case 'O': ftst = OP_FTROWNED; break;
4949 case 'e': ftst = OP_FTIS; break;
4950 case 'z': ftst = OP_FTZERO; break;
4951 case 's': ftst = OP_FTSIZE; break;
4952 case 'f': ftst = OP_FTFILE; break;
4953 case 'd': ftst = OP_FTDIR; break;
4954 case 'l': ftst = OP_FTLINK; break;
4955 case 'p': ftst = OP_FTPIPE; break;
4956 case 'S': ftst = OP_FTSOCK; break;
4957 case 'u': ftst = OP_FTSUID; break;
4958 case 'g': ftst = OP_FTSGID; break;
4959 case 'k': ftst = OP_FTSVTX; break;
4960 case 'b': ftst = OP_FTBLK; break;
4961 case 'c': ftst = OP_FTCHR; break;
4962 case 't': ftst = OP_FTTTY; break;
4963 case 'T': ftst = OP_FTTEXT; break;
4964 case 'B': ftst = OP_FTBINARY; break;
4965 case 'M': case 'A': case 'C':
4966 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
4968 case 'M': ftst = OP_FTMTIME; break;
4969 case 'A': ftst = OP_FTATIME; break;
4970 case 'C': ftst = OP_FTCTIME; break;
4978 PL_last_lop_op = (OPCODE)ftst;
4979 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4980 "### Saw file test %c\n", (int)tmp);
4985 /* Assume it was a minus followed by a one-letter named
4986 * subroutine call (or a -bareword), then. */
4987 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4988 "### '-%c' looked like a file test but was not\n",
4995 const char tmp = *s++;
4998 if (PL_expect == XOPERATOR)
5003 else if (*s == '>') {
5006 if (isIDFIRST_lazy_if(s,UTF)) {
5007 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5015 if (PL_expect == XOPERATOR)
5018 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5020 OPERATOR('-'); /* unary minus */
5026 const char tmp = *s++;
5029 if (PL_expect == XOPERATOR)
5034 if (PL_expect == XOPERATOR)
5037 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5044 if (PL_expect != XOPERATOR) {
5045 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5046 PL_expect = XOPERATOR;
5047 force_ident(PL_tokenbuf, '*');
5060 if (PL_expect == XOPERATOR) {
5064 PL_tokenbuf[0] = '%';
5065 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5066 sizeof PL_tokenbuf - 1, FALSE);
5067 if (!PL_tokenbuf[1]) {
5070 PL_pending_ident = '%';
5079 const char tmp = *s++;
5084 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5091 const char tmp = *s++;
5097 goto just_a_word_zero_gv;
5100 switch (PL_expect) {
5106 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5108 PL_bufptr = s; /* update in case we back off */
5110 deprecate(":= for an empty attribute list");
5117 PL_expect = XTERMBLOCK;
5120 stuffstart = s - SvPVX(PL_linestr) - 1;
5124 while (isIDFIRST_lazy_if(s,UTF)) {
5127 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5128 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5129 if (tmp < 0) tmp = -tmp;
5144 sv = newSVpvn(s, len);
5146 d = scan_str(d,TRUE,TRUE);
5148 /* MUST advance bufptr here to avoid bogus
5149 "at end of line" context messages from yyerror().
5151 PL_bufptr = s + len;
5152 yyerror("Unterminated attribute parameter in attribute list");
5156 return REPORT(0); /* EOF indicator */
5160 sv_catsv(sv, PL_lex_stuff);
5161 attrs = append_elem(OP_LIST, attrs,
5162 newSVOP(OP_CONST, 0, sv));
5163 SvREFCNT_dec(PL_lex_stuff);
5164 PL_lex_stuff = NULL;
5167 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5169 if (PL_in_my == KEY_our) {
5170 deprecate(":unique");
5173 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5176 /* NOTE: any CV attrs applied here need to be part of
5177 the CVf_BUILTIN_ATTRS define in cv.h! */
5178 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5180 CvLVALUE_on(PL_compcv);
5182 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5184 deprecate(":locked");
5186 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5188 CvMETHOD_on(PL_compcv);
5190 /* After we've set the flags, it could be argued that
5191 we don't need to do the attributes.pm-based setting
5192 process, and shouldn't bother appending recognized
5193 flags. To experiment with that, uncomment the
5194 following "else". (Note that's already been
5195 uncommented. That keeps the above-applied built-in
5196 attributes from being intercepted (and possibly
5197 rejected) by a package's attribute routines, but is
5198 justified by the performance win for the common case
5199 of applying only built-in attributes.) */
5201 attrs = append_elem(OP_LIST, attrs,
5202 newSVOP(OP_CONST, 0,
5206 if (*s == ':' && s[1] != ':')
5209 break; /* require real whitespace or :'s */
5210 /* XXX losing whitespace on sequential attributes here */
5214 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5215 if (*s != ';' && *s != '}' && *s != tmp
5216 && (tmp != '=' || *s != ')')) {
5217 const char q = ((*s == '\'') ? '"' : '\'');
5218 /* If here for an expression, and parsed no attrs, back
5220 if (tmp == '=' && !attrs) {
5224 /* MUST advance bufptr here to avoid bogus "at end of line"
5225 context messages from yyerror().
5228 yyerror( (const char *)
5230 ? Perl_form(aTHX_ "Invalid separator character "
5231 "%c%c%c in attribute list", q, *s, q)
5232 : "Unterminated attribute list" ) );
5240 start_force(PL_curforce);
5241 NEXTVAL_NEXTTOKE.opval = attrs;
5242 CURMAD('_', PL_nextwhite);
5247 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5248 (s - SvPVX(PL_linestr)) - stuffstart);
5256 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5257 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5265 const char tmp = *s++;
5270 const char tmp = *s++;
5278 if (PL_lex_brackets <= 0)
5279 yyerror("Unmatched right square bracket");
5282 if (PL_lex_state == LEX_INTERPNORMAL) {
5283 if (PL_lex_brackets == 0) {
5284 if (*s == '-' && s[1] == '>')
5285 PL_lex_state = LEX_INTERPENDMAYBE;
5286 else if (*s != '[' && *s != '{')
5287 PL_lex_state = LEX_INTERPEND;
5294 if (PL_lex_brackets > 100) {
5295 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5297 switch (PL_expect) {
5299 if (PL_lex_formbrack) {
5303 if (PL_oldoldbufptr == PL_last_lop)
5304 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5306 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5307 OPERATOR(HASHBRACK);
5309 while (s < PL_bufend && SPACE_OR_TAB(*s))
5312 PL_tokenbuf[0] = '\0';
5313 if (d < PL_bufend && *d == '-') {
5314 PL_tokenbuf[0] = '-';
5316 while (d < PL_bufend && SPACE_OR_TAB(*d))
5319 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5320 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5322 while (d < PL_bufend && SPACE_OR_TAB(*d))
5325 const char minus = (PL_tokenbuf[0] == '-');
5326 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5334 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5339 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5344 if (PL_oldoldbufptr == PL_last_lop)
5345 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5347 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5350 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5352 /* This hack is to get the ${} in the message. */
5354 yyerror("syntax error");
5357 OPERATOR(HASHBRACK);
5359 /* This hack serves to disambiguate a pair of curlies
5360 * as being a block or an anon hash. Normally, expectation
5361 * determines that, but in cases where we're not in a
5362 * position to expect anything in particular (like inside
5363 * eval"") we have to resolve the ambiguity. This code
5364 * covers the case where the first term in the curlies is a
5365 * quoted string. Most other cases need to be explicitly
5366 * disambiguated by prepending a "+" before the opening
5367 * curly in order to force resolution as an anon hash.
5369 * XXX should probably propagate the outer expectation
5370 * into eval"" to rely less on this hack, but that could
5371 * potentially break current behavior of eval"".
5375 if (*s == '\'' || *s == '"' || *s == '`') {
5376 /* common case: get past first string, handling escapes */
5377 for (t++; t < PL_bufend && *t != *s;)
5378 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5382 else if (*s == 'q') {
5385 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5388 /* skip q//-like construct */
5390 char open, close, term;
5393 while (t < PL_bufend && isSPACE(*t))
5395 /* check for q => */
5396 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5397 OPERATOR(HASHBRACK);
5401 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5405 for (t++; t < PL_bufend; t++) {
5406 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5408 else if (*t == open)
5412 for (t++; t < PL_bufend; t++) {
5413 if (*t == '\\' && t+1 < PL_bufend)
5415 else if (*t == close && --brackets <= 0)
5417 else if (*t == open)
5424 /* skip plain q word */
5425 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5428 else if (isALNUM_lazy_if(t,UTF)) {
5430 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5433 while (t < PL_bufend && isSPACE(*t))
5435 /* if comma follows first term, call it an anon hash */
5436 /* XXX it could be a comma expression with loop modifiers */
5437 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5438 || (*t == '=' && t[1] == '>')))
5439 OPERATOR(HASHBRACK);
5440 if (PL_expect == XREF)
5443 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5449 pl_yylval.ival = CopLINE(PL_curcop);
5450 if (isSPACE(*s) || *s == '#')
5451 PL_copline = NOLINE; /* invalidate current command line number */
5456 if (PL_lex_brackets <= 0)
5457 yyerror("Unmatched right curly bracket");
5459 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5460 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5461 PL_lex_formbrack = 0;
5462 if (PL_lex_state == LEX_INTERPNORMAL) {
5463 if (PL_lex_brackets == 0) {
5464 if (PL_expect & XFAKEBRACK) {
5465 PL_expect &= XENUMMASK;
5466 PL_lex_state = LEX_INTERPEND;
5471 PL_thiswhite = newSVpvs("");
5472 sv_catpvs(PL_thiswhite,"}");
5475 return yylex(); /* ignore fake brackets */
5477 if (*s == '-' && s[1] == '>')
5478 PL_lex_state = LEX_INTERPENDMAYBE;
5479 else if (*s != '[' && *s != '{')
5480 PL_lex_state = LEX_INTERPEND;
5483 if (PL_expect & XFAKEBRACK) {
5484 PL_expect &= XENUMMASK;
5486 return yylex(); /* ignore fake brackets */
5488 start_force(PL_curforce);
5490 curmad('X', newSVpvn(s-1,1));
5491 CURMAD('_', PL_thiswhite);
5496 PL_thistoken = newSVpvs("");
5504 if (PL_expect == XOPERATOR) {
5505 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5506 && isIDFIRST_lazy_if(s,UTF))
5508 CopLINE_dec(PL_curcop);
5509 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5510 CopLINE_inc(PL_curcop);
5515 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5517 PL_expect = XOPERATOR;
5518 force_ident(PL_tokenbuf, '&');
5522 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5534 const char tmp = *s++;
5541 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5542 && strchr("+-*/%.^&|<",tmp))
5543 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5544 "Reversed %c= operator",(int)tmp);
5546 if (PL_expect == XSTATE && isALPHA(tmp) &&
5547 (s == PL_linestart+1 || s[-2] == '\n') )
5549 if (PL_in_eval && !PL_rsfp) {
5554 if (strnEQ(s,"=cut",4)) {
5570 PL_thiswhite = newSVpvs("");
5571 sv_catpvn(PL_thiswhite, PL_linestart,
5572 PL_bufend - PL_linestart);
5576 PL_doextract = TRUE;
5580 if (PL_lex_brackets < PL_lex_formbrack) {
5582 #ifdef PERL_STRICT_CR
5583 while (SPACE_OR_TAB(*t))
5585 while (SPACE_OR_TAB(*t) || *t == '\r')
5588 if (*t == '\n' || *t == '#') {
5599 const char tmp = *s++;
5601 /* was this !=~ where !~ was meant?
5602 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5604 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5605 const char *t = s+1;
5607 while (t < PL_bufend && isSPACE(*t))
5610 if (*t == '/' || *t == '?' ||
5611 ((*t == 'm' || *t == 's' || *t == 'y')
5612 && !isALNUM(t[1])) ||
5613 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5614 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5615 "!=~ should be !~");
5625 if (PL_expect != XOPERATOR) {
5626 if (s[1] != '<' && !strchr(s,'>'))
5629 s = scan_heredoc(s);
5631 s = scan_inputsymbol(s);
5632 TERM(sublex_start());
5638 SHop(OP_LEFT_SHIFT);
5652 const char tmp = *s++;
5654 SHop(OP_RIGHT_SHIFT);
5655 else if (tmp == '=')
5664 if (PL_expect == XOPERATOR) {
5665 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5666 return deprecate_commaless_var_list();
5670 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
5671 PL_tokenbuf[0] = '@';
5672 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
5673 sizeof PL_tokenbuf - 1, FALSE);
5674 if (PL_expect == XOPERATOR)
5675 no_op("Array length", s);
5676 if (!PL_tokenbuf[1])
5678 PL_expect = XOPERATOR;
5679 PL_pending_ident = '#';
5683 PL_tokenbuf[0] = '$';
5684 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5685 sizeof PL_tokenbuf - 1, FALSE);
5686 if (PL_expect == XOPERATOR)
5688 if (!PL_tokenbuf[1]) {
5690 yyerror("Final $ should be \\$ or $name");
5694 /* This kludge not intended to be bulletproof. */
5695 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
5696 pl_yylval.opval = newSVOP(OP_CONST, 0,
5697 newSViv(CopARYBASE_get(&PL_compiling)));
5698 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
5704 const char tmp = *s;
5705 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
5708 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
5709 && intuit_more(s)) {
5711 PL_tokenbuf[0] = '@';
5712 if (ckWARN(WARN_SYNTAX)) {
5715 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
5718 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5719 while (t < PL_bufend && *t != ']')
5721 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5722 "Multidimensional syntax %.*s not supported",
5723 (int)((t - PL_bufptr) + 1), PL_bufptr);
5727 else if (*s == '{') {
5729 PL_tokenbuf[0] = '%';
5730 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
5731 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
5733 char tmpbuf[sizeof PL_tokenbuf];
5736 } while (isSPACE(*t));
5737 if (isIDFIRST_lazy_if(t,UTF)) {
5739 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
5743 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
5744 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5745 "You need to quote \"%s\"",
5752 PL_expect = XOPERATOR;
5753 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
5754 const bool islop = (PL_last_lop == PL_oldoldbufptr);
5755 if (!islop || PL_last_lop_op == OP_GREPSTART)
5756 PL_expect = XOPERATOR;
5757 else if (strchr("$@\"'`q", *s))
5758 PL_expect = XTERM; /* e.g. print $fh "foo" */
5759 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
5760 PL_expect = XTERM; /* e.g. print $fh &sub */
5761 else if (isIDFIRST_lazy_if(s,UTF)) {
5762 char tmpbuf[sizeof PL_tokenbuf];
5764 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5765 if ((t2 = keyword(tmpbuf, len, 0))) {
5766 /* binary operators exclude handle interpretations */
5778 PL_expect = XTERM; /* e.g. print $fh length() */
5783 PL_expect = XTERM; /* e.g. print $fh subr() */
5786 else if (isDIGIT(*s))
5787 PL_expect = XTERM; /* e.g. print $fh 3 */
5788 else if (*s == '.' && isDIGIT(s[1]))
5789 PL_expect = XTERM; /* e.g. print $fh .3 */
5790 else if ((*s == '?' || *s == '-' || *s == '+')
5791 && !isSPACE(s[1]) && s[1] != '=')
5792 PL_expect = XTERM; /* e.g. print $fh -1 */
5793 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5795 PL_expect = XTERM; /* e.g. print $fh /.../
5796 XXX except DORDOR operator
5798 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5800 PL_expect = XTERM; /* print $fh <<"EOF" */
5803 PL_pending_ident = '$';
5807 if (PL_expect == XOPERATOR)
5809 PL_tokenbuf[0] = '@';
5810 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5811 if (!PL_tokenbuf[1]) {
5814 if (PL_lex_state == LEX_NORMAL)
5816 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5818 PL_tokenbuf[0] = '%';
5820 /* Warn about @ where they meant $. */
5821 if (*s == '[' || *s == '{') {
5822 if (ckWARN(WARN_SYNTAX)) {
5823 const char *t = s + 1;
5824 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5826 if (*t == '}' || *t == ']') {
5828 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5829 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5830 "Scalar value %.*s better written as $%.*s",
5831 (int)(t-PL_bufptr), PL_bufptr,
5832 (int)(t-PL_bufptr-1), PL_bufptr+1);
5837 PL_pending_ident = '@';
5840 case '/': /* may be division, defined-or, or pattern */
5841 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5845 case '?': /* may either be conditional or pattern */
5846 if (PL_expect == XOPERATOR) {
5854 /* A // operator. */
5864 /* Disable warning on "study /blah/" */
5865 if (PL_oldoldbufptr == PL_last_uni
5866 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5867 || memNE(PL_last_uni, "study", 5)
5868 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5871 s = scan_pat(s,OP_MATCH);
5872 TERM(sublex_start());
5876 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5877 #ifdef PERL_STRICT_CR
5880 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5882 && (s == PL_linestart || s[-1] == '\n') )
5884 PL_lex_formbrack = 0;
5888 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5892 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5898 pl_yylval.ival = OPf_SPECIAL;
5907 case '0': case '1': case '2': case '3': case '4':
5908 case '5': case '6': case '7': case '8': case '9':
5909 s = scan_num(s, &pl_yylval);
5910 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5911 if (PL_expect == XOPERATOR)
5916 s = scan_str(s,!!PL_madskills,FALSE);
5917 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5918 if (PL_expect == XOPERATOR) {
5919 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5920 return deprecate_commaless_var_list();
5927 pl_yylval.ival = OP_CONST;
5928 TERM(sublex_start());
5931 s = scan_str(s,!!PL_madskills,FALSE);
5932 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5933 if (PL_expect == XOPERATOR) {
5934 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5935 return deprecate_commaless_var_list();
5942 pl_yylval.ival = OP_CONST;
5943 /* FIXME. I think that this can be const if char *d is replaced by
5944 more localised variables. */
5945 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5946 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5947 pl_yylval.ival = OP_STRINGIFY;
5951 TERM(sublex_start());
5954 s = scan_str(s,!!PL_madskills,FALSE);
5955 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
5956 if (PL_expect == XOPERATOR)
5957 no_op("Backticks",s);
5960 readpipe_override();
5961 TERM(sublex_start());
5965 if (PL_lex_inwhat && isDIGIT(*s))
5966 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
5968 if (PL_expect == XOPERATOR)
5969 no_op("Backslash",s);
5973 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
5974 char *start = s + 2;
5975 while (isDIGIT(*start) || *start == '_')
5977 if (*start == '.' && isDIGIT(start[1])) {
5978 s = scan_num(s, &pl_yylval);
5981 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
5982 else if (!isALPHA(*start) && (PL_expect == XTERM
5983 || PL_expect == XREF || PL_expect == XSTATE
5984 || PL_expect == XTERMORDORDOR)) {
5985 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
5987 s = scan_num(s, &pl_yylval);
5994 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6037 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6039 /* Some keywords can be followed by any delimiter, including ':' */
6040 anydelim = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
6041 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
6042 (PL_tokenbuf[0] == 'q' &&
6043 strchr("qwxr", PL_tokenbuf[1])))));
6045 /* x::* is just a word, unless x is "CORE" */
6046 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6050 while (d < PL_bufend && isSPACE(*d))
6051 d++; /* no comments skipped here, or s### is misparsed */
6053 /* Is this a word before a => operator? */
6054 if (*d == '=' && d[1] == '>') {
6057 = (OP*)newSVOP(OP_CONST, 0,
6058 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6059 pl_yylval.opval->op_private = OPpCONST_BARE;
6063 /* Check for plugged-in keyword */
6067 char *saved_bufptr = PL_bufptr;
6069 result = CALL_FPTR(PL_keyword_plugin)(aTHX_ PL_tokenbuf, len, &o);
6071 if (result == KEYWORD_PLUGIN_DECLINE) {
6072 /* not a plugged-in keyword */
6073 PL_bufptr = saved_bufptr;
6074 } else if (result == KEYWORD_PLUGIN_STMT) {
6075 pl_yylval.opval = o;
6078 return REPORT(PLUGSTMT);
6079 } else if (result == KEYWORD_PLUGIN_EXPR) {
6080 pl_yylval.opval = o;
6082 PL_expect = XOPERATOR;
6083 return REPORT(PLUGEXPR);
6085 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6090 /* Check for built-in keyword */
6091 tmp = keyword(PL_tokenbuf, len, 0);
6093 /* Is this a label? */
6094 if (!anydelim && PL_expect == XSTATE
6095 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6097 Perl_croak(aTHX_ "Can't use keyword '%s' as a label", PL_tokenbuf);
6099 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
6104 if (tmp < 0) { /* second-class keyword? */
6105 GV *ogv = NULL; /* override (winner) */
6106 GV *hgv = NULL; /* hidden (loser) */
6107 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6109 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
6112 if (GvIMPORTED_CV(gv))
6114 else if (! CvMETHOD(cv))
6118 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
6119 (gv = *gvp) && isGV_with_GP(gv) &&
6120 GvCVu(gv) && GvIMPORTED_CV(gv))
6127 tmp = 0; /* overridden by import or by GLOBAL */
6130 && -tmp==KEY_lock /* XXX generalizable kludge */
6133 tmp = 0; /* any sub overrides "weak" keyword */
6135 else { /* no override */
6137 if (tmp == KEY_dump) {
6138 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6139 "dump() better written as CORE::dump()");
6143 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6144 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6145 "Ambiguous call resolved as CORE::%s(), %s",
6146 GvENAME(hgv), "qualify as such or use &");
6153 default: /* not a keyword */
6154 /* Trade off - by using this evil construction we can pull the
6155 variable gv into the block labelled keylookup. If not, then
6156 we have to give it function scope so that the goto from the
6157 earlier ':' case doesn't bypass the initialisation. */
6159 just_a_word_zero_gv:
6167 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6171 SV *nextPL_nextwhite = 0;
6175 /* Get the rest if it looks like a package qualifier */
6177 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6179 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6182 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
6183 *s == '\'' ? "'" : "::");
6188 if (PL_expect == XOPERATOR) {
6189 if (PL_bufptr == PL_linestart) {
6190 CopLINE_dec(PL_curcop);
6191 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6192 CopLINE_inc(PL_curcop);
6195 no_op("Bareword",s);
6198 /* Look for a subroutine with this name in current package,
6199 unless name is "Foo::", in which case Foo is a bearword
6200 (and a package name). */
6202 if (len > 2 && !PL_madskills &&
6203 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6205 if (ckWARN(WARN_BAREWORD)
6206 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
6207 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6208 "Bareword \"%s\" refers to nonexistent package",
6211 PL_tokenbuf[len] = '\0';
6217 /* Mustn't actually add anything to a symbol table.
6218 But also don't want to "initialise" any placeholder
6219 constants that might already be there into full
6220 blown PVGVs with attached PVCV. */
6221 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6222 GV_NOADD_NOINIT, SVt_PVCV);
6227 /* if we saw a global override before, get the right name */
6230 sv = newSVpvs("CORE::GLOBAL::");
6231 sv_catpv(sv,PL_tokenbuf);
6234 /* If len is 0, newSVpv does strlen(), which is correct.
6235 If len is non-zero, then it will be the true length,
6236 and so the scalar will be created correctly. */
6237 sv = newSVpv(PL_tokenbuf,len);
6240 if (PL_madskills && !PL_thistoken) {
6241 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6242 PL_thistoken = newSVpvn(start,s - start);
6243 PL_realtokenstart = s - SvPVX(PL_linestr);
6247 /* Presume this is going to be a bareword of some sort. */
6250 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6251 pl_yylval.opval->op_private = OPpCONST_BARE;
6252 /* UTF-8 package name? */
6253 if (UTF && !IN_BYTES &&
6254 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
6257 /* And if "Foo::", then that's what it certainly is. */
6264 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc(sv));
6265 const_op->op_private = OPpCONST_BARE;
6266 rv2cv_op = newCVREF(0, const_op);
6268 if (rv2cv_op->op_type == OP_RV2CV &&
6269 (rv2cv_op->op_flags & OPf_KIDS)) {
6270 OP *rv_op = cUNOPx(rv2cv_op)->op_first;
6271 switch (rv_op->op_type) {
6273 SV *sv = cSVOPx_sv(rv_op);
6274 if (SvROK(sv) && SvTYPE(SvRV(sv)) == SVt_PVCV)
6278 GV *gv = cGVOPx_gv(rv_op);
6279 CV *maybe_cv = GvCVu(gv);
6280 if (maybe_cv && SvTYPE((SV*)maybe_cv) == SVt_PVCV)
6286 /* See if it's the indirect object for a list operator. */
6288 if (PL_oldoldbufptr &&
6289 PL_oldoldbufptr < PL_bufptr &&
6290 (PL_oldoldbufptr == PL_last_lop
6291 || PL_oldoldbufptr == PL_last_uni) &&
6292 /* NO SKIPSPACE BEFORE HERE! */
6293 (PL_expect == XREF ||
6294 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6296 bool immediate_paren = *s == '(';
6298 /* (Now we can afford to cross potential line boundary.) */
6299 s = SKIPSPACE2(s,nextPL_nextwhite);
6301 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6304 /* Two barewords in a row may indicate method call. */
6306 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6307 (tmp = intuit_method(s, gv, cv))) {
6312 /* If not a declared subroutine, it's an indirect object. */
6313 /* (But it's an indir obj regardless for sort.) */
6314 /* Also, if "_" follows a filetest operator, it's a bareword */
6317 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6319 (PL_last_lop_op != OP_MAPSTART &&
6320 PL_last_lop_op != OP_GREPSTART))))
6321 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6322 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6325 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6330 PL_expect = XOPERATOR;
6333 s = SKIPSPACE2(s,nextPL_nextwhite);
6334 PL_nextwhite = nextPL_nextwhite;
6339 /* Is this a word before a => operator? */
6340 if (*s == '=' && s[1] == '>' && !pkgname) {
6343 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6344 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6345 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6349 /* If followed by a paren, it's certainly a subroutine. */
6354 while (SPACE_OR_TAB(*d))
6356 if (*d == ')' && (sv = cv_const_sv(cv))) {
6363 PL_nextwhite = PL_thiswhite;
6366 start_force(PL_curforce);
6368 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6369 PL_expect = XOPERATOR;
6372 PL_nextwhite = nextPL_nextwhite;
6373 curmad('X', PL_thistoken);
6374 PL_thistoken = newSVpvs("");
6383 /* If followed by var or block, call it a method (unless sub) */
6385 if ((*s == '$' || *s == '{') && !cv) {
6387 PL_last_lop = PL_oldbufptr;
6388 PL_last_lop_op = OP_METHOD;
6392 /* If followed by a bareword, see if it looks like indir obj. */
6395 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6396 && (tmp = intuit_method(s, gv, cv))) {
6401 /* Not a method, so call it a subroutine (if defined) */
6404 if (lastchar == '-')
6405 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6406 "Ambiguous use of -%s resolved as -&%s()",
6407 PL_tokenbuf, PL_tokenbuf);
6408 /* Check for a constant sub */
6409 if ((sv = cv_const_sv(cv))) {
6412 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6413 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6414 pl_yylval.opval->op_private = 0;
6418 op_free(pl_yylval.opval);
6419 pl_yylval.opval = rv2cv_op;
6420 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6421 PL_last_lop = PL_oldbufptr;
6422 PL_last_lop_op = OP_ENTERSUB;
6423 /* Is there a prototype? */
6431 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
6434 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
6436 while (*proto == ';')
6438 if (*proto == '&' && *s == '{') {
6440 sv_setpvs(PL_subname, "__ANON__");
6442 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6449 PL_nextwhite = PL_thiswhite;
6452 start_force(PL_curforce);
6453 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6456 PL_nextwhite = nextPL_nextwhite;
6457 curmad('X', PL_thistoken);
6458 PL_thistoken = newSVpvs("");
6465 /* Guess harder when madskills require "best effort". */
6466 if (PL_madskills && (!gv || !GvCVu(gv))) {
6467 int probable_sub = 0;
6468 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6470 else if (isALPHA(*s)) {
6474 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6475 if (!keyword(tmpbuf, tmplen, 0))
6478 while (d < PL_bufend && isSPACE(*d))
6480 if (*d == '=' && d[1] == '>')
6485 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
6486 op_free(pl_yylval.opval);
6487 pl_yylval.opval = rv2cv_op;
6488 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6489 PL_last_lop = PL_oldbufptr;
6490 PL_last_lop_op = OP_ENTERSUB;
6491 PL_nextwhite = PL_thiswhite;
6493 start_force(PL_curforce);
6494 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6496 PL_nextwhite = nextPL_nextwhite;
6497 curmad('X', PL_thistoken);
6498 PL_thistoken = newSVpvs("");
6503 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6510 /* Call it a bare word */
6512 if (PL_hints & HINT_STRICT_SUBS)
6513 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6516 /* after "print" and similar functions (corresponding to
6517 * "F? L" in opcode.pl), whatever wasn't already parsed as
6518 * a filehandle should be subject to "strict subs".
6519 * Likewise for the optional indirect-object argument to system
6520 * or exec, which can't be a bareword */
6521 if ((PL_last_lop_op == OP_PRINT
6522 || PL_last_lop_op == OP_PRTF
6523 || PL_last_lop_op == OP_SAY
6524 || PL_last_lop_op == OP_SYSTEM
6525 || PL_last_lop_op == OP_EXEC)
6526 && (PL_hints & HINT_STRICT_SUBS))
6527 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6528 if (lastchar != '-') {
6529 if (ckWARN(WARN_RESERVED)) {
6533 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
6534 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6542 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
6543 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6544 "Operator or semicolon missing before %c%s",
6545 lastchar, PL_tokenbuf);
6546 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6547 "Ambiguous use of %c resolved as operator %c",
6548 lastchar, lastchar);
6554 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6555 newSVpv(CopFILE(PL_curcop),0));
6559 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6560 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
6563 case KEY___PACKAGE__:
6564 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6566 ? newSVhek(HvNAME_HEK(PL_curstash))
6573 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6574 const char *pname = "main";
6575 if (PL_tokenbuf[2] == 'D')
6576 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
6577 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
6581 GvIOp(gv) = newIO();
6582 IoIFP(GvIOp(gv)) = PL_rsfp;
6583 #if defined(HAS_FCNTL) && defined(F_SETFD)
6585 const int fd = PerlIO_fileno(PL_rsfp);
6586 fcntl(fd,F_SETFD,fd >= 3);
6589 /* Mark this internal pseudo-handle as clean */
6590 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6591 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6592 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6594 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6595 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6596 /* if the script was opened in binmode, we need to revert
6597 * it to text mode for compatibility; but only iff it has CRs
6598 * XXX this is a questionable hack at best. */
6599 if (PL_bufend-PL_bufptr > 2
6600 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6603 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6604 loc = PerlIO_tell(PL_rsfp);
6605 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6608 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6610 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6611 #endif /* NETWARE */
6612 #ifdef PERLIO_IS_STDIO /* really? */
6613 # if defined(__BORLANDC__)
6614 /* XXX see note in do_binmode() */
6615 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
6619 PerlIO_seek(PL_rsfp, loc, 0);
6623 #ifdef PERLIO_LAYERS
6626 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6627 else if (PL_encoding) {
6634 XPUSHs(PL_encoding);
6636 call_method("name", G_SCALAR);
6640 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
6641 Perl_form(aTHX_ ":encoding(%"SVf")",
6650 if (PL_realtokenstart >= 0) {
6651 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6653 PL_endwhite = newSVpvs("");
6654 sv_catsv(PL_endwhite, PL_thiswhite);
6656 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
6657 PL_realtokenstart = -1;
6659 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
6675 if (PL_expect == XSTATE) {
6682 if (*s == ':' && s[1] == ':') {
6685 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6686 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
6687 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
6690 else if (tmp == KEY_require || tmp == KEY_do)
6691 /* that's a way to remember we saw "CORE::" */
6704 LOP(OP_ACCEPT,XTERM);
6710 LOP(OP_ATAN2,XTERM);
6716 LOP(OP_BINMODE,XTERM);
6719 LOP(OP_BLESS,XTERM);
6728 /* When 'use switch' is in effect, continue has a dual
6729 life as a control operator. */
6731 if (!FEATURE_IS_ENABLED("switch"))
6734 /* We have to disambiguate the two senses of
6735 "continue". If the next token is a '{' then
6736 treat it as the start of a continue block;
6737 otherwise treat it as a control operator.
6749 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
6766 if (!PL_cryptseen) {
6767 PL_cryptseen = TRUE;
6771 LOP(OP_CRYPT,XTERM);
6774 LOP(OP_CHMOD,XTERM);
6777 LOP(OP_CHOWN,XTERM);
6780 LOP(OP_CONNECT,XTERM);
6799 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6800 if (orig_keyword == KEY_do) {
6809 PL_hints |= HINT_BLOCK_SCOPE;
6819 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
6820 LOP(OP_DBMOPEN,XTERM);
6826 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6833 pl_yylval.ival = CopLINE(PL_curcop);
6849 if (*s == '{') { /* block eval */
6850 PL_expect = XTERMBLOCK;
6851 UNIBRACK(OP_ENTERTRY);
6853 else { /* string eval */
6855 UNIBRACK(OP_ENTEREVAL);
6870 case KEY_endhostent:
6876 case KEY_endservent:
6879 case KEY_endprotoent:
6890 pl_yylval.ival = CopLINE(PL_curcop);
6892 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6895 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6898 if ((PL_bufend - p) >= 3 &&
6899 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6901 else if ((PL_bufend - p) >= 4 &&
6902 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6905 if (isIDFIRST_lazy_if(p,UTF)) {
6906 p = scan_ident(p, PL_bufend,
6907 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6911 Perl_croak(aTHX_ "Missing $ on loop variable");
6913 s = SvPVX(PL_linestr) + soff;
6919 LOP(OP_FORMLINE,XTERM);
6925 LOP(OP_FCNTL,XTERM);
6931 LOP(OP_FLOCK,XTERM);
6940 LOP(OP_GREPSTART, XREF);
6943 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6958 case KEY_getpriority:
6959 LOP(OP_GETPRIORITY,XTERM);
6961 case KEY_getprotobyname:
6964 case KEY_getprotobynumber:
6965 LOP(OP_GPBYNUMBER,XTERM);
6967 case KEY_getprotoent:
6979 case KEY_getpeername:
6980 UNI(OP_GETPEERNAME);
6982 case KEY_gethostbyname:
6985 case KEY_gethostbyaddr:
6986 LOP(OP_GHBYADDR,XTERM);
6988 case KEY_gethostent:
6991 case KEY_getnetbyname:
6994 case KEY_getnetbyaddr:
6995 LOP(OP_GNBYADDR,XTERM);
7000 case KEY_getservbyname:
7001 LOP(OP_GSBYNAME,XTERM);
7003 case KEY_getservbyport:
7004 LOP(OP_GSBYPORT,XTERM);
7006 case KEY_getservent:
7009 case KEY_getsockname:
7010 UNI(OP_GETSOCKNAME);
7012 case KEY_getsockopt:
7013 LOP(OP_GSOCKOPT,XTERM);
7028 pl_yylval.ival = CopLINE(PL_curcop);
7038 pl_yylval.ival = CopLINE(PL_curcop);
7042 LOP(OP_INDEX,XTERM);
7048 LOP(OP_IOCTL,XTERM);
7060 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7092 LOP(OP_LISTEN,XTERM);
7101 s = scan_pat(s,OP_MATCH);
7102 TERM(sublex_start());
7105 LOP(OP_MAPSTART, XREF);
7108 LOP(OP_MKDIR,XTERM);
7111 LOP(OP_MSGCTL,XTERM);
7114 LOP(OP_MSGGET,XTERM);
7117 LOP(OP_MSGRCV,XTERM);
7120 LOP(OP_MSGSND,XTERM);
7125 PL_in_my = (U16)tmp;
7127 if (isIDFIRST_lazy_if(s,UTF)) {
7131 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7132 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7134 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7135 if (!PL_in_my_stash) {
7138 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7142 if (PL_madskills) { /* just add type to declarator token */
7143 sv_catsv(PL_thistoken, PL_nextwhite);
7145 sv_catpvn(PL_thistoken, start, s - start);
7153 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7160 s = tokenize_use(0, s);
7164 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7171 if (isIDFIRST_lazy_if(s,UTF)) {
7173 for (d = s; isALNUM_lazy_if(d,UTF);)
7175 for (t=d; isSPACE(*t);)
7177 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7179 && !(t[0] == '=' && t[1] == '>')
7181 int parms_len = (int)(d-s);
7182 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7183 "Precedence problem: open %.*s should be open(%.*s)",
7184 parms_len, s, parms_len, s);
7190 pl_yylval.ival = OP_OR;
7200 LOP(OP_OPEN_DIR,XTERM);
7203 checkcomma(s,PL_tokenbuf,"filehandle");
7207 checkcomma(s,PL_tokenbuf,"filehandle");
7226 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7228 s = force_strict_version(s);
7232 LOP(OP_PIPE_OP,XTERM);
7235 s = scan_str(s,!!PL_madskills,FALSE);
7238 pl_yylval.ival = OP_CONST;
7239 TERM(sublex_start());
7245 s = scan_str(s,!!PL_madskills,FALSE);
7248 PL_expect = XOPERATOR;
7250 if (SvCUR(PL_lex_stuff)) {
7253 d = SvPV_force(PL_lex_stuff, len);
7255 for (; isSPACE(*d) && len; --len, ++d)
7260 if (!warned && ckWARN(WARN_QW)) {
7261 for (; !isSPACE(*d) && len; --len, ++d) {
7263 Perl_warner(aTHX_ packWARN(WARN_QW),
7264 "Possible attempt to separate words with commas");
7267 else if (*d == '#') {
7268 Perl_warner(aTHX_ packWARN(WARN_QW),
7269 "Possible attempt to put comments in qw() list");
7275 for (; !isSPACE(*d) && len; --len, ++d)
7278 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7279 words = append_elem(OP_LIST, words,
7280 newSVOP(OP_CONST, 0, tokeq(sv)));
7284 start_force(PL_curforce);
7285 NEXTVAL_NEXTTOKE.opval = words;
7290 SvREFCNT_dec(PL_lex_stuff);
7291 PL_lex_stuff = NULL;
7297 s = scan_str(s,!!PL_madskills,FALSE);
7300 pl_yylval.ival = OP_STRINGIFY;
7301 if (SvIVX(PL_lex_stuff) == '\'')
7302 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
7303 TERM(sublex_start());
7306 s = scan_pat(s,OP_QR);
7307 TERM(sublex_start());
7310 s = scan_str(s,!!PL_madskills,FALSE);
7313 readpipe_override();
7314 TERM(sublex_start());
7322 s = force_version(s, FALSE);
7324 else if (*s != 'v' || !isDIGIT(s[1])
7325 || (s = force_version(s, TRUE), *s == 'v'))
7327 *PL_tokenbuf = '\0';
7328 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7329 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7330 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7332 yyerror("<> should be quotes");
7334 if (orig_keyword == KEY_require) {
7342 PL_last_uni = PL_oldbufptr;
7343 PL_last_lop_op = OP_REQUIRE;
7345 return REPORT( (int)REQUIRE );
7351 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7355 LOP(OP_RENAME,XTERM);
7364 LOP(OP_RINDEX,XTERM);
7373 UNIDOR(OP_READLINE);
7376 UNIDOR(OP_BACKTICK);
7385 LOP(OP_REVERSE,XTERM);
7388 UNIDOR(OP_READLINK);
7395 if (pl_yylval.opval)
7396 TERM(sublex_start());
7398 TOKEN(1); /* force error */
7401 checkcomma(s,PL_tokenbuf,"filehandle");
7411 LOP(OP_SELECT,XTERM);
7417 LOP(OP_SEMCTL,XTERM);
7420 LOP(OP_SEMGET,XTERM);
7423 LOP(OP_SEMOP,XTERM);
7429 LOP(OP_SETPGRP,XTERM);
7431 case KEY_setpriority:
7432 LOP(OP_SETPRIORITY,XTERM);
7434 case KEY_sethostent:
7440 case KEY_setservent:
7443 case KEY_setprotoent:
7453 LOP(OP_SEEKDIR,XTERM);
7455 case KEY_setsockopt:
7456 LOP(OP_SSOCKOPT,XTERM);
7462 LOP(OP_SHMCTL,XTERM);
7465 LOP(OP_SHMGET,XTERM);
7468 LOP(OP_SHMREAD,XTERM);
7471 LOP(OP_SHMWRITE,XTERM);
7474 LOP(OP_SHUTDOWN,XTERM);
7483 LOP(OP_SOCKET,XTERM);
7485 case KEY_socketpair:
7486 LOP(OP_SOCKPAIR,XTERM);
7489 checkcomma(s,PL_tokenbuf,"subroutine name");
7491 if (*s == ';' || *s == ')') /* probably a close */
7492 Perl_croak(aTHX_ "sort is now a reserved word");
7494 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7498 LOP(OP_SPLIT,XTERM);
7501 LOP(OP_SPRINTF,XTERM);
7504 LOP(OP_SPLICE,XTERM);
7519 LOP(OP_SUBSTR,XTERM);
7525 char tmpbuf[sizeof PL_tokenbuf];
7526 SSize_t tboffset = 0;
7527 expectation attrful;
7528 bool have_name, have_proto;
7529 const int key = tmp;
7534 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7535 SV *subtoken = newSVpvn(tstart, s - tstart);
7539 s = SKIPSPACE2(s,tmpwhite);
7544 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7545 (*s == ':' && s[1] == ':'))
7548 SV *nametoke = NULL;
7552 attrful = XATTRBLOCK;
7553 /* remember buffer pos'n for later force_word */
7554 tboffset = s - PL_oldbufptr;
7555 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7558 nametoke = newSVpvn(s, d - s);
7560 if (memchr(tmpbuf, ':', len))
7561 sv_setpvn(PL_subname, tmpbuf, len);
7563 sv_setsv(PL_subname,PL_curstname);
7564 sv_catpvs(PL_subname,"::");
7565 sv_catpvn(PL_subname,tmpbuf,len);
7572 CURMAD('X', nametoke);
7573 CURMAD('_', tmpwhite);
7574 (void) force_word(PL_oldbufptr + tboffset, WORD,
7577 s = SKIPSPACE2(d,tmpwhite);
7584 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7585 PL_expect = XTERMBLOCK;
7586 attrful = XATTRTERM;
7587 sv_setpvs(PL_subname,"?");
7591 if (key == KEY_format) {
7593 PL_lex_formbrack = PL_lex_brackets + 1;
7595 PL_thistoken = subtoken;
7599 (void) force_word(PL_oldbufptr + tboffset, WORD,
7605 /* Look for a prototype */
7608 bool bad_proto = FALSE;
7609 bool in_brackets = FALSE;
7610 char greedy_proto = ' ';
7611 bool proto_after_greedy_proto = FALSE;
7612 bool must_be_last = FALSE;
7613 bool underscore = FALSE;
7614 bool seen_underscore = FALSE;
7615 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
7617 s = scan_str(s,!!PL_madskills,FALSE);
7619 Perl_croak(aTHX_ "Prototype not terminated");
7620 /* strip spaces and check for bad characters */
7621 d = SvPVX(PL_lex_stuff);
7623 for (p = d; *p; ++p) {
7627 if (warnillegalproto) {
7629 proto_after_greedy_proto = TRUE;
7630 if (!strchr("$@%*;[]&\\_", *p)) {
7642 else if ( *p == ']' ) {
7643 in_brackets = FALSE;
7645 else if ( (*p == '@' || *p == '%') &&
7646 ( tmp < 2 || d[tmp-2] != '\\' ) &&
7648 must_be_last = TRUE;
7651 else if ( *p == '_' ) {
7652 underscore = seen_underscore = TRUE;
7659 if (proto_after_greedy_proto)
7660 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7661 "Prototype after '%c' for %"SVf" : %s",
7662 greedy_proto, SVfARG(PL_subname), d);
7664 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7665 "Illegal character %sin prototype for %"SVf" : %s",
7666 seen_underscore ? "after '_' " : "",
7667 SVfARG(PL_subname), d);
7668 SvCUR_set(PL_lex_stuff, tmp);
7673 CURMAD('q', PL_thisopen);
7674 CURMAD('_', tmpwhite);
7675 CURMAD('=', PL_thisstuff);
7676 CURMAD('Q', PL_thisclose);
7677 NEXTVAL_NEXTTOKE.opval =
7678 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7679 PL_lex_stuff = NULL;
7682 s = SKIPSPACE2(s,tmpwhite);
7690 if (*s == ':' && s[1] != ':')
7691 PL_expect = attrful;
7692 else if (*s != '{' && key == KEY_sub) {
7694 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
7695 else if (*s != ';' && *s != '}')
7696 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
7703 curmad('^', newSVpvs(""));
7704 CURMAD('_', tmpwhite);
7708 PL_thistoken = subtoken;
7711 NEXTVAL_NEXTTOKE.opval =
7712 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7713 PL_lex_stuff = NULL;
7719 sv_setpvs(PL_subname, "__ANON__");
7721 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7725 (void) force_word(PL_oldbufptr + tboffset, WORD,
7734 LOP(OP_SYSTEM,XREF);
7737 LOP(OP_SYMLINK,XTERM);
7740 LOP(OP_SYSCALL,XTERM);
7743 LOP(OP_SYSOPEN,XTERM);
7746 LOP(OP_SYSSEEK,XTERM);
7749 LOP(OP_SYSREAD,XTERM);
7752 LOP(OP_SYSWRITE,XTERM);
7756 TERM(sublex_start());
7777 LOP(OP_TRUNCATE,XTERM);
7789 pl_yylval.ival = CopLINE(PL_curcop);
7793 pl_yylval.ival = CopLINE(PL_curcop);
7797 LOP(OP_UNLINK,XTERM);
7803 LOP(OP_UNPACK,XTERM);
7806 LOP(OP_UTIME,XTERM);
7812 LOP(OP_UNSHIFT,XTERM);
7815 s = tokenize_use(1, s);
7825 pl_yylval.ival = CopLINE(PL_curcop);
7829 pl_yylval.ival = CopLINE(PL_curcop);
7833 PL_hints |= HINT_BLOCK_SCOPE;
7840 LOP(OP_WAITPID,XTERM);
7849 ctl_l[0] = toCTRL('L');
7851 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7854 /* Make sure $^L is defined */
7855 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7860 if (PL_expect == XOPERATOR)
7866 pl_yylval.ival = OP_XOR;
7871 TERM(sublex_start());
7876 #pragma segment Main
7880 S_pending_ident(pTHX)
7885 /* pit holds the identifier we read and pending_ident is reset */
7886 char pit = PL_pending_ident;
7887 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7888 /* All routes through this function want to know if there is a colon. */
7889 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7890 PL_pending_ident = 0;
7892 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7893 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7894 "### Pending identifier '%s'\n", PL_tokenbuf); });
7896 /* if we're in a my(), we can't allow dynamics here.
7897 $foo'bar has already been turned into $foo::bar, so
7898 just check for colons.
7900 if it's a legal name, the OP is a PADANY.
7903 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7905 yyerror(Perl_form(aTHX_ "No package name allowed for "
7906 "variable %s in \"our\"",
7908 tmp = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7912 yyerror(Perl_form(aTHX_ PL_no_myglob,
7913 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7915 pl_yylval.opval = newOP(OP_PADANY, 0);
7916 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7922 build the ops for accesses to a my() variable.
7924 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7925 then used in a comparison. This catches most, but not
7926 all cases. For instance, it catches
7927 sort { my($a); $a <=> $b }
7929 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7930 (although why you'd do that is anyone's guess).
7935 tmp = pad_findmy(PL_tokenbuf, tokenbuf_len, 0);
7936 if (tmp != NOT_IN_PAD) {
7937 /* might be an "our" variable" */
7938 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7939 /* build ops for a bareword */
7940 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7941 HEK * const stashname = HvNAME_HEK(stash);
7942 SV * const sym = newSVhek(stashname);
7943 sv_catpvs(sym, "::");
7944 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7945 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7946 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7949 ? (GV_ADDMULTI | GV_ADDINEVAL)
7952 ((PL_tokenbuf[0] == '$') ? SVt_PV
7953 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
7958 /* if it's a sort block and they're naming $a or $b */
7959 if (PL_last_lop_op == OP_SORT &&
7960 PL_tokenbuf[0] == '$' &&
7961 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
7964 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
7965 d < PL_bufend && *d != '\n';
7968 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
7969 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
7975 pl_yylval.opval = newOP(OP_PADANY, 0);
7976 pl_yylval.opval->op_targ = tmp;
7982 Whine if they've said @foo in a doublequoted string,
7983 and @foo isn't a variable we can find in the symbol
7986 if (ckWARN(WARN_AMBIGUOUS) &&
7987 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
7988 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
7990 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
7991 /* DO NOT warn for @- and @+ */
7992 && !( PL_tokenbuf[2] == '\0' &&
7993 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
7996 /* Downgraded from fatal to warning 20000522 mjd */
7997 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
7998 "Possible unintended interpolation of %s in string",
8003 /* build ops for a bareword */
8004 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
8006 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8008 PL_tokenbuf + 1, tokenbuf_len - 1,
8009 /* If the identifier refers to a stash, don't autovivify it.
8010 * Change 24660 had the side effect of causing symbol table
8011 * hashes to always be defined, even if they were freshly
8012 * created and the only reference in the entire program was
8013 * the single statement with the defined %foo::bar:: test.
8014 * It appears that all code in the wild doing this actually
8015 * wants to know whether sub-packages have been loaded, so
8016 * by avoiding auto-vivifying symbol tables, we ensure that
8017 * defined %foo::bar:: continues to be false, and the existing
8018 * tests still give the expected answers, even though what
8019 * they're actually testing has now changed subtly.
8021 (*PL_tokenbuf == '%'
8022 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
8025 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
8026 ((PL_tokenbuf[0] == '$') ? SVt_PV
8027 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8033 * The following code was generated by perl_keyword.pl.
8037 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
8041 PERL_ARGS_ASSERT_KEYWORD;
8045 case 1: /* 5 tokens of length 1 */
8077 case 2: /* 18 tokens of length 2 */
8223 case 3: /* 29 tokens of length 3 */
8227 if (name[1] == 'N' &&
8290 if (name[1] == 'i' &&
8322 if (name[1] == 'o' &&
8331 if (name[1] == 'e' &&
8340 if (name[1] == 'n' &&
8349 if (name[1] == 'o' &&
8358 if (name[1] == 'a' &&
8367 if (name[1] == 'o' &&
8429 if (name[1] == 'e' &&
8443 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
8469 if (name[1] == 'i' &&
8478 if (name[1] == 's' &&
8487 if (name[1] == 'e' &&
8496 if (name[1] == 'o' &&
8508 case 4: /* 41 tokens of length 4 */
8512 if (name[1] == 'O' &&
8522 if (name[1] == 'N' &&
8532 if (name[1] == 'i' &&
8542 if (name[1] == 'h' &&
8552 if (name[1] == 'u' &&
8565 if (name[2] == 'c' &&
8574 if (name[2] == 's' &&
8583 if (name[2] == 'a' &&
8619 if (name[1] == 'o' &&
8632 if (name[2] == 't' &&
8641 if (name[2] == 'o' &&
8650 if (name[2] == 't' &&
8659 if (name[2] == 'e' &&
8672 if (name[1] == 'o' &&
8685 if (name[2] == 'y' &&
8694 if (name[2] == 'l' &&
8710 if (name[2] == 's' &&
8719 if (name[2] == 'n' &&
8728 if (name[2] == 'c' &&
8741 if (name[1] == 'e' &&
8751 if (name[1] == 'p' &&
8764 if (name[2] == 'c' &&
8773 if (name[2] == 'p' &&
8782 if (name[2] == 's' &&
8798 if (name[2] == 'n' &&
8868 if (name[2] == 'r' &&
8877 if (name[2] == 'r' &&
8886 if (name[2] == 'a' &&
8902 if (name[2] == 'l' &&
8964 if (name[2] == 'e' &&
8967 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
8980 case 5: /* 39 tokens of length 5 */
8984 if (name[1] == 'E' &&
8995 if (name[1] == 'H' &&
9009 if (name[2] == 'a' &&
9019 if (name[2] == 'a' &&
9036 if (name[2] == 'e' &&
9046 if (name[2] == 'e' &&
9050 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
9066 if (name[3] == 'i' &&
9075 if (name[3] == 'o' &&
9111 if (name[2] == 'o' &&
9121 if (name[2] == 'y' &&
9135 if (name[1] == 'l' &&
9149 if (name[2] == 'n' &&
9159 if (name[2] == 'o' &&
9173 if (name[1] == 'i' &&
9178 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
9187 if (name[2] == 'd' &&
9197 if (name[2] == 'c' &&
9214 if (name[2] == 'c' &&
9224 if (name[2] == 't' &&
9238 if (name[1] == 'k' &&
9249 if (name[1] == 'r' &&
9263 if (name[2] == 's' &&
9273 if (name[2] == 'd' &&
9290 if (name[2] == 'm' &&
9300 if (name[2] == 'i' &&
9310 if (name[2] == 'e' &&
9320 if (name[2] == 'l' &&
9330 if (name[2] == 'a' &&
9343 if (name[3] == 't' &&
9346 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
9352 if (name[3] == 'd' &&
9369 if (name[1] == 'i' &&
9383 if (name[2] == 'a' &&
9396 if (name[3] == 'e' &&
9431 if (name[2] == 'i' &&
9448 if (name[2] == 'i' &&
9458 if (name[2] == 'i' &&
9475 case 6: /* 33 tokens of length 6 */
9479 if (name[1] == 'c' &&
9494 if (name[2] == 'l' &&
9505 if (name[2] == 'r' &&
9520 if (name[1] == 'e' &&
9535 if (name[2] == 's' &&
9540 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
9546 if (name[2] == 'i' &&
9564 if (name[2] == 'l' &&
9575 if (name[2] == 'r' &&
9590 if (name[1] == 'm' &&
9605 if (name[2] == 'n' &&
9616 if (name[2] == 's' &&
9631 if (name[1] == 's' &&
9637 if (name[4] == 't' &&
9646 if (name[4] == 'e' &&
9655 if (name[4] == 'c' &&
9664 if (name[4] == 'n' &&
9680 if (name[1] == 'r' &&
9698 if (name[3] == 'a' &&
9708 if (name[3] == 'u' &&
9722 if (name[2] == 'n' &&
9740 if (name[2] == 'a' &&
9754 if (name[3] == 'e' &&
9767 if (name[4] == 't' &&
9776 if (name[4] == 'e' &&
9798 if (name[4] == 't' &&
9807 if (name[4] == 'e' &&
9823 if (name[2] == 'c' &&
9834 if (name[2] == 'l' &&
9845 if (name[2] == 'b' &&
9856 if (name[2] == 's' &&
9879 if (name[4] == 's' &&
9888 if (name[4] == 'n' &&
9901 if (name[3] == 'a' &&
9918 if (name[1] == 'a' &&
9933 case 7: /* 29 tokens of length 7 */
9937 if (name[1] == 'E' &&
9950 if (name[1] == '_' &&
9963 if (name[1] == 'i' &&
9970 return -KEY_binmode;
9976 if (name[1] == 'o' &&
9983 return -KEY_connect;
9992 if (name[2] == 'm' &&
9998 return -KEY_dbmopen;
10004 if (name[2] == 'f')
10009 if (name[4] == 'u' &&
10013 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
10019 if (name[4] == 'n' &&
10023 return KEY_defined;
10040 if (name[1] == 'o' &&
10047 return KEY_foreach;
10053 if (name[1] == 'e' &&
10060 if (name[5] == 'r' &&
10063 return -KEY_getpgrp;
10069 if (name[5] == 'i' &&
10072 return -KEY_getppid;
10085 if (name[1] == 'c' &&
10092 return -KEY_lcfirst;
10098 if (name[1] == 'p' &&
10105 return -KEY_opendir;
10111 if (name[1] == 'a' &&
10118 return KEY_package;
10124 if (name[1] == 'e')
10129 if (name[3] == 'd' &&
10134 return -KEY_readdir;
10140 if (name[3] == 'u' &&
10145 return KEY_require;
10151 if (name[3] == 'e' &&
10156 return -KEY_reverse;
10175 if (name[3] == 'k' &&
10180 return -KEY_seekdir;
10186 if (name[3] == 'p' &&
10191 return -KEY_setpgrp;
10201 if (name[2] == 'm' &&
10207 return -KEY_shmread;
10213 if (name[2] == 'r' &&
10219 return -KEY_sprintf;
10228 if (name[3] == 'l' &&
10233 return -KEY_symlink;
10242 if (name[4] == 'a' &&
10246 return -KEY_syscall;
10252 if (name[4] == 'p' &&
10256 return -KEY_sysopen;
10262 if (name[4] == 'e' &&
10266 return -KEY_sysread;
10272 if (name[4] == 'e' &&
10276 return -KEY_sysseek;
10294 if (name[1] == 'e' &&
10301 return -KEY_telldir;
10310 if (name[2] == 'f' &&
10316 return -KEY_ucfirst;
10322 if (name[2] == 's' &&
10328 return -KEY_unshift;
10338 if (name[1] == 'a' &&
10345 return -KEY_waitpid;
10354 case 8: /* 26 tokens of length 8 */
10358 if (name[1] == 'U' &&
10366 return KEY_AUTOLOAD;
10372 if (name[1] == '_')
10377 if (name[3] == 'A' &&
10383 return KEY___DATA__;
10389 if (name[3] == 'I' &&
10395 return -KEY___FILE__;
10401 if (name[3] == 'I' &&
10407 return -KEY___LINE__;
10423 if (name[2] == 'o' &&
10430 return -KEY_closedir;
10436 if (name[2] == 'n' &&
10443 return -KEY_continue;
10453 if (name[1] == 'b' &&
10461 return -KEY_dbmclose;
10467 if (name[1] == 'n' &&
10473 if (name[4] == 'r' &&
10478 return -KEY_endgrent;
10484 if (name[4] == 'w' &&
10489 return -KEY_endpwent;
10502 if (name[1] == 'o' &&
10510 return -KEY_formline;
10516 if (name[1] == 'e' &&
10522 if (name[4] == 'r')
10527 if (name[6] == 'n' &&
10530 return -KEY_getgrent;
10536 if (name[6] == 'i' &&
10539 return -KEY_getgrgid;
10545 if (name[6] == 'a' &&
10548 return -KEY_getgrnam;
10561 if (name[4] == 'o' &&
10566 return -KEY_getlogin;
10572 if (name[4] == 'w')
10577 if (name[6] == 'n' &&
10580 return -KEY_getpwent;
10586 if (name[6] == 'a' &&
10589 return -KEY_getpwnam;
10595 if (name[6] == 'i' &&
10598 return -KEY_getpwuid;
10618 if (name[1] == 'e' &&
10625 if (name[5] == 'i' &&
10632 return -KEY_readline;
10637 return -KEY_readlink;
10648 if (name[5] == 'i' &&
10652 return -KEY_readpipe;
10668 if (name[2] == 't')
10673 if (name[4] == 'r' &&
10678 return -KEY_setgrent;
10684 if (name[4] == 'w' &&
10689 return -KEY_setpwent;
10705 if (name[3] == 'w' &&
10711 return -KEY_shmwrite;
10717 if (name[3] == 't' &&
10723 return -KEY_shutdown;
10733 if (name[2] == 's' &&
10740 return -KEY_syswrite;
10750 if (name[1] == 'r' &&
10758 return -KEY_truncate;
10767 case 9: /* 9 tokens of length 9 */
10771 if (name[1] == 'N' &&
10780 return KEY_UNITCHECK;
10786 if (name[1] == 'n' &&
10795 return -KEY_endnetent;
10801 if (name[1] == 'e' &&
10810 return -KEY_getnetent;
10816 if (name[1] == 'o' &&
10825 return -KEY_localtime;
10831 if (name[1] == 'r' &&
10840 return KEY_prototype;
10846 if (name[1] == 'u' &&
10855 return -KEY_quotemeta;
10861 if (name[1] == 'e' &&
10870 return -KEY_rewinddir;
10876 if (name[1] == 'e' &&
10885 return -KEY_setnetent;
10891 if (name[1] == 'a' &&
10900 return -KEY_wantarray;
10909 case 10: /* 9 tokens of length 10 */
10913 if (name[1] == 'n' &&
10919 if (name[4] == 'o' &&
10926 return -KEY_endhostent;
10932 if (name[4] == 'e' &&
10939 return -KEY_endservent;
10952 if (name[1] == 'e' &&
10958 if (name[4] == 'o' &&
10965 return -KEY_gethostent;
10974 if (name[5] == 'r' &&
10980 return -KEY_getservent;
10986 if (name[5] == 'c' &&
10992 return -KEY_getsockopt;
11012 if (name[2] == 't')
11017 if (name[4] == 'o' &&
11024 return -KEY_sethostent;
11033 if (name[5] == 'r' &&
11039 return -KEY_setservent;
11045 if (name[5] == 'c' &&
11051 return -KEY_setsockopt;
11068 if (name[2] == 'c' &&
11077 return -KEY_socketpair;
11090 case 11: /* 8 tokens of length 11 */
11094 if (name[1] == '_' &&
11104 { /* __PACKAGE__ */
11105 return -KEY___PACKAGE__;
11111 if (name[1] == 'n' &&
11121 { /* endprotoent */
11122 return -KEY_endprotoent;
11128 if (name[1] == 'e' &&
11137 if (name[5] == 'e' &&
11143 { /* getpeername */
11144 return -KEY_getpeername;
11153 if (name[6] == 'o' &&
11158 { /* getpriority */
11159 return -KEY_getpriority;
11165 if (name[6] == 't' &&
11170 { /* getprotoent */
11171 return -KEY_getprotoent;
11185 if (name[4] == 'o' &&
11192 { /* getsockname */
11193 return -KEY_getsockname;
11206 if (name[1] == 'e' &&
11214 if (name[6] == 'o' &&
11219 { /* setpriority */
11220 return -KEY_setpriority;
11226 if (name[6] == 't' &&
11231 { /* setprotoent */
11232 return -KEY_setprotoent;
11248 case 12: /* 2 tokens of length 12 */
11249 if (name[0] == 'g' &&
11261 if (name[9] == 'd' &&
11264 { /* getnetbyaddr */
11265 return -KEY_getnetbyaddr;
11271 if (name[9] == 'a' &&
11274 { /* getnetbyname */
11275 return -KEY_getnetbyname;
11287 case 13: /* 4 tokens of length 13 */
11288 if (name[0] == 'g' &&
11295 if (name[4] == 'o' &&
11304 if (name[10] == 'd' &&
11307 { /* gethostbyaddr */
11308 return -KEY_gethostbyaddr;
11314 if (name[10] == 'a' &&
11317 { /* gethostbyname */
11318 return -KEY_gethostbyname;
11331 if (name[4] == 'e' &&
11340 if (name[10] == 'a' &&
11343 { /* getservbyname */
11344 return -KEY_getservbyname;
11350 if (name[10] == 'o' &&
11353 { /* getservbyport */
11354 return -KEY_getservbyport;
11373 case 14: /* 1 tokens of length 14 */
11374 if (name[0] == 'g' &&
11388 { /* getprotobyname */
11389 return -KEY_getprotobyname;
11394 case 16: /* 1 tokens of length 16 */
11395 if (name[0] == 'g' &&
11411 { /* getprotobynumber */
11412 return -KEY_getprotobynumber;
11426 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
11430 PERL_ARGS_ASSERT_CHECKCOMMA;
11432 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
11433 if (ckWARN(WARN_SYNTAX)) {
11436 for (w = s+2; *w && level; w++) {
11439 else if (*w == ')')
11442 while (isSPACE(*w))
11444 /* the list of chars below is for end of statements or
11445 * block / parens, boolean operators (&&, ||, //) and branch
11446 * constructs (or, and, if, until, unless, while, err, for).
11447 * Not a very solid hack... */
11448 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
11449 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11450 "%s (...) interpreted as function",name);
11453 while (s < PL_bufend && isSPACE(*s))
11457 while (s < PL_bufend && isSPACE(*s))
11459 if (isIDFIRST_lazy_if(s,UTF)) {
11460 const char * const w = s++;
11461 while (isALNUM_lazy_if(s,UTF))
11463 while (s < PL_bufend && isSPACE(*s))
11467 if (keyword(w, s - w, 0))
11470 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
11471 if (gv && GvCVu(gv))
11473 Perl_croak(aTHX_ "No comma allowed after %s", what);
11478 /* Either returns sv, or mortalizes sv and returns a new SV*.
11479 Best used as sv=new_constant(..., sv, ...).
11480 If s, pv are NULL, calls subroutine with one argument,
11481 and type is used with error messages only. */
11484 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
11485 SV *sv, SV *pv, const char *type, STRLEN typelen)
11488 HV * const table = GvHV(PL_hintgv); /* ^H */
11492 const char *why1 = "", *why2 = "", *why3 = "";
11494 PERL_ARGS_ASSERT_NEW_CONSTANT;
11496 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
11499 why2 = (const char *)
11500 (strEQ(key,"charnames")
11501 ? "(possibly a missing \"use charnames ...\")"
11503 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
11504 (type ? type: "undef"), why2);
11506 /* This is convoluted and evil ("goto considered harmful")
11507 * but I do not understand the intricacies of all the different
11508 * failure modes of %^H in here. The goal here is to make
11509 * the most probable error message user-friendly. --jhi */
11514 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
11515 (type ? type: "undef"), why1, why2, why3);
11517 yyerror(SvPVX_const(msg));
11522 /* charnames doesn't work well if there have been errors found */
11523 if (PL_error_count > 0 && strEQ(key,"charnames")) return res;
11525 cvp = hv_fetch(table, key, keylen, FALSE);
11526 if (!cvp || !SvOK(*cvp)) {
11529 why3 = "} is not defined";
11532 sv_2mortal(sv); /* Parent created it permanently */
11535 pv = newSVpvn_flags(s, len, SVs_TEMP);
11537 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
11539 typesv = &PL_sv_undef;
11541 PUSHSTACKi(PERLSI_OVERLOAD);
11553 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
11557 /* Check the eval first */
11558 if (!PL_in_eval && SvTRUE(ERRSV)) {
11559 sv_catpvs(ERRSV, "Propagated");
11560 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
11562 res = SvREFCNT_inc_simple(sv);
11566 SvREFCNT_inc_simple_void(res);
11575 why1 = "Call to &{$^H{";
11577 why3 = "}} did not return a defined value";
11585 /* Returns a NUL terminated string, with the length of the string written to
11589 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
11592 register char *d = dest;
11593 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11595 PERL_ARGS_ASSERT_SCAN_WORD;
11599 Perl_croak(aTHX_ ident_too_long);
11600 if (isALNUM(*s)) /* UTF handled below */
11602 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
11607 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
11611 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11612 char *t = s + UTF8SKIP(s);
11614 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11618 Perl_croak(aTHX_ ident_too_long);
11619 Copy(s, d, len, char);
11632 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
11635 char *bracket = NULL;
11637 register char *d = dest;
11638 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11640 PERL_ARGS_ASSERT_SCAN_IDENT;
11645 while (isDIGIT(*s)) {
11647 Perl_croak(aTHX_ ident_too_long);
11654 Perl_croak(aTHX_ ident_too_long);
11655 if (isALNUM(*s)) /* UTF handled below */
11657 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
11662 else if (*s == ':' && s[1] == ':') {
11666 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11667 char *t = s + UTF8SKIP(s);
11668 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11670 if (d + (t - s) > e)
11671 Perl_croak(aTHX_ ident_too_long);
11672 Copy(s, d, t - s, char);
11683 if (PL_lex_state != LEX_NORMAL)
11684 PL_lex_state = LEX_INTERPENDMAYBE;
11687 if (*s == '$' && s[1] &&
11688 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
11701 if (*d == '^' && *s && isCONTROLVAR(*s)) {
11706 if (isSPACE(s[-1])) {
11708 const char ch = *s++;
11709 if (!SPACE_OR_TAB(ch)) {
11715 if (isIDFIRST_lazy_if(d,UTF)) {
11719 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
11720 end += UTF8SKIP(end);
11721 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
11722 end += UTF8SKIP(end);
11724 Copy(s, d, end - s, char);
11729 while ((isALNUM(*s) || *s == ':') && d < e)
11732 Perl_croak(aTHX_ ident_too_long);
11735 while (s < send && SPACE_OR_TAB(*s))
11737 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
11738 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
11739 const char * const brack =
11741 ((*s == '[') ? "[...]" : "{...}");
11742 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11743 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
11744 funny, dest, brack, funny, dest, brack);
11747 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
11751 /* Handle extended ${^Foo} variables
11752 * 1999-02-27 mjd-perl-patch@plover.com */
11753 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
11757 while (isALNUM(*s) && d < e) {
11761 Perl_croak(aTHX_ ident_too_long);
11766 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
11767 PL_lex_state = LEX_INTERPEND;
11770 if (PL_lex_state == LEX_NORMAL) {
11771 if (ckWARN(WARN_AMBIGUOUS) &&
11772 (keyword(dest, d - dest, 0)
11773 || get_cvn_flags(dest, d - dest, 0)))
11777 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11778 "Ambiguous use of %c{%s} resolved to %c%s",
11779 funny, dest, funny, dest);
11784 s = bracket; /* let the parser handle it */
11788 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
11789 PL_lex_state = LEX_INTERPEND;
11794 S_pmflag(U32 pmfl, const char ch) {
11796 CASE_STD_PMMOD_FLAGS_PARSE_SET(&pmfl);
11797 case GLOBAL_PAT_MOD: pmfl |= PMf_GLOBAL; break;
11798 case CONTINUE_PAT_MOD: pmfl |= PMf_CONTINUE; break;
11799 case ONCE_PAT_MOD: pmfl |= PMf_KEEP; break;
11800 case KEEPCOPY_PAT_MOD: pmfl |= PMf_KEEPCOPY; break;
11806 Perl_pmflag(pTHX_ U32* pmfl, int ch)
11808 PERL_ARGS_ASSERT_PMFLAG;
11810 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
11811 "Perl_pmflag() is deprecated, and will be removed from the XS API");
11814 *pmfl = S_pmflag(*pmfl, (char)ch);
11819 S_scan_pat(pTHX_ char *start, I32 type)
11823 char *s = scan_str(start,!!PL_madskills,FALSE);
11824 const char * const valid_flags =
11825 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
11830 PERL_ARGS_ASSERT_SCAN_PAT;
11833 const char * const delimiter = skipspace(start);
11837 ? "Search pattern not terminated or ternary operator parsed as search pattern"
11838 : "Search pattern not terminated" ));
11841 pm = (PMOP*)newPMOP(type, 0);
11842 if (PL_multi_open == '?') {
11843 /* This is the only point in the code that sets PMf_ONCE: */
11844 pm->op_pmflags |= PMf_ONCE;
11846 /* Hence it's safe to do this bit of PMOP book-keeping here, which
11847 allows us to restrict the list needed by reset to just the ??
11849 assert(type != OP_TRANS);
11851 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11854 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11857 elements = mg->mg_len / sizeof(PMOP**);
11858 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11859 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11860 mg->mg_len = elements * sizeof(PMOP**);
11861 PmopSTASH_set(pm,PL_curstash);
11867 while (*s && strchr(valid_flags, *s))
11868 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11870 if (PL_madskills && modstart != s) {
11871 SV* tmptoken = newSVpvn(modstart, s - modstart);
11872 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11875 /* issue a warning if /c is specified,but /g is not */
11876 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11878 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11879 "Use of /c modifier is meaningless without /g" );
11882 PL_lex_op = (OP*)pm;
11883 pl_yylval.ival = OP_MATCH;
11888 S_scan_subst(pTHX_ char *start)
11899 PERL_ARGS_ASSERT_SCAN_SUBST;
11901 pl_yylval.ival = OP_NULL;
11903 s = scan_str(start,!!PL_madskills,FALSE);
11906 Perl_croak(aTHX_ "Substitution pattern not terminated");
11908 if (s[-1] == PL_multi_open)
11911 if (PL_madskills) {
11912 CURMAD('q', PL_thisopen);
11913 CURMAD('_', PL_thiswhite);
11914 CURMAD('E', PL_thisstuff);
11915 CURMAD('Q', PL_thisclose);
11916 PL_realtokenstart = s - SvPVX(PL_linestr);
11920 first_start = PL_multi_start;
11921 s = scan_str(s,!!PL_madskills,FALSE);
11923 if (PL_lex_stuff) {
11924 SvREFCNT_dec(PL_lex_stuff);
11925 PL_lex_stuff = NULL;
11927 Perl_croak(aTHX_ "Substitution replacement not terminated");
11929 PL_multi_start = first_start; /* so whole substitution is taken together */
11931 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11934 if (PL_madskills) {
11935 CURMAD('z', PL_thisopen);
11936 CURMAD('R', PL_thisstuff);
11937 CURMAD('Z', PL_thisclose);
11943 if (*s == EXEC_PAT_MOD) {
11947 else if (strchr(S_PAT_MODS, *s))
11948 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11954 if (PL_madskills) {
11956 curmad('m', newSVpvn(modstart, s - modstart));
11957 append_madprops(PL_thismad, (OP*)pm, 0);
11961 if ((pm->op_pmflags & PMf_CONTINUE)) {
11962 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
11966 SV * const repl = newSVpvs("");
11968 PL_sublex_info.super_bufptr = s;
11969 PL_sublex_info.super_bufend = PL_bufend;
11971 pm->op_pmflags |= PMf_EVAL;
11974 sv_catpvs(repl, "eval ");
11976 sv_catpvs(repl, "do ");
11978 sv_catpvs(repl, "{");
11979 sv_catsv(repl, PL_lex_repl);
11980 if (strchr(SvPVX(PL_lex_repl), '#'))
11981 sv_catpvs(repl, "\n");
11982 sv_catpvs(repl, "}");
11984 SvREFCNT_dec(PL_lex_repl);
11985 PL_lex_repl = repl;
11988 PL_lex_op = (OP*)pm;
11989 pl_yylval.ival = OP_SUBST;
11994 S_scan_trans(pTHX_ char *start)
12007 PERL_ARGS_ASSERT_SCAN_TRANS;
12009 pl_yylval.ival = OP_NULL;
12011 s = scan_str(start,!!PL_madskills,FALSE);
12013 Perl_croak(aTHX_ "Transliteration pattern not terminated");
12015 if (s[-1] == PL_multi_open)
12018 if (PL_madskills) {
12019 CURMAD('q', PL_thisopen);
12020 CURMAD('_', PL_thiswhite);
12021 CURMAD('E', PL_thisstuff);
12022 CURMAD('Q', PL_thisclose);
12023 PL_realtokenstart = s - SvPVX(PL_linestr);
12027 s = scan_str(s,!!PL_madskills,FALSE);
12029 if (PL_lex_stuff) {
12030 SvREFCNT_dec(PL_lex_stuff);
12031 PL_lex_stuff = NULL;
12033 Perl_croak(aTHX_ "Transliteration replacement not terminated");
12035 if (PL_madskills) {
12036 CURMAD('z', PL_thisopen);
12037 CURMAD('R', PL_thisstuff);
12038 CURMAD('Z', PL_thisclose);
12041 complement = del = squash = 0;
12048 complement = OPpTRANS_COMPLEMENT;
12051 del = OPpTRANS_DELETE;
12054 squash = OPpTRANS_SQUASH;
12063 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
12064 o = newPVOP(OP_TRANS, 0, (char*)tbl);
12065 o->op_private &= ~OPpTRANS_ALL;
12066 o->op_private |= del|squash|complement|
12067 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
12068 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
12071 pl_yylval.ival = OP_TRANS;
12074 if (PL_madskills) {
12076 curmad('m', newSVpvn(modstart, s - modstart));
12077 append_madprops(PL_thismad, o, 0);
12086 S_scan_heredoc(pTHX_ register char *s)
12090 I32 op_type = OP_SCALAR;
12094 const char *found_newline;
12098 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
12100 I32 stuffstart = s - SvPVX(PL_linestr);
12103 PL_realtokenstart = -1;
12106 PERL_ARGS_ASSERT_SCAN_HEREDOC;
12110 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
12114 while (SPACE_OR_TAB(*peek))
12116 if (*peek == '`' || *peek == '\'' || *peek =='"') {
12119 s = delimcpy(d, e, s, PL_bufend, term, &len);
12129 if (!isALNUM_lazy_if(s,UTF))
12130 deprecate("bare << to mean <<\"\"");
12131 for (; isALNUM_lazy_if(s,UTF); s++) {
12136 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
12137 Perl_croak(aTHX_ "Delimiter for here document is too long");
12140 len = d - PL_tokenbuf;
12143 if (PL_madskills) {
12144 tstart = PL_tokenbuf + !outer;
12145 PL_thisclose = newSVpvn(tstart, len - !outer);
12146 tstart = SvPVX(PL_linestr) + stuffstart;
12147 PL_thisopen = newSVpvn(tstart, s - tstart);
12148 stuffstart = s - SvPVX(PL_linestr);
12151 #ifndef PERL_STRICT_CR
12152 d = strchr(s, '\r');
12154 char * const olds = s;
12156 while (s < PL_bufend) {
12162 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
12171 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12178 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
12179 herewas = newSVpvn(s,PL_bufend-s);
12183 herewas = newSVpvn(s-1,found_newline-s+1);
12186 herewas = newSVpvn(s,found_newline-s);
12190 if (PL_madskills) {
12191 tstart = SvPVX(PL_linestr) + stuffstart;
12193 sv_catpvn(PL_thisstuff, tstart, s - tstart);
12195 PL_thisstuff = newSVpvn(tstart, s - tstart);
12198 s += SvCUR(herewas);
12201 stuffstart = s - SvPVX(PL_linestr);
12207 tmpstr = newSV_type(SVt_PVIV);
12208 SvGROW(tmpstr, 80);
12209 if (term == '\'') {
12210 op_type = OP_CONST;
12211 SvIV_set(tmpstr, -1);
12213 else if (term == '`') {
12214 op_type = OP_BACKTICK;
12215 SvIV_set(tmpstr, '\\');
12219 PL_multi_start = CopLINE(PL_curcop);
12220 PL_multi_open = PL_multi_close = '<';
12221 term = *PL_tokenbuf;
12222 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
12223 char * const bufptr = PL_sublex_info.super_bufptr;
12224 char * const bufend = PL_sublex_info.super_bufend;
12225 char * const olds = s - SvCUR(herewas);
12226 s = strchr(bufptr, '\n');
12230 while (s < bufend &&
12231 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12233 CopLINE_inc(PL_curcop);
12236 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12237 missingterm(PL_tokenbuf);
12239 sv_setpvn(herewas,bufptr,d-bufptr+1);
12240 sv_setpvn(tmpstr,d+1,s-d);
12242 sv_catpvn(herewas,s,bufend-s);
12243 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
12250 while (s < PL_bufend &&
12251 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12253 CopLINE_inc(PL_curcop);
12255 if (s >= PL_bufend) {
12256 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12257 missingterm(PL_tokenbuf);
12259 sv_setpvn(tmpstr,d+1,s-d);
12261 if (PL_madskills) {
12263 sv_catpvn(PL_thisstuff, d + 1, s - d);
12265 PL_thisstuff = newSVpvn(d + 1, s - d);
12266 stuffstart = s - SvPVX(PL_linestr);
12270 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
12272 sv_catpvn(herewas,s,PL_bufend-s);
12273 sv_setsv(PL_linestr,herewas);
12274 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
12275 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12276 PL_last_lop = PL_last_uni = NULL;
12279 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
12280 while (s >= PL_bufend) { /* multiple line string? */
12282 if (PL_madskills) {
12283 tstart = SvPVX(PL_linestr) + stuffstart;
12285 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12287 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12291 CopLINE_inc(PL_curcop);
12292 if (!outer || !lex_next_chunk(0)) {
12293 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12294 missingterm(PL_tokenbuf);
12296 CopLINE_dec(PL_curcop);
12299 stuffstart = s - SvPVX(PL_linestr);
12301 CopLINE_inc(PL_curcop);
12302 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12303 PL_last_lop = PL_last_uni = NULL;
12304 #ifndef PERL_STRICT_CR
12305 if (PL_bufend - PL_linestart >= 2) {
12306 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
12307 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
12309 PL_bufend[-2] = '\n';
12311 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12313 else if (PL_bufend[-1] == '\r')
12314 PL_bufend[-1] = '\n';
12316 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
12317 PL_bufend[-1] = '\n';
12319 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
12320 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
12321 *(SvPVX(PL_linestr) + off ) = ' ';
12322 sv_catsv(PL_linestr,herewas);
12323 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12324 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
12328 sv_catsv(tmpstr,PL_linestr);
12333 PL_multi_end = CopLINE(PL_curcop);
12334 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
12335 SvPV_shrink_to_cur(tmpstr);
12337 SvREFCNT_dec(herewas);
12339 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
12341 else if (PL_encoding)
12342 sv_recode_to_utf8(tmpstr, PL_encoding);
12344 PL_lex_stuff = tmpstr;
12345 pl_yylval.ival = op_type;
12349 /* scan_inputsymbol
12350 takes: current position in input buffer
12351 returns: new position in input buffer
12352 side-effects: pl_yylval and lex_op are set.
12357 <FH> read from filehandle
12358 <pkg::FH> read from package qualified filehandle
12359 <pkg'FH> read from package qualified filehandle
12360 <$fh> read from filehandle in $fh
12361 <*.h> filename glob
12366 S_scan_inputsymbol(pTHX_ char *start)
12369 register char *s = start; /* current position in buffer */
12372 char *d = PL_tokenbuf; /* start of temp holding space */
12373 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
12375 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
12377 end = strchr(s, '\n');
12380 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
12382 /* die if we didn't have space for the contents of the <>,
12383 or if it didn't end, or if we see a newline
12386 if (len >= (I32)sizeof PL_tokenbuf)
12387 Perl_croak(aTHX_ "Excessively long <> operator");
12389 Perl_croak(aTHX_ "Unterminated <> operator");
12394 Remember, only scalar variables are interpreted as filehandles by
12395 this code. Anything more complex (e.g., <$fh{$num}>) will be
12396 treated as a glob() call.
12397 This code makes use of the fact that except for the $ at the front,
12398 a scalar variable and a filehandle look the same.
12400 if (*d == '$' && d[1]) d++;
12402 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
12403 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
12406 /* If we've tried to read what we allow filehandles to look like, and
12407 there's still text left, then it must be a glob() and not a getline.
12408 Use scan_str to pull out the stuff between the <> and treat it
12409 as nothing more than a string.
12412 if (d - PL_tokenbuf != len) {
12413 pl_yylval.ival = OP_GLOB;
12414 s = scan_str(start,!!PL_madskills,FALSE);
12416 Perl_croak(aTHX_ "Glob not terminated");
12420 bool readline_overriden = FALSE;
12423 /* we're in a filehandle read situation */
12426 /* turn <> into <ARGV> */
12428 Copy("ARGV",d,5,char);
12430 /* Check whether readline() is overriden */
12431 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
12433 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
12435 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
12436 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
12437 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
12438 readline_overriden = TRUE;
12440 /* if <$fh>, create the ops to turn the variable into a
12444 /* try to find it in the pad for this block, otherwise find
12445 add symbol table ops
12447 const PADOFFSET tmp = pad_findmy(d, len, 0);
12448 if (tmp != NOT_IN_PAD) {
12449 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
12450 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
12451 HEK * const stashname = HvNAME_HEK(stash);
12452 SV * const sym = sv_2mortal(newSVhek(stashname));
12453 sv_catpvs(sym, "::");
12454 sv_catpv(sym, d+1);
12459 OP * const o = newOP(OP_PADSV, 0);
12461 PL_lex_op = readline_overriden
12462 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12463 append_elem(OP_LIST, o,
12464 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
12465 : (OP*)newUNOP(OP_READLINE, 0, o);
12474 ? (GV_ADDMULTI | GV_ADDINEVAL)
12477 PL_lex_op = readline_overriden
12478 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12479 append_elem(OP_LIST,
12480 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
12481 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12482 : (OP*)newUNOP(OP_READLINE, 0,
12483 newUNOP(OP_RV2SV, 0,
12484 newGVOP(OP_GV, 0, gv)));
12486 if (!readline_overriden)
12487 PL_lex_op->op_flags |= OPf_SPECIAL;
12488 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
12489 pl_yylval.ival = OP_NULL;
12492 /* If it's none of the above, it must be a literal filehandle
12493 (<Foo::BAR> or <FOO>) so build a simple readline OP */
12495 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
12496 PL_lex_op = readline_overriden
12497 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12498 append_elem(OP_LIST,
12499 newGVOP(OP_GV, 0, gv),
12500 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12501 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
12502 pl_yylval.ival = OP_NULL;
12511 takes: start position in buffer
12512 keep_quoted preserve \ on the embedded delimiter(s)
12513 keep_delims preserve the delimiters around the string
12514 returns: position to continue reading from buffer
12515 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
12516 updates the read buffer.
12518 This subroutine pulls a string out of the input. It is called for:
12519 q single quotes q(literal text)
12520 ' single quotes 'literal text'
12521 qq double quotes qq(interpolate $here please)
12522 " double quotes "interpolate $here please"
12523 qx backticks qx(/bin/ls -l)
12524 ` backticks `/bin/ls -l`
12525 qw quote words @EXPORT_OK = qw( func() $spam )
12526 m// regexp match m/this/
12527 s/// regexp substitute s/this/that/
12528 tr/// string transliterate tr/this/that/
12529 y/// string transliterate y/this/that/
12530 ($*@) sub prototypes sub foo ($)
12531 (stuff) sub attr parameters sub foo : attr(stuff)
12532 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
12534 In most of these cases (all but <>, patterns and transliterate)
12535 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
12536 calls scan_str(). s/// makes yylex() call scan_subst() which calls
12537 scan_str(). tr/// and y/// make yylex() call scan_trans() which
12540 It skips whitespace before the string starts, and treats the first
12541 character as the delimiter. If the delimiter is one of ([{< then
12542 the corresponding "close" character )]}> is used as the closing
12543 delimiter. It allows quoting of delimiters, and if the string has
12544 balanced delimiters ([{<>}]) it allows nesting.
12546 On success, the SV with the resulting string is put into lex_stuff or,
12547 if that is already non-NULL, into lex_repl. The second case occurs only
12548 when parsing the RHS of the special constructs s/// and tr/// (y///).
12549 For convenience, the terminating delimiter character is stuffed into
12554 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
12557 SV *sv; /* scalar value: string */
12558 const char *tmps; /* temp string, used for delimiter matching */
12559 register char *s = start; /* current position in the buffer */
12560 register char term; /* terminating character */
12561 register char *to; /* current position in the sv's data */
12562 I32 brackets = 1; /* bracket nesting level */
12563 bool has_utf8 = FALSE; /* is there any utf8 content? */
12564 I32 termcode; /* terminating char. code */
12565 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
12566 STRLEN termlen; /* length of terminating string */
12567 int last_off = 0; /* last position for nesting bracket */
12573 PERL_ARGS_ASSERT_SCAN_STR;
12575 /* skip space before the delimiter */
12581 if (PL_realtokenstart >= 0) {
12582 stuffstart = PL_realtokenstart;
12583 PL_realtokenstart = -1;
12586 stuffstart = start - SvPVX(PL_linestr);
12588 /* mark where we are, in case we need to report errors */
12591 /* after skipping whitespace, the next character is the terminator */
12594 termcode = termstr[0] = term;
12598 termcode = utf8_to_uvchr((U8*)s, &termlen);
12599 Copy(s, termstr, termlen, U8);
12600 if (!UTF8_IS_INVARIANT(term))
12604 /* mark where we are */
12605 PL_multi_start = CopLINE(PL_curcop);
12606 PL_multi_open = term;
12608 /* find corresponding closing delimiter */
12609 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
12610 termcode = termstr[0] = term = tmps[5];
12612 PL_multi_close = term;
12614 /* create a new SV to hold the contents. 79 is the SV's initial length.
12615 What a random number. */
12616 sv = newSV_type(SVt_PVIV);
12618 SvIV_set(sv, termcode);
12619 (void)SvPOK_only(sv); /* validate pointer */
12621 /* move past delimiter and try to read a complete string */
12623 sv_catpvn(sv, s, termlen);
12626 tstart = SvPVX(PL_linestr) + stuffstart;
12627 if (!PL_thisopen && !keep_delims) {
12628 PL_thisopen = newSVpvn(tstart, s - tstart);
12629 stuffstart = s - SvPVX(PL_linestr);
12633 if (PL_encoding && !UTF) {
12637 int offset = s - SvPVX_const(PL_linestr);
12638 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
12639 &offset, (char*)termstr, termlen);
12640 const char * const ns = SvPVX_const(PL_linestr) + offset;
12641 char * const svlast = SvEND(sv) - 1;
12643 for (; s < ns; s++) {
12644 if (*s == '\n' && !PL_rsfp)
12645 CopLINE_inc(PL_curcop);
12648 goto read_more_line;
12650 /* handle quoted delimiters */
12651 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
12653 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
12655 if ((svlast-1 - t) % 2) {
12656 if (!keep_quoted) {
12657 *(svlast-1) = term;
12659 SvCUR_set(sv, SvCUR(sv) - 1);
12664 if (PL_multi_open == PL_multi_close) {
12670 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
12671 /* At here, all closes are "was quoted" one,
12672 so we don't check PL_multi_close. */
12674 if (!keep_quoted && *(t+1) == PL_multi_open)
12679 else if (*t == PL_multi_open)
12687 SvCUR_set(sv, w - SvPVX_const(sv));
12689 last_off = w - SvPVX(sv);
12690 if (--brackets <= 0)
12695 if (!keep_delims) {
12696 SvCUR_set(sv, SvCUR(sv) - 1);
12702 /* extend sv if need be */
12703 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
12704 /* set 'to' to the next character in the sv's string */
12705 to = SvPVX(sv)+SvCUR(sv);
12707 /* if open delimiter is the close delimiter read unbridle */
12708 if (PL_multi_open == PL_multi_close) {
12709 for (; s < PL_bufend; s++,to++) {
12710 /* embedded newlines increment the current line number */
12711 if (*s == '\n' && !PL_rsfp)
12712 CopLINE_inc(PL_curcop);
12713 /* handle quoted delimiters */
12714 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
12715 if (!keep_quoted && s[1] == term)
12717 /* any other quotes are simply copied straight through */
12721 /* terminate when run out of buffer (the for() condition), or
12722 have found the terminator */
12723 else if (*s == term) {
12726 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
12729 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12735 /* if the terminator isn't the same as the start character (e.g.,
12736 matched brackets), we have to allow more in the quoting, and
12737 be prepared for nested brackets.
12740 /* read until we run out of string, or we find the terminator */
12741 for (; s < PL_bufend; s++,to++) {
12742 /* embedded newlines increment the line count */
12743 if (*s == '\n' && !PL_rsfp)
12744 CopLINE_inc(PL_curcop);
12745 /* backslashes can escape the open or closing characters */
12746 if (*s == '\\' && s+1 < PL_bufend) {
12747 if (!keep_quoted &&
12748 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
12753 /* allow nested opens and closes */
12754 else if (*s == PL_multi_close && --brackets <= 0)
12756 else if (*s == PL_multi_open)
12758 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12763 /* terminate the copied string and update the sv's end-of-string */
12765 SvCUR_set(sv, to - SvPVX_const(sv));
12768 * this next chunk reads more into the buffer if we're not done yet
12772 break; /* handle case where we are done yet :-) */
12774 #ifndef PERL_STRICT_CR
12775 if (to - SvPVX_const(sv) >= 2) {
12776 if ((to[-2] == '\r' && to[-1] == '\n') ||
12777 (to[-2] == '\n' && to[-1] == '\r'))
12781 SvCUR_set(sv, to - SvPVX_const(sv));
12783 else if (to[-1] == '\r')
12786 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
12791 /* if we're out of file, or a read fails, bail and reset the current
12792 line marker so we can report where the unterminated string began
12795 if (PL_madskills) {
12796 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12798 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12800 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12803 CopLINE_inc(PL_curcop);
12804 PL_bufptr = PL_bufend;
12805 if (!lex_next_chunk(0)) {
12807 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12816 /* at this point, we have successfully read the delimited string */
12818 if (!PL_encoding || UTF) {
12820 if (PL_madskills) {
12821 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12822 const int len = s - tstart;
12824 sv_catpvn(PL_thisstuff, tstart, len);
12826 PL_thisstuff = newSVpvn(tstart, len);
12827 if (!PL_thisclose && !keep_delims)
12828 PL_thisclose = newSVpvn(s,termlen);
12833 sv_catpvn(sv, s, termlen);
12838 if (PL_madskills) {
12839 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12840 const int len = s - tstart - termlen;
12842 sv_catpvn(PL_thisstuff, tstart, len);
12844 PL_thisstuff = newSVpvn(tstart, len);
12845 if (!PL_thisclose && !keep_delims)
12846 PL_thisclose = newSVpvn(s - termlen,termlen);
12850 if (has_utf8 || PL_encoding)
12853 PL_multi_end = CopLINE(PL_curcop);
12855 /* if we allocated too much space, give some back */
12856 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12857 SvLEN_set(sv, SvCUR(sv) + 1);
12858 SvPV_renew(sv, SvLEN(sv));
12861 /* decide whether this is the first or second quoted string we've read
12874 takes: pointer to position in buffer
12875 returns: pointer to new position in buffer
12876 side-effects: builds ops for the constant in pl_yylval.op
12878 Read a number in any of the formats that Perl accepts:
12880 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12881 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12884 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12886 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12889 If it reads a number without a decimal point or an exponent, it will
12890 try converting the number to an integer and see if it can do so
12891 without loss of precision.
12895 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12898 register const char *s = start; /* current position in buffer */
12899 register char *d; /* destination in temp buffer */
12900 register char *e; /* end of temp buffer */
12901 NV nv; /* number read, as a double */
12902 SV *sv = NULL; /* place to put the converted number */
12903 bool floatit; /* boolean: int or float? */
12904 const char *lastub = NULL; /* position of last underbar */
12905 static char const number_too_long[] = "Number too long";
12907 PERL_ARGS_ASSERT_SCAN_NUM;
12909 /* We use the first character to decide what type of number this is */
12913 Perl_croak(aTHX_ "panic: scan_num");
12915 /* if it starts with a 0, it could be an octal number, a decimal in
12916 0.13 disguise, or a hexadecimal number, or a binary number. */
12920 u holds the "number so far"
12921 shift the power of 2 of the base
12922 (hex == 4, octal == 3, binary == 1)
12923 overflowed was the number more than we can hold?
12925 Shift is used when we add a digit. It also serves as an "are
12926 we in octal/hex/binary?" indicator to disallow hex characters
12927 when in octal mode.
12932 bool overflowed = FALSE;
12933 bool just_zero = TRUE; /* just plain 0 or binary number? */
12934 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12935 static const char* const bases[5] =
12936 { "", "binary", "", "octal", "hexadecimal" };
12937 static const char* const Bases[5] =
12938 { "", "Binary", "", "Octal", "Hexadecimal" };
12939 static const char* const maxima[5] =
12941 "0b11111111111111111111111111111111",
12945 const char *base, *Base, *max;
12947 /* check for hex */
12952 } else if (s[1] == 'b') {
12957 /* check for a decimal in disguise */
12958 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
12960 /* so it must be octal */
12967 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12968 "Misplaced _ in number");
12972 base = bases[shift];
12973 Base = Bases[shift];
12974 max = maxima[shift];
12976 /* read the rest of the number */
12978 /* x is used in the overflow test,
12979 b is the digit we're adding on. */
12984 /* if we don't mention it, we're done */
12988 /* _ are ignored -- but warned about if consecutive */
12990 if (lastub && s == lastub + 1)
12991 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
12992 "Misplaced _ in number");
12996 /* 8 and 9 are not octal */
12997 case '8': case '9':
12999 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
13003 case '2': case '3': case '4':
13004 case '5': case '6': case '7':
13006 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
13009 case '0': case '1':
13010 b = *s++ & 15; /* ASCII digit -> value of digit */
13014 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
13015 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
13016 /* make sure they said 0x */
13019 b = (*s++ & 7) + 9;
13021 /* Prepare to put the digit we have onto the end
13022 of the number so far. We check for overflows.
13028 x = u << shift; /* make room for the digit */
13030 if ((x >> shift) != u
13031 && !(PL_hints & HINT_NEW_BINARY)) {
13034 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13035 "Integer overflow in %s number",
13038 u = x | b; /* add the digit to the end */
13041 n *= nvshift[shift];
13042 /* If an NV has not enough bits in its
13043 * mantissa to represent an UV this summing of
13044 * small low-order numbers is a waste of time
13045 * (because the NV cannot preserve the
13046 * low-order bits anyway): we could just
13047 * remember when did we overflow and in the
13048 * end just multiply n by the right
13056 /* if we get here, we had success: make a scalar value from
13061 /* final misplaced underbar check */
13062 if (s[-1] == '_') {
13063 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13068 if (n > 4294967295.0)
13069 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13070 "%s number > %s non-portable",
13076 if (u > 0xffffffff)
13077 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13078 "%s number > %s non-portable",
13083 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
13084 sv = new_constant(start, s - start, "integer",
13085 sv, NULL, NULL, 0);
13086 else if (PL_hints & HINT_NEW_BINARY)
13087 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
13092 handle decimal numbers.
13093 we're also sent here when we read a 0 as the first digit
13095 case '1': case '2': case '3': case '4': case '5':
13096 case '6': case '7': case '8': case '9': case '.':
13099 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
13102 /* read next group of digits and _ and copy into d */
13103 while (isDIGIT(*s) || *s == '_') {
13104 /* skip underscores, checking for misplaced ones
13108 if (lastub && s == lastub + 1)
13109 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13110 "Misplaced _ in number");
13114 /* check for end of fixed-length buffer */
13116 Perl_croak(aTHX_ number_too_long);
13117 /* if we're ok, copy the character */
13122 /* final misplaced underbar check */
13123 if (lastub && s == lastub + 1) {
13124 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13127 /* read a decimal portion if there is one. avoid
13128 3..5 being interpreted as the number 3. followed
13131 if (*s == '.' && s[1] != '.') {
13136 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13137 "Misplaced _ in number");
13141 /* copy, ignoring underbars, until we run out of digits.
13143 for (; isDIGIT(*s) || *s == '_'; s++) {
13144 /* fixed length buffer check */
13146 Perl_croak(aTHX_ number_too_long);
13148 if (lastub && s == lastub + 1)
13149 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13150 "Misplaced _ in number");
13156 /* fractional part ending in underbar? */
13157 if (s[-1] == '_') {
13158 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13159 "Misplaced _ in number");
13161 if (*s == '.' && isDIGIT(s[1])) {
13162 /* oops, it's really a v-string, but without the "v" */
13168 /* read exponent part, if present */
13169 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
13173 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
13174 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
13176 /* stray preinitial _ */
13178 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13179 "Misplaced _ in number");
13183 /* allow positive or negative exponent */
13184 if (*s == '+' || *s == '-')
13187 /* stray initial _ */
13189 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13190 "Misplaced _ in number");
13194 /* read digits of exponent */
13195 while (isDIGIT(*s) || *s == '_') {
13198 Perl_croak(aTHX_ number_too_long);
13202 if (((lastub && s == lastub + 1) ||
13203 (!isDIGIT(s[1]) && s[1] != '_')))
13204 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13205 "Misplaced _ in number");
13212 /* make an sv from the string */
13216 We try to do an integer conversion first if no characters
13217 indicating "float" have been found.
13222 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
13224 if (flags == IS_NUMBER_IN_UV) {
13226 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
13229 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
13230 if (uv <= (UV) IV_MIN)
13231 sv_setiv(sv, -(IV)uv);
13238 /* terminate the string */
13240 nv = Atof(PL_tokenbuf);
13245 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
13246 const char *const key = floatit ? "float" : "integer";
13247 const STRLEN keylen = floatit ? 5 : 7;
13248 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
13249 key, keylen, sv, NULL, NULL, 0);
13253 /* if it starts with a v, it could be a v-string */
13256 sv = newSV(5); /* preallocate storage space */
13257 s = scan_vstring(s, PL_bufend, sv);
13261 /* make the op for the constant and return */
13264 lvalp->opval = newSVOP(OP_CONST, 0, sv);
13266 lvalp->opval = NULL;
13272 S_scan_formline(pTHX_ register char *s)
13275 register char *eol;
13277 SV * const stuff = newSVpvs("");
13278 bool needargs = FALSE;
13279 bool eofmt = FALSE;
13281 char *tokenstart = s;
13282 SV* savewhite = NULL;
13284 if (PL_madskills) {
13285 savewhite = PL_thiswhite;
13290 PERL_ARGS_ASSERT_SCAN_FORMLINE;
13292 while (!needargs) {
13295 #ifdef PERL_STRICT_CR
13296 while (SPACE_OR_TAB(*t))
13299 while (SPACE_OR_TAB(*t) || *t == '\r')
13302 if (*t == '\n' || t == PL_bufend) {
13307 if (PL_in_eval && !PL_rsfp) {
13308 eol = (char *) memchr(s,'\n',PL_bufend-s);
13313 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
13315 for (t = s; t < eol; t++) {
13316 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
13318 goto enough; /* ~~ must be first line in formline */
13320 if (*t == '@' || *t == '^')
13324 sv_catpvn(stuff, s, eol-s);
13325 #ifndef PERL_STRICT_CR
13326 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
13327 char *end = SvPVX(stuff) + SvCUR(stuff);
13330 SvCUR_set(stuff, SvCUR(stuff) - 1);
13341 if (PL_madskills) {
13343 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
13345 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
13348 PL_bufptr = PL_bufend;
13349 CopLINE_inc(PL_curcop);
13350 got_some = lex_next_chunk(0);
13351 CopLINE_dec(PL_curcop);
13354 tokenstart = PL_bufptr;
13362 if (SvCUR(stuff)) {
13365 PL_lex_state = LEX_NORMAL;
13366 start_force(PL_curforce);
13367 NEXTVAL_NEXTTOKE.ival = 0;
13371 PL_lex_state = LEX_FORMLINE;
13373 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
13375 else if (PL_encoding)
13376 sv_recode_to_utf8(stuff, PL_encoding);
13378 start_force(PL_curforce);
13379 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
13381 start_force(PL_curforce);
13382 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
13386 SvREFCNT_dec(stuff);
13388 PL_lex_formbrack = 0;
13392 if (PL_madskills) {
13394 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
13396 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
13397 PL_thiswhite = savewhite;
13404 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
13407 const I32 oldsavestack_ix = PL_savestack_ix;
13408 CV* const outsidecv = PL_compcv;
13411 assert(SvTYPE(PL_compcv) == SVt_PVCV);
13413 SAVEI32(PL_subline);
13414 save_item(PL_subname);
13415 SAVESPTR(PL_compcv);
13417 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
13418 CvFLAGS(PL_compcv) |= flags;
13420 PL_subline = CopLINE(PL_curcop);
13421 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
13422 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
13423 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
13425 return oldsavestack_ix;
13429 #pragma segment Perl_yylex
13432 S_yywarn(pTHX_ const char *const s)
13436 PERL_ARGS_ASSERT_YYWARN;
13438 PL_in_eval |= EVAL_WARNONLY;
13440 PL_in_eval &= ~EVAL_WARNONLY;
13445 Perl_yyerror(pTHX_ const char *const s)
13448 const char *where = NULL;
13449 const char *context = NULL;
13452 int yychar = PL_parser->yychar;
13454 PERL_ARGS_ASSERT_YYERROR;
13456 if (!yychar || (yychar == ';' && !PL_rsfp))
13458 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
13459 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
13460 PL_oldbufptr != PL_bufptr) {
13463 The code below is removed for NetWare because it abends/crashes on NetWare
13464 when the script has error such as not having the closing quotes like:
13465 if ($var eq "value)
13466 Checking of white spaces is anyway done in NetWare code.
13469 while (isSPACE(*PL_oldoldbufptr))
13472 context = PL_oldoldbufptr;
13473 contlen = PL_bufptr - PL_oldoldbufptr;
13475 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
13476 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
13479 The code below is removed for NetWare because it abends/crashes on NetWare
13480 when the script has error such as not having the closing quotes like:
13481 if ($var eq "value)
13482 Checking of white spaces is anyway done in NetWare code.
13485 while (isSPACE(*PL_oldbufptr))
13488 context = PL_oldbufptr;
13489 contlen = PL_bufptr - PL_oldbufptr;
13491 else if (yychar > 255)
13492 where = "next token ???";
13493 else if (yychar == -2) { /* YYEMPTY */
13494 if (PL_lex_state == LEX_NORMAL ||
13495 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
13496 where = "at end of line";
13497 else if (PL_lex_inpat)
13498 where = "within pattern";
13500 where = "within string";
13503 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
13505 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
13506 else if (isPRINT_LC(yychar)) {
13507 const char string = yychar;
13508 sv_catpvn(where_sv, &string, 1);
13511 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
13512 where = SvPVX_const(where_sv);
13514 msg = sv_2mortal(newSVpv(s, 0));
13515 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
13516 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
13518 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
13520 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
13521 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
13522 Perl_sv_catpvf(aTHX_ msg,
13523 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
13524 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
13527 if (PL_in_eval & EVAL_WARNONLY) {
13528 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
13532 if (PL_error_count >= 10) {
13533 if (PL_in_eval && SvCUR(ERRSV))
13534 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
13535 SVfARG(ERRSV), OutCopFILE(PL_curcop));
13537 Perl_croak(aTHX_ "%s has too many errors.\n",
13538 OutCopFILE(PL_curcop));
13541 PL_in_my_stash = NULL;
13545 #pragma segment Main
13549 S_swallow_bom(pTHX_ U8 *s)
13552 const STRLEN slen = SvCUR(PL_linestr);
13554 PERL_ARGS_ASSERT_SWALLOW_BOM;
13558 if (s[1] == 0xFE) {
13559 /* UTF-16 little-endian? (or UTF-32LE?) */
13560 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
13561 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
13562 #ifndef PERL_NO_UTF16_FILTER
13563 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
13565 if (PL_bufend > (char*)s) {
13566 s = add_utf16_textfilter(s, TRUE);
13569 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13574 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
13575 #ifndef PERL_NO_UTF16_FILTER
13576 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
13578 if (PL_bufend > (char *)s) {
13579 s = add_utf16_textfilter(s, FALSE);
13582 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13587 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
13588 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13589 s += 3; /* UTF-8 */
13595 if (s[2] == 0xFE && s[3] == 0xFF) {
13596 /* UTF-32 big-endian */
13597 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
13600 else if (s[2] == 0 && s[3] != 0) {
13603 * are a good indicator of UTF-16BE. */
13604 #ifndef PERL_NO_UTF16_FILTER
13605 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
13606 s = add_utf16_textfilter(s, FALSE);
13608 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13614 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
13615 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13616 s += 4; /* UTF-8 */
13622 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
13625 * are a good indicator of UTF-16LE. */
13626 #ifndef PERL_NO_UTF16_FILTER
13627 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
13628 s = add_utf16_textfilter(s, TRUE);
13630 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13638 #ifndef PERL_NO_UTF16_FILTER
13640 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
13643 SV *const filter = FILTER_DATA(idx);
13644 /* We re-use this each time round, throwing the contents away before we
13646 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
13647 SV *const utf8_buffer = filter;
13648 IV status = IoPAGE(filter);
13649 const bool reverse = (bool) IoLINES(filter);
13652 /* As we're automatically added, at the lowest level, and hence only called
13653 from this file, we can be sure that we're not called in block mode. Hence
13654 don't bother writing code to deal with block mode. */
13656 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
13659 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
13661 DEBUG_P(PerlIO_printf(Perl_debug_log,
13662 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13663 FPTR2DPTR(void *, S_utf16_textfilter),
13664 reverse ? 'l' : 'b', idx, maxlen, status,
13665 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13672 /* First, look in our buffer of existing UTF-8 data: */
13673 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
13677 } else if (status == 0) {
13679 IoPAGE(filter) = 0;
13680 nl = SvEND(utf8_buffer);
13683 STRLEN got = nl - SvPVX(utf8_buffer);
13684 /* Did we have anything to append? */
13686 sv_catpvn(sv, SvPVX(utf8_buffer), got);
13687 /* Everything else in this code works just fine if SVp_POK isn't
13688 set. This, however, needs it, and we need it to work, else
13689 we loop infinitely because the buffer is never consumed. */
13690 sv_chop(utf8_buffer, nl);
13694 /* OK, not a complete line there, so need to read some more UTF-16.
13695 Read an extra octect if the buffer currently has an odd number. */
13699 if (SvCUR(utf16_buffer) >= 2) {
13700 /* Location of the high octet of the last complete code point.
13701 Gosh, UTF-16 is a pain. All the benefits of variable length,
13702 *coupled* with all the benefits of partial reads and
13704 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
13705 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
13707 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
13711 /* We have the first half of a surrogate. Read more. */
13712 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
13715 status = FILTER_READ(idx + 1, utf16_buffer,
13716 160 + (SvCUR(utf16_buffer) & 1));
13717 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
13718 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
13721 IoPAGE(filter) = status;
13726 chars = SvCUR(utf16_buffer) >> 1;
13727 have = SvCUR(utf8_buffer);
13728 SvGROW(utf8_buffer, have + chars * 3 + 1);
13731 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
13732 (U8*)SvPVX_const(utf8_buffer) + have,
13733 chars * 2, &newlen);
13735 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
13736 (U8*)SvPVX_const(utf8_buffer) + have,
13737 chars * 2, &newlen);
13739 SvCUR_set(utf8_buffer, have + newlen);
13742 /* No need to keep this SV "well-formed" with a '\0' after the end, as
13743 it's private to us, and utf16_to_utf8{,reversed} take a
13744 (pointer,length) pair, rather than a NUL-terminated string. */
13745 if(SvCUR(utf16_buffer) & 1) {
13746 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
13747 SvCUR_set(utf16_buffer, 1);
13749 SvCUR_set(utf16_buffer, 0);
13752 DEBUG_P(PerlIO_printf(Perl_debug_log,
13753 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13755 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13756 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
13761 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
13763 SV *filter = filter_add(S_utf16_textfilter, NULL);
13765 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
13766 sv_setpvs(filter, "");
13767 IoLINES(filter) = reversed;
13768 IoPAGE(filter) = 1; /* Not EOF */
13770 /* Sadly, we have to return a valid pointer, come what may, so we have to
13771 ignore any error return from this. */
13772 SvCUR_set(PL_linestr, 0);
13773 if (FILTER_READ(0, PL_linestr, 0)) {
13774 SvUTF8_on(PL_linestr);
13776 SvUTF8_on(PL_linestr);
13778 PL_bufend = SvEND(PL_linestr);
13779 return (U8*)SvPVX(PL_linestr);
13784 Returns a pointer to the next character after the parsed
13785 vstring, as well as updating the passed in sv.
13787 Function must be called like
13790 s = scan_vstring(s,e,sv);
13792 where s and e are the start and end of the string.
13793 The sv should already be large enough to store the vstring
13794 passed in, for performance reasons.
13799 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
13802 const char *pos = s;
13803 const char *start = s;
13805 PERL_ARGS_ASSERT_SCAN_VSTRING;
13807 if (*pos == 'v') pos++; /* get past 'v' */
13808 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13810 if ( *pos != '.') {
13811 /* this may not be a v-string if followed by => */
13812 const char *next = pos;
13813 while (next < e && isSPACE(*next))
13815 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
13816 /* return string not v-string */
13817 sv_setpvn(sv,(char *)s,pos-s);
13818 return (char *)pos;
13822 if (!isALPHA(*pos)) {
13823 U8 tmpbuf[UTF8_MAXBYTES+1];
13826 s++; /* get past 'v' */
13831 /* this is atoi() that tolerates underscores */
13834 const char *end = pos;
13836 while (--end >= s) {
13838 const UV orev = rev;
13839 rev += (*end - '0') * mult;
13842 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13843 "Integer overflow in decimal number");
13847 if (rev > 0x7FFFFFFF)
13848 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
13850 /* Append native character for the rev point */
13851 tmpend = uvchr_to_utf8(tmpbuf, rev);
13852 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
13853 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
13855 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
13861 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13865 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
13872 Perl_keyword_plugin_standard(pTHX_
13873 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
13875 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
13876 PERL_UNUSED_CONTEXT;
13877 PERL_UNUSED_ARG(keyword_ptr);
13878 PERL_UNUSED_ARG(keyword_len);
13879 PERL_UNUSED_ARG(op_ptr);
13880 return KEYWORD_PLUGIN_DECLINE;
13885 * c-indentation-style: bsd
13886 * c-basic-offset: 4
13887 * indent-tabs-mode: t
13890 * ex: set ts=8 sts=4 sw=4 noet: