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)
943 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
944 if (flags & ~(LEX_STUFF_UTF8))
945 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
947 if (flags & LEX_STUFF_UTF8) {
951 char *p, *e = pv+len;
952 for (p = pv; p != e; p++)
953 highhalf += !!(((U8)*p) & 0x80);
956 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
957 bufptr = PL_parser->bufptr;
958 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
959 PL_parser->bufend += len+highhalf;
960 for (p = pv; p != e; p++) {
963 *bufptr++ = (char)(0xc0 | (c >> 6));
964 *bufptr++ = (char)(0x80 | (c & 0x3f));
971 if (flags & LEX_STUFF_UTF8) {
973 char *p, *e = pv+len;
974 for (p = pv; p != e; p++) {
977 Perl_croak(aTHX_ "Lexing code attempted to stuff "
978 "non-Latin-1 character into Latin-1 input");
979 } else if (c >= 0xc2 && p+1 != e &&
980 (((U8)p[1]) & 0xc0) == 0x80) {
983 } else if (c >= 0x80) {
984 /* malformed UTF-8 */
986 SAVESPTR(PL_warnhook);
987 PL_warnhook = PERL_WARNHOOK_FATAL;
988 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
994 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
995 bufptr = PL_parser->bufptr;
996 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
997 PL_parser->bufend += len-highhalf;
998 for (p = pv; p != e; p++) {
1001 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1004 *bufptr++ = (char)c;
1009 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1010 bufptr = PL_parser->bufptr;
1011 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1012 PL_parser->bufend += len;
1013 Copy(pv, bufptr, len, char);
1019 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1021 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1022 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1023 reallocating the buffer if necessary. This means that lexing code that
1024 runs later will see the characters as if they had appeared in the input.
1025 It is not recommended to do this as part of normal parsing, and most
1026 uses of this facility run the risk of the inserted characters being
1027 interpreted in an unintended manner.
1029 The string to be inserted is the string value of I<sv>. The characters
1030 are recoded for the lexer buffer, according to how the buffer is currently
1031 being interpreted (L</lex_bufutf8>). If a string to be interpreted is
1032 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1033 need to construct a scalar.
1039 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1043 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1045 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1047 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1051 =for apidoc Amx|void|lex_unstuff|char *ptr
1053 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1054 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1055 This hides the discarded text from any lexing code that runs later,
1056 as if the text had never appeared.
1058 This is not the normal way to consume lexed text. For that, use
1065 Perl_lex_unstuff(pTHX_ char *ptr)
1069 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1070 buf = PL_parser->bufptr;
1072 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1075 bufend = PL_parser->bufend;
1077 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1078 unstuff_len = ptr - buf;
1079 Move(ptr, buf, bufend+1-ptr, char);
1080 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1081 PL_parser->bufend = bufend - unstuff_len;
1085 =for apidoc Amx|void|lex_read_to|char *ptr
1087 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1088 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1089 performing the correct bookkeeping whenever a newline character is passed.
1090 This is the normal way to consume lexed text.
1092 Interpretation of the buffer's octets can be abstracted out by
1093 using the slightly higher-level functions L</lex_peek_unichar> and
1094 L</lex_read_unichar>.
1100 Perl_lex_read_to(pTHX_ char *ptr)
1103 PERL_ARGS_ASSERT_LEX_READ_TO;
1104 s = PL_parser->bufptr;
1105 if (ptr < s || ptr > PL_parser->bufend)
1106 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1107 for (; s != ptr; s++)
1109 CopLINE_inc(PL_curcop);
1110 PL_parser->linestart = s+1;
1112 PL_parser->bufptr = ptr;
1116 =for apidoc Amx|void|lex_discard_to|char *ptr
1118 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1119 up to I<ptr>. The remaining content of the buffer will be moved, and
1120 all pointers into the buffer updated appropriately. I<ptr> must not
1121 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1122 it is not permitted to discard text that has yet to be lexed.
1124 Normally it is not necessarily to do this directly, because it suffices to
1125 use the implicit discarding behaviour of L</lex_next_chunk> and things
1126 based on it. However, if a token stretches across multiple lines,
1127 and the lexing code has kept multiple lines of text in the buffer fof
1128 that purpose, then after completion of the token it would be wise to
1129 explicitly discard the now-unneeded earlier lines, to avoid future
1130 multi-line tokens growing the buffer without bound.
1136 Perl_lex_discard_to(pTHX_ char *ptr)
1140 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1141 buf = SvPVX(PL_parser->linestr);
1143 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1146 if (ptr > PL_parser->bufptr)
1147 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1148 discard_len = ptr - buf;
1149 if (PL_parser->oldbufptr < ptr)
1150 PL_parser->oldbufptr = ptr;
1151 if (PL_parser->oldoldbufptr < ptr)
1152 PL_parser->oldoldbufptr = ptr;
1153 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1154 PL_parser->last_uni = NULL;
1155 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1156 PL_parser->last_lop = NULL;
1157 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1158 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1159 PL_parser->bufend -= discard_len;
1160 PL_parser->bufptr -= discard_len;
1161 PL_parser->oldbufptr -= discard_len;
1162 PL_parser->oldoldbufptr -= discard_len;
1163 if (PL_parser->last_uni)
1164 PL_parser->last_uni -= discard_len;
1165 if (PL_parser->last_lop)
1166 PL_parser->last_lop -= discard_len;
1170 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1172 Reads in the next chunk of text to be lexed, appending it to
1173 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1174 looked to the end of the current chunk and wants to know more. It is
1175 usual, but not necessary, for lexing to have consumed the entirety of
1176 the current chunk at this time.
1178 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1179 chunk (i.e., the current chunk has been entirely consumed), normally the
1180 current chunk will be discarded at the same time that the new chunk is
1181 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1182 will not be discarded. If the current chunk has not been entirely
1183 consumed, then it will not be discarded regardless of the flag.
1185 Returns true if some new text was added to the buffer, or false if the
1186 buffer has reached the end of the input text.
1191 #define LEX_FAKE_EOF 0x80000000
1194 Perl_lex_next_chunk(pTHX_ U32 flags)
1198 STRLEN old_bufend_pos, new_bufend_pos;
1199 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1200 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1201 bool got_some_for_debugger = 0;
1203 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1204 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1205 linestr = PL_parser->linestr;
1206 buf = SvPVX(linestr);
1207 if (!(flags & LEX_KEEP_PREVIOUS) &&
1208 PL_parser->bufptr == PL_parser->bufend) {
1209 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1211 if (PL_parser->last_uni != PL_parser->bufend)
1212 PL_parser->last_uni = NULL;
1213 if (PL_parser->last_lop != PL_parser->bufend)
1214 PL_parser->last_lop = NULL;
1215 last_uni_pos = last_lop_pos = 0;
1219 old_bufend_pos = PL_parser->bufend - buf;
1220 bufptr_pos = PL_parser->bufptr - buf;
1221 oldbufptr_pos = PL_parser->oldbufptr - buf;
1222 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1223 linestart_pos = PL_parser->linestart - buf;
1224 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1225 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1227 if (flags & LEX_FAKE_EOF) {
1229 } else if (!PL_parser->rsfp) {
1231 } else if (filter_gets(linestr, old_bufend_pos)) {
1233 got_some_for_debugger = 1;
1235 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1236 sv_setpvs(linestr, "");
1238 /* End of real input. Close filehandle (unless it was STDIN),
1239 * then add implicit termination.
1241 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1242 PerlIO_clearerr(PL_parser->rsfp);
1243 else if (PL_parser->rsfp)
1244 (void)PerlIO_close(PL_parser->rsfp);
1245 PL_parser->rsfp = NULL;
1246 PL_doextract = FALSE;
1248 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1251 if (!PL_in_eval && PL_minus_p) {
1253 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1254 PL_minus_n = PL_minus_p = 0;
1255 } else if (!PL_in_eval && PL_minus_n) {
1256 sv_catpvs(linestr, /*{*/";}");
1259 sv_catpvs(linestr, ";");
1262 buf = SvPVX(linestr);
1263 new_bufend_pos = SvCUR(linestr);
1264 PL_parser->bufend = buf + new_bufend_pos;
1265 PL_parser->bufptr = buf + bufptr_pos;
1266 PL_parser->oldbufptr = buf + oldbufptr_pos;
1267 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1268 PL_parser->linestart = buf + linestart_pos;
1269 if (PL_parser->last_uni)
1270 PL_parser->last_uni = buf + last_uni_pos;
1271 if (PL_parser->last_lop)
1272 PL_parser->last_lop = buf + last_lop_pos;
1273 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1274 PL_curstash != PL_debstash) {
1275 /* debugger active and we're not compiling the debugger code,
1276 * so store the line into the debugger's array of lines
1278 update_debugger_info(NULL, buf+old_bufend_pos,
1279 new_bufend_pos-old_bufend_pos);
1285 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1287 Looks ahead one (Unicode) character in the text currently being lexed.
1288 Returns the codepoint (unsigned integer value) of the next character,
1289 or -1 if lexing has reached the end of the input text. To consume the
1290 peeked character, use L</lex_read_unichar>.
1292 If the next character is in (or extends into) the next chunk of input
1293 text, the next chunk will be read in. Normally the current chunk will be
1294 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1295 then the current chunk will not be discarded.
1297 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1298 is encountered, an exception is generated.
1304 Perl_lex_peek_unichar(pTHX_ U32 flags)
1308 if (flags & ~(LEX_KEEP_PREVIOUS))
1309 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1310 s = PL_parser->bufptr;
1311 bufend = PL_parser->bufend;
1317 if (!lex_next_chunk(flags))
1319 s = PL_parser->bufptr;
1320 bufend = PL_parser->bufend;
1326 len = PL_utf8skip[head];
1327 while ((STRLEN)(bufend-s) < len) {
1328 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1330 s = PL_parser->bufptr;
1331 bufend = PL_parser->bufend;
1334 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1335 if (retlen == (STRLEN)-1) {
1336 /* malformed UTF-8 */
1338 SAVESPTR(PL_warnhook);
1339 PL_warnhook = PERL_WARNHOOK_FATAL;
1340 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1346 if (!lex_next_chunk(flags))
1348 s = PL_parser->bufptr;
1355 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1357 Reads the next (Unicode) character in the text currently being lexed.
1358 Returns the codepoint (unsigned integer value) of the character read,
1359 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1360 if lexing has reached the end of the input text. To non-destructively
1361 examine the next character, use L</lex_peek_unichar> instead.
1363 If the next character is in (or extends into) the next chunk of input
1364 text, the next chunk will be read in. Normally the current chunk will be
1365 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1366 then the current chunk will not be discarded.
1368 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1369 is encountered, an exception is generated.
1375 Perl_lex_read_unichar(pTHX_ U32 flags)
1378 if (flags & ~(LEX_KEEP_PREVIOUS))
1379 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1380 c = lex_peek_unichar(flags);
1383 CopLINE_inc(PL_curcop);
1384 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1390 =for apidoc Amx|void|lex_read_space|U32 flags
1392 Reads optional spaces, in Perl style, in the text currently being
1393 lexed. The spaces may include ordinary whitespace characters and
1394 Perl-style comments. C<#line> directives are processed if encountered.
1395 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1396 at a non-space character (or the end of the input text).
1398 If spaces extend into the next chunk of input text, the next chunk will
1399 be read in. Normally the current chunk will be discarded at the same
1400 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1401 chunk will not be discarded.
1406 #define LEX_NO_NEXT_CHUNK 0x80000000
1409 Perl_lex_read_space(pTHX_ U32 flags)
1412 bool need_incline = 0;
1413 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1414 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1417 sv_free(PL_skipwhite);
1418 PL_skipwhite = NULL;
1421 PL_skipwhite = newSVpvs("");
1422 #endif /* PERL_MAD */
1423 s = PL_parser->bufptr;
1424 bufend = PL_parser->bufend;
1430 } while (!(c == '\n' || (c == 0 && s == bufend)));
1431 } else if (c == '\n') {
1433 PL_parser->linestart = s;
1438 } else if (isSPACE(c)) {
1440 } else if (c == 0 && s == bufend) {
1444 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1445 #endif /* PERL_MAD */
1446 if (flags & LEX_NO_NEXT_CHUNK)
1448 PL_parser->bufptr = s;
1449 CopLINE_inc(PL_curcop);
1450 got_more = lex_next_chunk(flags);
1451 CopLINE_dec(PL_curcop);
1452 s = PL_parser->bufptr;
1453 bufend = PL_parser->bufend;
1456 if (need_incline && PL_parser->rsfp) {
1466 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1467 #endif /* PERL_MAD */
1468 PL_parser->bufptr = s;
1473 * This subroutine has nothing to do with tilting, whether at windmills
1474 * or pinball tables. Its name is short for "increment line". It
1475 * increments the current line number in CopLINE(PL_curcop) and checks
1476 * to see whether the line starts with a comment of the form
1477 * # line 500 "foo.pm"
1478 * If so, it sets the current line number and file to the values in the comment.
1482 S_incline(pTHX_ const char *s)
1489 PERL_ARGS_ASSERT_INCLINE;
1491 CopLINE_inc(PL_curcop);
1494 while (SPACE_OR_TAB(*s))
1496 if (strnEQ(s, "line", 4))
1500 if (SPACE_OR_TAB(*s))
1504 while (SPACE_OR_TAB(*s))
1512 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1514 while (SPACE_OR_TAB(*s))
1516 if (*s == '"' && (t = strchr(s+1, '"'))) {
1522 while (!isSPACE(*t))
1526 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1528 if (*e != '\n' && *e != '\0')
1529 return; /* false alarm */
1532 const STRLEN len = t - s;
1533 #ifndef USE_ITHREADS
1534 SV *const temp_sv = CopFILESV(PL_curcop);
1539 cf = SvPVX(temp_sv);
1540 tmplen = SvCUR(temp_sv);
1546 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
1547 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1548 * to *{"::_<newfilename"} */
1549 /* However, the long form of evals is only turned on by the
1550 debugger - usually they're "(eval %lu)" */
1554 STRLEN tmplen2 = len;
1555 if (tmplen + 2 <= sizeof smallbuf)
1558 Newx(tmpbuf, tmplen + 2, char);
1561 memcpy(tmpbuf + 2, cf, tmplen);
1563 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1568 if (tmplen2 + 2 <= sizeof smallbuf)
1571 Newx(tmpbuf2, tmplen2 + 2, char);
1573 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1574 /* Either they malloc'd it, or we malloc'd it,
1575 so no prefix is present in ours. */
1580 memcpy(tmpbuf2 + 2, s, tmplen2);
1583 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1585 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1586 /* adjust ${"::_<newfilename"} to store the new file name */
1587 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1588 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1589 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1592 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1594 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1597 CopFILE_free(PL_curcop);
1598 CopFILE_setn(PL_curcop, s, len);
1600 CopLINE_set(PL_curcop, atoi(n)-1);
1604 /* skip space before PL_thistoken */
1607 S_skipspace0(pTHX_ register char *s)
1609 PERL_ARGS_ASSERT_SKIPSPACE0;
1616 PL_thiswhite = newSVpvs("");
1617 sv_catsv(PL_thiswhite, PL_skipwhite);
1618 sv_free(PL_skipwhite);
1621 PL_realtokenstart = s - SvPVX(PL_linestr);
1625 /* skip space after PL_thistoken */
1628 S_skipspace1(pTHX_ register char *s)
1630 const char *start = s;
1631 I32 startoff = start - SvPVX(PL_linestr);
1633 PERL_ARGS_ASSERT_SKIPSPACE1;
1638 start = SvPVX(PL_linestr) + startoff;
1639 if (!PL_thistoken && PL_realtokenstart >= 0) {
1640 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1641 PL_thistoken = newSVpvn(tstart, start - tstart);
1643 PL_realtokenstart = -1;
1646 PL_nextwhite = newSVpvs("");
1647 sv_catsv(PL_nextwhite, PL_skipwhite);
1648 sv_free(PL_skipwhite);
1655 S_skipspace2(pTHX_ register char *s, SV **svp)
1658 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1659 const I32 startoff = s - SvPVX(PL_linestr);
1661 PERL_ARGS_ASSERT_SKIPSPACE2;
1664 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1665 if (!PL_madskills || !svp)
1667 start = SvPVX(PL_linestr) + startoff;
1668 if (!PL_thistoken && PL_realtokenstart >= 0) {
1669 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1670 PL_thistoken = newSVpvn(tstart, start - tstart);
1671 PL_realtokenstart = -1;
1675 *svp = newSVpvs("");
1676 sv_setsv(*svp, PL_skipwhite);
1677 sv_free(PL_skipwhite);
1686 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1688 AV *av = CopFILEAVx(PL_curcop);
1690 SV * const sv = newSV_type(SVt_PVMG);
1692 sv_setsv(sv, orig_sv);
1694 sv_setpvn(sv, buf, len);
1697 av_store(av, (I32)CopLINE(PL_curcop), sv);
1703 * Called to gobble the appropriate amount and type of whitespace.
1704 * Skips comments as well.
1708 S_skipspace(pTHX_ register char *s)
1712 #endif /* PERL_MAD */
1713 PERL_ARGS_ASSERT_SKIPSPACE;
1716 sv_free(PL_skipwhite);
1717 PL_skipwhite = NULL;
1719 #endif /* PERL_MAD */
1720 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1721 while (s < PL_bufend && SPACE_OR_TAB(*s))
1724 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1726 lex_read_space(LEX_KEEP_PREVIOUS |
1727 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1728 LEX_NO_NEXT_CHUNK : 0));
1730 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1731 if (PL_linestart > PL_bufptr)
1732 PL_bufptr = PL_linestart;
1737 PL_skipwhite = newSVpvn(start, s-start);
1738 #endif /* PERL_MAD */
1744 * Check the unary operators to ensure there's no ambiguity in how they're
1745 * used. An ambiguous piece of code would be:
1747 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1748 * the +5 is its argument.
1758 if (PL_oldoldbufptr != PL_last_uni)
1760 while (isSPACE(*PL_last_uni))
1763 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1765 if ((t = strchr(s, '(')) && t < PL_bufptr)
1768 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1769 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1770 (int)(s - PL_last_uni), PL_last_uni);
1774 * LOP : macro to build a list operator. Its behaviour has been replaced
1775 * with a subroutine, S_lop() for which LOP is just another name.
1778 #define LOP(f,x) return lop(f,x,s)
1782 * Build a list operator (or something that might be one). The rules:
1783 * - if we have a next token, then it's a list operator [why?]
1784 * - if the next thing is an opening paren, then it's a function
1785 * - else it's a list operator
1789 S_lop(pTHX_ I32 f, int x, char *s)
1793 PERL_ARGS_ASSERT_LOP;
1799 PL_last_lop = PL_oldbufptr;
1800 PL_last_lop_op = (OPCODE)f;
1803 return REPORT(LSTOP);
1806 return REPORT(LSTOP);
1809 return REPORT(FUNC);
1812 return REPORT(FUNC);
1814 return REPORT(LSTOP);
1820 * Sets up for an eventual force_next(). start_force(0) basically does
1821 * an unshift, while start_force(-1) does a push. yylex removes items
1826 S_start_force(pTHX_ int where)
1830 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1831 where = PL_lasttoke;
1832 assert(PL_curforce < 0 || PL_curforce == where);
1833 if (PL_curforce != where) {
1834 for (i = PL_lasttoke; i > where; --i) {
1835 PL_nexttoke[i] = PL_nexttoke[i-1];
1839 if (PL_curforce < 0) /* in case of duplicate start_force() */
1840 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1841 PL_curforce = where;
1844 curmad('^', newSVpvs(""));
1845 CURMAD('_', PL_nextwhite);
1850 S_curmad(pTHX_ char slot, SV *sv)
1856 if (PL_curforce < 0)
1857 where = &PL_thismad;
1859 where = &PL_nexttoke[PL_curforce].next_mad;
1865 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1867 else if (PL_encoding) {
1868 sv_recode_to_utf8(sv, PL_encoding);
1873 /* keep a slot open for the head of the list? */
1874 if (slot != '_' && *where && (*where)->mad_key == '^') {
1875 (*where)->mad_key = slot;
1876 sv_free(MUTABLE_SV(((*where)->mad_val)));
1877 (*where)->mad_val = (void*)sv;
1880 addmad(newMADsv(slot, sv), where, 0);
1883 # define start_force(where) NOOP
1884 # define curmad(slot, sv) NOOP
1889 * When the lexer realizes it knows the next token (for instance,
1890 * it is reordering tokens for the parser) then it can call S_force_next
1891 * to know what token to return the next time the lexer is called. Caller
1892 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1893 * and possibly PL_expect to ensure the lexer handles the token correctly.
1897 S_force_next(pTHX_ I32 type)
1902 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1903 tokereport(type, &NEXTVAL_NEXTTOKE);
1907 if (PL_curforce < 0)
1908 start_force(PL_lasttoke);
1909 PL_nexttoke[PL_curforce].next_type = type;
1910 if (PL_lex_state != LEX_KNOWNEXT)
1911 PL_lex_defer = PL_lex_state;
1912 PL_lex_state = LEX_KNOWNEXT;
1913 PL_lex_expect = PL_expect;
1916 PL_nexttype[PL_nexttoke] = type;
1918 if (PL_lex_state != LEX_KNOWNEXT) {
1919 PL_lex_defer = PL_lex_state;
1920 PL_lex_expect = PL_expect;
1921 PL_lex_state = LEX_KNOWNEXT;
1927 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
1930 SV * const sv = newSVpvn_utf8(start, len,
1933 && !is_ascii_string((const U8*)start, len)
1934 && is_utf8_string((const U8*)start, len));
1940 * When the lexer knows the next thing is a word (for instance, it has
1941 * just seen -> and it knows that the next char is a word char, then
1942 * it calls S_force_word to stick the next word into the PL_nexttoke/val
1946 * char *start : buffer position (must be within PL_linestr)
1947 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
1948 * int check_keyword : if true, Perl checks to make sure the word isn't
1949 * a keyword (do this if the word is a label, e.g. goto FOO)
1950 * int allow_pack : if true, : characters will also be allowed (require,
1951 * use, etc. do this)
1952 * int allow_initial_tick : used by the "sub" lexer only.
1956 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
1962 PERL_ARGS_ASSERT_FORCE_WORD;
1964 start = SKIPSPACE1(start);
1966 if (isIDFIRST_lazy_if(s,UTF) ||
1967 (allow_pack && *s == ':') ||
1968 (allow_initial_tick && *s == '\'') )
1970 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
1971 if (check_keyword && keyword(PL_tokenbuf, len, 0))
1973 start_force(PL_curforce);
1975 curmad('X', newSVpvn(start,s-start));
1976 if (token == METHOD) {
1981 PL_expect = XOPERATOR;
1985 curmad('g', newSVpvs( "forced" ));
1986 NEXTVAL_NEXTTOKE.opval
1987 = (OP*)newSVOP(OP_CONST,0,
1988 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
1989 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
1997 * Called when the lexer wants $foo *foo &foo etc, but the program
1998 * text only contains the "foo" portion. The first argument is a pointer
1999 * to the "foo", and the second argument is the type symbol to prefix.
2000 * Forces the next token to be a "WORD".
2001 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2005 S_force_ident(pTHX_ register const char *s, int kind)
2009 PERL_ARGS_ASSERT_FORCE_IDENT;
2012 const STRLEN len = strlen(s);
2013 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
2014 start_force(PL_curforce);
2015 NEXTVAL_NEXTTOKE.opval = o;
2018 o->op_private = OPpCONST_ENTERED;
2019 /* XXX see note in pp_entereval() for why we forgo typo
2020 warnings if the symbol must be introduced in an eval.
2022 gv_fetchpvn_flags(s, len,
2023 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2025 kind == '$' ? SVt_PV :
2026 kind == '@' ? SVt_PVAV :
2027 kind == '%' ? SVt_PVHV :
2035 Perl_str_to_version(pTHX_ SV *sv)
2040 const char *start = SvPV_const(sv,len);
2041 const char * const end = start + len;
2042 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2044 PERL_ARGS_ASSERT_STR_TO_VERSION;
2046 while (start < end) {
2050 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2055 retval += ((NV)n)/nshift;
2064 * Forces the next token to be a version number.
2065 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2066 * and if "guessing" is TRUE, then no new token is created (and the caller
2067 * must use an alternative parsing method).
2071 S_force_version(pTHX_ char *s, int guessing)
2077 I32 startoff = s - SvPVX(PL_linestr);
2080 PERL_ARGS_ASSERT_FORCE_VERSION;
2088 while (isDIGIT(*d) || *d == '_' || *d == '.')
2092 start_force(PL_curforce);
2093 curmad('X', newSVpvn(s,d-s));
2096 if (*d == ';' || isSPACE(*d) || *d == '}' || !*d) {
2098 #ifdef USE_LOCALE_NUMERIC
2099 char *loc = setlocale(LC_NUMERIC, "C");
2101 s = scan_num(s, &pl_yylval);
2102 #ifdef USE_LOCALE_NUMERIC
2103 setlocale(LC_NUMERIC, loc);
2105 version = pl_yylval.opval;
2106 ver = cSVOPx(version)->op_sv;
2107 if (SvPOK(ver) && !SvNIOK(ver)) {
2108 SvUPGRADE(ver, SVt_PVNV);
2109 SvNV_set(ver, str_to_version(ver));
2110 SvNOK_on(ver); /* hint that it is a version */
2113 else if (guessing) {
2116 sv_free(PL_nextwhite); /* let next token collect whitespace */
2118 s = SvPVX(PL_linestr) + startoff;
2126 if (PL_madskills && !version) {
2127 sv_free(PL_nextwhite); /* let next token collect whitespace */
2129 s = SvPVX(PL_linestr) + startoff;
2132 /* NOTE: The parser sees the package name and the VERSION swapped */
2133 start_force(PL_curforce);
2134 NEXTVAL_NEXTTOKE.opval = version;
2141 * S_force_strict_version
2142 * Forces the next token to be a version number using strict syntax rules.
2146 S_force_strict_version(pTHX_ char *s)
2151 I32 startoff = s - SvPVX(PL_linestr);
2153 const char *errstr = NULL;
2155 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2157 while (isSPACE(*s)) /* leading whitespace */
2160 if (is_STRICT_VERSION(s,&errstr)) {
2162 s = (char *)scan_version(s, ver, 0);
2163 version = newSVOP(OP_CONST, 0, ver);
2165 else if ( (*s != ';' && *s != '}' ) && (s = SKIPSPACE1(s), (*s != ';' && *s !='}' ))) {
2168 yyerror(errstr); /* version required */
2173 if (PL_madskills && !version) {
2174 sv_free(PL_nextwhite); /* let next token collect whitespace */
2176 s = SvPVX(PL_linestr) + startoff;
2179 /* NOTE: The parser sees the package name and the VERSION swapped */
2180 start_force(PL_curforce);
2181 NEXTVAL_NEXTTOKE.opval = version;
2189 * Tokenize a quoted string passed in as an SV. It finds the next
2190 * chunk, up to end of string or a backslash. It may make a new
2191 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2196 S_tokeq(pTHX_ SV *sv)
2200 register char *send;
2205 PERL_ARGS_ASSERT_TOKEQ;
2210 s = SvPV_force(sv, len);
2211 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2214 while (s < send && *s != '\\')
2219 if ( PL_hints & HINT_NEW_STRING ) {
2220 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2224 if (s + 1 < send && (s[1] == '\\'))
2225 s++; /* all that, just for this */
2230 SvCUR_set(sv, d - SvPVX_const(sv));
2232 if ( PL_hints & HINT_NEW_STRING )
2233 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2238 * Now come three functions related to double-quote context,
2239 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2240 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2241 * interact with PL_lex_state, and create fake ( ... ) argument lists
2242 * to handle functions and concatenation.
2243 * They assume that whoever calls them will be setting up a fake
2244 * join call, because each subthing puts a ',' after it. This lets
2247 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2249 * (I'm not sure whether the spurious commas at the end of lcfirst's
2250 * arguments and join's arguments are created or not).
2255 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2257 * Pattern matching will set PL_lex_op to the pattern-matching op to
2258 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2260 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2262 * Everything else becomes a FUNC.
2264 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2265 * had an OP_CONST or OP_READLINE). This just sets us up for a
2266 * call to S_sublex_push().
2270 S_sublex_start(pTHX)
2273 register const I32 op_type = pl_yylval.ival;
2275 if (op_type == OP_NULL) {
2276 pl_yylval.opval = PL_lex_op;
2280 if (op_type == OP_CONST || op_type == OP_READLINE) {
2281 SV *sv = tokeq(PL_lex_stuff);
2283 if (SvTYPE(sv) == SVt_PVIV) {
2284 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2286 const char * const p = SvPV_const(sv, len);
2287 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2291 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2292 PL_lex_stuff = NULL;
2293 /* Allow <FH> // "foo" */
2294 if (op_type == OP_READLINE)
2295 PL_expect = XTERMORDORDOR;
2298 else if (op_type == OP_BACKTICK && PL_lex_op) {
2299 /* readpipe() vas overriden */
2300 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2301 pl_yylval.opval = PL_lex_op;
2303 PL_lex_stuff = NULL;
2307 PL_sublex_info.super_state = PL_lex_state;
2308 PL_sublex_info.sub_inwhat = (U16)op_type;
2309 PL_sublex_info.sub_op = PL_lex_op;
2310 PL_lex_state = LEX_INTERPPUSH;
2314 pl_yylval.opval = PL_lex_op;
2324 * Create a new scope to save the lexing state. The scope will be
2325 * ended in S_sublex_done. Returns a '(', starting the function arguments
2326 * to the uc, lc, etc. found before.
2327 * Sets PL_lex_state to LEX_INTERPCONCAT.
2336 PL_lex_state = PL_sublex_info.super_state;
2337 SAVEBOOL(PL_lex_dojoin);
2338 SAVEI32(PL_lex_brackets);
2339 SAVEI32(PL_lex_casemods);
2340 SAVEI32(PL_lex_starts);
2341 SAVEI8(PL_lex_state);
2342 SAVEVPTR(PL_lex_inpat);
2343 SAVEI16(PL_lex_inwhat);
2344 SAVECOPLINE(PL_curcop);
2345 SAVEPPTR(PL_bufptr);
2346 SAVEPPTR(PL_bufend);
2347 SAVEPPTR(PL_oldbufptr);
2348 SAVEPPTR(PL_oldoldbufptr);
2349 SAVEPPTR(PL_last_lop);
2350 SAVEPPTR(PL_last_uni);
2351 SAVEPPTR(PL_linestart);
2352 SAVESPTR(PL_linestr);
2353 SAVEGENERICPV(PL_lex_brackstack);
2354 SAVEGENERICPV(PL_lex_casestack);
2356 PL_linestr = PL_lex_stuff;
2357 PL_lex_stuff = NULL;
2359 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2360 = SvPVX(PL_linestr);
2361 PL_bufend += SvCUR(PL_linestr);
2362 PL_last_lop = PL_last_uni = NULL;
2363 SAVEFREESV(PL_linestr);
2365 PL_lex_dojoin = FALSE;
2366 PL_lex_brackets = 0;
2367 Newx(PL_lex_brackstack, 120, char);
2368 Newx(PL_lex_casestack, 12, char);
2369 PL_lex_casemods = 0;
2370 *PL_lex_casestack = '\0';
2372 PL_lex_state = LEX_INTERPCONCAT;
2373 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2375 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2376 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2377 PL_lex_inpat = PL_sublex_info.sub_op;
2379 PL_lex_inpat = NULL;
2386 * Restores lexer state after a S_sublex_push.
2393 if (!PL_lex_starts++) {
2394 SV * const sv = newSVpvs("");
2395 if (SvUTF8(PL_linestr))
2397 PL_expect = XOPERATOR;
2398 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2402 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2403 PL_lex_state = LEX_INTERPCASEMOD;
2407 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2408 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2409 PL_linestr = PL_lex_repl;
2411 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2412 PL_bufend += SvCUR(PL_linestr);
2413 PL_last_lop = PL_last_uni = NULL;
2414 SAVEFREESV(PL_linestr);
2415 PL_lex_dojoin = FALSE;
2416 PL_lex_brackets = 0;
2417 PL_lex_casemods = 0;
2418 *PL_lex_casestack = '\0';
2420 if (SvEVALED(PL_lex_repl)) {
2421 PL_lex_state = LEX_INTERPNORMAL;
2423 /* we don't clear PL_lex_repl here, so that we can check later
2424 whether this is an evalled subst; that means we rely on the
2425 logic to ensure sublex_done() is called again only via the
2426 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2429 PL_lex_state = LEX_INTERPCONCAT;
2439 PL_endwhite = newSVpvs("");
2440 sv_catsv(PL_endwhite, PL_thiswhite);
2444 sv_setpvs(PL_thistoken,"");
2446 PL_realtokenstart = -1;
2450 PL_bufend = SvPVX(PL_linestr);
2451 PL_bufend += SvCUR(PL_linestr);
2452 PL_expect = XOPERATOR;
2453 PL_sublex_info.sub_inwhat = 0;
2461 Extracts a pattern, double-quoted string, or transliteration. This
2464 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2465 processing a pattern (PL_lex_inpat is true), a transliteration
2466 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2468 Returns a pointer to the character scanned up to. If this is
2469 advanced from the start pointer supplied (i.e. if anything was
2470 successfully parsed), will leave an OP for the substring scanned
2471 in pl_yylval. Caller must intuit reason for not parsing further
2472 by looking at the next characters herself.
2476 constants: \N{NAME} only
2477 case and quoting: \U \Q \E
2478 stops on @ and $, but not for $ as tail anchor
2480 In transliterations:
2481 characters are VERY literal, except for - not at the start or end
2482 of the string, which indicates a range. If the range is in bytes,
2483 scan_const expands the range to the full set of intermediate
2484 characters. If the range is in utf8, the hyphen is replaced with
2485 a certain range mark which will be handled by pmtrans() in op.c.
2487 In double-quoted strings:
2489 double-quoted style: \r and \n
2490 constants: \x31, etc.
2491 deprecated backrefs: \1 (in substitution replacements)
2492 case and quoting: \U \Q \E
2495 scan_const does *not* construct ops to handle interpolated strings.
2496 It stops processing as soon as it finds an embedded $ or @ variable
2497 and leaves it to the caller to work out what's going on.
2499 embedded arrays (whether in pattern or not) could be:
2500 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2502 $ in double-quoted strings must be the symbol of an embedded scalar.
2504 $ in pattern could be $foo or could be tail anchor. Assumption:
2505 it's a tail anchor if $ is the last thing in the string, or if it's
2506 followed by one of "()| \r\n\t"
2508 \1 (backreferences) are turned into $1
2510 The structure of the code is
2511 while (there's a character to process) {
2512 handle transliteration ranges
2513 skip regexp comments /(?#comment)/ and codes /(?{code})/
2514 skip #-initiated comments in //x patterns
2515 check for embedded arrays
2516 check for embedded scalars
2518 deprecate \1 in substitution replacements
2519 handle string-changing backslashes \l \U \Q \E, etc.
2520 switch (what was escaped) {
2521 handle \- in a transliteration (becomes a literal -)
2522 if a pattern and not \N{, go treat as regular character
2523 handle \132 (octal characters)
2524 handle \x15 and \x{1234} (hex characters)
2525 handle \N{name} (named characters, also \N{3,5} in a pattern)
2526 handle \cV (control characters)
2527 handle printf-style backslashes (\f, \r, \n, etc)
2530 } (end if backslash)
2531 handle regular character
2532 } (end while character to read)
2537 S_scan_const(pTHX_ char *start)
2540 register char *send = PL_bufend; /* end of the constant */
2541 SV *sv = newSV(send - start); /* sv for the constant. See
2542 note below on sizing. */
2543 register char *s = start; /* start of the constant */
2544 register char *d = SvPVX(sv); /* destination for copies */
2545 bool dorange = FALSE; /* are we in a translit range? */
2546 bool didrange = FALSE; /* did we just finish a range? */
2547 I32 has_utf8 = FALSE; /* Output constant is UTF8 */
2548 I32 this_utf8 = UTF; /* Is the source string assumed
2549 to be UTF8? But, this can
2550 show as true when the source
2551 isn't utf8, as for example
2552 when it is entirely composed
2555 /* Note on sizing: The scanned constant is placed into sv, which is
2556 * initialized by newSV() assuming one byte of output for every byte of
2557 * input. This routine expects newSV() to allocate an extra byte for a
2558 * trailing NUL, which this routine will append if it gets to the end of
2559 * the input. There may be more bytes of input than output (eg., \N{LATIN
2560 * CAPITAL LETTER A}), or more output than input if the constant ends up
2561 * recoded to utf8, but each time a construct is found that might increase
2562 * the needed size, SvGROW() is called. Its size parameter each time is
2563 * based on the best guess estimate at the time, namely the length used so
2564 * far, plus the length the current construct will occupy, plus room for
2565 * the trailing NUL, plus one byte for every input byte still unscanned */
2569 UV literal_endpoint = 0;
2570 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2573 PERL_ARGS_ASSERT_SCAN_CONST;
2575 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2576 /* If we are doing a trans and we know we want UTF8 set expectation */
2577 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2578 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2582 while (s < send || dorange) {
2584 /* get transliterations out of the way (they're most literal) */
2585 if (PL_lex_inwhat == OP_TRANS) {
2586 /* expand a range A-Z to the full set of characters. AIE! */
2588 I32 i; /* current expanded character */
2589 I32 min; /* first character in range */
2590 I32 max; /* last character in range */
2601 char * const c = (char*)utf8_hop((U8*)d, -1);
2605 *c = (char)UTF_TO_NATIVE(0xff);
2606 /* mark the range as done, and continue */
2612 i = d - SvPVX_const(sv); /* remember current offset */
2615 SvLEN(sv) + (has_utf8 ?
2616 (512 - UTF_CONTINUATION_MARK +
2619 /* How many two-byte within 0..255: 128 in UTF-8,
2620 * 96 in UTF-8-mod. */
2622 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2624 d = SvPVX(sv) + i; /* refresh d after realloc */
2628 for (j = 0; j <= 1; j++) {
2629 char * const c = (char*)utf8_hop((U8*)d, -1);
2630 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2636 max = (U8)0xff; /* only to \xff */
2637 uvmax = uv; /* \x{100} to uvmax */
2639 d = c; /* eat endpoint chars */
2644 d -= 2; /* eat the first char and the - */
2645 min = (U8)*d; /* first char in range */
2646 max = (U8)d[1]; /* last char in range */
2653 "Invalid range \"%c-%c\" in transliteration operator",
2654 (char)min, (char)max);
2658 if (literal_endpoint == 2 &&
2659 ((isLOWER(min) && isLOWER(max)) ||
2660 (isUPPER(min) && isUPPER(max)))) {
2662 for (i = min; i <= max; i++)
2664 *d++ = NATIVE_TO_NEED(has_utf8,i);
2666 for (i = min; i <= max; i++)
2668 *d++ = NATIVE_TO_NEED(has_utf8,i);
2673 for (i = min; i <= max; i++)
2676 const U8 ch = (U8)NATIVE_TO_UTF(i);
2677 if (UNI_IS_INVARIANT(ch))
2680 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2681 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2690 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2692 *d++ = (char)UTF_TO_NATIVE(0xff);
2694 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2698 /* mark the range as done, and continue */
2702 literal_endpoint = 0;
2707 /* range begins (ignore - as first or last char) */
2708 else if (*s == '-' && s+1 < send && s != start) {
2710 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2717 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2727 literal_endpoint = 0;
2728 native_range = TRUE;
2733 /* if we get here, we're not doing a transliteration */
2735 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2736 except for the last char, which will be done separately. */
2737 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2739 while (s+1 < send && *s != ')')
2740 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2742 else if (s[2] == '{' /* This should match regcomp.c */
2743 || (s[2] == '?' && s[3] == '{'))
2746 char *regparse = s + (s[2] == '{' ? 3 : 4);
2749 while (count && (c = *regparse)) {
2750 if (c == '\\' && regparse[1])
2758 if (*regparse != ')')
2759 regparse--; /* Leave one char for continuation. */
2760 while (s < regparse)
2761 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2765 /* likewise skip #-initiated comments in //x patterns */
2766 else if (*s == '#' && PL_lex_inpat &&
2767 ((PMOP*)PL_lex_inpat)->op_pmflags & PMf_EXTENDED) {
2768 while (s+1 < send && *s != '\n')
2769 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2772 /* check for embedded arrays
2773 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2775 else if (*s == '@' && s[1]) {
2776 if (isALNUM_lazy_if(s+1,UTF))
2778 if (strchr(":'{$", s[1]))
2780 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2781 break; /* in regexp, neither @+ nor @- are interpolated */
2784 /* check for embedded scalars. only stop if we're sure it's a
2787 else if (*s == '$') {
2788 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2790 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2792 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2793 "Possible unintended interpolation of $\\ in regex");
2795 break; /* in regexp, $ might be tail anchor */
2799 /* End of else if chain - OP_TRANS rejoin rest */
2802 if (*s == '\\' && s+1 < send) {
2803 char* e; /* Can be used for ending '}', etc. */
2807 /* deprecate \1 in strings and substitution replacements */
2808 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2809 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2811 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2816 /* string-change backslash escapes */
2817 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2821 /* In a pattern, process \N, but skip any other backslash escapes.
2822 * This is because we don't want to translate an escape sequence
2823 * into a meta symbol and have the regex compiler use the meta
2824 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2825 * in spite of this, we do have to process \N here while the proper
2826 * charnames handler is in scope. See bugs #56444 and #62056.
2827 * There is a complication because \N in a pattern may also stand
2828 * for 'match a non-nl', and not mean a charname, in which case its
2829 * processing should be deferred to the regex compiler. To be a
2830 * charname it must be followed immediately by a '{', and not look
2831 * like \N followed by a curly quantifier, i.e., not something like
2832 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2834 else if (PL_lex_inpat
2837 || regcurly(s + 1)))
2839 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2840 goto default_action;
2845 /* quoted - in transliterations */
2847 if (PL_lex_inwhat == OP_TRANS) {
2854 if ((isALPHA(*s) || isDIGIT(*s)))
2855 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2856 "Unrecognized escape \\%c passed through",
2858 /* default action is to copy the quoted character */
2859 goto default_action;
2862 /* eg. \132 indicates the octal constant 0x132 */
2863 case '0': case '1': case '2': case '3':
2864 case '4': case '5': case '6': case '7':
2868 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2871 goto NUM_ESCAPE_INSERT;
2873 /* eg. \x24 indicates the hex constant 0x24 */
2877 char* const e = strchr(s, '}');
2878 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2879 PERL_SCAN_DISALLOW_PREFIX;
2884 yyerror("Missing right brace on \\x{}");
2888 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2894 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2895 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2901 /* Insert oct or hex escaped character. There will always be
2902 * enough room in sv since such escapes will be longer than any
2903 * UTF-8 sequence they can end up as, except if they force us
2904 * to recode the rest of the string into utf8 */
2906 /* Here uv is the ordinal of the next character being added in
2907 * unicode (converted from native). */
2908 if (!UNI_IS_INVARIANT(uv)) {
2909 if (!has_utf8 && uv > 255) {
2910 /* Might need to recode whatever we have accumulated so
2911 * far if it contains any chars variant in utf8 or
2914 SvCUR_set(sv, d - SvPVX_const(sv));
2917 /* See Note on sizing above. */
2918 sv_utf8_upgrade_flags_grow(sv,
2919 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
2920 UNISKIP(uv) + (STRLEN)(send - s) + 1);
2921 d = SvPVX(sv) + SvCUR(sv);
2926 d = (char*)uvuni_to_utf8((U8*)d, uv);
2927 if (PL_lex_inwhat == OP_TRANS &&
2928 PL_sublex_info.sub_op) {
2929 PL_sublex_info.sub_op->op_private |=
2930 (PL_lex_repl ? OPpTRANS_FROM_UTF
2934 if (uv > 255 && !dorange)
2935 native_range = FALSE;
2948 /* In a non-pattern \N must be a named character, like \N{LATIN
2949 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
2950 * mean to match a non-newline. For non-patterns, named
2951 * characters are converted to their string equivalents. In
2952 * patterns, named characters are not converted to their
2953 * ultimate forms for the same reasons that other escapes
2954 * aren't. Instead, they are converted to the \N{U+...} form
2955 * to get the value from the charnames that is in effect right
2956 * now, while preserving the fact that it was a named character
2957 * so that the regex compiler knows this */
2959 /* This section of code doesn't generally use the
2960 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
2961 * a close examination of this macro and determined it is a
2962 * no-op except on utfebcdic variant characters. Every
2963 * character generated by this that would normally need to be
2964 * enclosed by this macro is invariant, so the macro is not
2965 * needed, and would complicate use of copy(). There are other
2966 * parts of this file where the macro is used inconsistently,
2967 * but are saved by it being a no-op */
2969 /* The structure of this section of code (besides checking for
2970 * errors and upgrading to utf8) is:
2971 * Further disambiguate between the two meanings of \N, and if
2972 * not a charname, go process it elsewhere
2973 * If of form \N{U+...}, pass it through if a pattern;
2974 * otherwise convert to utf8
2975 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
2976 * pattern; otherwise convert to utf8 */
2978 /* Here, s points to the 'N'; the test below is guaranteed to
2979 * succeed if we are being called on a pattern as we already
2980 * know from a test above that the next character is a '{'.
2981 * On a non-pattern \N must mean 'named sequence, which
2982 * requires braces */
2985 yyerror("Missing braces on \\N{}");
2990 /* If there is no matching '}', it is an error. */
2991 if (! (e = strchr(s, '}'))) {
2992 if (! PL_lex_inpat) {
2993 yyerror("Missing right brace on \\N{}");
2995 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3000 /* Here it looks like a named character */
3004 /* XXX This block is temporary code. \N{} implies that the
3005 * pattern is to have Unicode semantics, and therefore
3006 * currently has to be encoded in utf8. By putting it in
3007 * utf8 now, we save a whole pass in the regular expression
3008 * compiler. Once that code is changed so Unicode
3009 * semantics doesn't necessarily have to be in utf8, this
3010 * block should be removed */
3012 SvCUR_set(sv, d - SvPVX_const(sv));
3015 /* See Note on sizing above. */
3016 sv_utf8_upgrade_flags_grow(sv,
3017 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3018 /* 5 = '\N{' + cur char + NUL */
3019 (STRLEN)(send - s) + 5);
3020 d = SvPVX(sv) + SvCUR(sv);
3025 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3026 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3027 | PERL_SCAN_DISALLOW_PREFIX;
3030 /* For \N{U+...}, the '...' is a unicode value even on
3031 * EBCDIC machines */
3032 s += 2; /* Skip to next char after the 'U+' */
3034 uv = grok_hex(s, &len, &flags, NULL);
3035 if (len == 0 || len != (STRLEN)(e - s)) {
3036 yyerror("Invalid hexadecimal number in \\N{U+...}");
3043 /* Pass through to the regex compiler unchanged. The
3044 * reason we evaluated the number above is to make sure
3045 * there wasn't a syntax error. */
3046 s -= 5; /* Include the '\N{U+' */
3047 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3050 else { /* Not a pattern: convert the hex to string */
3052 /* If destination is not in utf8, unconditionally
3053 * recode it to be so. This is because \N{} implies
3054 * Unicode semantics, and scalars have to be in utf8
3055 * to guarantee those semantics */
3057 SvCUR_set(sv, d - SvPVX_const(sv));
3060 /* See Note on sizing above. */
3061 sv_utf8_upgrade_flags_grow(
3063 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3064 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3065 d = SvPVX(sv) + SvCUR(sv);
3069 /* Add the string to the output */
3070 if (UNI_IS_INVARIANT(uv)) {
3073 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3076 else { /* Here is \N{NAME} but not \N{U+...}. */
3078 SV *res; /* result from charnames */
3079 const char *str; /* the string in 'res' */
3080 STRLEN len; /* its length */
3082 /* Get the value for NAME */
3083 res = newSVpvn(s, e - s);
3084 res = new_constant( NULL, 0, "charnames",
3085 /* includes all of: \N{...} */
3086 res, NULL, s - 3, e - s + 4 );
3088 /* Most likely res will be in utf8 already since the
3089 * standard charnames uses pack U, but a custom translator
3090 * can leave it otherwise, so make sure. XXX This can be
3091 * revisited to not have charnames use utf8 for characters
3092 * that don't need it when regexes don't have to be in utf8
3093 * for Unicode semantics. If doing so, remember EBCDIC */
3094 sv_utf8_upgrade(res);
3095 str = SvPV_const(res, len);
3097 /* Don't accept malformed input */
3098 if (! is_utf8_string((U8 *) str, len)) {
3099 yyerror("Malformed UTF-8 returned by \\N");
3101 else if (PL_lex_inpat) {
3103 if (! len) { /* The name resolved to an empty string */
3104 Copy("\\N{}", d, 4, char);
3108 /* In order to not lose information for the regex
3109 * compiler, pass the result in the specially made
3110 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3111 * the code points in hex of each character
3112 * returned by charnames */
3114 const char *str_end = str + len;
3115 STRLEN char_length; /* cur char's byte length */
3116 STRLEN output_length; /* and the number of bytes
3117 after this is translated
3119 const STRLEN off = d - SvPVX_const(sv);
3121 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3122 * max('U+', '.'); and 1 for NUL */
3123 char hex_string[2 * UTF8_MAXBYTES + 5];
3125 /* Get the first character of the result. */
3126 U32 uv = utf8n_to_uvuni((U8 *) str,
3131 /* The call to is_utf8_string() above hopefully
3132 * guarantees that there won't be an error. But
3133 * it's easy here to make sure. The function just
3134 * above warns and returns 0 if invalid utf8, but
3135 * it can also return 0 if the input is validly a
3136 * NUL. Disambiguate */
3137 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3138 uv = UNICODE_REPLACEMENT;
3141 /* Convert first code point to hex, including the
3142 * boiler plate before it */
3143 sprintf(hex_string, "\\N{U+%X", (unsigned int) uv);
3144 output_length = strlen(hex_string);
3146 /* Make sure there is enough space to hold it */
3147 d = off + SvGROW(sv, off
3149 + (STRLEN)(send - e)
3150 + 2); /* '}' + NUL */
3152 Copy(hex_string, d, output_length, char);
3155 /* For each subsequent character, append dot and
3156 * its ordinal in hex */
3157 while ((str += char_length) < str_end) {
3158 const STRLEN off = d - SvPVX_const(sv);
3159 U32 uv = utf8n_to_uvuni((U8 *) str,
3163 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3164 uv = UNICODE_REPLACEMENT;
3167 sprintf(hex_string, ".%X", (unsigned int) uv);
3168 output_length = strlen(hex_string);
3170 d = off + SvGROW(sv, off
3172 + (STRLEN)(send - e)
3173 + 2); /* '}' + NUL */
3174 Copy(hex_string, d, output_length, char);
3178 *d++ = '}'; /* Done. Add the trailing brace */
3181 else { /* Here, not in a pattern. Convert the name to a
3184 /* If destination is not in utf8, unconditionally
3185 * recode it to be so. This is because \N{} implies
3186 * Unicode semantics, and scalars have to be in utf8
3187 * to guarantee those semantics */
3189 SvCUR_set(sv, d - SvPVX_const(sv));
3192 /* See Note on sizing above. */
3193 sv_utf8_upgrade_flags_grow(sv,
3194 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3195 len + (STRLEN)(send - s) + 1);
3196 d = SvPVX(sv) + SvCUR(sv);
3198 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3200 /* See Note on sizing above. (NOTE: SvCUR() is not
3201 * set correctly here). */
3202 const STRLEN off = d - SvPVX_const(sv);
3203 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3205 Copy(str, d, len, char);
3210 /* Deprecate non-approved name syntax */
3211 if (ckWARN_d(WARN_DEPRECATED)) {
3212 bool problematic = FALSE;
3215 /* For non-ut8 input, look to see that the first
3216 * character is an alpha, then loop through the rest
3217 * checking that each is a continuation */
3219 if (! isALPHAU(*i)) problematic = TRUE;
3220 else for (i = s + 1; i < e; i++) {
3221 if (isCHARNAME_CONT(*i)) continue;
3227 /* Similarly for utf8. For invariants can check
3228 * directly. We accept anything above the latin1
3229 * range because it is immaterial to Perl if it is
3230 * correct or not, and is expensive to check. But
3231 * it is fairly easy in the latin1 range to convert
3232 * the variants into a single character and check
3234 if (UTF8_IS_INVARIANT(*i)) {
3235 if (! isALPHAU(*i)) problematic = TRUE;
3236 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3237 if (! isALPHAU(UNI_TO_NATIVE(UTF8_ACCUMULATE(*i,
3243 if (! problematic) for (i = s + UTF8SKIP(s);
3247 if (UTF8_IS_INVARIANT(*i)) {
3248 if (isCHARNAME_CONT(*i)) continue;
3249 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3251 } else if (isCHARNAME_CONT(
3253 UTF8_ACCUMULATE(*i, *(i+1)))))
3263 Newx(string, e - i + 1, char);
3264 Copy(i, string, e - i, char);
3265 string[e - i] = '\0';
3266 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3267 "Deprecated character(s) in \\N{...} starting at '%s'",
3272 } /* End \N{NAME} */
3275 native_range = FALSE; /* \N{} is defined to be Unicode */
3277 s = e + 1; /* Point to just after the '}' */
3280 /* \c is a control character */
3289 *d++ = NATIVE_TO_NEED(has_utf8,toCTRL(c));
3292 yyerror("Missing control char name in \\c");
3296 /* printf-style backslashes, formfeeds, newlines, etc */
3298 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3301 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3304 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3307 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3310 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3313 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3316 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3322 } /* end if (backslash) */
3329 /* If we started with encoded form, or already know we want it,
3330 then encode the next character */
3331 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3335 /* One might think that it is wasted effort in the case of the
3336 * source being utf8 (this_utf8 == TRUE) to take the next character
3337 * in the source, convert it to an unsigned value, and then convert
3338 * it back again. But the source has not been validated here. The
3339 * routine that does the conversion checks for errors like
3342 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3343 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3345 SvCUR_set(sv, d - SvPVX_const(sv));
3348 /* See Note on sizing above. */
3349 sv_utf8_upgrade_flags_grow(sv,
3350 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3351 need + (STRLEN)(send - s) + 1);
3352 d = SvPVX(sv) + SvCUR(sv);
3354 } else if (need > len) {
3355 /* encoded value larger than old, may need extra space (NOTE:
3356 * SvCUR() is not set correctly here). See Note on sizing
3358 const STRLEN off = d - SvPVX_const(sv);
3359 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3363 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3365 if (uv > 255 && !dorange)
3366 native_range = FALSE;
3370 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3372 } /* while loop to process each character */
3374 /* terminate the string and set up the sv */
3376 SvCUR_set(sv, d - SvPVX_const(sv));
3377 if (SvCUR(sv) >= SvLEN(sv))
3378 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3381 if (PL_encoding && !has_utf8) {
3382 sv_recode_to_utf8(sv, PL_encoding);
3388 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3389 PL_sublex_info.sub_op->op_private |=
3390 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3394 /* shrink the sv if we allocated more than we used */
3395 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3396 SvPV_shrink_to_cur(sv);
3399 /* return the substring (via pl_yylval) only if we parsed anything */
3400 if (s > PL_bufptr) {
3401 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3402 const char *const key = PL_lex_inpat ? "qr" : "q";
3403 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3407 if (PL_lex_inwhat == OP_TRANS) {
3410 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3418 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3421 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3428 * Returns TRUE if there's more to the expression (e.g., a subscript),
3431 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3433 * ->[ and ->{ return TRUE
3434 * { and [ outside a pattern are always subscripts, so return TRUE
3435 * if we're outside a pattern and it's not { or [, then return FALSE
3436 * if we're in a pattern and the first char is a {
3437 * {4,5} (any digits around the comma) returns FALSE
3438 * if we're in a pattern and the first char is a [
3440 * [SOMETHING] has a funky algorithm to decide whether it's a
3441 * character class or not. It has to deal with things like
3442 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3443 * anything else returns TRUE
3446 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3449 S_intuit_more(pTHX_ register char *s)
3453 PERL_ARGS_ASSERT_INTUIT_MORE;
3455 if (PL_lex_brackets)
3457 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3459 if (*s != '{' && *s != '[')
3464 /* In a pattern, so maybe we have {n,m}. */
3481 /* On the other hand, maybe we have a character class */
3484 if (*s == ']' || *s == '^')
3487 /* this is terrifying, and it works */
3488 int weight = 2; /* let's weigh the evidence */
3490 unsigned char un_char = 255, last_un_char;
3491 const char * const send = strchr(s,']');
3492 char tmpbuf[sizeof PL_tokenbuf * 4];
3494 if (!send) /* has to be an expression */
3497 Zero(seen,256,char);
3500 else if (isDIGIT(*s)) {
3502 if (isDIGIT(s[1]) && s[2] == ']')
3508 for (; s < send; s++) {
3509 last_un_char = un_char;
3510 un_char = (unsigned char)*s;
3515 weight -= seen[un_char] * 10;
3516 if (isALNUM_lazy_if(s+1,UTF)) {
3518 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3519 len = (int)strlen(tmpbuf);
3520 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
3525 else if (*s == '$' && s[1] &&
3526 strchr("[#!%*<>()-=",s[1])) {
3527 if (/*{*/ strchr("])} =",s[2]))
3536 if (strchr("wds]",s[1]))
3538 else if (seen[(U8)'\''] || seen[(U8)'"'])
3540 else if (strchr("rnftbxcav",s[1]))
3542 else if (isDIGIT(s[1])) {
3544 while (s[1] && isDIGIT(s[1]))
3554 if (strchr("aA01! ",last_un_char))
3556 if (strchr("zZ79~",s[1]))
3558 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3559 weight -= 5; /* cope with negative subscript */
3562 if (!isALNUM(last_un_char)
3563 && !(last_un_char == '$' || last_un_char == '@'
3564 || last_un_char == '&')
3565 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3570 if (keyword(tmpbuf, d - tmpbuf, 0))
3573 if (un_char == last_un_char + 1)
3575 weight -= seen[un_char];
3580 if (weight >= 0) /* probably a character class */
3590 * Does all the checking to disambiguate
3592 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3593 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3595 * First argument is the stuff after the first token, e.g. "bar".
3597 * Not a method if bar is a filehandle.
3598 * Not a method if foo is a subroutine prototyped to take a filehandle.
3599 * Not a method if it's really "Foo $bar"
3600 * Method if it's "foo $bar"
3601 * Not a method if it's really "print foo $bar"
3602 * Method if it's really "foo package::" (interpreted as package->foo)
3603 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3604 * Not a method if bar is a filehandle or package, but is quoted with
3609 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3612 char *s = start + (*start == '$');
3613 char tmpbuf[sizeof PL_tokenbuf];
3620 PERL_ARGS_ASSERT_INTUIT_METHOD;
3623 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3627 const char *proto = SvPVX_const(cv);
3638 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3639 /* start is the beginning of the possible filehandle/object,
3640 * and s is the end of it
3641 * tmpbuf is a copy of it
3644 if (*start == '$') {
3645 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3646 isUPPER(*PL_tokenbuf))
3649 len = start - SvPVX(PL_linestr);
3653 start = SvPVX(PL_linestr) + len;
3657 return *s == '(' ? FUNCMETH : METHOD;
3659 if (!keyword(tmpbuf, len, 0)) {
3660 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3664 soff = s - SvPVX(PL_linestr);
3668 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
3669 if (indirgv && GvCVu(indirgv))
3671 /* filehandle or package name makes it a method */
3672 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
3674 soff = s - SvPVX(PL_linestr);
3677 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3678 return 0; /* no assumptions -- "=>" quotes bearword */
3680 start_force(PL_curforce);
3681 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3682 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3683 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3685 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3690 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3692 return *s == '(' ? FUNCMETH : METHOD;
3698 /* Encoded script support. filter_add() effectively inserts a
3699 * 'pre-processing' function into the current source input stream.
3700 * Note that the filter function only applies to the current source file
3701 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3703 * The datasv parameter (which may be NULL) can be used to pass
3704 * private data to this instance of the filter. The filter function
3705 * can recover the SV using the FILTER_DATA macro and use it to
3706 * store private buffers and state information.
3708 * The supplied datasv parameter is upgraded to a PVIO type
3709 * and the IoDIRP/IoANY field is used to store the function pointer,
3710 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3711 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3712 * private use must be set using malloc'd pointers.
3716 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3725 if (!PL_rsfp_filters)
3726 PL_rsfp_filters = newAV();
3729 SvUPGRADE(datasv, SVt_PVIO);
3730 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3731 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3732 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3733 FPTR2DPTR(void *, IoANY(datasv)),
3734 SvPV_nolen(datasv)));
3735 av_unshift(PL_rsfp_filters, 1);
3736 av_store(PL_rsfp_filters, 0, datasv) ;
3741 /* Delete most recently added instance of this filter function. */
3743 Perl_filter_del(pTHX_ filter_t funcp)
3748 PERL_ARGS_ASSERT_FILTER_DEL;
3751 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3752 FPTR2DPTR(void*, funcp)));
3754 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3756 /* if filter is on top of stack (usual case) just pop it off */
3757 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3758 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3759 IoFLAGS(datasv) &= ~IOf_FAKE_DIRP;
3760 IoANY(datasv) = (void *)NULL;
3761 sv_free(av_pop(PL_rsfp_filters));
3765 /* we need to search for the correct entry and clear it */
3766 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3770 /* Invoke the idxth filter function for the current rsfp. */
3771 /* maxlen 0 = read one text line */
3773 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3778 /* This API is bad. It should have been using unsigned int for maxlen.
3779 Not sure if we want to change the API, but if not we should sanity
3780 check the value here. */
3781 const unsigned int correct_length
3790 PERL_ARGS_ASSERT_FILTER_READ;
3792 if (!PL_parser || !PL_rsfp_filters)
3794 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3795 /* Provide a default input filter to make life easy. */
3796 /* Note that we append to the line. This is handy. */
3797 DEBUG_P(PerlIO_printf(Perl_debug_log,
3798 "filter_read %d: from rsfp\n", idx));
3799 if (correct_length) {
3802 const int old_len = SvCUR(buf_sv);
3804 /* ensure buf_sv is large enough */
3805 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3806 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3807 correct_length)) <= 0) {
3808 if (PerlIO_error(PL_rsfp))
3809 return -1; /* error */
3811 return 0 ; /* end of file */
3813 SvCUR_set(buf_sv, old_len + len) ;
3814 SvPVX(buf_sv)[old_len + len] = '\0';
3817 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3818 if (PerlIO_error(PL_rsfp))
3819 return -1; /* error */
3821 return 0 ; /* end of file */
3824 return SvCUR(buf_sv);
3826 /* Skip this filter slot if filter has been deleted */
3827 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3828 DEBUG_P(PerlIO_printf(Perl_debug_log,
3829 "filter_read %d: skipped (filter deleted)\n",
3831 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
3833 /* Get function pointer hidden within datasv */
3834 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
3835 DEBUG_P(PerlIO_printf(Perl_debug_log,
3836 "filter_read %d: via function %p (%s)\n",
3837 idx, (void*)datasv, SvPV_nolen_const(datasv)));
3838 /* Call function. The function is expected to */
3839 /* call "FILTER_READ(idx+1, buf_sv)" first. */
3840 /* Return: <0:error, =0:eof, >0:not eof */
3841 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
3845 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
3849 PERL_ARGS_ASSERT_FILTER_GETS;
3851 #ifdef PERL_CR_FILTER
3852 if (!PL_rsfp_filters) {
3853 filter_add(S_cr_textfilter,NULL);
3856 if (PL_rsfp_filters) {
3858 SvCUR_set(sv, 0); /* start with empty line */
3859 if (FILTER_READ(0, sv, 0) > 0)
3860 return ( SvPVX(sv) ) ;
3865 return (sv_gets(sv, PL_rsfp, append));
3869 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3874 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3876 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3880 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3881 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3883 return GvHV(gv); /* Foo:: */
3886 /* use constant CLASS => 'MyClass' */
3887 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3888 if (gv && GvCV(gv)) {
3889 SV * const sv = cv_const_sv(GvCV(gv));
3891 pkgname = SvPV_const(sv, len);
3894 return gv_stashpvn(pkgname, len, 0);
3898 * S_readpipe_override
3899 * Check whether readpipe() is overriden, and generates the appropriate
3900 * optree, provided sublex_start() is called afterwards.
3903 S_readpipe_override(pTHX)
3906 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
3907 pl_yylval.ival = OP_BACKTICK;
3909 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
3911 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
3912 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
3913 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
3915 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
3916 append_elem(OP_LIST,
3917 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
3918 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
3925 * The intent of this yylex wrapper is to minimize the changes to the
3926 * tokener when we aren't interested in collecting madprops. It remains
3927 * to be seen how successful this strategy will be...
3934 char *s = PL_bufptr;
3936 /* make sure PL_thiswhite is initialized */
3940 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
3941 if (PL_pending_ident)
3942 return S_pending_ident(aTHX);
3944 /* previous token ate up our whitespace? */
3945 if (!PL_lasttoke && PL_nextwhite) {
3946 PL_thiswhite = PL_nextwhite;
3950 /* isolate the token, and figure out where it is without whitespace */
3951 PL_realtokenstart = -1;
3955 assert(PL_curforce < 0);
3957 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
3958 if (!PL_thistoken) {
3959 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
3960 PL_thistoken = newSVpvs("");
3962 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
3963 PL_thistoken = newSVpvn(tstart, s - tstart);
3966 if (PL_thismad) /* install head */
3967 CURMAD('X', PL_thistoken);
3970 /* last whitespace of a sublex? */
3971 if (optype == ')' && PL_endwhite) {
3972 CURMAD('X', PL_endwhite);
3977 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
3978 if (!PL_thiswhite && !PL_endwhite && !optype) {
3979 sv_free(PL_thistoken);
3984 /* put off final whitespace till peg */
3985 if (optype == ';' && !PL_rsfp) {
3986 PL_nextwhite = PL_thiswhite;
3989 else if (PL_thisopen) {
3990 CURMAD('q', PL_thisopen);
3992 sv_free(PL_thistoken);
3996 /* Store actual token text as madprop X */
3997 CURMAD('X', PL_thistoken);
4001 /* add preceding whitespace as madprop _ */
4002 CURMAD('_', PL_thiswhite);
4006 /* add quoted material as madprop = */
4007 CURMAD('=', PL_thisstuff);
4011 /* add terminating quote as madprop Q */
4012 CURMAD('Q', PL_thisclose);
4016 /* special processing based on optype */
4020 /* opval doesn't need a TOKEN since it can already store mp */
4030 if (pl_yylval.opval)
4031 append_madprops(PL_thismad, pl_yylval.opval, 0);
4039 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4048 /* remember any fake bracket that lexer is about to discard */
4049 if (PL_lex_brackets == 1 &&
4050 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4053 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4056 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4057 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4060 break; /* don't bother looking for trailing comment */
4069 /* attach a trailing comment to its statement instead of next token */
4073 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4075 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4077 if (*s == '\n' || *s == '#') {
4078 while (s < PL_bufend && *s != '\n')
4082 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4083 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4100 /* Create new token struct. Note: opvals return early above. */
4101 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4108 S_tokenize_use(pTHX_ int is_use, char *s) {
4111 PERL_ARGS_ASSERT_TOKENIZE_USE;
4113 if (PL_expect != XSTATE)
4114 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4115 is_use ? "use" : "no"));
4117 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4118 s = force_version(s, TRUE);
4119 if (*s == ';' || *s == '}'
4120 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4121 start_force(PL_curforce);
4122 NEXTVAL_NEXTTOKE.opval = NULL;
4125 else if (*s == 'v') {
4126 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4127 s = force_version(s, FALSE);
4131 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4132 s = force_version(s, FALSE);
4134 pl_yylval.ival = is_use;
4138 static const char* const exp_name[] =
4139 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4140 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4147 Works out what to call the token just pulled out of the input
4148 stream. The yacc parser takes care of taking the ops we return and
4149 stitching them into a tree.
4155 if read an identifier
4156 if we're in a my declaration
4157 croak if they tried to say my($foo::bar)
4158 build the ops for a my() declaration
4159 if it's an access to a my() variable
4160 are we in a sort block?
4161 croak if my($a); $a <=> $b
4162 build ops for access to a my() variable
4163 if in a dq string, and they've said @foo and we can't find @foo
4165 build ops for a bareword
4166 if we already built the token before, use it.
4171 #pragma segment Perl_yylex
4177 register char *s = PL_bufptr;
4183 /* orig_keyword, gvp, and gv are initialized here because
4184 * jump to the label just_a_word_zero can bypass their
4185 * initialization later. */
4186 I32 orig_keyword = 0;
4191 SV* tmp = newSVpvs("");
4192 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4193 (IV)CopLINE(PL_curcop),
4194 lex_state_names[PL_lex_state],
4195 exp_name[PL_expect],
4196 pv_display(tmp, s, strlen(s), 0, 60));
4199 /* check if there's an identifier for us to look at */
4200 if (PL_pending_ident)
4201 return REPORT(S_pending_ident(aTHX));
4203 /* no identifier pending identification */
4205 switch (PL_lex_state) {
4207 case LEX_NORMAL: /* Some compilers will produce faster */
4208 case LEX_INTERPNORMAL: /* code if we comment these out. */
4212 /* when we've already built the next token, just pull it out of the queue */
4216 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4218 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4219 PL_nexttoke[PL_lasttoke].next_mad = 0;
4220 if (PL_thismad && PL_thismad->mad_key == '_') {
4221 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4222 PL_thismad->mad_val = 0;
4223 mad_free(PL_thismad);
4228 PL_lex_state = PL_lex_defer;
4229 PL_expect = PL_lex_expect;
4230 PL_lex_defer = LEX_NORMAL;
4231 if (!PL_nexttoke[PL_lasttoke].next_type)
4236 pl_yylval = PL_nextval[PL_nexttoke];
4238 PL_lex_state = PL_lex_defer;
4239 PL_expect = PL_lex_expect;
4240 PL_lex_defer = LEX_NORMAL;
4244 /* FIXME - can these be merged? */
4245 return(PL_nexttoke[PL_lasttoke].next_type);
4247 return REPORT(PL_nexttype[PL_nexttoke]);
4250 /* interpolated case modifiers like \L \U, including \Q and \E.
4251 when we get here, PL_bufptr is at the \
4253 case LEX_INTERPCASEMOD:
4255 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4256 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
4258 /* handle \E or end of string */
4259 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4261 if (PL_lex_casemods) {
4262 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4263 PL_lex_casestack[PL_lex_casemods] = '\0';
4265 if (PL_bufptr != PL_bufend
4266 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
4268 PL_lex_state = LEX_INTERPCONCAT;
4271 PL_thistoken = newSVpvs("\\E");
4277 while (PL_bufptr != PL_bufend &&
4278 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4280 PL_thiswhite = newSVpvs("");
4281 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4285 if (PL_bufptr != PL_bufend)
4288 PL_lex_state = LEX_INTERPCONCAT;
4292 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4293 "### Saw case modifier\n"); });
4295 if (s[1] == '\\' && s[2] == 'E') {
4298 PL_thiswhite = newSVpvs("");
4299 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4302 PL_lex_state = LEX_INTERPCONCAT;
4307 if (!PL_madskills) /* when just compiling don't need correct */
4308 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4309 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4310 if ((*s == 'L' || *s == 'U') &&
4311 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4312 PL_lex_casestack[--PL_lex_casemods] = '\0';
4315 if (PL_lex_casemods > 10)
4316 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4317 PL_lex_casestack[PL_lex_casemods++] = *s;
4318 PL_lex_casestack[PL_lex_casemods] = '\0';
4319 PL_lex_state = LEX_INTERPCONCAT;
4320 start_force(PL_curforce);
4321 NEXTVAL_NEXTTOKE.ival = 0;
4323 start_force(PL_curforce);
4325 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4327 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4329 NEXTVAL_NEXTTOKE.ival = OP_LC;
4331 NEXTVAL_NEXTTOKE.ival = OP_UC;
4333 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4335 Perl_croak(aTHX_ "panic: yylex");
4337 SV* const tmpsv = newSVpvs("\\ ");
4338 /* replace the space with the character we want to escape
4340 SvPVX(tmpsv)[1] = *s;
4346 if (PL_lex_starts) {
4352 sv_free(PL_thistoken);
4353 PL_thistoken = newSVpvs("");
4356 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4357 if (PL_lex_casemods == 1 && PL_lex_inpat)
4366 case LEX_INTERPPUSH:
4367 return REPORT(sublex_push());
4369 case LEX_INTERPSTART:
4370 if (PL_bufptr == PL_bufend)
4371 return REPORT(sublex_done());
4372 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4373 "### Interpolated variable\n"); });
4375 PL_lex_dojoin = (*PL_bufptr == '@');
4376 PL_lex_state = LEX_INTERPNORMAL;
4377 if (PL_lex_dojoin) {
4378 start_force(PL_curforce);
4379 NEXTVAL_NEXTTOKE.ival = 0;
4381 start_force(PL_curforce);
4382 force_ident("\"", '$');
4383 start_force(PL_curforce);
4384 NEXTVAL_NEXTTOKE.ival = 0;
4386 start_force(PL_curforce);
4387 NEXTVAL_NEXTTOKE.ival = 0;
4389 start_force(PL_curforce);
4390 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4393 if (PL_lex_starts++) {
4398 sv_free(PL_thistoken);
4399 PL_thistoken = newSVpvs("");
4402 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4403 if (!PL_lex_casemods && PL_lex_inpat)
4410 case LEX_INTERPENDMAYBE:
4411 if (intuit_more(PL_bufptr)) {
4412 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4418 if (PL_lex_dojoin) {
4419 PL_lex_dojoin = FALSE;
4420 PL_lex_state = LEX_INTERPCONCAT;
4424 sv_free(PL_thistoken);
4425 PL_thistoken = newSVpvs("");
4430 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4431 && SvEVALED(PL_lex_repl))
4433 if (PL_bufptr != PL_bufend)
4434 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4438 case LEX_INTERPCONCAT:
4440 if (PL_lex_brackets)
4441 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4443 if (PL_bufptr == PL_bufend)
4444 return REPORT(sublex_done());
4446 if (SvIVX(PL_linestr) == '\'') {
4447 SV *sv = newSVsv(PL_linestr);
4450 else if ( PL_hints & HINT_NEW_RE )
4451 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4452 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4456 s = scan_const(PL_bufptr);
4458 PL_lex_state = LEX_INTERPCASEMOD;
4460 PL_lex_state = LEX_INTERPSTART;
4463 if (s != PL_bufptr) {
4464 start_force(PL_curforce);
4466 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4468 NEXTVAL_NEXTTOKE = pl_yylval;
4471 if (PL_lex_starts++) {
4475 sv_free(PL_thistoken);
4476 PL_thistoken = newSVpvs("");
4479 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4480 if (!PL_lex_casemods && PL_lex_inpat)
4493 PL_lex_state = LEX_NORMAL;
4494 s = scan_formline(PL_bufptr);
4495 if (!PL_lex_formbrack)
4501 PL_oldoldbufptr = PL_oldbufptr;
4507 sv_free(PL_thistoken);
4510 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4514 if (isIDFIRST_lazy_if(s,UTF))
4517 unsigned char c = *s;
4518 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4519 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4520 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4525 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4529 goto fake_eof; /* emulate EOF on ^D or ^Z */
4538 if (PL_lex_brackets) {
4539 yyerror((const char *)
4541 ? "Format not terminated"
4542 : "Missing right curly or square bracket"));
4544 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4545 "### Tokener got EOF\n");
4549 if (s++ < PL_bufend)
4550 goto retry; /* ignore stray nulls */
4553 if (!PL_in_eval && !PL_preambled) {
4554 PL_preambled = TRUE;
4560 /* Generate a string of Perl code to load the debugger.
4561 * If PERL5DB is set, it will return the contents of that,
4562 * otherwise a compile-time require of perl5db.pl. */
4564 const char * const pdb = PerlEnv_getenv("PERL5DB");
4567 sv_setpv(PL_linestr, pdb);
4568 sv_catpvs(PL_linestr,";");
4570 SETERRNO(0,SS_NORMAL);
4571 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4574 sv_setpvs(PL_linestr,"");
4575 if (PL_preambleav) {
4576 SV **svp = AvARRAY(PL_preambleav);
4577 SV **const end = svp + AvFILLp(PL_preambleav);
4579 sv_catsv(PL_linestr, *svp);
4581 sv_catpvs(PL_linestr, ";");
4583 sv_free(MUTABLE_SV(PL_preambleav));
4584 PL_preambleav = NULL;
4587 sv_catpvs(PL_linestr,
4588 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4589 if (PL_minus_n || PL_minus_p) {
4590 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4592 sv_catpvs(PL_linestr,"chomp;");
4595 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4596 || *PL_splitstr == '"')
4597 && strchr(PL_splitstr + 1, *PL_splitstr))
4598 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4600 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4601 bytes can be used as quoting characters. :-) */
4602 const char *splits = PL_splitstr;
4603 sv_catpvs(PL_linestr, "our @F=split(q\0");
4606 if (*splits == '\\')
4607 sv_catpvn(PL_linestr, splits, 1);
4608 sv_catpvn(PL_linestr, splits, 1);
4609 } while (*splits++);
4610 /* This loop will embed the trailing NUL of
4611 PL_linestr as the last thing it does before
4613 sv_catpvs(PL_linestr, ");");
4617 sv_catpvs(PL_linestr,"our @F=split(' ');");
4620 sv_catpvs(PL_linestr, "\n");
4621 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4622 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4623 PL_last_lop = PL_last_uni = NULL;
4624 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4625 update_debugger_info(PL_linestr, NULL, 0);
4630 bof = PL_rsfp ? TRUE : FALSE;
4633 fake_eof = LEX_FAKE_EOF;
4635 PL_bufptr = PL_bufend;
4636 CopLINE_inc(PL_curcop);
4637 if (!lex_next_chunk(fake_eof)) {
4638 CopLINE_dec(PL_curcop);
4640 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4642 CopLINE_dec(PL_curcop);
4645 PL_realtokenstart = -1;
4648 /* If it looks like the start of a BOM or raw UTF-16,
4649 * check if it in fact is. */
4650 if (bof && PL_rsfp &&
4655 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
4657 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4658 s = swallow_bom((U8*)s);
4662 /* Incest with pod. */
4665 sv_catsv(PL_thiswhite, PL_linestr);
4667 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4668 sv_setpvs(PL_linestr, "");
4669 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4670 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4671 PL_last_lop = PL_last_uni = NULL;
4672 PL_doextract = FALSE;
4677 } while (PL_doextract);
4678 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4679 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4680 PL_last_lop = PL_last_uni = NULL;
4681 if (CopLINE(PL_curcop) == 1) {
4682 while (s < PL_bufend && isSPACE(*s))
4684 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4688 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4692 if (*s == '#' && *(s+1) == '!')
4694 #ifdef ALTERNATE_SHEBANG
4696 static char const as[] = ALTERNATE_SHEBANG;
4697 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4698 d = s + (sizeof(as) - 1);
4700 #endif /* ALTERNATE_SHEBANG */
4709 while (*d && !isSPACE(*d))
4713 #ifdef ARG_ZERO_IS_SCRIPT
4714 if (ipathend > ipath) {
4716 * HP-UX (at least) sets argv[0] to the script name,
4717 * which makes $^X incorrect. And Digital UNIX and Linux,
4718 * at least, set argv[0] to the basename of the Perl
4719 * interpreter. So, having found "#!", we'll set it right.
4721 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4723 assert(SvPOK(x) || SvGMAGICAL(x));
4724 if (sv_eq(x, CopFILESV(PL_curcop))) {
4725 sv_setpvn(x, ipath, ipathend - ipath);
4731 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4732 const char * const lstart = SvPV_const(x,llen);
4734 bstart += blen - llen;
4735 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4736 sv_setpvn(x, ipath, ipathend - ipath);
4741 TAINT_NOT; /* $^X is always tainted, but that's OK */
4743 #endif /* ARG_ZERO_IS_SCRIPT */
4748 d = instr(s,"perl -");
4750 d = instr(s,"perl");
4752 /* avoid getting into infinite loops when shebang
4753 * line contains "Perl" rather than "perl" */
4755 for (d = ipathend-4; d >= ipath; --d) {
4756 if ((*d == 'p' || *d == 'P')
4757 && !ibcmp(d, "perl", 4))
4767 #ifdef ALTERNATE_SHEBANG
4769 * If the ALTERNATE_SHEBANG on this system starts with a
4770 * character that can be part of a Perl expression, then if
4771 * we see it but not "perl", we're probably looking at the
4772 * start of Perl code, not a request to hand off to some
4773 * other interpreter. Similarly, if "perl" is there, but
4774 * not in the first 'word' of the line, we assume the line
4775 * contains the start of the Perl program.
4777 if (d && *s != '#') {
4778 const char *c = ipath;
4779 while (*c && !strchr("; \t\r\n\f\v#", *c))
4782 d = NULL; /* "perl" not in first word; ignore */
4784 *s = '#'; /* Don't try to parse shebang line */
4786 #endif /* ALTERNATE_SHEBANG */
4791 !instr(s,"indir") &&
4792 instr(PL_origargv[0],"perl"))
4799 while (s < PL_bufend && isSPACE(*s))
4801 if (s < PL_bufend) {
4802 Newx(newargv,PL_origargc+3,char*);
4804 while (s < PL_bufend && !isSPACE(*s))
4807 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4810 newargv = PL_origargv;
4813 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4815 Perl_croak(aTHX_ "Can't exec %s", ipath);
4818 while (*d && !isSPACE(*d))
4820 while (SPACE_OR_TAB(*d))
4824 const bool switches_done = PL_doswitches;
4825 const U32 oldpdb = PL_perldb;
4826 const bool oldn = PL_minus_n;
4827 const bool oldp = PL_minus_p;
4831 bool baduni = FALSE;
4833 const char *d2 = d1 + 1;
4834 if (parse_unicode_opts((const char **)&d2)
4838 if (baduni || *d1 == 'M' || *d1 == 'm') {
4839 const char * const m = d1;
4840 while (*d1 && !isSPACE(*d1))
4842 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4845 d1 = moreswitches(d1);
4847 if (PL_doswitches && !switches_done) {
4848 int argc = PL_origargc;
4849 char **argv = PL_origargv;
4852 } while (argc && argv[0][0] == '-' && argv[0][1]);
4853 init_argv_symbols(argc,argv);
4855 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4856 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4857 /* if we have already added "LINE: while (<>) {",
4858 we must not do it again */
4860 sv_setpvs(PL_linestr, "");
4861 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4862 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4863 PL_last_lop = PL_last_uni = NULL;
4864 PL_preambled = FALSE;
4865 if (PERLDB_LINE || PERLDB_SAVESRC)
4866 (void)gv_fetchfile(PL_origfilename);
4873 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4875 PL_lex_state = LEX_FORMLINE;
4880 #ifdef PERL_STRICT_CR
4881 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
4883 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
4885 case ' ': case '\t': case '\f': case 013:
4887 PL_realtokenstart = -1;
4889 PL_thiswhite = newSVpvs("");
4890 sv_catpvn(PL_thiswhite, s, 1);
4897 PL_realtokenstart = -1;
4901 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
4902 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
4903 /* handle eval qq[#line 1 "foo"\n ...] */
4904 CopLINE_dec(PL_curcop);
4907 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
4909 if (!PL_in_eval || PL_rsfp)
4914 while (d < PL_bufend && *d != '\n')
4918 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4919 Perl_croak(aTHX_ "panic: input overflow");
4922 PL_thiswhite = newSVpvn(s, d - s);
4927 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
4929 PL_lex_state = LEX_FORMLINE;
4935 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
4936 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
4939 TOKEN(PEG); /* make sure any #! line is accessible */
4944 /* if (PL_madskills && PL_lex_formbrack) { */
4946 while (d < PL_bufend && *d != '\n')
4950 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
4951 Perl_croak(aTHX_ "panic: input overflow");
4952 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
4954 PL_thiswhite = newSVpvs("");
4955 if (CopLINE(PL_curcop) == 1) {
4956 sv_setpvs(PL_thiswhite, "");
4959 sv_catpvn(PL_thiswhite, s, d - s);
4973 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
4981 while (s < PL_bufend && SPACE_OR_TAB(*s))
4984 if (strnEQ(s,"=>",2)) {
4985 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
4986 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
4987 OPERATOR('-'); /* unary minus */
4989 PL_last_uni = PL_oldbufptr;
4991 case 'r': ftst = OP_FTEREAD; break;
4992 case 'w': ftst = OP_FTEWRITE; break;
4993 case 'x': ftst = OP_FTEEXEC; break;
4994 case 'o': ftst = OP_FTEOWNED; break;
4995 case 'R': ftst = OP_FTRREAD; break;
4996 case 'W': ftst = OP_FTRWRITE; break;
4997 case 'X': ftst = OP_FTREXEC; break;
4998 case 'O': ftst = OP_FTROWNED; break;
4999 case 'e': ftst = OP_FTIS; break;
5000 case 'z': ftst = OP_FTZERO; break;
5001 case 's': ftst = OP_FTSIZE; break;
5002 case 'f': ftst = OP_FTFILE; break;
5003 case 'd': ftst = OP_FTDIR; break;
5004 case 'l': ftst = OP_FTLINK; break;
5005 case 'p': ftst = OP_FTPIPE; break;
5006 case 'S': ftst = OP_FTSOCK; break;
5007 case 'u': ftst = OP_FTSUID; break;
5008 case 'g': ftst = OP_FTSGID; break;
5009 case 'k': ftst = OP_FTSVTX; break;
5010 case 'b': ftst = OP_FTBLK; break;
5011 case 'c': ftst = OP_FTCHR; break;
5012 case 't': ftst = OP_FTTTY; break;
5013 case 'T': ftst = OP_FTTEXT; break;
5014 case 'B': ftst = OP_FTBINARY; break;
5015 case 'M': case 'A': case 'C':
5016 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5018 case 'M': ftst = OP_FTMTIME; break;
5019 case 'A': ftst = OP_FTATIME; break;
5020 case 'C': ftst = OP_FTCTIME; break;
5028 PL_last_lop_op = (OPCODE)ftst;
5029 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5030 "### Saw file test %c\n", (int)tmp);
5035 /* Assume it was a minus followed by a one-letter named
5036 * subroutine call (or a -bareword), then. */
5037 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5038 "### '-%c' looked like a file test but was not\n",
5045 const char tmp = *s++;
5048 if (PL_expect == XOPERATOR)
5053 else if (*s == '>') {
5056 if (isIDFIRST_lazy_if(s,UTF)) {
5057 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5065 if (PL_expect == XOPERATOR)
5068 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5070 OPERATOR('-'); /* unary minus */
5076 const char tmp = *s++;
5079 if (PL_expect == XOPERATOR)
5084 if (PL_expect == XOPERATOR)
5087 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5094 if (PL_expect != XOPERATOR) {
5095 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5096 PL_expect = XOPERATOR;
5097 force_ident(PL_tokenbuf, '*');
5110 if (PL_expect == XOPERATOR) {
5114 PL_tokenbuf[0] = '%';
5115 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5116 sizeof PL_tokenbuf - 1, FALSE);
5117 if (!PL_tokenbuf[1]) {
5120 PL_pending_ident = '%';
5129 const char tmp = *s++;
5134 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5141 const char tmp = *s++;
5147 goto just_a_word_zero_gv;
5150 switch (PL_expect) {
5156 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5158 PL_bufptr = s; /* update in case we back off */
5160 deprecate(":= for an empty attribute list");
5167 PL_expect = XTERMBLOCK;
5170 stuffstart = s - SvPVX(PL_linestr) - 1;
5174 while (isIDFIRST_lazy_if(s,UTF)) {
5177 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5178 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5179 if (tmp < 0) tmp = -tmp;
5194 sv = newSVpvn(s, len);
5196 d = scan_str(d,TRUE,TRUE);
5198 /* MUST advance bufptr here to avoid bogus
5199 "at end of line" context messages from yyerror().
5201 PL_bufptr = s + len;
5202 yyerror("Unterminated attribute parameter in attribute list");
5206 return REPORT(0); /* EOF indicator */
5210 sv_catsv(sv, PL_lex_stuff);
5211 attrs = append_elem(OP_LIST, attrs,
5212 newSVOP(OP_CONST, 0, sv));
5213 SvREFCNT_dec(PL_lex_stuff);
5214 PL_lex_stuff = NULL;
5217 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5219 if (PL_in_my == KEY_our) {
5220 deprecate(":unique");
5223 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5226 /* NOTE: any CV attrs applied here need to be part of
5227 the CVf_BUILTIN_ATTRS define in cv.h! */
5228 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5230 CvLVALUE_on(PL_compcv);
5232 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5234 deprecate(":locked");
5236 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5238 CvMETHOD_on(PL_compcv);
5240 /* After we've set the flags, it could be argued that
5241 we don't need to do the attributes.pm-based setting
5242 process, and shouldn't bother appending recognized
5243 flags. To experiment with that, uncomment the
5244 following "else". (Note that's already been
5245 uncommented. That keeps the above-applied built-in
5246 attributes from being intercepted (and possibly
5247 rejected) by a package's attribute routines, but is
5248 justified by the performance win for the common case
5249 of applying only built-in attributes.) */
5251 attrs = append_elem(OP_LIST, attrs,
5252 newSVOP(OP_CONST, 0,
5256 if (*s == ':' && s[1] != ':')
5259 break; /* require real whitespace or :'s */
5260 /* XXX losing whitespace on sequential attributes here */
5264 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5265 if (*s != ';' && *s != '}' && *s != tmp
5266 && (tmp != '=' || *s != ')')) {
5267 const char q = ((*s == '\'') ? '"' : '\'');
5268 /* If here for an expression, and parsed no attrs, back
5270 if (tmp == '=' && !attrs) {
5274 /* MUST advance bufptr here to avoid bogus "at end of line"
5275 context messages from yyerror().
5278 yyerror( (const char *)
5280 ? Perl_form(aTHX_ "Invalid separator character "
5281 "%c%c%c in attribute list", q, *s, q)
5282 : "Unterminated attribute list" ) );
5290 start_force(PL_curforce);
5291 NEXTVAL_NEXTTOKE.opval = attrs;
5292 CURMAD('_', PL_nextwhite);
5297 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5298 (s - SvPVX(PL_linestr)) - stuffstart);
5306 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5307 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5315 const char tmp = *s++;
5320 const char tmp = *s++;
5328 if (PL_lex_brackets <= 0)
5329 yyerror("Unmatched right square bracket");
5332 if (PL_lex_state == LEX_INTERPNORMAL) {
5333 if (PL_lex_brackets == 0) {
5334 if (*s == '-' && s[1] == '>')
5335 PL_lex_state = LEX_INTERPENDMAYBE;
5336 else if (*s != '[' && *s != '{')
5337 PL_lex_state = LEX_INTERPEND;
5344 if (PL_lex_brackets > 100) {
5345 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5347 switch (PL_expect) {
5349 if (PL_lex_formbrack) {
5353 if (PL_oldoldbufptr == PL_last_lop)
5354 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5356 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5357 OPERATOR(HASHBRACK);
5359 while (s < PL_bufend && SPACE_OR_TAB(*s))
5362 PL_tokenbuf[0] = '\0';
5363 if (d < PL_bufend && *d == '-') {
5364 PL_tokenbuf[0] = '-';
5366 while (d < PL_bufend && SPACE_OR_TAB(*d))
5369 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5370 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5372 while (d < PL_bufend && SPACE_OR_TAB(*d))
5375 const char minus = (PL_tokenbuf[0] == '-');
5376 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5384 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5389 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5394 if (PL_oldoldbufptr == PL_last_lop)
5395 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5397 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5400 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5402 /* This hack is to get the ${} in the message. */
5404 yyerror("syntax error");
5407 OPERATOR(HASHBRACK);
5409 /* This hack serves to disambiguate a pair of curlies
5410 * as being a block or an anon hash. Normally, expectation
5411 * determines that, but in cases where we're not in a
5412 * position to expect anything in particular (like inside
5413 * eval"") we have to resolve the ambiguity. This code
5414 * covers the case where the first term in the curlies is a
5415 * quoted string. Most other cases need to be explicitly
5416 * disambiguated by prepending a "+" before the opening
5417 * curly in order to force resolution as an anon hash.
5419 * XXX should probably propagate the outer expectation
5420 * into eval"" to rely less on this hack, but that could
5421 * potentially break current behavior of eval"".
5425 if (*s == '\'' || *s == '"' || *s == '`') {
5426 /* common case: get past first string, handling escapes */
5427 for (t++; t < PL_bufend && *t != *s;)
5428 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5432 else if (*s == 'q') {
5435 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5438 /* skip q//-like construct */
5440 char open, close, term;
5443 while (t < PL_bufend && isSPACE(*t))
5445 /* check for q => */
5446 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5447 OPERATOR(HASHBRACK);
5451 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5455 for (t++; t < PL_bufend; t++) {
5456 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5458 else if (*t == open)
5462 for (t++; t < PL_bufend; t++) {
5463 if (*t == '\\' && t+1 < PL_bufend)
5465 else if (*t == close && --brackets <= 0)
5467 else if (*t == open)
5474 /* skip plain q word */
5475 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5478 else if (isALNUM_lazy_if(t,UTF)) {
5480 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5483 while (t < PL_bufend && isSPACE(*t))
5485 /* if comma follows first term, call it an anon hash */
5486 /* XXX it could be a comma expression with loop modifiers */
5487 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5488 || (*t == '=' && t[1] == '>')))
5489 OPERATOR(HASHBRACK);
5490 if (PL_expect == XREF)
5493 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5499 pl_yylval.ival = CopLINE(PL_curcop);
5500 if (isSPACE(*s) || *s == '#')
5501 PL_copline = NOLINE; /* invalidate current command line number */
5506 if (PL_lex_brackets <= 0)
5507 yyerror("Unmatched right curly bracket");
5509 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5510 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5511 PL_lex_formbrack = 0;
5512 if (PL_lex_state == LEX_INTERPNORMAL) {
5513 if (PL_lex_brackets == 0) {
5514 if (PL_expect & XFAKEBRACK) {
5515 PL_expect &= XENUMMASK;
5516 PL_lex_state = LEX_INTERPEND;
5521 PL_thiswhite = newSVpvs("");
5522 sv_catpvs(PL_thiswhite,"}");
5525 return yylex(); /* ignore fake brackets */
5527 if (*s == '-' && s[1] == '>')
5528 PL_lex_state = LEX_INTERPENDMAYBE;
5529 else if (*s != '[' && *s != '{')
5530 PL_lex_state = LEX_INTERPEND;
5533 if (PL_expect & XFAKEBRACK) {
5534 PL_expect &= XENUMMASK;
5536 return yylex(); /* ignore fake brackets */
5538 start_force(PL_curforce);
5540 curmad('X', newSVpvn(s-1,1));
5541 CURMAD('_', PL_thiswhite);
5546 PL_thistoken = newSVpvs("");
5554 if (PL_expect == XOPERATOR) {
5555 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5556 && isIDFIRST_lazy_if(s,UTF))
5558 CopLINE_dec(PL_curcop);
5559 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5560 CopLINE_inc(PL_curcop);
5565 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5567 PL_expect = XOPERATOR;
5568 force_ident(PL_tokenbuf, '&');
5572 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5584 const char tmp = *s++;
5591 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5592 && strchr("+-*/%.^&|<",tmp))
5593 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5594 "Reversed %c= operator",(int)tmp);
5596 if (PL_expect == XSTATE && isALPHA(tmp) &&
5597 (s == PL_linestart+1 || s[-2] == '\n') )
5599 if (PL_in_eval && !PL_rsfp) {
5604 if (strnEQ(s,"=cut",4)) {
5620 PL_thiswhite = newSVpvs("");
5621 sv_catpvn(PL_thiswhite, PL_linestart,
5622 PL_bufend - PL_linestart);
5626 PL_doextract = TRUE;
5630 if (PL_lex_brackets < PL_lex_formbrack) {
5632 #ifdef PERL_STRICT_CR
5633 while (SPACE_OR_TAB(*t))
5635 while (SPACE_OR_TAB(*t) || *t == '\r')
5638 if (*t == '\n' || *t == '#') {
5649 const char tmp = *s++;
5651 /* was this !=~ where !~ was meant?
5652 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5654 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5655 const char *t = s+1;
5657 while (t < PL_bufend && isSPACE(*t))
5660 if (*t == '/' || *t == '?' ||
5661 ((*t == 'm' || *t == 's' || *t == 'y')
5662 && !isALNUM(t[1])) ||
5663 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5664 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5665 "!=~ should be !~");
5675 if (PL_expect != XOPERATOR) {
5676 if (s[1] != '<' && !strchr(s,'>'))
5679 s = scan_heredoc(s);
5681 s = scan_inputsymbol(s);
5682 TERM(sublex_start());
5688 SHop(OP_LEFT_SHIFT);
5702 const char tmp = *s++;
5704 SHop(OP_RIGHT_SHIFT);
5705 else if (tmp == '=')
5714 if (PL_expect == XOPERATOR) {
5715 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5716 return deprecate_commaless_var_list();
5720 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-", s[2]))) {
5721 PL_tokenbuf[0] = '@';
5722 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
5723 sizeof PL_tokenbuf - 1, FALSE);
5724 if (PL_expect == XOPERATOR)
5725 no_op("Array length", s);
5726 if (!PL_tokenbuf[1])
5728 PL_expect = XOPERATOR;
5729 PL_pending_ident = '#';
5733 PL_tokenbuf[0] = '$';
5734 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5735 sizeof PL_tokenbuf - 1, FALSE);
5736 if (PL_expect == XOPERATOR)
5738 if (!PL_tokenbuf[1]) {
5740 yyerror("Final $ should be \\$ or $name");
5744 /* This kludge not intended to be bulletproof. */
5745 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
5746 pl_yylval.opval = newSVOP(OP_CONST, 0,
5747 newSViv(CopARYBASE_get(&PL_compiling)));
5748 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
5754 const char tmp = *s;
5755 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
5758 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
5759 && intuit_more(s)) {
5761 PL_tokenbuf[0] = '@';
5762 if (ckWARN(WARN_SYNTAX)) {
5765 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
5768 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5769 while (t < PL_bufend && *t != ']')
5771 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5772 "Multidimensional syntax %.*s not supported",
5773 (int)((t - PL_bufptr) + 1), PL_bufptr);
5777 else if (*s == '{') {
5779 PL_tokenbuf[0] = '%';
5780 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
5781 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
5783 char tmpbuf[sizeof PL_tokenbuf];
5786 } while (isSPACE(*t));
5787 if (isIDFIRST_lazy_if(t,UTF)) {
5789 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
5793 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
5794 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5795 "You need to quote \"%s\"",
5802 PL_expect = XOPERATOR;
5803 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
5804 const bool islop = (PL_last_lop == PL_oldoldbufptr);
5805 if (!islop || PL_last_lop_op == OP_GREPSTART)
5806 PL_expect = XOPERATOR;
5807 else if (strchr("$@\"'`q", *s))
5808 PL_expect = XTERM; /* e.g. print $fh "foo" */
5809 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
5810 PL_expect = XTERM; /* e.g. print $fh &sub */
5811 else if (isIDFIRST_lazy_if(s,UTF)) {
5812 char tmpbuf[sizeof PL_tokenbuf];
5814 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
5815 if ((t2 = keyword(tmpbuf, len, 0))) {
5816 /* binary operators exclude handle interpretations */
5828 PL_expect = XTERM; /* e.g. print $fh length() */
5833 PL_expect = XTERM; /* e.g. print $fh subr() */
5836 else if (isDIGIT(*s))
5837 PL_expect = XTERM; /* e.g. print $fh 3 */
5838 else if (*s == '.' && isDIGIT(s[1]))
5839 PL_expect = XTERM; /* e.g. print $fh .3 */
5840 else if ((*s == '?' || *s == '-' || *s == '+')
5841 && !isSPACE(s[1]) && s[1] != '=')
5842 PL_expect = XTERM; /* e.g. print $fh -1 */
5843 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
5845 PL_expect = XTERM; /* e.g. print $fh /.../
5846 XXX except DORDOR operator
5848 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
5850 PL_expect = XTERM; /* print $fh <<"EOF" */
5853 PL_pending_ident = '$';
5857 if (PL_expect == XOPERATOR)
5859 PL_tokenbuf[0] = '@';
5860 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
5861 if (!PL_tokenbuf[1]) {
5864 if (PL_lex_state == LEX_NORMAL)
5866 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
5868 PL_tokenbuf[0] = '%';
5870 /* Warn about @ where they meant $. */
5871 if (*s == '[' || *s == '{') {
5872 if (ckWARN(WARN_SYNTAX)) {
5873 const char *t = s + 1;
5874 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
5876 if (*t == '}' || *t == ']') {
5878 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
5879 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5880 "Scalar value %.*s better written as $%.*s",
5881 (int)(t-PL_bufptr), PL_bufptr,
5882 (int)(t-PL_bufptr-1), PL_bufptr+1);
5887 PL_pending_ident = '@';
5890 case '/': /* may be division, defined-or, or pattern */
5891 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
5895 case '?': /* may either be conditional or pattern */
5896 if (PL_expect == XOPERATOR) {
5904 /* A // operator. */
5914 /* Disable warning on "study /blah/" */
5915 if (PL_oldoldbufptr == PL_last_uni
5916 && (*PL_last_uni != 's' || s - PL_last_uni < 5
5917 || memNE(PL_last_uni, "study", 5)
5918 || isALNUM_lazy_if(PL_last_uni+5,UTF)
5921 s = scan_pat(s,OP_MATCH);
5922 TERM(sublex_start());
5926 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
5927 #ifdef PERL_STRICT_CR
5930 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
5932 && (s == PL_linestart || s[-1] == '\n') )
5934 PL_lex_formbrack = 0;
5938 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
5942 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
5948 pl_yylval.ival = OPf_SPECIAL;
5957 case '0': case '1': case '2': case '3': case '4':
5958 case '5': case '6': case '7': case '8': case '9':
5959 s = scan_num(s, &pl_yylval);
5960 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
5961 if (PL_expect == XOPERATOR)
5966 s = scan_str(s,!!PL_madskills,FALSE);
5967 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5968 if (PL_expect == XOPERATOR) {
5969 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5970 return deprecate_commaless_var_list();
5977 pl_yylval.ival = OP_CONST;
5978 TERM(sublex_start());
5981 s = scan_str(s,!!PL_madskills,FALSE);
5982 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
5983 if (PL_expect == XOPERATOR) {
5984 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5985 return deprecate_commaless_var_list();
5992 pl_yylval.ival = OP_CONST;
5993 /* FIXME. I think that this can be const if char *d is replaced by
5994 more localised variables. */
5995 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
5996 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
5997 pl_yylval.ival = OP_STRINGIFY;
6001 TERM(sublex_start());
6004 s = scan_str(s,!!PL_madskills,FALSE);
6005 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6006 if (PL_expect == XOPERATOR)
6007 no_op("Backticks",s);
6010 readpipe_override();
6011 TERM(sublex_start());
6015 if (PL_lex_inwhat && isDIGIT(*s))
6016 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6018 if (PL_expect == XOPERATOR)
6019 no_op("Backslash",s);
6023 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6024 char *start = s + 2;
6025 while (isDIGIT(*start) || *start == '_')
6027 if (*start == '.' && isDIGIT(start[1])) {
6028 s = scan_num(s, &pl_yylval);
6031 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6032 else if (!isALPHA(*start) && (PL_expect == XTERM
6033 || PL_expect == XREF || PL_expect == XSTATE
6034 || PL_expect == XTERMORDORDOR)) {
6035 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
6037 s = scan_num(s, &pl_yylval);
6044 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6087 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6089 /* Some keywords can be followed by any delimiter, including ':' */
6090 anydelim = ((len == 1 && strchr("msyq", PL_tokenbuf[0])) ||
6091 (len == 2 && ((PL_tokenbuf[0] == 't' && PL_tokenbuf[1] == 'r') ||
6092 (PL_tokenbuf[0] == 'q' &&
6093 strchr("qwxr", PL_tokenbuf[1])))));
6095 /* x::* is just a word, unless x is "CORE" */
6096 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6100 while (d < PL_bufend && isSPACE(*d))
6101 d++; /* no comments skipped here, or s### is misparsed */
6103 /* Is this a word before a => operator? */
6104 if (*d == '=' && d[1] == '>') {
6107 = (OP*)newSVOP(OP_CONST, 0,
6108 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6109 pl_yylval.opval->op_private = OPpCONST_BARE;
6113 /* Check for plugged-in keyword */
6117 char *saved_bufptr = PL_bufptr;
6119 result = CALL_FPTR(PL_keyword_plugin)(aTHX_ PL_tokenbuf, len, &o);
6121 if (result == KEYWORD_PLUGIN_DECLINE) {
6122 /* not a plugged-in keyword */
6123 PL_bufptr = saved_bufptr;
6124 } else if (result == KEYWORD_PLUGIN_STMT) {
6125 pl_yylval.opval = o;
6128 return REPORT(PLUGSTMT);
6129 } else if (result == KEYWORD_PLUGIN_EXPR) {
6130 pl_yylval.opval = o;
6132 PL_expect = XOPERATOR;
6133 return REPORT(PLUGEXPR);
6135 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6140 /* Check for built-in keyword */
6141 tmp = keyword(PL_tokenbuf, len, 0);
6143 /* Is this a label? */
6144 if (!anydelim && PL_expect == XSTATE
6145 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6147 Perl_croak(aTHX_ "Can't use keyword '%s' as a label", PL_tokenbuf);
6149 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
6154 if (tmp < 0) { /* second-class keyword? */
6155 GV *ogv = NULL; /* override (winner) */
6156 GV *hgv = NULL; /* hidden (loser) */
6157 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6159 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
6162 if (GvIMPORTED_CV(gv))
6164 else if (! CvMETHOD(cv))
6168 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
6169 (gv = *gvp) && isGV_with_GP(gv) &&
6170 GvCVu(gv) && GvIMPORTED_CV(gv))
6177 tmp = 0; /* overridden by import or by GLOBAL */
6180 && -tmp==KEY_lock /* XXX generalizable kludge */
6183 tmp = 0; /* any sub overrides "weak" keyword */
6185 else { /* no override */
6187 if (tmp == KEY_dump) {
6188 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6189 "dump() better written as CORE::dump()");
6193 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6194 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6195 "Ambiguous call resolved as CORE::%s(), %s",
6196 GvENAME(hgv), "qualify as such or use &");
6203 default: /* not a keyword */
6204 /* Trade off - by using this evil construction we can pull the
6205 variable gv into the block labelled keylookup. If not, then
6206 we have to give it function scope so that the goto from the
6207 earlier ':' case doesn't bypass the initialisation. */
6209 just_a_word_zero_gv:
6217 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6221 SV *nextPL_nextwhite = 0;
6225 /* Get the rest if it looks like a package qualifier */
6227 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6229 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6232 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
6233 *s == '\'' ? "'" : "::");
6238 if (PL_expect == XOPERATOR) {
6239 if (PL_bufptr == PL_linestart) {
6240 CopLINE_dec(PL_curcop);
6241 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6242 CopLINE_inc(PL_curcop);
6245 no_op("Bareword",s);
6248 /* Look for a subroutine with this name in current package,
6249 unless name is "Foo::", in which case Foo is a bearword
6250 (and a package name). */
6252 if (len > 2 && !PL_madskills &&
6253 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6255 if (ckWARN(WARN_BAREWORD)
6256 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
6257 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6258 "Bareword \"%s\" refers to nonexistent package",
6261 PL_tokenbuf[len] = '\0';
6267 /* Mustn't actually add anything to a symbol table.
6268 But also don't want to "initialise" any placeholder
6269 constants that might already be there into full
6270 blown PVGVs with attached PVCV. */
6271 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6272 GV_NOADD_NOINIT, SVt_PVCV);
6277 /* if we saw a global override before, get the right name */
6280 sv = newSVpvs("CORE::GLOBAL::");
6281 sv_catpv(sv,PL_tokenbuf);
6284 /* If len is 0, newSVpv does strlen(), which is correct.
6285 If len is non-zero, then it will be the true length,
6286 and so the scalar will be created correctly. */
6287 sv = newSVpv(PL_tokenbuf,len);
6290 if (PL_madskills && !PL_thistoken) {
6291 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6292 PL_thistoken = newSVpvn(start,s - start);
6293 PL_realtokenstart = s - SvPVX(PL_linestr);
6297 /* Presume this is going to be a bareword of some sort. */
6300 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6301 pl_yylval.opval->op_private = OPpCONST_BARE;
6302 /* UTF-8 package name? */
6303 if (UTF && !IN_BYTES &&
6304 is_utf8_string((U8*)SvPVX_const(sv), SvCUR(sv)))
6307 /* And if "Foo::", then that's what it certainly is. */
6314 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc(sv));
6315 const_op->op_private = OPpCONST_BARE;
6316 rv2cv_op = newCVREF(0, const_op);
6318 if (rv2cv_op->op_type == OP_RV2CV &&
6319 (rv2cv_op->op_flags & OPf_KIDS)) {
6320 OP *rv_op = cUNOPx(rv2cv_op)->op_first;
6321 switch (rv_op->op_type) {
6323 SV *sv = cSVOPx_sv(rv_op);
6324 if (SvROK(sv) && SvTYPE(SvRV(sv)) == SVt_PVCV)
6328 GV *gv = cGVOPx_gv(rv_op);
6329 CV *maybe_cv = GvCVu(gv);
6330 if (maybe_cv && SvTYPE((SV*)maybe_cv) == SVt_PVCV)
6336 /* See if it's the indirect object for a list operator. */
6338 if (PL_oldoldbufptr &&
6339 PL_oldoldbufptr < PL_bufptr &&
6340 (PL_oldoldbufptr == PL_last_lop
6341 || PL_oldoldbufptr == PL_last_uni) &&
6342 /* NO SKIPSPACE BEFORE HERE! */
6343 (PL_expect == XREF ||
6344 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6346 bool immediate_paren = *s == '(';
6348 /* (Now we can afford to cross potential line boundary.) */
6349 s = SKIPSPACE2(s,nextPL_nextwhite);
6351 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6354 /* Two barewords in a row may indicate method call. */
6356 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6357 (tmp = intuit_method(s, gv, cv))) {
6362 /* If not a declared subroutine, it's an indirect object. */
6363 /* (But it's an indir obj regardless for sort.) */
6364 /* Also, if "_" follows a filetest operator, it's a bareword */
6367 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6369 (PL_last_lop_op != OP_MAPSTART &&
6370 PL_last_lop_op != OP_GREPSTART))))
6371 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6372 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6375 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6380 PL_expect = XOPERATOR;
6383 s = SKIPSPACE2(s,nextPL_nextwhite);
6384 PL_nextwhite = nextPL_nextwhite;
6389 /* Is this a word before a => operator? */
6390 if (*s == '=' && s[1] == '>' && !pkgname) {
6393 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6394 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6395 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6399 /* If followed by a paren, it's certainly a subroutine. */
6404 while (SPACE_OR_TAB(*d))
6406 if (*d == ')' && (sv = cv_const_sv(cv))) {
6413 PL_nextwhite = PL_thiswhite;
6416 start_force(PL_curforce);
6418 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6419 PL_expect = XOPERATOR;
6422 PL_nextwhite = nextPL_nextwhite;
6423 curmad('X', PL_thistoken);
6424 PL_thistoken = newSVpvs("");
6433 /* If followed by var or block, call it a method (unless sub) */
6435 if ((*s == '$' || *s == '{') && !cv) {
6437 PL_last_lop = PL_oldbufptr;
6438 PL_last_lop_op = OP_METHOD;
6442 /* If followed by a bareword, see if it looks like indir obj. */
6445 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6446 && (tmp = intuit_method(s, gv, cv))) {
6451 /* Not a method, so call it a subroutine (if defined) */
6454 if (lastchar == '-')
6455 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6456 "Ambiguous use of -%s resolved as -&%s()",
6457 PL_tokenbuf, PL_tokenbuf);
6458 /* Check for a constant sub */
6459 if ((sv = cv_const_sv(cv))) {
6462 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6463 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6464 pl_yylval.opval->op_private = 0;
6468 op_free(pl_yylval.opval);
6469 pl_yylval.opval = rv2cv_op;
6470 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6471 PL_last_lop = PL_oldbufptr;
6472 PL_last_lop_op = OP_ENTERSUB;
6473 /* Is there a prototype? */
6481 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
6484 if ((*proto == '$' || *proto == '_') && proto[1] == '\0')
6486 while (*proto == ';')
6488 if (*proto == '&' && *s == '{') {
6490 sv_setpvs(PL_subname, "__ANON__");
6492 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6499 PL_nextwhite = PL_thiswhite;
6502 start_force(PL_curforce);
6503 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6506 PL_nextwhite = nextPL_nextwhite;
6507 curmad('X', PL_thistoken);
6508 PL_thistoken = newSVpvs("");
6515 /* Guess harder when madskills require "best effort". */
6516 if (PL_madskills && (!gv || !GvCVu(gv))) {
6517 int probable_sub = 0;
6518 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6520 else if (isALPHA(*s)) {
6524 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6525 if (!keyword(tmpbuf, tmplen, 0))
6528 while (d < PL_bufend && isSPACE(*d))
6530 if (*d == '=' && d[1] == '>')
6535 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
6536 op_free(pl_yylval.opval);
6537 pl_yylval.opval = rv2cv_op;
6538 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6539 PL_last_lop = PL_oldbufptr;
6540 PL_last_lop_op = OP_ENTERSUB;
6541 PL_nextwhite = PL_thiswhite;
6543 start_force(PL_curforce);
6544 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6546 PL_nextwhite = nextPL_nextwhite;
6547 curmad('X', PL_thistoken);
6548 PL_thistoken = newSVpvs("");
6553 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6560 /* Call it a bare word */
6562 if (PL_hints & HINT_STRICT_SUBS)
6563 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6566 /* after "print" and similar functions (corresponding to
6567 * "F? L" in opcode.pl), whatever wasn't already parsed as
6568 * a filehandle should be subject to "strict subs".
6569 * Likewise for the optional indirect-object argument to system
6570 * or exec, which can't be a bareword */
6571 if ((PL_last_lop_op == OP_PRINT
6572 || PL_last_lop_op == OP_PRTF
6573 || PL_last_lop_op == OP_SAY
6574 || PL_last_lop_op == OP_SYSTEM
6575 || PL_last_lop_op == OP_EXEC)
6576 && (PL_hints & HINT_STRICT_SUBS))
6577 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6578 if (lastchar != '-') {
6579 if (ckWARN(WARN_RESERVED)) {
6583 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
6584 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6592 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
6593 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6594 "Operator or semicolon missing before %c%s",
6595 lastchar, PL_tokenbuf);
6596 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6597 "Ambiguous use of %c resolved as operator %c",
6598 lastchar, lastchar);
6604 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6605 newSVpv(CopFILE(PL_curcop),0));
6609 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6610 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
6613 case KEY___PACKAGE__:
6614 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6616 ? newSVhek(HvNAME_HEK(PL_curstash))
6623 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6624 const char *pname = "main";
6625 if (PL_tokenbuf[2] == 'D')
6626 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
6627 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
6631 GvIOp(gv) = newIO();
6632 IoIFP(GvIOp(gv)) = PL_rsfp;
6633 #if defined(HAS_FCNTL) && defined(F_SETFD)
6635 const int fd = PerlIO_fileno(PL_rsfp);
6636 fcntl(fd,F_SETFD,fd >= 3);
6639 /* Mark this internal pseudo-handle as clean */
6640 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6641 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6642 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6644 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6645 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6646 /* if the script was opened in binmode, we need to revert
6647 * it to text mode for compatibility; but only iff it has CRs
6648 * XXX this is a questionable hack at best. */
6649 if (PL_bufend-PL_bufptr > 2
6650 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6653 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6654 loc = PerlIO_tell(PL_rsfp);
6655 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6658 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6660 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6661 #endif /* NETWARE */
6662 #ifdef PERLIO_IS_STDIO /* really? */
6663 # if defined(__BORLANDC__)
6664 /* XXX see note in do_binmode() */
6665 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
6669 PerlIO_seek(PL_rsfp, loc, 0);
6673 #ifdef PERLIO_LAYERS
6676 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6677 else if (PL_encoding) {
6684 XPUSHs(PL_encoding);
6686 call_method("name", G_SCALAR);
6690 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
6691 Perl_form(aTHX_ ":encoding(%"SVf")",
6700 if (PL_realtokenstart >= 0) {
6701 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
6703 PL_endwhite = newSVpvs("");
6704 sv_catsv(PL_endwhite, PL_thiswhite);
6706 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
6707 PL_realtokenstart = -1;
6709 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
6725 if (PL_expect == XSTATE) {
6732 if (*s == ':' && s[1] == ':') {
6735 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6736 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
6737 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
6740 else if (tmp == KEY_require || tmp == KEY_do)
6741 /* that's a way to remember we saw "CORE::" */
6754 LOP(OP_ACCEPT,XTERM);
6760 LOP(OP_ATAN2,XTERM);
6766 LOP(OP_BINMODE,XTERM);
6769 LOP(OP_BLESS,XTERM);
6778 /* When 'use switch' is in effect, continue has a dual
6779 life as a control operator. */
6781 if (!FEATURE_IS_ENABLED("switch"))
6784 /* We have to disambiguate the two senses of
6785 "continue". If the next token is a '{' then
6786 treat it as the start of a continue block;
6787 otherwise treat it as a control operator.
6799 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
6816 if (!PL_cryptseen) {
6817 PL_cryptseen = TRUE;
6821 LOP(OP_CRYPT,XTERM);
6824 LOP(OP_CHMOD,XTERM);
6827 LOP(OP_CHOWN,XTERM);
6830 LOP(OP_CONNECT,XTERM);
6849 s = force_word(s,WORD,TRUE,TRUE,FALSE);
6850 if (orig_keyword == KEY_do) {
6859 PL_hints |= HINT_BLOCK_SCOPE;
6869 gv_fetchpvs("AnyDBM_File::ISA", GV_ADDMULTI, SVt_PVAV);
6870 LOP(OP_DBMOPEN,XTERM);
6876 s = force_word(s,WORD,TRUE,FALSE,FALSE);
6883 pl_yylval.ival = CopLINE(PL_curcop);
6899 if (*s == '{') { /* block eval */
6900 PL_expect = XTERMBLOCK;
6901 UNIBRACK(OP_ENTERTRY);
6903 else { /* string eval */
6905 UNIBRACK(OP_ENTEREVAL);
6920 case KEY_endhostent:
6926 case KEY_endservent:
6929 case KEY_endprotoent:
6940 pl_yylval.ival = CopLINE(PL_curcop);
6942 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
6945 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
6948 if ((PL_bufend - p) >= 3 &&
6949 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
6951 else if ((PL_bufend - p) >= 4 &&
6952 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
6955 if (isIDFIRST_lazy_if(p,UTF)) {
6956 p = scan_ident(p, PL_bufend,
6957 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
6961 Perl_croak(aTHX_ "Missing $ on loop variable");
6963 s = SvPVX(PL_linestr) + soff;
6969 LOP(OP_FORMLINE,XTERM);
6975 LOP(OP_FCNTL,XTERM);
6981 LOP(OP_FLOCK,XTERM);
6990 LOP(OP_GREPSTART, XREF);
6993 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7008 case KEY_getpriority:
7009 LOP(OP_GETPRIORITY,XTERM);
7011 case KEY_getprotobyname:
7014 case KEY_getprotobynumber:
7015 LOP(OP_GPBYNUMBER,XTERM);
7017 case KEY_getprotoent:
7029 case KEY_getpeername:
7030 UNI(OP_GETPEERNAME);
7032 case KEY_gethostbyname:
7035 case KEY_gethostbyaddr:
7036 LOP(OP_GHBYADDR,XTERM);
7038 case KEY_gethostent:
7041 case KEY_getnetbyname:
7044 case KEY_getnetbyaddr:
7045 LOP(OP_GNBYADDR,XTERM);
7050 case KEY_getservbyname:
7051 LOP(OP_GSBYNAME,XTERM);
7053 case KEY_getservbyport:
7054 LOP(OP_GSBYPORT,XTERM);
7056 case KEY_getservent:
7059 case KEY_getsockname:
7060 UNI(OP_GETSOCKNAME);
7062 case KEY_getsockopt:
7063 LOP(OP_GSOCKOPT,XTERM);
7078 pl_yylval.ival = CopLINE(PL_curcop);
7088 pl_yylval.ival = CopLINE(PL_curcop);
7092 LOP(OP_INDEX,XTERM);
7098 LOP(OP_IOCTL,XTERM);
7110 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7142 LOP(OP_LISTEN,XTERM);
7151 s = scan_pat(s,OP_MATCH);
7152 TERM(sublex_start());
7155 LOP(OP_MAPSTART, XREF);
7158 LOP(OP_MKDIR,XTERM);
7161 LOP(OP_MSGCTL,XTERM);
7164 LOP(OP_MSGGET,XTERM);
7167 LOP(OP_MSGRCV,XTERM);
7170 LOP(OP_MSGSND,XTERM);
7175 PL_in_my = (U16)tmp;
7177 if (isIDFIRST_lazy_if(s,UTF)) {
7181 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7182 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7184 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7185 if (!PL_in_my_stash) {
7188 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7192 if (PL_madskills) { /* just add type to declarator token */
7193 sv_catsv(PL_thistoken, PL_nextwhite);
7195 sv_catpvn(PL_thistoken, start, s - start);
7203 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7210 s = tokenize_use(0, s);
7214 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7221 if (isIDFIRST_lazy_if(s,UTF)) {
7223 for (d = s; isALNUM_lazy_if(d,UTF);)
7225 for (t=d; isSPACE(*t);)
7227 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7229 && !(t[0] == '=' && t[1] == '>')
7231 int parms_len = (int)(d-s);
7232 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7233 "Precedence problem: open %.*s should be open(%.*s)",
7234 parms_len, s, parms_len, s);
7240 pl_yylval.ival = OP_OR;
7250 LOP(OP_OPEN_DIR,XTERM);
7253 checkcomma(s,PL_tokenbuf,"filehandle");
7257 checkcomma(s,PL_tokenbuf,"filehandle");
7276 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7278 s = force_strict_version(s);
7282 LOP(OP_PIPE_OP,XTERM);
7285 s = scan_str(s,!!PL_madskills,FALSE);
7288 pl_yylval.ival = OP_CONST;
7289 TERM(sublex_start());
7295 s = scan_str(s,!!PL_madskills,FALSE);
7298 PL_expect = XOPERATOR;
7300 if (SvCUR(PL_lex_stuff)) {
7303 d = SvPV_force(PL_lex_stuff, len);
7305 for (; isSPACE(*d) && len; --len, ++d)
7310 if (!warned && ckWARN(WARN_QW)) {
7311 for (; !isSPACE(*d) && len; --len, ++d) {
7313 Perl_warner(aTHX_ packWARN(WARN_QW),
7314 "Possible attempt to separate words with commas");
7317 else if (*d == '#') {
7318 Perl_warner(aTHX_ packWARN(WARN_QW),
7319 "Possible attempt to put comments in qw() list");
7325 for (; !isSPACE(*d) && len; --len, ++d)
7328 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7329 words = append_elem(OP_LIST, words,
7330 newSVOP(OP_CONST, 0, tokeq(sv)));
7334 start_force(PL_curforce);
7335 NEXTVAL_NEXTTOKE.opval = words;
7340 SvREFCNT_dec(PL_lex_stuff);
7341 PL_lex_stuff = NULL;
7347 s = scan_str(s,!!PL_madskills,FALSE);
7350 pl_yylval.ival = OP_STRINGIFY;
7351 if (SvIVX(PL_lex_stuff) == '\'')
7352 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should intepolate */
7353 TERM(sublex_start());
7356 s = scan_pat(s,OP_QR);
7357 TERM(sublex_start());
7360 s = scan_str(s,!!PL_madskills,FALSE);
7363 readpipe_override();
7364 TERM(sublex_start());
7372 s = force_version(s, FALSE);
7374 else if (*s != 'v' || !isDIGIT(s[1])
7375 || (s = force_version(s, TRUE), *s == 'v'))
7377 *PL_tokenbuf = '\0';
7378 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7379 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7380 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7382 yyerror("<> should be quotes");
7384 if (orig_keyword == KEY_require) {
7392 PL_last_uni = PL_oldbufptr;
7393 PL_last_lop_op = OP_REQUIRE;
7395 return REPORT( (int)REQUIRE );
7401 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7405 LOP(OP_RENAME,XTERM);
7414 LOP(OP_RINDEX,XTERM);
7423 UNIDOR(OP_READLINE);
7426 UNIDOR(OP_BACKTICK);
7435 LOP(OP_REVERSE,XTERM);
7438 UNIDOR(OP_READLINK);
7445 if (pl_yylval.opval)
7446 TERM(sublex_start());
7448 TOKEN(1); /* force error */
7451 checkcomma(s,PL_tokenbuf,"filehandle");
7461 LOP(OP_SELECT,XTERM);
7467 LOP(OP_SEMCTL,XTERM);
7470 LOP(OP_SEMGET,XTERM);
7473 LOP(OP_SEMOP,XTERM);
7479 LOP(OP_SETPGRP,XTERM);
7481 case KEY_setpriority:
7482 LOP(OP_SETPRIORITY,XTERM);
7484 case KEY_sethostent:
7490 case KEY_setservent:
7493 case KEY_setprotoent:
7503 LOP(OP_SEEKDIR,XTERM);
7505 case KEY_setsockopt:
7506 LOP(OP_SSOCKOPT,XTERM);
7512 LOP(OP_SHMCTL,XTERM);
7515 LOP(OP_SHMGET,XTERM);
7518 LOP(OP_SHMREAD,XTERM);
7521 LOP(OP_SHMWRITE,XTERM);
7524 LOP(OP_SHUTDOWN,XTERM);
7533 LOP(OP_SOCKET,XTERM);
7535 case KEY_socketpair:
7536 LOP(OP_SOCKPAIR,XTERM);
7539 checkcomma(s,PL_tokenbuf,"subroutine name");
7541 if (*s == ';' || *s == ')') /* probably a close */
7542 Perl_croak(aTHX_ "sort is now a reserved word");
7544 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7548 LOP(OP_SPLIT,XTERM);
7551 LOP(OP_SPRINTF,XTERM);
7554 LOP(OP_SPLICE,XTERM);
7569 LOP(OP_SUBSTR,XTERM);
7575 char tmpbuf[sizeof PL_tokenbuf];
7576 SSize_t tboffset = 0;
7577 expectation attrful;
7578 bool have_name, have_proto;
7579 const int key = tmp;
7584 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7585 SV *subtoken = newSVpvn(tstart, s - tstart);
7589 s = SKIPSPACE2(s,tmpwhite);
7594 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7595 (*s == ':' && s[1] == ':'))
7598 SV *nametoke = NULL;
7602 attrful = XATTRBLOCK;
7603 /* remember buffer pos'n for later force_word */
7604 tboffset = s - PL_oldbufptr;
7605 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7608 nametoke = newSVpvn(s, d - s);
7610 if (memchr(tmpbuf, ':', len))
7611 sv_setpvn(PL_subname, tmpbuf, len);
7613 sv_setsv(PL_subname,PL_curstname);
7614 sv_catpvs(PL_subname,"::");
7615 sv_catpvn(PL_subname,tmpbuf,len);
7622 CURMAD('X', nametoke);
7623 CURMAD('_', tmpwhite);
7624 (void) force_word(PL_oldbufptr + tboffset, WORD,
7627 s = SKIPSPACE2(d,tmpwhite);
7634 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7635 PL_expect = XTERMBLOCK;
7636 attrful = XATTRTERM;
7637 sv_setpvs(PL_subname,"?");
7641 if (key == KEY_format) {
7643 PL_lex_formbrack = PL_lex_brackets + 1;
7645 PL_thistoken = subtoken;
7649 (void) force_word(PL_oldbufptr + tboffset, WORD,
7655 /* Look for a prototype */
7658 bool bad_proto = FALSE;
7659 bool in_brackets = FALSE;
7660 char greedy_proto = ' ';
7661 bool proto_after_greedy_proto = FALSE;
7662 bool must_be_last = FALSE;
7663 bool underscore = FALSE;
7664 bool seen_underscore = FALSE;
7665 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
7667 s = scan_str(s,!!PL_madskills,FALSE);
7669 Perl_croak(aTHX_ "Prototype not terminated");
7670 /* strip spaces and check for bad characters */
7671 d = SvPVX(PL_lex_stuff);
7673 for (p = d; *p; ++p) {
7677 if (warnillegalproto) {
7679 proto_after_greedy_proto = TRUE;
7680 if (!strchr("$@%*;[]&\\_", *p)) {
7692 else if ( *p == ']' ) {
7693 in_brackets = FALSE;
7695 else if ( (*p == '@' || *p == '%') &&
7696 ( tmp < 2 || d[tmp-2] != '\\' ) &&
7698 must_be_last = TRUE;
7701 else if ( *p == '_' ) {
7702 underscore = seen_underscore = TRUE;
7709 if (proto_after_greedy_proto)
7710 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7711 "Prototype after '%c' for %"SVf" : %s",
7712 greedy_proto, SVfARG(PL_subname), d);
7714 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
7715 "Illegal character %sin prototype for %"SVf" : %s",
7716 seen_underscore ? "after '_' " : "",
7717 SVfARG(PL_subname), d);
7718 SvCUR_set(PL_lex_stuff, tmp);
7723 CURMAD('q', PL_thisopen);
7724 CURMAD('_', tmpwhite);
7725 CURMAD('=', PL_thisstuff);
7726 CURMAD('Q', PL_thisclose);
7727 NEXTVAL_NEXTTOKE.opval =
7728 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7729 PL_lex_stuff = NULL;
7732 s = SKIPSPACE2(s,tmpwhite);
7740 if (*s == ':' && s[1] != ':')
7741 PL_expect = attrful;
7742 else if (*s != '{' && key == KEY_sub) {
7744 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
7745 else if (*s != ';' && *s != '}')
7746 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
7753 curmad('^', newSVpvs(""));
7754 CURMAD('_', tmpwhite);
7758 PL_thistoken = subtoken;
7761 NEXTVAL_NEXTTOKE.opval =
7762 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
7763 PL_lex_stuff = NULL;
7769 sv_setpvs(PL_subname, "__ANON__");
7771 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7775 (void) force_word(PL_oldbufptr + tboffset, WORD,
7784 LOP(OP_SYSTEM,XREF);
7787 LOP(OP_SYMLINK,XTERM);
7790 LOP(OP_SYSCALL,XTERM);
7793 LOP(OP_SYSOPEN,XTERM);
7796 LOP(OP_SYSSEEK,XTERM);
7799 LOP(OP_SYSREAD,XTERM);
7802 LOP(OP_SYSWRITE,XTERM);
7806 TERM(sublex_start());
7827 LOP(OP_TRUNCATE,XTERM);
7839 pl_yylval.ival = CopLINE(PL_curcop);
7843 pl_yylval.ival = CopLINE(PL_curcop);
7847 LOP(OP_UNLINK,XTERM);
7853 LOP(OP_UNPACK,XTERM);
7856 LOP(OP_UTIME,XTERM);
7862 LOP(OP_UNSHIFT,XTERM);
7865 s = tokenize_use(1, s);
7875 pl_yylval.ival = CopLINE(PL_curcop);
7879 pl_yylval.ival = CopLINE(PL_curcop);
7883 PL_hints |= HINT_BLOCK_SCOPE;
7890 LOP(OP_WAITPID,XTERM);
7899 ctl_l[0] = toCTRL('L');
7901 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
7904 /* Make sure $^L is defined */
7905 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
7910 if (PL_expect == XOPERATOR)
7916 pl_yylval.ival = OP_XOR;
7921 TERM(sublex_start());
7926 #pragma segment Main
7930 S_pending_ident(pTHX)
7935 /* pit holds the identifier we read and pending_ident is reset */
7936 char pit = PL_pending_ident;
7937 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
7938 /* All routes through this function want to know if there is a colon. */
7939 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
7940 PL_pending_ident = 0;
7942 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
7943 DEBUG_T({ PerlIO_printf(Perl_debug_log,
7944 "### Pending identifier '%s'\n", PL_tokenbuf); });
7946 /* if we're in a my(), we can't allow dynamics here.
7947 $foo'bar has already been turned into $foo::bar, so
7948 just check for colons.
7950 if it's a legal name, the OP is a PADANY.
7953 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
7955 yyerror(Perl_form(aTHX_ "No package name allowed for "
7956 "variable %s in \"our\"",
7958 tmp = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7962 yyerror(Perl_form(aTHX_ PL_no_myglob,
7963 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
7965 pl_yylval.opval = newOP(OP_PADANY, 0);
7966 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len, 0);
7972 build the ops for accesses to a my() variable.
7974 Deny my($a) or my($b) in a sort block, *if* $a or $b is
7975 then used in a comparison. This catches most, but not
7976 all cases. For instance, it catches
7977 sort { my($a); $a <=> $b }
7979 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
7980 (although why you'd do that is anyone's guess).
7985 tmp = pad_findmy(PL_tokenbuf, tokenbuf_len, 0);
7986 if (tmp != NOT_IN_PAD) {
7987 /* might be an "our" variable" */
7988 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
7989 /* build ops for a bareword */
7990 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
7991 HEK * const stashname = HvNAME_HEK(stash);
7992 SV * const sym = newSVhek(stashname);
7993 sv_catpvs(sym, "::");
7994 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
7995 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
7996 pl_yylval.opval->op_private = OPpCONST_ENTERED;
7999 ? (GV_ADDMULTI | GV_ADDINEVAL)
8002 ((PL_tokenbuf[0] == '$') ? SVt_PV
8003 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8008 /* if it's a sort block and they're naming $a or $b */
8009 if (PL_last_lop_op == OP_SORT &&
8010 PL_tokenbuf[0] == '$' &&
8011 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
8014 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
8015 d < PL_bufend && *d != '\n';
8018 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
8019 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
8025 pl_yylval.opval = newOP(OP_PADANY, 0);
8026 pl_yylval.opval->op_targ = tmp;
8032 Whine if they've said @foo in a doublequoted string,
8033 and @foo isn't a variable we can find in the symbol
8036 if (ckWARN(WARN_AMBIGUOUS) &&
8037 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8038 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
8040 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8041 /* DO NOT warn for @- and @+ */
8042 && !( PL_tokenbuf[2] == '\0' &&
8043 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8046 /* Downgraded from fatal to warning 20000522 mjd */
8047 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8048 "Possible unintended interpolation of %s in string",
8053 /* build ops for a bareword */
8054 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
8056 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8058 PL_tokenbuf + 1, tokenbuf_len - 1,
8059 /* If the identifier refers to a stash, don't autovivify it.
8060 * Change 24660 had the side effect of causing symbol table
8061 * hashes to always be defined, even if they were freshly
8062 * created and the only reference in the entire program was
8063 * the single statement with the defined %foo::bar:: test.
8064 * It appears that all code in the wild doing this actually
8065 * wants to know whether sub-packages have been loaded, so
8066 * by avoiding auto-vivifying symbol tables, we ensure that
8067 * defined %foo::bar:: continues to be false, and the existing
8068 * tests still give the expected answers, even though what
8069 * they're actually testing has now changed subtly.
8071 (*PL_tokenbuf == '%'
8072 && *(d = PL_tokenbuf + tokenbuf_len - 1) == ':'
8075 : PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD),
8076 ((PL_tokenbuf[0] == '$') ? SVt_PV
8077 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8083 * The following code was generated by perl_keyword.pl.
8087 Perl_keyword (pTHX_ const char *name, I32 len, bool all_keywords)
8091 PERL_ARGS_ASSERT_KEYWORD;
8095 case 1: /* 5 tokens of length 1 */
8127 case 2: /* 18 tokens of length 2 */
8273 case 3: /* 29 tokens of length 3 */
8277 if (name[1] == 'N' &&
8340 if (name[1] == 'i' &&
8372 if (name[1] == 'o' &&
8381 if (name[1] == 'e' &&
8390 if (name[1] == 'n' &&
8399 if (name[1] == 'o' &&
8408 if (name[1] == 'a' &&
8417 if (name[1] == 'o' &&
8479 if (name[1] == 'e' &&
8493 return (all_keywords || FEATURE_IS_ENABLED("say") ? KEY_say : 0);
8519 if (name[1] == 'i' &&
8528 if (name[1] == 's' &&
8537 if (name[1] == 'e' &&
8546 if (name[1] == 'o' &&
8558 case 4: /* 41 tokens of length 4 */
8562 if (name[1] == 'O' &&
8572 if (name[1] == 'N' &&
8582 if (name[1] == 'i' &&
8592 if (name[1] == 'h' &&
8602 if (name[1] == 'u' &&
8615 if (name[2] == 'c' &&
8624 if (name[2] == 's' &&
8633 if (name[2] == 'a' &&
8669 if (name[1] == 'o' &&
8682 if (name[2] == 't' &&
8691 if (name[2] == 'o' &&
8700 if (name[2] == 't' &&
8709 if (name[2] == 'e' &&
8722 if (name[1] == 'o' &&
8735 if (name[2] == 'y' &&
8744 if (name[2] == 'l' &&
8760 if (name[2] == 's' &&
8769 if (name[2] == 'n' &&
8778 if (name[2] == 'c' &&
8791 if (name[1] == 'e' &&
8801 if (name[1] == 'p' &&
8814 if (name[2] == 'c' &&
8823 if (name[2] == 'p' &&
8832 if (name[2] == 's' &&
8848 if (name[2] == 'n' &&
8918 if (name[2] == 'r' &&
8927 if (name[2] == 'r' &&
8936 if (name[2] == 'a' &&
8952 if (name[2] == 'l' &&
9014 if (name[2] == 'e' &&
9017 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_when : 0);
9030 case 5: /* 39 tokens of length 5 */
9034 if (name[1] == 'E' &&
9045 if (name[1] == 'H' &&
9059 if (name[2] == 'a' &&
9069 if (name[2] == 'a' &&
9086 if (name[2] == 'e' &&
9096 if (name[2] == 'e' &&
9100 return (all_keywords || FEATURE_IS_ENABLED("switch") ? -KEY_break : 0);
9116 if (name[3] == 'i' &&
9125 if (name[3] == 'o' &&
9161 if (name[2] == 'o' &&
9171 if (name[2] == 'y' &&
9185 if (name[1] == 'l' &&
9199 if (name[2] == 'n' &&
9209 if (name[2] == 'o' &&
9223 if (name[1] == 'i' &&
9228 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_given : 0);
9237 if (name[2] == 'd' &&
9247 if (name[2] == 'c' &&
9264 if (name[2] == 'c' &&
9274 if (name[2] == 't' &&
9288 if (name[1] == 'k' &&
9299 if (name[1] == 'r' &&
9313 if (name[2] == 's' &&
9323 if (name[2] == 'd' &&
9340 if (name[2] == 'm' &&
9350 if (name[2] == 'i' &&
9360 if (name[2] == 'e' &&
9370 if (name[2] == 'l' &&
9380 if (name[2] == 'a' &&
9393 if (name[3] == 't' &&
9396 return (all_keywords || FEATURE_IS_ENABLED("state") ? KEY_state : 0);
9402 if (name[3] == 'd' &&
9419 if (name[1] == 'i' &&
9433 if (name[2] == 'a' &&
9446 if (name[3] == 'e' &&
9481 if (name[2] == 'i' &&
9498 if (name[2] == 'i' &&
9508 if (name[2] == 'i' &&
9525 case 6: /* 33 tokens of length 6 */
9529 if (name[1] == 'c' &&
9544 if (name[2] == 'l' &&
9555 if (name[2] == 'r' &&
9570 if (name[1] == 'e' &&
9585 if (name[2] == 's' &&
9590 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "elseif should be elsif");
9596 if (name[2] == 'i' &&
9614 if (name[2] == 'l' &&
9625 if (name[2] == 'r' &&
9640 if (name[1] == 'm' &&
9655 if (name[2] == 'n' &&
9666 if (name[2] == 's' &&
9681 if (name[1] == 's' &&
9687 if (name[4] == 't' &&
9696 if (name[4] == 'e' &&
9705 if (name[4] == 'c' &&
9714 if (name[4] == 'n' &&
9730 if (name[1] == 'r' &&
9748 if (name[3] == 'a' &&
9758 if (name[3] == 'u' &&
9772 if (name[2] == 'n' &&
9790 if (name[2] == 'a' &&
9804 if (name[3] == 'e' &&
9817 if (name[4] == 't' &&
9826 if (name[4] == 'e' &&
9848 if (name[4] == 't' &&
9857 if (name[4] == 'e' &&
9873 if (name[2] == 'c' &&
9884 if (name[2] == 'l' &&
9895 if (name[2] == 'b' &&
9906 if (name[2] == 's' &&
9929 if (name[4] == 's' &&
9938 if (name[4] == 'n' &&
9951 if (name[3] == 'a' &&
9968 if (name[1] == 'a' &&
9983 case 7: /* 29 tokens of length 7 */
9987 if (name[1] == 'E' &&
10000 if (name[1] == '_' &&
10007 return KEY___END__;
10013 if (name[1] == 'i' &&
10020 return -KEY_binmode;
10026 if (name[1] == 'o' &&
10033 return -KEY_connect;
10042 if (name[2] == 'm' &&
10048 return -KEY_dbmopen;
10054 if (name[2] == 'f')
10059 if (name[4] == 'u' &&
10063 return (all_keywords || FEATURE_IS_ENABLED("switch") ? KEY_default : 0);
10069 if (name[4] == 'n' &&
10073 return KEY_defined;
10090 if (name[1] == 'o' &&
10097 return KEY_foreach;
10103 if (name[1] == 'e' &&
10110 if (name[5] == 'r' &&
10113 return -KEY_getpgrp;
10119 if (name[5] == 'i' &&
10122 return -KEY_getppid;
10135 if (name[1] == 'c' &&
10142 return -KEY_lcfirst;
10148 if (name[1] == 'p' &&
10155 return -KEY_opendir;
10161 if (name[1] == 'a' &&
10168 return KEY_package;
10174 if (name[1] == 'e')
10179 if (name[3] == 'd' &&
10184 return -KEY_readdir;
10190 if (name[3] == 'u' &&
10195 return KEY_require;
10201 if (name[3] == 'e' &&
10206 return -KEY_reverse;
10225 if (name[3] == 'k' &&
10230 return -KEY_seekdir;
10236 if (name[3] == 'p' &&
10241 return -KEY_setpgrp;
10251 if (name[2] == 'm' &&
10257 return -KEY_shmread;
10263 if (name[2] == 'r' &&
10269 return -KEY_sprintf;
10278 if (name[3] == 'l' &&
10283 return -KEY_symlink;
10292 if (name[4] == 'a' &&
10296 return -KEY_syscall;
10302 if (name[4] == 'p' &&
10306 return -KEY_sysopen;
10312 if (name[4] == 'e' &&
10316 return -KEY_sysread;
10322 if (name[4] == 'e' &&
10326 return -KEY_sysseek;
10344 if (name[1] == 'e' &&
10351 return -KEY_telldir;
10360 if (name[2] == 'f' &&
10366 return -KEY_ucfirst;
10372 if (name[2] == 's' &&
10378 return -KEY_unshift;
10388 if (name[1] == 'a' &&
10395 return -KEY_waitpid;
10404 case 8: /* 26 tokens of length 8 */
10408 if (name[1] == 'U' &&
10416 return KEY_AUTOLOAD;
10422 if (name[1] == '_')
10427 if (name[3] == 'A' &&
10433 return KEY___DATA__;
10439 if (name[3] == 'I' &&
10445 return -KEY___FILE__;
10451 if (name[3] == 'I' &&
10457 return -KEY___LINE__;
10473 if (name[2] == 'o' &&
10480 return -KEY_closedir;
10486 if (name[2] == 'n' &&
10493 return -KEY_continue;
10503 if (name[1] == 'b' &&
10511 return -KEY_dbmclose;
10517 if (name[1] == 'n' &&
10523 if (name[4] == 'r' &&
10528 return -KEY_endgrent;
10534 if (name[4] == 'w' &&
10539 return -KEY_endpwent;
10552 if (name[1] == 'o' &&
10560 return -KEY_formline;
10566 if (name[1] == 'e' &&
10572 if (name[4] == 'r')
10577 if (name[6] == 'n' &&
10580 return -KEY_getgrent;
10586 if (name[6] == 'i' &&
10589 return -KEY_getgrgid;
10595 if (name[6] == 'a' &&
10598 return -KEY_getgrnam;
10611 if (name[4] == 'o' &&
10616 return -KEY_getlogin;
10622 if (name[4] == 'w')
10627 if (name[6] == 'n' &&
10630 return -KEY_getpwent;
10636 if (name[6] == 'a' &&
10639 return -KEY_getpwnam;
10645 if (name[6] == 'i' &&
10648 return -KEY_getpwuid;
10668 if (name[1] == 'e' &&
10675 if (name[5] == 'i' &&
10682 return -KEY_readline;
10687 return -KEY_readlink;
10698 if (name[5] == 'i' &&
10702 return -KEY_readpipe;
10718 if (name[2] == 't')
10723 if (name[4] == 'r' &&
10728 return -KEY_setgrent;
10734 if (name[4] == 'w' &&
10739 return -KEY_setpwent;
10755 if (name[3] == 'w' &&
10761 return -KEY_shmwrite;
10767 if (name[3] == 't' &&
10773 return -KEY_shutdown;
10783 if (name[2] == 's' &&
10790 return -KEY_syswrite;
10800 if (name[1] == 'r' &&
10808 return -KEY_truncate;
10817 case 9: /* 9 tokens of length 9 */
10821 if (name[1] == 'N' &&
10830 return KEY_UNITCHECK;
10836 if (name[1] == 'n' &&
10845 return -KEY_endnetent;
10851 if (name[1] == 'e' &&
10860 return -KEY_getnetent;
10866 if (name[1] == 'o' &&
10875 return -KEY_localtime;
10881 if (name[1] == 'r' &&
10890 return KEY_prototype;
10896 if (name[1] == 'u' &&
10905 return -KEY_quotemeta;
10911 if (name[1] == 'e' &&
10920 return -KEY_rewinddir;
10926 if (name[1] == 'e' &&
10935 return -KEY_setnetent;
10941 if (name[1] == 'a' &&
10950 return -KEY_wantarray;
10959 case 10: /* 9 tokens of length 10 */
10963 if (name[1] == 'n' &&
10969 if (name[4] == 'o' &&
10976 return -KEY_endhostent;
10982 if (name[4] == 'e' &&
10989 return -KEY_endservent;
11002 if (name[1] == 'e' &&
11008 if (name[4] == 'o' &&
11015 return -KEY_gethostent;
11024 if (name[5] == 'r' &&
11030 return -KEY_getservent;
11036 if (name[5] == 'c' &&
11042 return -KEY_getsockopt;
11062 if (name[2] == 't')
11067 if (name[4] == 'o' &&
11074 return -KEY_sethostent;
11083 if (name[5] == 'r' &&
11089 return -KEY_setservent;
11095 if (name[5] == 'c' &&
11101 return -KEY_setsockopt;
11118 if (name[2] == 'c' &&
11127 return -KEY_socketpair;
11140 case 11: /* 8 tokens of length 11 */
11144 if (name[1] == '_' &&
11154 { /* __PACKAGE__ */
11155 return -KEY___PACKAGE__;
11161 if (name[1] == 'n' &&
11171 { /* endprotoent */
11172 return -KEY_endprotoent;
11178 if (name[1] == 'e' &&
11187 if (name[5] == 'e' &&
11193 { /* getpeername */
11194 return -KEY_getpeername;
11203 if (name[6] == 'o' &&
11208 { /* getpriority */
11209 return -KEY_getpriority;
11215 if (name[6] == 't' &&
11220 { /* getprotoent */
11221 return -KEY_getprotoent;
11235 if (name[4] == 'o' &&
11242 { /* getsockname */
11243 return -KEY_getsockname;
11256 if (name[1] == 'e' &&
11264 if (name[6] == 'o' &&
11269 { /* setpriority */
11270 return -KEY_setpriority;
11276 if (name[6] == 't' &&
11281 { /* setprotoent */
11282 return -KEY_setprotoent;
11298 case 12: /* 2 tokens of length 12 */
11299 if (name[0] == 'g' &&
11311 if (name[9] == 'd' &&
11314 { /* getnetbyaddr */
11315 return -KEY_getnetbyaddr;
11321 if (name[9] == 'a' &&
11324 { /* getnetbyname */
11325 return -KEY_getnetbyname;
11337 case 13: /* 4 tokens of length 13 */
11338 if (name[0] == 'g' &&
11345 if (name[4] == 'o' &&
11354 if (name[10] == 'd' &&
11357 { /* gethostbyaddr */
11358 return -KEY_gethostbyaddr;
11364 if (name[10] == 'a' &&
11367 { /* gethostbyname */
11368 return -KEY_gethostbyname;
11381 if (name[4] == 'e' &&
11390 if (name[10] == 'a' &&
11393 { /* getservbyname */
11394 return -KEY_getservbyname;
11400 if (name[10] == 'o' &&
11403 { /* getservbyport */
11404 return -KEY_getservbyport;
11423 case 14: /* 1 tokens of length 14 */
11424 if (name[0] == 'g' &&
11438 { /* getprotobyname */
11439 return -KEY_getprotobyname;
11444 case 16: /* 1 tokens of length 16 */
11445 if (name[0] == 'g' &&
11461 { /* getprotobynumber */
11462 return -KEY_getprotobynumber;
11476 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
11480 PERL_ARGS_ASSERT_CHECKCOMMA;
11482 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
11483 if (ckWARN(WARN_SYNTAX)) {
11486 for (w = s+2; *w && level; w++) {
11489 else if (*w == ')')
11492 while (isSPACE(*w))
11494 /* the list of chars below is for end of statements or
11495 * block / parens, boolean operators (&&, ||, //) and branch
11496 * constructs (or, and, if, until, unless, while, err, for).
11497 * Not a very solid hack... */
11498 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
11499 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
11500 "%s (...) interpreted as function",name);
11503 while (s < PL_bufend && isSPACE(*s))
11507 while (s < PL_bufend && isSPACE(*s))
11509 if (isIDFIRST_lazy_if(s,UTF)) {
11510 const char * const w = s++;
11511 while (isALNUM_lazy_if(s,UTF))
11513 while (s < PL_bufend && isSPACE(*s))
11517 if (keyword(w, s - w, 0))
11520 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
11521 if (gv && GvCVu(gv))
11523 Perl_croak(aTHX_ "No comma allowed after %s", what);
11528 /* Either returns sv, or mortalizes sv and returns a new SV*.
11529 Best used as sv=new_constant(..., sv, ...).
11530 If s, pv are NULL, calls subroutine with one argument,
11531 and type is used with error messages only. */
11534 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
11535 SV *sv, SV *pv, const char *type, STRLEN typelen)
11538 HV * const table = GvHV(PL_hintgv); /* ^H */
11542 const char *why1 = "", *why2 = "", *why3 = "";
11544 PERL_ARGS_ASSERT_NEW_CONSTANT;
11546 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
11549 why2 = (const char *)
11550 (strEQ(key,"charnames")
11551 ? "(possibly a missing \"use charnames ...\")"
11553 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
11554 (type ? type: "undef"), why2);
11556 /* This is convoluted and evil ("goto considered harmful")
11557 * but I do not understand the intricacies of all the different
11558 * failure modes of %^H in here. The goal here is to make
11559 * the most probable error message user-friendly. --jhi */
11564 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
11565 (type ? type: "undef"), why1, why2, why3);
11567 yyerror(SvPVX_const(msg));
11572 /* charnames doesn't work well if there have been errors found */
11573 if (PL_error_count > 0 && strEQ(key,"charnames"))
11574 return &PL_sv_undef;
11576 cvp = hv_fetch(table, key, keylen, FALSE);
11577 if (!cvp || !SvOK(*cvp)) {
11580 why3 = "} is not defined";
11583 sv_2mortal(sv); /* Parent created it permanently */
11586 pv = newSVpvn_flags(s, len, SVs_TEMP);
11588 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
11590 typesv = &PL_sv_undef;
11592 PUSHSTACKi(PERLSI_OVERLOAD);
11604 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
11608 /* Check the eval first */
11609 if (!PL_in_eval && SvTRUE(ERRSV)) {
11610 sv_catpvs(ERRSV, "Propagated");
11611 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
11613 res = SvREFCNT_inc_simple(sv);
11617 SvREFCNT_inc_simple_void(res);
11626 why1 = "Call to &{$^H{";
11628 why3 = "}} did not return a defined value";
11636 /* Returns a NUL terminated string, with the length of the string written to
11640 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
11643 register char *d = dest;
11644 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11646 PERL_ARGS_ASSERT_SCAN_WORD;
11650 Perl_croak(aTHX_ ident_too_long);
11651 if (isALNUM(*s)) /* UTF handled below */
11653 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
11658 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
11662 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11663 char *t = s + UTF8SKIP(s);
11665 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11669 Perl_croak(aTHX_ ident_too_long);
11670 Copy(s, d, len, char);
11683 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
11686 char *bracket = NULL;
11688 register char *d = dest;
11689 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
11691 PERL_ARGS_ASSERT_SCAN_IDENT;
11696 while (isDIGIT(*s)) {
11698 Perl_croak(aTHX_ ident_too_long);
11705 Perl_croak(aTHX_ ident_too_long);
11706 if (isALNUM(*s)) /* UTF handled below */
11708 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
11713 else if (*s == ':' && s[1] == ':') {
11717 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
11718 char *t = s + UTF8SKIP(s);
11719 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
11721 if (d + (t - s) > e)
11722 Perl_croak(aTHX_ ident_too_long);
11723 Copy(s, d, t - s, char);
11734 if (PL_lex_state != LEX_NORMAL)
11735 PL_lex_state = LEX_INTERPENDMAYBE;
11738 if (*s == '$' && s[1] &&
11739 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
11752 if (*d == '^' && *s && isCONTROLVAR(*s)) {
11757 if (isSPACE(s[-1])) {
11759 const char ch = *s++;
11760 if (!SPACE_OR_TAB(ch)) {
11766 if (isIDFIRST_lazy_if(d,UTF)) {
11770 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
11771 end += UTF8SKIP(end);
11772 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
11773 end += UTF8SKIP(end);
11775 Copy(s, d, end - s, char);
11780 while ((isALNUM(*s) || *s == ':') && d < e)
11783 Perl_croak(aTHX_ ident_too_long);
11786 while (s < send && SPACE_OR_TAB(*s))
11788 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
11789 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
11790 const char * const brack =
11792 ((*s == '[') ? "[...]" : "{...}");
11793 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11794 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
11795 funny, dest, brack, funny, dest, brack);
11798 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
11802 /* Handle extended ${^Foo} variables
11803 * 1999-02-27 mjd-perl-patch@plover.com */
11804 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
11808 while (isALNUM(*s) && d < e) {
11812 Perl_croak(aTHX_ ident_too_long);
11817 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
11818 PL_lex_state = LEX_INTERPEND;
11821 if (PL_lex_state == LEX_NORMAL) {
11822 if (ckWARN(WARN_AMBIGUOUS) &&
11823 (keyword(dest, d - dest, 0)
11824 || get_cvn_flags(dest, d - dest, 0)))
11828 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
11829 "Ambiguous use of %c{%s} resolved to %c%s",
11830 funny, dest, funny, dest);
11835 s = bracket; /* let the parser handle it */
11839 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
11840 PL_lex_state = LEX_INTERPEND;
11845 S_pmflag(U32 pmfl, const char ch) {
11847 CASE_STD_PMMOD_FLAGS_PARSE_SET(&pmfl);
11848 case GLOBAL_PAT_MOD: pmfl |= PMf_GLOBAL; break;
11849 case CONTINUE_PAT_MOD: pmfl |= PMf_CONTINUE; break;
11850 case ONCE_PAT_MOD: pmfl |= PMf_KEEP; break;
11851 case KEEPCOPY_PAT_MOD: pmfl |= PMf_KEEPCOPY; break;
11857 Perl_pmflag(pTHX_ U32* pmfl, int ch)
11859 PERL_ARGS_ASSERT_PMFLAG;
11861 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEPRECATED),
11862 "Perl_pmflag() is deprecated, and will be removed from the XS API");
11865 *pmfl = S_pmflag(*pmfl, (char)ch);
11870 S_scan_pat(pTHX_ char *start, I32 type)
11874 char *s = scan_str(start,!!PL_madskills,FALSE);
11875 const char * const valid_flags =
11876 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
11881 PERL_ARGS_ASSERT_SCAN_PAT;
11884 const char * const delimiter = skipspace(start);
11888 ? "Search pattern not terminated or ternary operator parsed as search pattern"
11889 : "Search pattern not terminated" ));
11892 pm = (PMOP*)newPMOP(type, 0);
11893 if (PL_multi_open == '?') {
11894 /* This is the only point in the code that sets PMf_ONCE: */
11895 pm->op_pmflags |= PMf_ONCE;
11897 /* Hence it's safe to do this bit of PMOP book-keeping here, which
11898 allows us to restrict the list needed by reset to just the ??
11900 assert(type != OP_TRANS);
11902 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
11905 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
11908 elements = mg->mg_len / sizeof(PMOP**);
11909 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
11910 ((PMOP**)mg->mg_ptr) [elements++] = pm;
11911 mg->mg_len = elements * sizeof(PMOP**);
11912 PmopSTASH_set(pm,PL_curstash);
11918 while (*s && strchr(valid_flags, *s))
11919 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
11921 if (PL_madskills && modstart != s) {
11922 SV* tmptoken = newSVpvn(modstart, s - modstart);
11923 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
11926 /* issue a warning if /c is specified,but /g is not */
11927 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
11929 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
11930 "Use of /c modifier is meaningless without /g" );
11933 PL_lex_op = (OP*)pm;
11934 pl_yylval.ival = OP_MATCH;
11939 S_scan_subst(pTHX_ char *start)
11950 PERL_ARGS_ASSERT_SCAN_SUBST;
11952 pl_yylval.ival = OP_NULL;
11954 s = scan_str(start,!!PL_madskills,FALSE);
11957 Perl_croak(aTHX_ "Substitution pattern not terminated");
11959 if (s[-1] == PL_multi_open)
11962 if (PL_madskills) {
11963 CURMAD('q', PL_thisopen);
11964 CURMAD('_', PL_thiswhite);
11965 CURMAD('E', PL_thisstuff);
11966 CURMAD('Q', PL_thisclose);
11967 PL_realtokenstart = s - SvPVX(PL_linestr);
11971 first_start = PL_multi_start;
11972 s = scan_str(s,!!PL_madskills,FALSE);
11974 if (PL_lex_stuff) {
11975 SvREFCNT_dec(PL_lex_stuff);
11976 PL_lex_stuff = NULL;
11978 Perl_croak(aTHX_ "Substitution replacement not terminated");
11980 PL_multi_start = first_start; /* so whole substitution is taken together */
11982 pm = (PMOP*)newPMOP(OP_SUBST, 0);
11985 if (PL_madskills) {
11986 CURMAD('z', PL_thisopen);
11987 CURMAD('R', PL_thisstuff);
11988 CURMAD('Z', PL_thisclose);
11994 if (*s == EXEC_PAT_MOD) {
11998 else if (strchr(S_PAT_MODS, *s))
11999 pm->op_pmflags = S_pmflag(pm->op_pmflags, *s++);
12005 if (PL_madskills) {
12007 curmad('m', newSVpvn(modstart, s - modstart));
12008 append_madprops(PL_thismad, (OP*)pm, 0);
12012 if ((pm->op_pmflags & PMf_CONTINUE)) {
12013 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
12017 SV * const repl = newSVpvs("");
12019 PL_sublex_info.super_bufptr = s;
12020 PL_sublex_info.super_bufend = PL_bufend;
12022 pm->op_pmflags |= PMf_EVAL;
12025 sv_catpvs(repl, "eval ");
12027 sv_catpvs(repl, "do ");
12029 sv_catpvs(repl, "{");
12030 sv_catsv(repl, PL_lex_repl);
12031 if (strchr(SvPVX(PL_lex_repl), '#'))
12032 sv_catpvs(repl, "\n");
12033 sv_catpvs(repl, "}");
12035 SvREFCNT_dec(PL_lex_repl);
12036 PL_lex_repl = repl;
12039 PL_lex_op = (OP*)pm;
12040 pl_yylval.ival = OP_SUBST;
12045 S_scan_trans(pTHX_ char *start)
12058 PERL_ARGS_ASSERT_SCAN_TRANS;
12060 pl_yylval.ival = OP_NULL;
12062 s = scan_str(start,!!PL_madskills,FALSE);
12064 Perl_croak(aTHX_ "Transliteration pattern not terminated");
12066 if (s[-1] == PL_multi_open)
12069 if (PL_madskills) {
12070 CURMAD('q', PL_thisopen);
12071 CURMAD('_', PL_thiswhite);
12072 CURMAD('E', PL_thisstuff);
12073 CURMAD('Q', PL_thisclose);
12074 PL_realtokenstart = s - SvPVX(PL_linestr);
12078 s = scan_str(s,!!PL_madskills,FALSE);
12080 if (PL_lex_stuff) {
12081 SvREFCNT_dec(PL_lex_stuff);
12082 PL_lex_stuff = NULL;
12084 Perl_croak(aTHX_ "Transliteration replacement not terminated");
12086 if (PL_madskills) {
12087 CURMAD('z', PL_thisopen);
12088 CURMAD('R', PL_thisstuff);
12089 CURMAD('Z', PL_thisclose);
12092 complement = del = squash = 0;
12099 complement = OPpTRANS_COMPLEMENT;
12102 del = OPpTRANS_DELETE;
12105 squash = OPpTRANS_SQUASH;
12114 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
12115 o = newPVOP(OP_TRANS, 0, (char*)tbl);
12116 o->op_private &= ~OPpTRANS_ALL;
12117 o->op_private |= del|squash|complement|
12118 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
12119 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
12122 pl_yylval.ival = OP_TRANS;
12125 if (PL_madskills) {
12127 curmad('m', newSVpvn(modstart, s - modstart));
12128 append_madprops(PL_thismad, o, 0);
12137 S_scan_heredoc(pTHX_ register char *s)
12141 I32 op_type = OP_SCALAR;
12145 const char *found_newline;
12149 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
12151 I32 stuffstart = s - SvPVX(PL_linestr);
12154 PL_realtokenstart = -1;
12157 PERL_ARGS_ASSERT_SCAN_HEREDOC;
12161 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
12165 while (SPACE_OR_TAB(*peek))
12167 if (*peek == '`' || *peek == '\'' || *peek =='"') {
12170 s = delimcpy(d, e, s, PL_bufend, term, &len);
12180 if (!isALNUM_lazy_if(s,UTF))
12181 deprecate("bare << to mean <<\"\"");
12182 for (; isALNUM_lazy_if(s,UTF); s++) {
12187 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
12188 Perl_croak(aTHX_ "Delimiter for here document is too long");
12191 len = d - PL_tokenbuf;
12194 if (PL_madskills) {
12195 tstart = PL_tokenbuf + !outer;
12196 PL_thisclose = newSVpvn(tstart, len - !outer);
12197 tstart = SvPVX(PL_linestr) + stuffstart;
12198 PL_thisopen = newSVpvn(tstart, s - tstart);
12199 stuffstart = s - SvPVX(PL_linestr);
12202 #ifndef PERL_STRICT_CR
12203 d = strchr(s, '\r');
12205 char * const olds = s;
12207 while (s < PL_bufend) {
12213 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
12222 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12229 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
12230 herewas = newSVpvn(s,PL_bufend-s);
12234 herewas = newSVpvn(s-1,found_newline-s+1);
12237 herewas = newSVpvn(s,found_newline-s);
12241 if (PL_madskills) {
12242 tstart = SvPVX(PL_linestr) + stuffstart;
12244 sv_catpvn(PL_thisstuff, tstart, s - tstart);
12246 PL_thisstuff = newSVpvn(tstart, s - tstart);
12249 s += SvCUR(herewas);
12252 stuffstart = s - SvPVX(PL_linestr);
12258 tmpstr = newSV_type(SVt_PVIV);
12259 SvGROW(tmpstr, 80);
12260 if (term == '\'') {
12261 op_type = OP_CONST;
12262 SvIV_set(tmpstr, -1);
12264 else if (term == '`') {
12265 op_type = OP_BACKTICK;
12266 SvIV_set(tmpstr, '\\');
12270 PL_multi_start = CopLINE(PL_curcop);
12271 PL_multi_open = PL_multi_close = '<';
12272 term = *PL_tokenbuf;
12273 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
12274 char * const bufptr = PL_sublex_info.super_bufptr;
12275 char * const bufend = PL_sublex_info.super_bufend;
12276 char * const olds = s - SvCUR(herewas);
12277 s = strchr(bufptr, '\n');
12281 while (s < bufend &&
12282 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12284 CopLINE_inc(PL_curcop);
12287 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12288 missingterm(PL_tokenbuf);
12290 sv_setpvn(herewas,bufptr,d-bufptr+1);
12291 sv_setpvn(tmpstr,d+1,s-d);
12293 sv_catpvn(herewas,s,bufend-s);
12294 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
12301 while (s < PL_bufend &&
12302 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
12304 CopLINE_inc(PL_curcop);
12306 if (s >= PL_bufend) {
12307 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12308 missingterm(PL_tokenbuf);
12310 sv_setpvn(tmpstr,d+1,s-d);
12312 if (PL_madskills) {
12314 sv_catpvn(PL_thisstuff, d + 1, s - d);
12316 PL_thisstuff = newSVpvn(d + 1, s - d);
12317 stuffstart = s - SvPVX(PL_linestr);
12321 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
12323 sv_catpvn(herewas,s,PL_bufend-s);
12324 sv_setsv(PL_linestr,herewas);
12325 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
12326 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12327 PL_last_lop = PL_last_uni = NULL;
12330 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
12331 while (s >= PL_bufend) { /* multiple line string? */
12333 if (PL_madskills) {
12334 tstart = SvPVX(PL_linestr) + stuffstart;
12336 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12338 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12342 CopLINE_inc(PL_curcop);
12343 if (!outer || !lex_next_chunk(0)) {
12344 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12345 missingterm(PL_tokenbuf);
12347 CopLINE_dec(PL_curcop);
12350 stuffstart = s - SvPVX(PL_linestr);
12352 CopLINE_inc(PL_curcop);
12353 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12354 PL_last_lop = PL_last_uni = NULL;
12355 #ifndef PERL_STRICT_CR
12356 if (PL_bufend - PL_linestart >= 2) {
12357 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
12358 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
12360 PL_bufend[-2] = '\n';
12362 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
12364 else if (PL_bufend[-1] == '\r')
12365 PL_bufend[-1] = '\n';
12367 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
12368 PL_bufend[-1] = '\n';
12370 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
12371 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
12372 *(SvPVX(PL_linestr) + off ) = ' ';
12373 sv_catsv(PL_linestr,herewas);
12374 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
12375 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
12379 sv_catsv(tmpstr,PL_linestr);
12384 PL_multi_end = CopLINE(PL_curcop);
12385 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
12386 SvPV_shrink_to_cur(tmpstr);
12388 SvREFCNT_dec(herewas);
12390 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
12392 else if (PL_encoding)
12393 sv_recode_to_utf8(tmpstr, PL_encoding);
12395 PL_lex_stuff = tmpstr;
12396 pl_yylval.ival = op_type;
12400 /* scan_inputsymbol
12401 takes: current position in input buffer
12402 returns: new position in input buffer
12403 side-effects: pl_yylval and lex_op are set.
12408 <FH> read from filehandle
12409 <pkg::FH> read from package qualified filehandle
12410 <pkg'FH> read from package qualified filehandle
12411 <$fh> read from filehandle in $fh
12412 <*.h> filename glob
12417 S_scan_inputsymbol(pTHX_ char *start)
12420 register char *s = start; /* current position in buffer */
12423 char *d = PL_tokenbuf; /* start of temp holding space */
12424 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
12426 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
12428 end = strchr(s, '\n');
12431 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
12433 /* die if we didn't have space for the contents of the <>,
12434 or if it didn't end, or if we see a newline
12437 if (len >= (I32)sizeof PL_tokenbuf)
12438 Perl_croak(aTHX_ "Excessively long <> operator");
12440 Perl_croak(aTHX_ "Unterminated <> operator");
12445 Remember, only scalar variables are interpreted as filehandles by
12446 this code. Anything more complex (e.g., <$fh{$num}>) will be
12447 treated as a glob() call.
12448 This code makes use of the fact that except for the $ at the front,
12449 a scalar variable and a filehandle look the same.
12451 if (*d == '$' && d[1]) d++;
12453 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
12454 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
12457 /* If we've tried to read what we allow filehandles to look like, and
12458 there's still text left, then it must be a glob() and not a getline.
12459 Use scan_str to pull out the stuff between the <> and treat it
12460 as nothing more than a string.
12463 if (d - PL_tokenbuf != len) {
12464 pl_yylval.ival = OP_GLOB;
12465 s = scan_str(start,!!PL_madskills,FALSE);
12467 Perl_croak(aTHX_ "Glob not terminated");
12471 bool readline_overriden = FALSE;
12474 /* we're in a filehandle read situation */
12477 /* turn <> into <ARGV> */
12479 Copy("ARGV",d,5,char);
12481 /* Check whether readline() is overriden */
12482 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
12484 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
12486 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
12487 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
12488 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
12489 readline_overriden = TRUE;
12491 /* if <$fh>, create the ops to turn the variable into a
12495 /* try to find it in the pad for this block, otherwise find
12496 add symbol table ops
12498 const PADOFFSET tmp = pad_findmy(d, len, 0);
12499 if (tmp != NOT_IN_PAD) {
12500 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
12501 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
12502 HEK * const stashname = HvNAME_HEK(stash);
12503 SV * const sym = sv_2mortal(newSVhek(stashname));
12504 sv_catpvs(sym, "::");
12505 sv_catpv(sym, d+1);
12510 OP * const o = newOP(OP_PADSV, 0);
12512 PL_lex_op = readline_overriden
12513 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12514 append_elem(OP_LIST, o,
12515 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
12516 : (OP*)newUNOP(OP_READLINE, 0, o);
12525 ? (GV_ADDMULTI | GV_ADDINEVAL)
12528 PL_lex_op = readline_overriden
12529 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12530 append_elem(OP_LIST,
12531 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
12532 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12533 : (OP*)newUNOP(OP_READLINE, 0,
12534 newUNOP(OP_RV2SV, 0,
12535 newGVOP(OP_GV, 0, gv)));
12537 if (!readline_overriden)
12538 PL_lex_op->op_flags |= OPf_SPECIAL;
12539 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
12540 pl_yylval.ival = OP_NULL;
12543 /* If it's none of the above, it must be a literal filehandle
12544 (<Foo::BAR> or <FOO>) so build a simple readline OP */
12546 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
12547 PL_lex_op = readline_overriden
12548 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
12549 append_elem(OP_LIST,
12550 newGVOP(OP_GV, 0, gv),
12551 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
12552 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
12553 pl_yylval.ival = OP_NULL;
12562 takes: start position in buffer
12563 keep_quoted preserve \ on the embedded delimiter(s)
12564 keep_delims preserve the delimiters around the string
12565 returns: position to continue reading from buffer
12566 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
12567 updates the read buffer.
12569 This subroutine pulls a string out of the input. It is called for:
12570 q single quotes q(literal text)
12571 ' single quotes 'literal text'
12572 qq double quotes qq(interpolate $here please)
12573 " double quotes "interpolate $here please"
12574 qx backticks qx(/bin/ls -l)
12575 ` backticks `/bin/ls -l`
12576 qw quote words @EXPORT_OK = qw( func() $spam )
12577 m// regexp match m/this/
12578 s/// regexp substitute s/this/that/
12579 tr/// string transliterate tr/this/that/
12580 y/// string transliterate y/this/that/
12581 ($*@) sub prototypes sub foo ($)
12582 (stuff) sub attr parameters sub foo : attr(stuff)
12583 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
12585 In most of these cases (all but <>, patterns and transliterate)
12586 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
12587 calls scan_str(). s/// makes yylex() call scan_subst() which calls
12588 scan_str(). tr/// and y/// make yylex() call scan_trans() which
12591 It skips whitespace before the string starts, and treats the first
12592 character as the delimiter. If the delimiter is one of ([{< then
12593 the corresponding "close" character )]}> is used as the closing
12594 delimiter. It allows quoting of delimiters, and if the string has
12595 balanced delimiters ([{<>}]) it allows nesting.
12597 On success, the SV with the resulting string is put into lex_stuff or,
12598 if that is already non-NULL, into lex_repl. The second case occurs only
12599 when parsing the RHS of the special constructs s/// and tr/// (y///).
12600 For convenience, the terminating delimiter character is stuffed into
12605 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
12608 SV *sv; /* scalar value: string */
12609 const char *tmps; /* temp string, used for delimiter matching */
12610 register char *s = start; /* current position in the buffer */
12611 register char term; /* terminating character */
12612 register char *to; /* current position in the sv's data */
12613 I32 brackets = 1; /* bracket nesting level */
12614 bool has_utf8 = FALSE; /* is there any utf8 content? */
12615 I32 termcode; /* terminating char. code */
12616 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
12617 STRLEN termlen; /* length of terminating string */
12618 int last_off = 0; /* last position for nesting bracket */
12624 PERL_ARGS_ASSERT_SCAN_STR;
12626 /* skip space before the delimiter */
12632 if (PL_realtokenstart >= 0) {
12633 stuffstart = PL_realtokenstart;
12634 PL_realtokenstart = -1;
12637 stuffstart = start - SvPVX(PL_linestr);
12639 /* mark where we are, in case we need to report errors */
12642 /* after skipping whitespace, the next character is the terminator */
12645 termcode = termstr[0] = term;
12649 termcode = utf8_to_uvchr((U8*)s, &termlen);
12650 Copy(s, termstr, termlen, U8);
12651 if (!UTF8_IS_INVARIANT(term))
12655 /* mark where we are */
12656 PL_multi_start = CopLINE(PL_curcop);
12657 PL_multi_open = term;
12659 /* find corresponding closing delimiter */
12660 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
12661 termcode = termstr[0] = term = tmps[5];
12663 PL_multi_close = term;
12665 /* create a new SV to hold the contents. 79 is the SV's initial length.
12666 What a random number. */
12667 sv = newSV_type(SVt_PVIV);
12669 SvIV_set(sv, termcode);
12670 (void)SvPOK_only(sv); /* validate pointer */
12672 /* move past delimiter and try to read a complete string */
12674 sv_catpvn(sv, s, termlen);
12677 tstart = SvPVX(PL_linestr) + stuffstart;
12678 if (!PL_thisopen && !keep_delims) {
12679 PL_thisopen = newSVpvn(tstart, s - tstart);
12680 stuffstart = s - SvPVX(PL_linestr);
12684 if (PL_encoding && !UTF) {
12688 int offset = s - SvPVX_const(PL_linestr);
12689 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
12690 &offset, (char*)termstr, termlen);
12691 const char * const ns = SvPVX_const(PL_linestr) + offset;
12692 char * const svlast = SvEND(sv) - 1;
12694 for (; s < ns; s++) {
12695 if (*s == '\n' && !PL_rsfp)
12696 CopLINE_inc(PL_curcop);
12699 goto read_more_line;
12701 /* handle quoted delimiters */
12702 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
12704 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
12706 if ((svlast-1 - t) % 2) {
12707 if (!keep_quoted) {
12708 *(svlast-1) = term;
12710 SvCUR_set(sv, SvCUR(sv) - 1);
12715 if (PL_multi_open == PL_multi_close) {
12721 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
12722 /* At here, all closes are "was quoted" one,
12723 so we don't check PL_multi_close. */
12725 if (!keep_quoted && *(t+1) == PL_multi_open)
12730 else if (*t == PL_multi_open)
12738 SvCUR_set(sv, w - SvPVX_const(sv));
12740 last_off = w - SvPVX(sv);
12741 if (--brackets <= 0)
12746 if (!keep_delims) {
12747 SvCUR_set(sv, SvCUR(sv) - 1);
12753 /* extend sv if need be */
12754 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
12755 /* set 'to' to the next character in the sv's string */
12756 to = SvPVX(sv)+SvCUR(sv);
12758 /* if open delimiter is the close delimiter read unbridle */
12759 if (PL_multi_open == PL_multi_close) {
12760 for (; s < PL_bufend; s++,to++) {
12761 /* embedded newlines increment the current line number */
12762 if (*s == '\n' && !PL_rsfp)
12763 CopLINE_inc(PL_curcop);
12764 /* handle quoted delimiters */
12765 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
12766 if (!keep_quoted && s[1] == term)
12768 /* any other quotes are simply copied straight through */
12772 /* terminate when run out of buffer (the for() condition), or
12773 have found the terminator */
12774 else if (*s == term) {
12777 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
12780 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12786 /* if the terminator isn't the same as the start character (e.g.,
12787 matched brackets), we have to allow more in the quoting, and
12788 be prepared for nested brackets.
12791 /* read until we run out of string, or we find the terminator */
12792 for (; s < PL_bufend; s++,to++) {
12793 /* embedded newlines increment the line count */
12794 if (*s == '\n' && !PL_rsfp)
12795 CopLINE_inc(PL_curcop);
12796 /* backslashes can escape the open or closing characters */
12797 if (*s == '\\' && s+1 < PL_bufend) {
12798 if (!keep_quoted &&
12799 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
12804 /* allow nested opens and closes */
12805 else if (*s == PL_multi_close && --brackets <= 0)
12807 else if (*s == PL_multi_open)
12809 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
12814 /* terminate the copied string and update the sv's end-of-string */
12816 SvCUR_set(sv, to - SvPVX_const(sv));
12819 * this next chunk reads more into the buffer if we're not done yet
12823 break; /* handle case where we are done yet :-) */
12825 #ifndef PERL_STRICT_CR
12826 if (to - SvPVX_const(sv) >= 2) {
12827 if ((to[-2] == '\r' && to[-1] == '\n') ||
12828 (to[-2] == '\n' && to[-1] == '\r'))
12832 SvCUR_set(sv, to - SvPVX_const(sv));
12834 else if (to[-1] == '\r')
12837 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
12842 /* if we're out of file, or a read fails, bail and reset the current
12843 line marker so we can report where the unterminated string began
12846 if (PL_madskills) {
12847 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12849 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
12851 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
12854 CopLINE_inc(PL_curcop);
12855 PL_bufptr = PL_bufend;
12856 if (!lex_next_chunk(0)) {
12858 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
12867 /* at this point, we have successfully read the delimited string */
12869 if (!PL_encoding || UTF) {
12871 if (PL_madskills) {
12872 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12873 const int len = s - tstart;
12875 sv_catpvn(PL_thisstuff, tstart, len);
12877 PL_thisstuff = newSVpvn(tstart, len);
12878 if (!PL_thisclose && !keep_delims)
12879 PL_thisclose = newSVpvn(s,termlen);
12884 sv_catpvn(sv, s, termlen);
12889 if (PL_madskills) {
12890 char * const tstart = SvPVX(PL_linestr) + stuffstart;
12891 const int len = s - tstart - termlen;
12893 sv_catpvn(PL_thisstuff, tstart, len);
12895 PL_thisstuff = newSVpvn(tstart, len);
12896 if (!PL_thisclose && !keep_delims)
12897 PL_thisclose = newSVpvn(s - termlen,termlen);
12901 if (has_utf8 || PL_encoding)
12904 PL_multi_end = CopLINE(PL_curcop);
12906 /* if we allocated too much space, give some back */
12907 if (SvCUR(sv) + 5 < SvLEN(sv)) {
12908 SvLEN_set(sv, SvCUR(sv) + 1);
12909 SvPV_renew(sv, SvLEN(sv));
12912 /* decide whether this is the first or second quoted string we've read
12925 takes: pointer to position in buffer
12926 returns: pointer to new position in buffer
12927 side-effects: builds ops for the constant in pl_yylval.op
12929 Read a number in any of the formats that Perl accepts:
12931 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
12932 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
12935 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
12937 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
12940 If it reads a number without a decimal point or an exponent, it will
12941 try converting the number to an integer and see if it can do so
12942 without loss of precision.
12946 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
12949 register const char *s = start; /* current position in buffer */
12950 register char *d; /* destination in temp buffer */
12951 register char *e; /* end of temp buffer */
12952 NV nv; /* number read, as a double */
12953 SV *sv = NULL; /* place to put the converted number */
12954 bool floatit; /* boolean: int or float? */
12955 const char *lastub = NULL; /* position of last underbar */
12956 static char const number_too_long[] = "Number too long";
12958 PERL_ARGS_ASSERT_SCAN_NUM;
12960 /* We use the first character to decide what type of number this is */
12964 Perl_croak(aTHX_ "panic: scan_num");
12966 /* if it starts with a 0, it could be an octal number, a decimal in
12967 0.13 disguise, or a hexadecimal number, or a binary number. */
12971 u holds the "number so far"
12972 shift the power of 2 of the base
12973 (hex == 4, octal == 3, binary == 1)
12974 overflowed was the number more than we can hold?
12976 Shift is used when we add a digit. It also serves as an "are
12977 we in octal/hex/binary?" indicator to disallow hex characters
12978 when in octal mode.
12983 bool overflowed = FALSE;
12984 bool just_zero = TRUE; /* just plain 0 or binary number? */
12985 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
12986 static const char* const bases[5] =
12987 { "", "binary", "", "octal", "hexadecimal" };
12988 static const char* const Bases[5] =
12989 { "", "Binary", "", "Octal", "Hexadecimal" };
12990 static const char* const maxima[5] =
12992 "0b11111111111111111111111111111111",
12996 const char *base, *Base, *max;
12998 /* check for hex */
13003 } else if (s[1] == 'b') {
13008 /* check for a decimal in disguise */
13009 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
13011 /* so it must be octal */
13018 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13019 "Misplaced _ in number");
13023 base = bases[shift];
13024 Base = Bases[shift];
13025 max = maxima[shift];
13027 /* read the rest of the number */
13029 /* x is used in the overflow test,
13030 b is the digit we're adding on. */
13035 /* if we don't mention it, we're done */
13039 /* _ are ignored -- but warned about if consecutive */
13041 if (lastub && s == lastub + 1)
13042 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13043 "Misplaced _ in number");
13047 /* 8 and 9 are not octal */
13048 case '8': case '9':
13050 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
13054 case '2': case '3': case '4':
13055 case '5': case '6': case '7':
13057 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
13060 case '0': case '1':
13061 b = *s++ & 15; /* ASCII digit -> value of digit */
13065 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
13066 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
13067 /* make sure they said 0x */
13070 b = (*s++ & 7) + 9;
13072 /* Prepare to put the digit we have onto the end
13073 of the number so far. We check for overflows.
13079 x = u << shift; /* make room for the digit */
13081 if ((x >> shift) != u
13082 && !(PL_hints & HINT_NEW_BINARY)) {
13085 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13086 "Integer overflow in %s number",
13089 u = x | b; /* add the digit to the end */
13092 n *= nvshift[shift];
13093 /* If an NV has not enough bits in its
13094 * mantissa to represent an UV this summing of
13095 * small low-order numbers is a waste of time
13096 * (because the NV cannot preserve the
13097 * low-order bits anyway): we could just
13098 * remember when did we overflow and in the
13099 * end just multiply n by the right
13107 /* if we get here, we had success: make a scalar value from
13112 /* final misplaced underbar check */
13113 if (s[-1] == '_') {
13114 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13119 if (n > 4294967295.0)
13120 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13121 "%s number > %s non-portable",
13127 if (u > 0xffffffff)
13128 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
13129 "%s number > %s non-portable",
13134 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
13135 sv = new_constant(start, s - start, "integer",
13136 sv, NULL, NULL, 0);
13137 else if (PL_hints & HINT_NEW_BINARY)
13138 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
13143 handle decimal numbers.
13144 we're also sent here when we read a 0 as the first digit
13146 case '1': case '2': case '3': case '4': case '5':
13147 case '6': case '7': case '8': case '9': case '.':
13150 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
13153 /* read next group of digits and _ and copy into d */
13154 while (isDIGIT(*s) || *s == '_') {
13155 /* skip underscores, checking for misplaced ones
13159 if (lastub && s == lastub + 1)
13160 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13161 "Misplaced _ in number");
13165 /* check for end of fixed-length buffer */
13167 Perl_croak(aTHX_ number_too_long);
13168 /* if we're ok, copy the character */
13173 /* final misplaced underbar check */
13174 if (lastub && s == lastub + 1) {
13175 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
13178 /* read a decimal portion if there is one. avoid
13179 3..5 being interpreted as the number 3. followed
13182 if (*s == '.' && s[1] != '.') {
13187 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13188 "Misplaced _ in number");
13192 /* copy, ignoring underbars, until we run out of digits.
13194 for (; isDIGIT(*s) || *s == '_'; s++) {
13195 /* fixed length buffer check */
13197 Perl_croak(aTHX_ number_too_long);
13199 if (lastub && s == lastub + 1)
13200 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13201 "Misplaced _ in number");
13207 /* fractional part ending in underbar? */
13208 if (s[-1] == '_') {
13209 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13210 "Misplaced _ in number");
13212 if (*s == '.' && isDIGIT(s[1])) {
13213 /* oops, it's really a v-string, but without the "v" */
13219 /* read exponent part, if present */
13220 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
13224 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
13225 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
13227 /* stray preinitial _ */
13229 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13230 "Misplaced _ in number");
13234 /* allow positive or negative exponent */
13235 if (*s == '+' || *s == '-')
13238 /* stray initial _ */
13240 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13241 "Misplaced _ in number");
13245 /* read digits of exponent */
13246 while (isDIGIT(*s) || *s == '_') {
13249 Perl_croak(aTHX_ number_too_long);
13253 if (((lastub && s == lastub + 1) ||
13254 (!isDIGIT(s[1]) && s[1] != '_')))
13255 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
13256 "Misplaced _ in number");
13263 /* make an sv from the string */
13267 We try to do an integer conversion first if no characters
13268 indicating "float" have been found.
13273 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
13275 if (flags == IS_NUMBER_IN_UV) {
13277 sv_setiv(sv, uv); /* Prefer IVs over UVs. */
13280 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
13281 if (uv <= (UV) IV_MIN)
13282 sv_setiv(sv, -(IV)uv);
13289 /* terminate the string */
13291 nv = Atof(PL_tokenbuf);
13296 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
13297 const char *const key = floatit ? "float" : "integer";
13298 const STRLEN keylen = floatit ? 5 : 7;
13299 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
13300 key, keylen, sv, NULL, NULL, 0);
13304 /* if it starts with a v, it could be a v-string */
13307 sv = newSV(5); /* preallocate storage space */
13308 s = scan_vstring(s, PL_bufend, sv);
13312 /* make the op for the constant and return */
13315 lvalp->opval = newSVOP(OP_CONST, 0, sv);
13317 lvalp->opval = NULL;
13323 S_scan_formline(pTHX_ register char *s)
13326 register char *eol;
13328 SV * const stuff = newSVpvs("");
13329 bool needargs = FALSE;
13330 bool eofmt = FALSE;
13332 char *tokenstart = s;
13333 SV* savewhite = NULL;
13335 if (PL_madskills) {
13336 savewhite = PL_thiswhite;
13341 PERL_ARGS_ASSERT_SCAN_FORMLINE;
13343 while (!needargs) {
13346 #ifdef PERL_STRICT_CR
13347 while (SPACE_OR_TAB(*t))
13350 while (SPACE_OR_TAB(*t) || *t == '\r')
13353 if (*t == '\n' || t == PL_bufend) {
13358 if (PL_in_eval && !PL_rsfp) {
13359 eol = (char *) memchr(s,'\n',PL_bufend-s);
13364 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
13366 for (t = s; t < eol; t++) {
13367 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
13369 goto enough; /* ~~ must be first line in formline */
13371 if (*t == '@' || *t == '^')
13375 sv_catpvn(stuff, s, eol-s);
13376 #ifndef PERL_STRICT_CR
13377 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
13378 char *end = SvPVX(stuff) + SvCUR(stuff);
13381 SvCUR_set(stuff, SvCUR(stuff) - 1);
13392 if (PL_madskills) {
13394 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
13396 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
13399 PL_bufptr = PL_bufend;
13400 CopLINE_inc(PL_curcop);
13401 got_some = lex_next_chunk(0);
13402 CopLINE_dec(PL_curcop);
13405 tokenstart = PL_bufptr;
13413 if (SvCUR(stuff)) {
13416 PL_lex_state = LEX_NORMAL;
13417 start_force(PL_curforce);
13418 NEXTVAL_NEXTTOKE.ival = 0;
13422 PL_lex_state = LEX_FORMLINE;
13424 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
13426 else if (PL_encoding)
13427 sv_recode_to_utf8(stuff, PL_encoding);
13429 start_force(PL_curforce);
13430 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
13432 start_force(PL_curforce);
13433 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
13437 SvREFCNT_dec(stuff);
13439 PL_lex_formbrack = 0;
13443 if (PL_madskills) {
13445 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
13447 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
13448 PL_thiswhite = savewhite;
13455 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
13458 const I32 oldsavestack_ix = PL_savestack_ix;
13459 CV* const outsidecv = PL_compcv;
13462 assert(SvTYPE(PL_compcv) == SVt_PVCV);
13464 SAVEI32(PL_subline);
13465 save_item(PL_subname);
13466 SAVESPTR(PL_compcv);
13468 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
13469 CvFLAGS(PL_compcv) |= flags;
13471 PL_subline = CopLINE(PL_curcop);
13472 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
13473 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
13474 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
13476 return oldsavestack_ix;
13480 #pragma segment Perl_yylex
13483 S_yywarn(pTHX_ const char *const s)
13487 PERL_ARGS_ASSERT_YYWARN;
13489 PL_in_eval |= EVAL_WARNONLY;
13491 PL_in_eval &= ~EVAL_WARNONLY;
13496 Perl_yyerror(pTHX_ const char *const s)
13499 const char *where = NULL;
13500 const char *context = NULL;
13503 int yychar = PL_parser->yychar;
13505 PERL_ARGS_ASSERT_YYERROR;
13507 if (!yychar || (yychar == ';' && !PL_rsfp))
13509 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
13510 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
13511 PL_oldbufptr != PL_bufptr) {
13514 The code below is removed for NetWare because it abends/crashes on NetWare
13515 when the script has error such as not having the closing quotes like:
13516 if ($var eq "value)
13517 Checking of white spaces is anyway done in NetWare code.
13520 while (isSPACE(*PL_oldoldbufptr))
13523 context = PL_oldoldbufptr;
13524 contlen = PL_bufptr - PL_oldoldbufptr;
13526 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
13527 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
13530 The code below is removed for NetWare because it abends/crashes on NetWare
13531 when the script has error such as not having the closing quotes like:
13532 if ($var eq "value)
13533 Checking of white spaces is anyway done in NetWare code.
13536 while (isSPACE(*PL_oldbufptr))
13539 context = PL_oldbufptr;
13540 contlen = PL_bufptr - PL_oldbufptr;
13542 else if (yychar > 255)
13543 where = "next token ???";
13544 else if (yychar == -2) { /* YYEMPTY */
13545 if (PL_lex_state == LEX_NORMAL ||
13546 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
13547 where = "at end of line";
13548 else if (PL_lex_inpat)
13549 where = "within pattern";
13551 where = "within string";
13554 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
13556 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
13557 else if (isPRINT_LC(yychar)) {
13558 const char string = yychar;
13559 sv_catpvn(where_sv, &string, 1);
13562 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
13563 where = SvPVX_const(where_sv);
13565 msg = sv_2mortal(newSVpv(s, 0));
13566 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
13567 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
13569 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
13571 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
13572 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
13573 Perl_sv_catpvf(aTHX_ msg,
13574 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
13575 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
13578 if (PL_in_eval & EVAL_WARNONLY) {
13579 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
13583 if (PL_error_count >= 10) {
13584 if (PL_in_eval && SvCUR(ERRSV))
13585 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
13586 SVfARG(ERRSV), OutCopFILE(PL_curcop));
13588 Perl_croak(aTHX_ "%s has too many errors.\n",
13589 OutCopFILE(PL_curcop));
13592 PL_in_my_stash = NULL;
13596 #pragma segment Main
13600 S_swallow_bom(pTHX_ U8 *s)
13603 const STRLEN slen = SvCUR(PL_linestr);
13605 PERL_ARGS_ASSERT_SWALLOW_BOM;
13609 if (s[1] == 0xFE) {
13610 /* UTF-16 little-endian? (or UTF-32LE?) */
13611 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
13612 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
13613 #ifndef PERL_NO_UTF16_FILTER
13614 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
13616 if (PL_bufend > (char*)s) {
13617 s = add_utf16_textfilter(s, TRUE);
13620 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13625 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
13626 #ifndef PERL_NO_UTF16_FILTER
13627 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
13629 if (PL_bufend > (char *)s) {
13630 s = add_utf16_textfilter(s, FALSE);
13633 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13638 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
13639 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13640 s += 3; /* UTF-8 */
13646 if (s[2] == 0xFE && s[3] == 0xFF) {
13647 /* UTF-32 big-endian */
13648 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
13651 else if (s[2] == 0 && s[3] != 0) {
13654 * are a good indicator of UTF-16BE. */
13655 #ifndef PERL_NO_UTF16_FILTER
13656 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
13657 s = add_utf16_textfilter(s, FALSE);
13659 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
13665 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
13666 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
13667 s += 4; /* UTF-8 */
13673 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
13676 * are a good indicator of UTF-16LE. */
13677 #ifndef PERL_NO_UTF16_FILTER
13678 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
13679 s = add_utf16_textfilter(s, TRUE);
13681 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
13689 #ifndef PERL_NO_UTF16_FILTER
13691 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
13694 SV *const filter = FILTER_DATA(idx);
13695 /* We re-use this each time round, throwing the contents away before we
13697 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
13698 SV *const utf8_buffer = filter;
13699 IV status = IoPAGE(filter);
13700 const bool reverse = (bool) IoLINES(filter);
13703 /* As we're automatically added, at the lowest level, and hence only called
13704 from this file, we can be sure that we're not called in block mode. Hence
13705 don't bother writing code to deal with block mode. */
13707 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
13710 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
13712 DEBUG_P(PerlIO_printf(Perl_debug_log,
13713 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13714 FPTR2DPTR(void *, S_utf16_textfilter),
13715 reverse ? 'l' : 'b', idx, maxlen, status,
13716 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13723 /* First, look in our buffer of existing UTF-8 data: */
13724 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
13728 } else if (status == 0) {
13730 IoPAGE(filter) = 0;
13731 nl = SvEND(utf8_buffer);
13734 STRLEN got = nl - SvPVX(utf8_buffer);
13735 /* Did we have anything to append? */
13737 sv_catpvn(sv, SvPVX(utf8_buffer), got);
13738 /* Everything else in this code works just fine if SVp_POK isn't
13739 set. This, however, needs it, and we need it to work, else
13740 we loop infinitely because the buffer is never consumed. */
13741 sv_chop(utf8_buffer, nl);
13745 /* OK, not a complete line there, so need to read some more UTF-16.
13746 Read an extra octect if the buffer currently has an odd number. */
13750 if (SvCUR(utf16_buffer) >= 2) {
13751 /* Location of the high octet of the last complete code point.
13752 Gosh, UTF-16 is a pain. All the benefits of variable length,
13753 *coupled* with all the benefits of partial reads and
13755 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
13756 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
13758 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
13762 /* We have the first half of a surrogate. Read more. */
13763 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
13766 status = FILTER_READ(idx + 1, utf16_buffer,
13767 160 + (SvCUR(utf16_buffer) & 1));
13768 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
13769 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
13772 IoPAGE(filter) = status;
13777 chars = SvCUR(utf16_buffer) >> 1;
13778 have = SvCUR(utf8_buffer);
13779 SvGROW(utf8_buffer, have + chars * 3 + 1);
13782 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
13783 (U8*)SvPVX_const(utf8_buffer) + have,
13784 chars * 2, &newlen);
13786 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
13787 (U8*)SvPVX_const(utf8_buffer) + have,
13788 chars * 2, &newlen);
13790 SvCUR_set(utf8_buffer, have + newlen);
13793 /* No need to keep this SV "well-formed" with a '\0' after the end, as
13794 it's private to us, and utf16_to_utf8{,reversed} take a
13795 (pointer,length) pair, rather than a NUL-terminated string. */
13796 if(SvCUR(utf16_buffer) & 1) {
13797 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
13798 SvCUR_set(utf16_buffer, 1);
13800 SvCUR_set(utf16_buffer, 0);
13803 DEBUG_P(PerlIO_printf(Perl_debug_log,
13804 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
13806 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
13807 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
13812 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
13814 SV *filter = filter_add(S_utf16_textfilter, NULL);
13816 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
13817 sv_setpvs(filter, "");
13818 IoLINES(filter) = reversed;
13819 IoPAGE(filter) = 1; /* Not EOF */
13821 /* Sadly, we have to return a valid pointer, come what may, so we have to
13822 ignore any error return from this. */
13823 SvCUR_set(PL_linestr, 0);
13824 if (FILTER_READ(0, PL_linestr, 0)) {
13825 SvUTF8_on(PL_linestr);
13827 SvUTF8_on(PL_linestr);
13829 PL_bufend = SvEND(PL_linestr);
13830 return (U8*)SvPVX(PL_linestr);
13835 Returns a pointer to the next character after the parsed
13836 vstring, as well as updating the passed in sv.
13838 Function must be called like
13841 s = scan_vstring(s,e,sv);
13843 where s and e are the start and end of the string.
13844 The sv should already be large enough to store the vstring
13845 passed in, for performance reasons.
13850 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
13853 const char *pos = s;
13854 const char *start = s;
13856 PERL_ARGS_ASSERT_SCAN_VSTRING;
13858 if (*pos == 'v') pos++; /* get past 'v' */
13859 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13861 if ( *pos != '.') {
13862 /* this may not be a v-string if followed by => */
13863 const char *next = pos;
13864 while (next < e && isSPACE(*next))
13866 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
13867 /* return string not v-string */
13868 sv_setpvn(sv,(char *)s,pos-s);
13869 return (char *)pos;
13873 if (!isALPHA(*pos)) {
13874 U8 tmpbuf[UTF8_MAXBYTES+1];
13877 s++; /* get past 'v' */
13882 /* this is atoi() that tolerates underscores */
13885 const char *end = pos;
13887 while (--end >= s) {
13889 const UV orev = rev;
13890 rev += (*end - '0') * mult;
13893 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
13894 "Integer overflow in decimal number");
13898 if (rev > 0x7FFFFFFF)
13899 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
13901 /* Append native character for the rev point */
13902 tmpend = uvchr_to_utf8(tmpbuf, rev);
13903 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
13904 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
13906 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
13912 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
13916 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
13923 Perl_keyword_plugin_standard(pTHX_
13924 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
13926 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
13927 PERL_UNUSED_CONTEXT;
13928 PERL_UNUSED_ARG(keyword_ptr);
13929 PERL_UNUSED_ARG(keyword_len);
13930 PERL_UNUSED_ARG(op_ptr);
13931 return KEYWORD_PLUGIN_DECLINE;
13936 * c-indentation-style: bsd
13937 * c-basic-offset: 4
13938 * indent-tabs-mode: t
13941 * ex: set ts=8 sts=4 sw=4 noet: