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
42 #include "dquote_static.c"
44 #define new_constant(a,b,c,d,e,f,g) \
45 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
47 #define pl_yylval (PL_parser->yylval)
49 /* XXX temporary backwards compatibility */
50 #define PL_lex_brackets (PL_parser->lex_brackets)
51 #define PL_lex_allbrackets (PL_parser->lex_allbrackets)
52 #define PL_lex_fakeeof (PL_parser->lex_fakeeof)
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 XENUMMASK 0x3f
130 #define XFAKEEOF 0x40
131 #define XFAKEBRACK 0x80
133 #ifdef USE_UTF8_SCRIPTS
134 # define UTF (!IN_BYTES)
136 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
139 /* The maximum number of characters preceding the unrecognized one to display */
140 #define UNRECOGNIZED_PRECEDE_COUNT 10
142 /* In variables named $^X, these are the legal values for X.
143 * 1999-02-27 mjd-perl-patch@plover.com */
144 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
146 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
148 /* LEX_* are values for PL_lex_state, the state of the lexer.
149 * They are arranged oddly so that the guard on the switch statement
150 * can get by with a single comparison (if the compiler is smart enough).
153 /* #define LEX_NOTPARSING 11 is done in perl.h. */
155 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
156 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
157 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
158 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
159 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
161 /* at end of code, eg "$x" followed by: */
162 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
163 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
165 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
166 string or after \E, $foo, etc */
167 #define LEX_INTERPCONST 2 /* NOT USED */
168 #define LEX_FORMLINE 1 /* expecting a format line */
169 #define LEX_KNOWNEXT 0 /* next token known; just return it */
173 static const char* const lex_state_names[] = {
192 #include "keywords.h"
194 /* CLINE is a macro that ensures PL_copline has a sane value */
199 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
202 # define SKIPSPACE0(s) skipspace0(s)
203 # define SKIPSPACE1(s) skipspace1(s)
204 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
205 # define PEEKSPACE(s) skipspace2(s,0)
207 # define SKIPSPACE0(s) skipspace(s)
208 # define SKIPSPACE1(s) skipspace(s)
209 # define SKIPSPACE2(s,tsv) skipspace(s)
210 # define PEEKSPACE(s) skipspace(s)
214 * Convenience functions to return different tokens and prime the
215 * lexer for the next token. They all take an argument.
217 * TOKEN : generic token (used for '(', DOLSHARP, etc)
218 * OPERATOR : generic operator
219 * AOPERATOR : assignment operator
220 * PREBLOCK : beginning the block after an if, while, foreach, ...
221 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
222 * PREREF : *EXPR where EXPR is not a simple identifier
223 * TERM : expression term
224 * LOOPX : loop exiting command (goto, last, dump, etc)
225 * FTST : file test operator
226 * FUN0 : zero-argument function
227 * FUN0OP : zero-argument function, with its op created in this file
228 * FUN1 : not used, except for not, which isn't a UNIOP
229 * BOop : bitwise or or xor
231 * SHop : shift operator
232 * PWop : power operator
233 * PMop : pattern-matching operator
234 * Aop : addition-level operator
235 * Mop : multiplication-level operator
236 * Eop : equality-testing operator
237 * Rop : relational operator <= != gt
239 * Also see LOP and lop() below.
242 #ifdef DEBUGGING /* Serve -DT. */
243 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
245 # define REPORT(retval) (retval)
248 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
249 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
250 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
251 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
252 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
253 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
254 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
255 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
256 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
257 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
258 #define FUN0OP(f) return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
259 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
260 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
261 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
262 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
263 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
264 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
265 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
266 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
267 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
268 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
270 /* This bit of chicanery makes a unary function followed by
271 * a parenthesis into a function with one argument, highest precedence.
272 * The UNIDOR macro is for unary functions that can be followed by the //
273 * operator (such as C<shift // 0>).
275 #define UNI2(f,x) { \
276 pl_yylval.ival = f; \
279 PL_last_uni = PL_oldbufptr; \
280 PL_last_lop_op = f; \
282 return REPORT( (int)FUNC1 ); \
284 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
286 #define UNI(f) UNI2(f,XTERM)
287 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
289 #define UNIBRACK(f) { \
290 pl_yylval.ival = f; \
292 PL_last_uni = PL_oldbufptr; \
294 return REPORT( (int)FUNC1 ); \
296 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
299 /* grandfather return to old style */
302 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
303 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
304 pl_yylval.ival = (f); \
312 /* how to interpret the pl_yylval associated with the token */
316 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
322 static struct debug_tokens {
324 enum token_type type;
326 } const debug_tokens[] =
328 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
329 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
330 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
331 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
332 { ARROW, TOKENTYPE_NONE, "ARROW" },
333 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
334 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
335 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
336 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
337 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
338 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
339 { DO, TOKENTYPE_NONE, "DO" },
340 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
341 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
342 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
343 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
344 { ELSE, TOKENTYPE_NONE, "ELSE" },
345 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
346 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
347 { FOR, TOKENTYPE_IVAL, "FOR" },
348 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
349 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
350 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
351 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
352 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
353 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
354 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
355 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
356 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
357 { IF, TOKENTYPE_IVAL, "IF" },
358 { LABEL, TOKENTYPE_PVAL, "LABEL" },
359 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
360 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
361 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
362 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
363 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
364 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
365 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
366 { MY, TOKENTYPE_IVAL, "MY" },
367 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
368 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
369 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
370 { OROP, TOKENTYPE_IVAL, "OROP" },
371 { OROR, TOKENTYPE_NONE, "OROR" },
372 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
373 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
374 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
375 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
376 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
377 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
378 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
379 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
380 { PREINC, TOKENTYPE_NONE, "PREINC" },
381 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
382 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
383 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
384 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
385 { SUB, TOKENTYPE_NONE, "SUB" },
386 { THING, TOKENTYPE_OPVAL, "THING" },
387 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
388 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
389 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
390 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
391 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
392 { USE, TOKENTYPE_IVAL, "USE" },
393 { WHEN, TOKENTYPE_IVAL, "WHEN" },
394 { WHILE, TOKENTYPE_IVAL, "WHILE" },
395 { WORD, TOKENTYPE_OPVAL, "WORD" },
396 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
397 { 0, TOKENTYPE_NONE, NULL }
400 /* dump the returned token in rv, plus any optional arg in pl_yylval */
403 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
407 PERL_ARGS_ASSERT_TOKEREPORT;
410 const char *name = NULL;
411 enum token_type type = TOKENTYPE_NONE;
412 const struct debug_tokens *p;
413 SV* const report = newSVpvs("<== ");
415 for (p = debug_tokens; p->token; p++) {
416 if (p->token == (int)rv) {
423 Perl_sv_catpv(aTHX_ report, name);
424 else if ((char)rv > ' ' && (char)rv < '~')
425 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
427 sv_catpvs(report, "EOF");
429 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
432 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
435 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
437 case TOKENTYPE_OPNUM:
438 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
439 PL_op_name[lvalp->ival]);
442 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
444 case TOKENTYPE_OPVAL:
446 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
447 PL_op_name[lvalp->opval->op_type]);
448 if (lvalp->opval->op_type == OP_CONST) {
449 Perl_sv_catpvf(aTHX_ report, " %s",
450 SvPEEK(cSVOPx_sv(lvalp->opval)));
455 sv_catpvs(report, "(opval=null)");
458 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
464 /* print the buffer with suitable escapes */
467 S_printbuf(pTHX_ const char *const fmt, const char *const s)
469 SV* const tmp = newSVpvs("");
471 PERL_ARGS_ASSERT_PRINTBUF;
473 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
480 S_deprecate_commaless_var_list(pTHX) {
482 deprecate("comma-less variable list");
483 return REPORT(','); /* grandfather non-comma-format format */
489 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
490 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
494 S_ao(pTHX_ int toketype)
497 if (*PL_bufptr == '=') {
499 if (toketype == ANDAND)
500 pl_yylval.ival = OP_ANDASSIGN;
501 else if (toketype == OROR)
502 pl_yylval.ival = OP_ORASSIGN;
503 else if (toketype == DORDOR)
504 pl_yylval.ival = OP_DORASSIGN;
512 * When Perl expects an operator and finds something else, no_op
513 * prints the warning. It always prints "<something> found where
514 * operator expected. It prints "Missing semicolon on previous line?"
515 * if the surprise occurs at the start of the line. "do you need to
516 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
517 * where the compiler doesn't know if foo is a method call or a function.
518 * It prints "Missing operator before end of line" if there's nothing
519 * after the missing operator, or "... before <...>" if there is something
520 * after the missing operator.
524 S_no_op(pTHX_ const char *const what, char *s)
527 char * const oldbp = PL_bufptr;
528 const bool is_first = (PL_oldbufptr == PL_linestart);
530 PERL_ARGS_ASSERT_NO_OP;
536 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
537 if (ckWARN_d(WARN_SYNTAX)) {
539 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
540 "\t(Missing semicolon on previous line?)\n");
541 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
543 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
545 if (t < PL_bufptr && isSPACE(*t))
546 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
547 "\t(Do you need to predeclare %.*s?)\n",
548 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
552 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
553 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
561 * Complain about missing quote/regexp/heredoc terminator.
562 * If it's called with NULL then it cauterizes the line buffer.
563 * If we're in a delimited string and the delimiter is a control
564 * character, it's reformatted into a two-char sequence like ^C.
569 S_missingterm(pTHX_ char *s)
575 char * const nl = strrchr(s,'\n');
579 else if (isCNTRL(PL_multi_close)) {
581 tmpbuf[1] = (char)toCTRL(PL_multi_close);
586 *tmpbuf = (char)PL_multi_close;
590 q = strchr(s,'"') ? '\'' : '"';
591 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
595 * Check whether the named feature is enabled.
598 Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
601 HV * const hinthv = GvHV(PL_hintgv);
602 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
604 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
606 if (namelen > MAX_FEATURE_LEN)
608 memcpy(&he_name[8], name, namelen);
610 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
614 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
615 * utf16-to-utf8-reversed.
618 #ifdef PERL_CR_FILTER
622 register const char *s = SvPVX_const(sv);
623 register const char * const e = s + SvCUR(sv);
625 PERL_ARGS_ASSERT_STRIP_RETURN;
627 /* outer loop optimized to do nothing if there are no CR-LFs */
629 if (*s++ == '\r' && *s == '\n') {
630 /* hit a CR-LF, need to copy the rest */
631 register char *d = s - 1;
634 if (*s == '\r' && s[1] == '\n')
645 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
647 const I32 count = FILTER_READ(idx+1, sv, maxlen);
648 if (count > 0 && !maxlen)
655 =for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
657 Creates and initialises a new lexer/parser state object, supplying
658 a context in which to lex and parse from a new source of Perl code.
659 A pointer to the new state object is placed in L</PL_parser>. An entry
660 is made on the save stack so that upon unwinding the new state object
661 will be destroyed and the former value of L</PL_parser> will be restored.
662 Nothing else need be done to clean up the parsing context.
664 The code to be parsed comes from I<line> and I<rsfp>. I<line>, if
665 non-null, provides a string (in SV form) containing code to be parsed.
666 A copy of the string is made, so subsequent modification of I<line>
667 does not affect parsing. I<rsfp>, if non-null, provides an input stream
668 from which code will be read to be parsed. If both are non-null, the
669 code in I<line> comes first and must consist of complete lines of input,
670 and I<rsfp> supplies the remainder of the source.
672 The I<flags> parameter is reserved for future use, and must always
673 be zero, except for one flag that is currently reserved for perl's internal
679 /* LEX_START_SAME_FILTER indicates that this is not a new file, so it
680 can share filters with the current parser. */
683 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
686 const char *s = NULL;
688 yy_parser *parser, *oparser;
689 if (flags && flags != LEX_START_SAME_FILTER)
690 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
692 /* create and initialise a parser */
694 Newxz(parser, 1, yy_parser);
695 parser->old_parser = oparser = PL_parser;
698 parser->stack = NULL;
700 parser->stack_size = 0;
702 /* on scope exit, free this parser and restore any outer one */
704 parser->saved_curcop = PL_curcop;
706 /* initialise lexer state */
709 parser->curforce = -1;
711 parser->nexttoke = 0;
713 parser->error_count = oparser ? oparser->error_count : 0;
714 parser->copline = NOLINE;
715 parser->lex_state = LEX_NORMAL;
716 parser->expect = XSTATE;
718 parser->rsfp_filters =
719 !(flags & LEX_START_SAME_FILTER) || !oparser
721 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
723 Newx(parser->lex_brackstack, 120, char);
724 Newx(parser->lex_casestack, 12, char);
725 *parser->lex_casestack = '\0';
728 s = SvPV_const(line, len);
734 parser->linestr = newSVpvs("\n;");
736 parser->linestr = newSVpvn_flags(s, len, SvUTF8(line));
738 sv_catpvs(parser->linestr, "\n;");
740 parser->oldoldbufptr =
743 parser->linestart = SvPVX(parser->linestr);
744 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
745 parser->last_lop = parser->last_uni = NULL;
751 /* delete a parser object */
754 Perl_parser_free(pTHX_ const yy_parser *parser)
756 PERL_ARGS_ASSERT_PARSER_FREE;
758 PL_curcop = parser->saved_curcop;
759 SvREFCNT_dec(parser->linestr);
761 if (parser->rsfp == PerlIO_stdin())
762 PerlIO_clearerr(parser->rsfp);
763 else if (parser->rsfp && (!parser->old_parser ||
764 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
765 PerlIO_close(parser->rsfp);
766 SvREFCNT_dec(parser->rsfp_filters);
768 Safefree(parser->lex_brackstack);
769 Safefree(parser->lex_casestack);
770 PL_parser = parser->old_parser;
776 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
778 Buffer scalar containing the chunk currently under consideration of the
779 text currently being lexed. This is always a plain string scalar (for
780 which C<SvPOK> is true). It is not intended to be used as a scalar by
781 normal scalar means; instead refer to the buffer directly by the pointer
782 variables described below.
784 The lexer maintains various C<char*> pointers to things in the
785 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
786 reallocated, all of these pointers must be updated. Don't attempt to
787 do this manually, but rather use L</lex_grow_linestr> if you need to
788 reallocate the buffer.
790 The content of the text chunk in the buffer is commonly exactly one
791 complete line of input, up to and including a newline terminator,
792 but there are situations where it is otherwise. The octets of the
793 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
794 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
795 flag on this scalar, which may disagree with it.
797 For direct examination of the buffer, the variable
798 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
799 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
800 of these pointers is usually preferable to examination of the scalar
801 through normal scalar means.
803 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
805 Direct pointer to the end of the chunk of text currently being lexed, the
806 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
807 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
808 always located at the end of the buffer, and does not count as part of
809 the buffer's contents.
811 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
813 Points to the current position of lexing inside the lexer buffer.
814 Characters around this point may be freely examined, within
815 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
816 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
817 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
819 Lexing code (whether in the Perl core or not) moves this pointer past
820 the characters that it consumes. It is also expected to perform some
821 bookkeeping whenever a newline character is consumed. This movement
822 can be more conveniently performed by the function L</lex_read_to>,
823 which handles newlines appropriately.
825 Interpretation of the buffer's octets can be abstracted out by
826 using the slightly higher-level functions L</lex_peek_unichar> and
827 L</lex_read_unichar>.
829 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
831 Points to the start of the current line inside the lexer buffer.
832 This is useful for indicating at which column an error occurred, and
833 not much else. This must be updated by any lexing code that consumes
834 a newline; the function L</lex_read_to> handles this detail.
840 =for apidoc Amx|bool|lex_bufutf8
842 Indicates whether the octets in the lexer buffer
843 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
844 of Unicode characters. If not, they should be interpreted as Latin-1
845 characters. This is analogous to the C<SvUTF8> flag for scalars.
847 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
848 contains valid UTF-8. Lexing code must be robust in the face of invalid
851 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
852 is significant, but not the whole story regarding the input character
853 encoding. Normally, when a file is being read, the scalar contains octets
854 and its C<SvUTF8> flag is off, but the octets should be interpreted as
855 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
856 however, the scalar may have the C<SvUTF8> flag on, and in this case its
857 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
858 is in effect. This logic may change in the future; use this function
859 instead of implementing the logic yourself.
865 Perl_lex_bufutf8(pTHX)
871 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
873 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
874 at least I<len> octets (including terminating NUL). Returns a
875 pointer to the reallocated buffer. This is necessary before making
876 any direct modification of the buffer that would increase its length.
877 L</lex_stuff_pvn> provides a more convenient way to insert text into
880 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
881 this function updates all of the lexer's variables that point directly
888 Perl_lex_grow_linestr(pTHX_ STRLEN len)
892 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
893 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
894 linestr = PL_parser->linestr;
895 buf = SvPVX(linestr);
896 if (len <= SvLEN(linestr))
898 bufend_pos = PL_parser->bufend - buf;
899 bufptr_pos = PL_parser->bufptr - buf;
900 oldbufptr_pos = PL_parser->oldbufptr - buf;
901 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
902 linestart_pos = PL_parser->linestart - buf;
903 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
904 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
905 buf = sv_grow(linestr, len);
906 PL_parser->bufend = buf + bufend_pos;
907 PL_parser->bufptr = buf + bufptr_pos;
908 PL_parser->oldbufptr = buf + oldbufptr_pos;
909 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
910 PL_parser->linestart = buf + linestart_pos;
911 if (PL_parser->last_uni)
912 PL_parser->last_uni = buf + last_uni_pos;
913 if (PL_parser->last_lop)
914 PL_parser->last_lop = buf + last_lop_pos;
919 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
921 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
922 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
923 reallocating the buffer if necessary. This means that lexing code that
924 runs later will see the characters as if they had appeared in the input.
925 It is not recommended to do this as part of normal parsing, and most
926 uses of this facility run the risk of the inserted characters being
927 interpreted in an unintended manner.
929 The string to be inserted is represented by I<len> octets starting
930 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
931 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
932 The characters are recoded for the lexer buffer, according to how the
933 buffer is currently being interpreted (L</lex_bufutf8>). If a string
934 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
935 function is more convenient.
941 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
945 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
946 if (flags & ~(LEX_STUFF_UTF8))
947 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
949 if (flags & LEX_STUFF_UTF8) {
953 const char *p, *e = pv+len;
954 for (p = pv; p != e; p++)
955 highhalf += !!(((U8)*p) & 0x80);
958 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
959 bufptr = PL_parser->bufptr;
960 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
961 SvCUR_set(PL_parser->linestr,
962 SvCUR(PL_parser->linestr) + len+highhalf);
963 PL_parser->bufend += len+highhalf;
964 for (p = pv; p != e; p++) {
967 *bufptr++ = (char)(0xc0 | (c >> 6));
968 *bufptr++ = (char)(0x80 | (c & 0x3f));
975 if (flags & LEX_STUFF_UTF8) {
977 const char *p, *e = pv+len;
978 for (p = pv; p != e; p++) {
981 Perl_croak(aTHX_ "Lexing code attempted to stuff "
982 "non-Latin-1 character into Latin-1 input");
983 } else if (c >= 0xc2 && p+1 != e &&
984 (((U8)p[1]) & 0xc0) == 0x80) {
987 } else if (c >= 0x80) {
988 /* malformed UTF-8 */
990 SAVESPTR(PL_warnhook);
991 PL_warnhook = PERL_WARNHOOK_FATAL;
992 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
998 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
999 bufptr = PL_parser->bufptr;
1000 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1001 SvCUR_set(PL_parser->linestr,
1002 SvCUR(PL_parser->linestr) + len-highhalf);
1003 PL_parser->bufend += len-highhalf;
1004 for (p = pv; p != e; p++) {
1007 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1010 *bufptr++ = (char)c;
1015 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1016 bufptr = PL_parser->bufptr;
1017 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1018 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1019 PL_parser->bufend += len;
1020 Copy(pv, bufptr, len, char);
1026 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1028 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1029 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1030 reallocating the buffer if necessary. This means that lexing code that
1031 runs later will see the characters as if they had appeared in the input.
1032 It is not recommended to do this as part of normal parsing, and most
1033 uses of this facility run the risk of the inserted characters being
1034 interpreted in an unintended manner.
1036 The string to be inserted is represented by octets starting at I<pv>
1037 and continuing to the first nul. These octets are interpreted as either
1038 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1039 in I<flags>. The characters are recoded for the lexer buffer, according
1040 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1041 If it is not convenient to nul-terminate a string to be inserted, the
1042 L</lex_stuff_pvn> function is more appropriate.
1048 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1050 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1051 lex_stuff_pvn(pv, strlen(pv), flags);
1055 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1057 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1058 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1059 reallocating the buffer if necessary. This means that lexing code that
1060 runs later will see the characters as if they had appeared in the input.
1061 It is not recommended to do this as part of normal parsing, and most
1062 uses of this facility run the risk of the inserted characters being
1063 interpreted in an unintended manner.
1065 The string to be inserted is the string value of I<sv>. The characters
1066 are recoded for the lexer buffer, according to how the buffer is currently
1067 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1068 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1069 need to construct a scalar.
1075 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1079 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1081 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1083 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1087 =for apidoc Amx|void|lex_unstuff|char *ptr
1089 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1090 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1091 This hides the discarded text from any lexing code that runs later,
1092 as if the text had never appeared.
1094 This is not the normal way to consume lexed text. For that, use
1101 Perl_lex_unstuff(pTHX_ char *ptr)
1105 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1106 buf = PL_parser->bufptr;
1108 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1111 bufend = PL_parser->bufend;
1113 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1114 unstuff_len = ptr - buf;
1115 Move(ptr, buf, bufend+1-ptr, char);
1116 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1117 PL_parser->bufend = bufend - unstuff_len;
1121 =for apidoc Amx|void|lex_read_to|char *ptr
1123 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1124 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1125 performing the correct bookkeeping whenever a newline character is passed.
1126 This is the normal way to consume lexed text.
1128 Interpretation of the buffer's octets can be abstracted out by
1129 using the slightly higher-level functions L</lex_peek_unichar> and
1130 L</lex_read_unichar>.
1136 Perl_lex_read_to(pTHX_ char *ptr)
1139 PERL_ARGS_ASSERT_LEX_READ_TO;
1140 s = PL_parser->bufptr;
1141 if (ptr < s || ptr > PL_parser->bufend)
1142 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1143 for (; s != ptr; s++)
1145 CopLINE_inc(PL_curcop);
1146 PL_parser->linestart = s+1;
1148 PL_parser->bufptr = ptr;
1152 =for apidoc Amx|void|lex_discard_to|char *ptr
1154 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1155 up to I<ptr>. The remaining content of the buffer will be moved, and
1156 all pointers into the buffer updated appropriately. I<ptr> must not
1157 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1158 it is not permitted to discard text that has yet to be lexed.
1160 Normally it is not necessarily to do this directly, because it suffices to
1161 use the implicit discarding behaviour of L</lex_next_chunk> and things
1162 based on it. However, if a token stretches across multiple lines,
1163 and the lexing code has kept multiple lines of text in the buffer for
1164 that purpose, then after completion of the token it would be wise to
1165 explicitly discard the now-unneeded earlier lines, to avoid future
1166 multi-line tokens growing the buffer without bound.
1172 Perl_lex_discard_to(pTHX_ char *ptr)
1176 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1177 buf = SvPVX(PL_parser->linestr);
1179 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1182 if (ptr > PL_parser->bufptr)
1183 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1184 discard_len = ptr - buf;
1185 if (PL_parser->oldbufptr < ptr)
1186 PL_parser->oldbufptr = ptr;
1187 if (PL_parser->oldoldbufptr < ptr)
1188 PL_parser->oldoldbufptr = ptr;
1189 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1190 PL_parser->last_uni = NULL;
1191 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1192 PL_parser->last_lop = NULL;
1193 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1194 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1195 PL_parser->bufend -= discard_len;
1196 PL_parser->bufptr -= discard_len;
1197 PL_parser->oldbufptr -= discard_len;
1198 PL_parser->oldoldbufptr -= discard_len;
1199 if (PL_parser->last_uni)
1200 PL_parser->last_uni -= discard_len;
1201 if (PL_parser->last_lop)
1202 PL_parser->last_lop -= discard_len;
1206 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1208 Reads in the next chunk of text to be lexed, appending it to
1209 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1210 looked to the end of the current chunk and wants to know more. It is
1211 usual, but not necessary, for lexing to have consumed the entirety of
1212 the current chunk at this time.
1214 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1215 chunk (i.e., the current chunk has been entirely consumed), normally the
1216 current chunk will be discarded at the same time that the new chunk is
1217 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1218 will not be discarded. If the current chunk has not been entirely
1219 consumed, then it will not be discarded regardless of the flag.
1221 Returns true if some new text was added to the buffer, or false if the
1222 buffer has reached the end of the input text.
1227 #define LEX_FAKE_EOF 0x80000000
1230 Perl_lex_next_chunk(pTHX_ U32 flags)
1234 STRLEN old_bufend_pos, new_bufend_pos;
1235 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1236 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1237 bool got_some_for_debugger = 0;
1239 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1240 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1241 linestr = PL_parser->linestr;
1242 buf = SvPVX(linestr);
1243 if (!(flags & LEX_KEEP_PREVIOUS) &&
1244 PL_parser->bufptr == PL_parser->bufend) {
1245 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1247 if (PL_parser->last_uni != PL_parser->bufend)
1248 PL_parser->last_uni = NULL;
1249 if (PL_parser->last_lop != PL_parser->bufend)
1250 PL_parser->last_lop = NULL;
1251 last_uni_pos = last_lop_pos = 0;
1255 old_bufend_pos = PL_parser->bufend - buf;
1256 bufptr_pos = PL_parser->bufptr - buf;
1257 oldbufptr_pos = PL_parser->oldbufptr - buf;
1258 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1259 linestart_pos = PL_parser->linestart - buf;
1260 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1261 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1263 if (flags & LEX_FAKE_EOF) {
1265 } else if (!PL_parser->rsfp) {
1267 } else if (filter_gets(linestr, old_bufend_pos)) {
1269 got_some_for_debugger = 1;
1271 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1272 sv_setpvs(linestr, "");
1274 /* End of real input. Close filehandle (unless it was STDIN),
1275 * then add implicit termination.
1277 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1278 PerlIO_clearerr(PL_parser->rsfp);
1279 else if (PL_parser->rsfp)
1280 (void)PerlIO_close(PL_parser->rsfp);
1281 PL_parser->rsfp = NULL;
1282 PL_parser->in_pod = 0;
1284 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1287 if (!PL_in_eval && PL_minus_p) {
1289 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1290 PL_minus_n = PL_minus_p = 0;
1291 } else if (!PL_in_eval && PL_minus_n) {
1292 sv_catpvs(linestr, /*{*/";}");
1295 sv_catpvs(linestr, ";");
1298 buf = SvPVX(linestr);
1299 new_bufend_pos = SvCUR(linestr);
1300 PL_parser->bufend = buf + new_bufend_pos;
1301 PL_parser->bufptr = buf + bufptr_pos;
1302 PL_parser->oldbufptr = buf + oldbufptr_pos;
1303 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1304 PL_parser->linestart = buf + linestart_pos;
1305 if (PL_parser->last_uni)
1306 PL_parser->last_uni = buf + last_uni_pos;
1307 if (PL_parser->last_lop)
1308 PL_parser->last_lop = buf + last_lop_pos;
1309 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1310 PL_curstash != PL_debstash) {
1311 /* debugger active and we're not compiling the debugger code,
1312 * so store the line into the debugger's array of lines
1314 update_debugger_info(NULL, buf+old_bufend_pos,
1315 new_bufend_pos-old_bufend_pos);
1321 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1323 Looks ahead one (Unicode) character in the text currently being lexed.
1324 Returns the codepoint (unsigned integer value) of the next character,
1325 or -1 if lexing has reached the end of the input text. To consume the
1326 peeked character, use L</lex_read_unichar>.
1328 If the next character is in (or extends into) the next chunk of input
1329 text, the next chunk will be read in. Normally the current chunk will be
1330 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1331 then the current chunk will not be discarded.
1333 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1334 is encountered, an exception is generated.
1340 Perl_lex_peek_unichar(pTHX_ U32 flags)
1344 if (flags & ~(LEX_KEEP_PREVIOUS))
1345 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1346 s = PL_parser->bufptr;
1347 bufend = PL_parser->bufend;
1353 if (!lex_next_chunk(flags))
1355 s = PL_parser->bufptr;
1356 bufend = PL_parser->bufend;
1362 len = PL_utf8skip[head];
1363 while ((STRLEN)(bufend-s) < len) {
1364 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1366 s = PL_parser->bufptr;
1367 bufend = PL_parser->bufend;
1370 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1371 if (retlen == (STRLEN)-1) {
1372 /* malformed UTF-8 */
1374 SAVESPTR(PL_warnhook);
1375 PL_warnhook = PERL_WARNHOOK_FATAL;
1376 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1382 if (!lex_next_chunk(flags))
1384 s = PL_parser->bufptr;
1391 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1393 Reads the next (Unicode) character in the text currently being lexed.
1394 Returns the codepoint (unsigned integer value) of the character read,
1395 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1396 if lexing has reached the end of the input text. To non-destructively
1397 examine the next character, use L</lex_peek_unichar> instead.
1399 If the next character is in (or extends into) the next chunk of input
1400 text, the next chunk will be read in. Normally the current chunk will be
1401 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1402 then the current chunk will not be discarded.
1404 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1405 is encountered, an exception is generated.
1411 Perl_lex_read_unichar(pTHX_ U32 flags)
1414 if (flags & ~(LEX_KEEP_PREVIOUS))
1415 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1416 c = lex_peek_unichar(flags);
1419 CopLINE_inc(PL_curcop);
1420 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1426 =for apidoc Amx|void|lex_read_space|U32 flags
1428 Reads optional spaces, in Perl style, in the text currently being
1429 lexed. The spaces may include ordinary whitespace characters and
1430 Perl-style comments. C<#line> directives are processed if encountered.
1431 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1432 at a non-space character (or the end of the input text).
1434 If spaces extend into the next chunk of input text, the next chunk will
1435 be read in. Normally the current chunk will be discarded at the same
1436 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1437 chunk will not be discarded.
1442 #define LEX_NO_NEXT_CHUNK 0x80000000
1445 Perl_lex_read_space(pTHX_ U32 flags)
1448 bool need_incline = 0;
1449 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1450 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1453 sv_free(PL_skipwhite);
1454 PL_skipwhite = NULL;
1457 PL_skipwhite = newSVpvs("");
1458 #endif /* PERL_MAD */
1459 s = PL_parser->bufptr;
1460 bufend = PL_parser->bufend;
1466 } while (!(c == '\n' || (c == 0 && s == bufend)));
1467 } else if (c == '\n') {
1469 PL_parser->linestart = s;
1474 } else if (isSPACE(c)) {
1476 } else if (c == 0 && s == bufend) {
1480 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1481 #endif /* PERL_MAD */
1482 if (flags & LEX_NO_NEXT_CHUNK)
1484 PL_parser->bufptr = s;
1485 CopLINE_inc(PL_curcop);
1486 got_more = lex_next_chunk(flags);
1487 CopLINE_dec(PL_curcop);
1488 s = PL_parser->bufptr;
1489 bufend = PL_parser->bufend;
1492 if (need_incline && PL_parser->rsfp) {
1502 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1503 #endif /* PERL_MAD */
1504 PL_parser->bufptr = s;
1509 * This subroutine has nothing to do with tilting, whether at windmills
1510 * or pinball tables. Its name is short for "increment line". It
1511 * increments the current line number in CopLINE(PL_curcop) and checks
1512 * to see whether the line starts with a comment of the form
1513 * # line 500 "foo.pm"
1514 * If so, it sets the current line number and file to the values in the comment.
1518 S_incline(pTHX_ const char *s)
1526 PERL_ARGS_ASSERT_INCLINE;
1528 CopLINE_inc(PL_curcop);
1531 while (SPACE_OR_TAB(*s))
1533 if (strnEQ(s, "line", 4))
1537 if (SPACE_OR_TAB(*s))
1541 while (SPACE_OR_TAB(*s))
1549 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1551 while (SPACE_OR_TAB(*s))
1553 if (*s == '"' && (t = strchr(s+1, '"'))) {
1559 while (!isSPACE(*t))
1563 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1565 if (*e != '\n' && *e != '\0')
1566 return; /* false alarm */
1568 line_num = atoi(n)-1;
1571 const STRLEN len = t - s;
1572 SV *const temp_sv = CopFILESV(PL_curcop);
1577 cf = SvPVX(temp_sv);
1578 tmplen = SvCUR(temp_sv);
1584 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
1585 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1586 * to *{"::_<newfilename"} */
1587 /* However, the long form of evals is only turned on by the
1588 debugger - usually they're "(eval %lu)" */
1592 STRLEN tmplen2 = len;
1593 if (tmplen + 2 <= sizeof smallbuf)
1596 Newx(tmpbuf, tmplen + 2, char);
1599 memcpy(tmpbuf + 2, cf, tmplen);
1601 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1606 if (tmplen2 + 2 <= sizeof smallbuf)
1609 Newx(tmpbuf2, tmplen2 + 2, char);
1611 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1612 /* Either they malloc'd it, or we malloc'd it,
1613 so no prefix is present in ours. */
1618 memcpy(tmpbuf2 + 2, s, tmplen2);
1621 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1623 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1624 /* adjust ${"::_<newfilename"} to store the new file name */
1625 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1626 /* The line number may differ. If that is the case,
1627 alias the saved lines that are in the array.
1628 Otherwise alias the whole array. */
1629 if (CopLINE(PL_curcop) == line_num) {
1630 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1631 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1633 else if (GvAV(*gvp)) {
1634 AV * const av = GvAV(*gvp);
1635 const I32 start = CopLINE(PL_curcop)+1;
1636 I32 items = AvFILLp(av) - start;
1638 AV * const av2 = GvAVn(gv2);
1639 SV **svp = AvARRAY(av) + start;
1640 I32 l = (I32)line_num+1;
1642 av_store(av2, l++, SvREFCNT_inc(*svp++));
1647 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1649 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1651 CopFILE_free(PL_curcop);
1652 CopFILE_setn(PL_curcop, s, len);
1654 CopLINE_set(PL_curcop, line_num);
1658 /* skip space before PL_thistoken */
1661 S_skipspace0(pTHX_ register char *s)
1663 PERL_ARGS_ASSERT_SKIPSPACE0;
1670 PL_thiswhite = newSVpvs("");
1671 sv_catsv(PL_thiswhite, PL_skipwhite);
1672 sv_free(PL_skipwhite);
1675 PL_realtokenstart = s - SvPVX(PL_linestr);
1679 /* skip space after PL_thistoken */
1682 S_skipspace1(pTHX_ register char *s)
1684 const char *start = s;
1685 I32 startoff = start - SvPVX(PL_linestr);
1687 PERL_ARGS_ASSERT_SKIPSPACE1;
1692 start = SvPVX(PL_linestr) + startoff;
1693 if (!PL_thistoken && PL_realtokenstart >= 0) {
1694 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1695 PL_thistoken = newSVpvn(tstart, start - tstart);
1697 PL_realtokenstart = -1;
1700 PL_nextwhite = newSVpvs("");
1701 sv_catsv(PL_nextwhite, PL_skipwhite);
1702 sv_free(PL_skipwhite);
1709 S_skipspace2(pTHX_ register char *s, SV **svp)
1712 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1713 const I32 startoff = s - SvPVX(PL_linestr);
1715 PERL_ARGS_ASSERT_SKIPSPACE2;
1718 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1719 if (!PL_madskills || !svp)
1721 start = SvPVX(PL_linestr) + startoff;
1722 if (!PL_thistoken && PL_realtokenstart >= 0) {
1723 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1724 PL_thistoken = newSVpvn(tstart, start - tstart);
1725 PL_realtokenstart = -1;
1729 *svp = newSVpvs("");
1730 sv_setsv(*svp, PL_skipwhite);
1731 sv_free(PL_skipwhite);
1740 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1742 AV *av = CopFILEAVx(PL_curcop);
1744 SV * const sv = newSV_type(SVt_PVMG);
1746 sv_setsv(sv, orig_sv);
1748 sv_setpvn(sv, buf, len);
1751 av_store(av, (I32)CopLINE(PL_curcop), sv);
1757 * Called to gobble the appropriate amount and type of whitespace.
1758 * Skips comments as well.
1762 S_skipspace(pTHX_ register char *s)
1766 #endif /* PERL_MAD */
1767 PERL_ARGS_ASSERT_SKIPSPACE;
1770 sv_free(PL_skipwhite);
1771 PL_skipwhite = NULL;
1773 #endif /* PERL_MAD */
1774 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1775 while (s < PL_bufend && SPACE_OR_TAB(*s))
1778 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1780 lex_read_space(LEX_KEEP_PREVIOUS |
1781 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1782 LEX_NO_NEXT_CHUNK : 0));
1784 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1785 if (PL_linestart > PL_bufptr)
1786 PL_bufptr = PL_linestart;
1791 PL_skipwhite = newSVpvn(start, s-start);
1792 #endif /* PERL_MAD */
1798 * Check the unary operators to ensure there's no ambiguity in how they're
1799 * used. An ambiguous piece of code would be:
1801 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1802 * the +5 is its argument.
1812 if (PL_oldoldbufptr != PL_last_uni)
1814 while (isSPACE(*PL_last_uni))
1817 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1819 if ((t = strchr(s, '(')) && t < PL_bufptr)
1822 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1823 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1824 (int)(s - PL_last_uni), PL_last_uni);
1828 * LOP : macro to build a list operator. Its behaviour has been replaced
1829 * with a subroutine, S_lop() for which LOP is just another name.
1832 #define LOP(f,x) return lop(f,x,s)
1836 * Build a list operator (or something that might be one). The rules:
1837 * - if we have a next token, then it's a list operator [why?]
1838 * - if the next thing is an opening paren, then it's a function
1839 * - else it's a list operator
1843 S_lop(pTHX_ I32 f, int x, char *s)
1847 PERL_ARGS_ASSERT_LOP;
1853 PL_last_lop = PL_oldbufptr;
1854 PL_last_lop_op = (OPCODE)f;
1863 return REPORT(FUNC);
1866 return REPORT(FUNC);
1869 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1870 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1871 return REPORT(LSTOP);
1878 * Sets up for an eventual force_next(). start_force(0) basically does
1879 * an unshift, while start_force(-1) does a push. yylex removes items
1884 S_start_force(pTHX_ int where)
1888 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1889 where = PL_lasttoke;
1890 assert(PL_curforce < 0 || PL_curforce == where);
1891 if (PL_curforce != where) {
1892 for (i = PL_lasttoke; i > where; --i) {
1893 PL_nexttoke[i] = PL_nexttoke[i-1];
1897 if (PL_curforce < 0) /* in case of duplicate start_force() */
1898 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1899 PL_curforce = where;
1902 curmad('^', newSVpvs(""));
1903 CURMAD('_', PL_nextwhite);
1908 S_curmad(pTHX_ char slot, SV *sv)
1914 if (PL_curforce < 0)
1915 where = &PL_thismad;
1917 where = &PL_nexttoke[PL_curforce].next_mad;
1923 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1925 else if (PL_encoding) {
1926 sv_recode_to_utf8(sv, PL_encoding);
1931 /* keep a slot open for the head of the list? */
1932 if (slot != '_' && *where && (*where)->mad_key == '^') {
1933 (*where)->mad_key = slot;
1934 sv_free(MUTABLE_SV(((*where)->mad_val)));
1935 (*where)->mad_val = (void*)sv;
1938 addmad(newMADsv(slot, sv), where, 0);
1941 # define start_force(where) NOOP
1942 # define curmad(slot, sv) NOOP
1947 * When the lexer realizes it knows the next token (for instance,
1948 * it is reordering tokens for the parser) then it can call S_force_next
1949 * to know what token to return the next time the lexer is called. Caller
1950 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1951 * and possibly PL_expect to ensure the lexer handles the token correctly.
1955 S_force_next(pTHX_ I32 type)
1960 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1961 tokereport(type, &NEXTVAL_NEXTTOKE);
1965 if (PL_curforce < 0)
1966 start_force(PL_lasttoke);
1967 PL_nexttoke[PL_curforce].next_type = type;
1968 if (PL_lex_state != LEX_KNOWNEXT)
1969 PL_lex_defer = PL_lex_state;
1970 PL_lex_state = LEX_KNOWNEXT;
1971 PL_lex_expect = PL_expect;
1974 PL_nexttype[PL_nexttoke] = type;
1976 if (PL_lex_state != LEX_KNOWNEXT) {
1977 PL_lex_defer = PL_lex_state;
1978 PL_lex_expect = PL_expect;
1979 PL_lex_state = LEX_KNOWNEXT;
1987 int yyc = PL_parser->yychar;
1988 if (yyc != YYEMPTY) {
1991 NEXTVAL_NEXTTOKE = PL_parser->yylval;
1992 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
1993 PL_lex_allbrackets--;
1995 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
1996 } else if (yyc == '('/*)*/) {
1997 PL_lex_allbrackets--;
2002 PL_parser->yychar = YYEMPTY;
2007 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
2010 SV * const sv = newSVpvn_utf8(start, len,
2013 && !is_ascii_string((const U8*)start, len)
2014 && is_utf8_string((const U8*)start, len));
2020 * When the lexer knows the next thing is a word (for instance, it has
2021 * just seen -> and it knows that the next char is a word char, then
2022 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2026 * char *start : buffer position (must be within PL_linestr)
2027 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2028 * int check_keyword : if true, Perl checks to make sure the word isn't
2029 * a keyword (do this if the word is a label, e.g. goto FOO)
2030 * int allow_pack : if true, : characters will also be allowed (require,
2031 * use, etc. do this)
2032 * int allow_initial_tick : used by the "sub" lexer only.
2036 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
2042 PERL_ARGS_ASSERT_FORCE_WORD;
2044 start = SKIPSPACE1(start);
2046 if (isIDFIRST_lazy_if(s,UTF) ||
2047 (allow_pack && *s == ':') ||
2048 (allow_initial_tick && *s == '\'') )
2050 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2051 if (check_keyword && keyword(PL_tokenbuf, len, 0))
2053 start_force(PL_curforce);
2055 curmad('X', newSVpvn(start,s-start));
2056 if (token == METHOD) {
2061 PL_expect = XOPERATOR;
2065 curmad('g', newSVpvs( "forced" ));
2066 NEXTVAL_NEXTTOKE.opval
2067 = (OP*)newSVOP(OP_CONST,0,
2068 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2069 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2077 * Called when the lexer wants $foo *foo &foo etc, but the program
2078 * text only contains the "foo" portion. The first argument is a pointer
2079 * to the "foo", and the second argument is the type symbol to prefix.
2080 * Forces the next token to be a "WORD".
2081 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2085 S_force_ident(pTHX_ register const char *s, int kind)
2089 PERL_ARGS_ASSERT_FORCE_IDENT;
2092 const STRLEN len = strlen(s);
2093 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
2094 start_force(PL_curforce);
2095 NEXTVAL_NEXTTOKE.opval = o;
2098 o->op_private = OPpCONST_ENTERED;
2099 /* XXX see note in pp_entereval() for why we forgo typo
2100 warnings if the symbol must be introduced in an eval.
2102 gv_fetchpvn_flags(s, len,
2103 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2105 kind == '$' ? SVt_PV :
2106 kind == '@' ? SVt_PVAV :
2107 kind == '%' ? SVt_PVHV :
2115 Perl_str_to_version(pTHX_ SV *sv)
2120 const char *start = SvPV_const(sv,len);
2121 const char * const end = start + len;
2122 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2124 PERL_ARGS_ASSERT_STR_TO_VERSION;
2126 while (start < end) {
2130 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2135 retval += ((NV)n)/nshift;
2144 * Forces the next token to be a version number.
2145 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2146 * and if "guessing" is TRUE, then no new token is created (and the caller
2147 * must use an alternative parsing method).
2151 S_force_version(pTHX_ char *s, int guessing)
2157 I32 startoff = s - SvPVX(PL_linestr);
2160 PERL_ARGS_ASSERT_FORCE_VERSION;
2168 while (isDIGIT(*d) || *d == '_' || *d == '.')
2172 start_force(PL_curforce);
2173 curmad('X', newSVpvn(s,d-s));
2176 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2178 #ifdef USE_LOCALE_NUMERIC
2179 char *loc = setlocale(LC_NUMERIC, "C");
2181 s = scan_num(s, &pl_yylval);
2182 #ifdef USE_LOCALE_NUMERIC
2183 setlocale(LC_NUMERIC, loc);
2185 version = pl_yylval.opval;
2186 ver = cSVOPx(version)->op_sv;
2187 if (SvPOK(ver) && !SvNIOK(ver)) {
2188 SvUPGRADE(ver, SVt_PVNV);
2189 SvNV_set(ver, str_to_version(ver));
2190 SvNOK_on(ver); /* hint that it is a version */
2193 else if (guessing) {
2196 sv_free(PL_nextwhite); /* let next token collect whitespace */
2198 s = SvPVX(PL_linestr) + startoff;
2206 if (PL_madskills && !version) {
2207 sv_free(PL_nextwhite); /* let next token collect whitespace */
2209 s = SvPVX(PL_linestr) + startoff;
2212 /* NOTE: The parser sees the package name and the VERSION swapped */
2213 start_force(PL_curforce);
2214 NEXTVAL_NEXTTOKE.opval = version;
2221 * S_force_strict_version
2222 * Forces the next token to be a version number using strict syntax rules.
2226 S_force_strict_version(pTHX_ char *s)
2231 I32 startoff = s - SvPVX(PL_linestr);
2233 const char *errstr = NULL;
2235 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2237 while (isSPACE(*s)) /* leading whitespace */
2240 if (is_STRICT_VERSION(s,&errstr)) {
2242 s = (char *)scan_version(s, ver, 0);
2243 version = newSVOP(OP_CONST, 0, ver);
2245 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2246 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2250 yyerror(errstr); /* version required */
2255 if (PL_madskills && !version) {
2256 sv_free(PL_nextwhite); /* let next token collect whitespace */
2258 s = SvPVX(PL_linestr) + startoff;
2261 /* NOTE: The parser sees the package name and the VERSION swapped */
2262 start_force(PL_curforce);
2263 NEXTVAL_NEXTTOKE.opval = version;
2271 * Tokenize a quoted string passed in as an SV. It finds the next
2272 * chunk, up to end of string or a backslash. It may make a new
2273 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2278 S_tokeq(pTHX_ SV *sv)
2282 register char *send;
2287 PERL_ARGS_ASSERT_TOKEQ;
2292 s = SvPV_force(sv, len);
2293 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2296 /* This is relying on the SV being "well formed" with a trailing '\0' */
2297 while (s < send && !(*s == '\\' && s[1] == '\\'))
2302 if ( PL_hints & HINT_NEW_STRING ) {
2303 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2307 if (s + 1 < send && (s[1] == '\\'))
2308 s++; /* all that, just for this */
2313 SvCUR_set(sv, d - SvPVX_const(sv));
2315 if ( PL_hints & HINT_NEW_STRING )
2316 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2321 * Now come three functions related to double-quote context,
2322 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2323 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2324 * interact with PL_lex_state, and create fake ( ... ) argument lists
2325 * to handle functions and concatenation.
2326 * They assume that whoever calls them will be setting up a fake
2327 * join call, because each subthing puts a ',' after it. This lets
2330 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2332 * (I'm not sure whether the spurious commas at the end of lcfirst's
2333 * arguments and join's arguments are created or not).
2338 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2340 * Pattern matching will set PL_lex_op to the pattern-matching op to
2341 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2343 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2345 * Everything else becomes a FUNC.
2347 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2348 * had an OP_CONST or OP_READLINE). This just sets us up for a
2349 * call to S_sublex_push().
2353 S_sublex_start(pTHX)
2356 register const I32 op_type = pl_yylval.ival;
2358 if (op_type == OP_NULL) {
2359 pl_yylval.opval = PL_lex_op;
2363 if (op_type == OP_CONST || op_type == OP_READLINE) {
2364 SV *sv = tokeq(PL_lex_stuff);
2366 if (SvTYPE(sv) == SVt_PVIV) {
2367 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2369 const char * const p = SvPV_const(sv, len);
2370 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2374 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2375 PL_lex_stuff = NULL;
2376 /* Allow <FH> // "foo" */
2377 if (op_type == OP_READLINE)
2378 PL_expect = XTERMORDORDOR;
2381 else if (op_type == OP_BACKTICK && PL_lex_op) {
2382 /* readpipe() vas overriden */
2383 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2384 pl_yylval.opval = PL_lex_op;
2386 PL_lex_stuff = NULL;
2390 PL_sublex_info.super_state = PL_lex_state;
2391 PL_sublex_info.sub_inwhat = (U16)op_type;
2392 PL_sublex_info.sub_op = PL_lex_op;
2393 PL_lex_state = LEX_INTERPPUSH;
2397 pl_yylval.opval = PL_lex_op;
2407 * Create a new scope to save the lexing state. The scope will be
2408 * ended in S_sublex_done. Returns a '(', starting the function arguments
2409 * to the uc, lc, etc. found before.
2410 * Sets PL_lex_state to LEX_INTERPCONCAT.
2419 PL_lex_state = PL_sublex_info.super_state;
2420 SAVEBOOL(PL_lex_dojoin);
2421 SAVEI32(PL_lex_brackets);
2422 SAVEI32(PL_lex_allbrackets);
2423 SAVEI8(PL_lex_fakeeof);
2424 SAVEI32(PL_lex_casemods);
2425 SAVEI32(PL_lex_starts);
2426 SAVEI8(PL_lex_state);
2427 SAVEVPTR(PL_lex_inpat);
2428 SAVEI16(PL_lex_inwhat);
2429 SAVECOPLINE(PL_curcop);
2430 SAVEPPTR(PL_bufptr);
2431 SAVEPPTR(PL_bufend);
2432 SAVEPPTR(PL_oldbufptr);
2433 SAVEPPTR(PL_oldoldbufptr);
2434 SAVEPPTR(PL_last_lop);
2435 SAVEPPTR(PL_last_uni);
2436 SAVEPPTR(PL_linestart);
2437 SAVESPTR(PL_linestr);
2438 SAVEGENERICPV(PL_lex_brackstack);
2439 SAVEGENERICPV(PL_lex_casestack);
2441 PL_linestr = PL_lex_stuff;
2442 PL_lex_stuff = NULL;
2444 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2445 = SvPVX(PL_linestr);
2446 PL_bufend += SvCUR(PL_linestr);
2447 PL_last_lop = PL_last_uni = NULL;
2448 SAVEFREESV(PL_linestr);
2450 PL_lex_dojoin = FALSE;
2451 PL_lex_brackets = 0;
2452 PL_lex_allbrackets = 0;
2453 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2454 Newx(PL_lex_brackstack, 120, char);
2455 Newx(PL_lex_casestack, 12, char);
2456 PL_lex_casemods = 0;
2457 *PL_lex_casestack = '\0';
2459 PL_lex_state = LEX_INTERPCONCAT;
2460 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2462 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2463 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2464 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2465 PL_lex_inpat = PL_sublex_info.sub_op;
2467 PL_lex_inpat = NULL;
2474 * Restores lexer state after a S_sublex_push.
2481 if (!PL_lex_starts++) {
2482 SV * const sv = newSVpvs("");
2483 if (SvUTF8(PL_linestr))
2485 PL_expect = XOPERATOR;
2486 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2490 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2491 PL_lex_state = LEX_INTERPCASEMOD;
2495 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2496 assert(PL_lex_inwhat != OP_TRANSR);
2497 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2498 PL_linestr = PL_lex_repl;
2500 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2501 PL_bufend += SvCUR(PL_linestr);
2502 PL_last_lop = PL_last_uni = NULL;
2503 SAVEFREESV(PL_linestr);
2504 PL_lex_dojoin = FALSE;
2505 PL_lex_brackets = 0;
2506 PL_lex_allbrackets = 0;
2507 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2508 PL_lex_casemods = 0;
2509 *PL_lex_casestack = '\0';
2511 if (SvEVALED(PL_lex_repl)) {
2512 PL_lex_state = LEX_INTERPNORMAL;
2514 /* we don't clear PL_lex_repl here, so that we can check later
2515 whether this is an evalled subst; that means we rely on the
2516 logic to ensure sublex_done() is called again only via the
2517 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2520 PL_lex_state = LEX_INTERPCONCAT;
2530 PL_endwhite = newSVpvs("");
2531 sv_catsv(PL_endwhite, PL_thiswhite);
2535 sv_setpvs(PL_thistoken,"");
2537 PL_realtokenstart = -1;
2541 PL_bufend = SvPVX(PL_linestr);
2542 PL_bufend += SvCUR(PL_linestr);
2543 PL_expect = XOPERATOR;
2544 PL_sublex_info.sub_inwhat = 0;
2552 Extracts a pattern, double-quoted string, or transliteration. This
2555 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2556 processing a pattern (PL_lex_inpat is true), a transliteration
2557 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2559 Returns a pointer to the character scanned up to. If this is
2560 advanced from the start pointer supplied (i.e. if anything was
2561 successfully parsed), will leave an OP for the substring scanned
2562 in pl_yylval. Caller must intuit reason for not parsing further
2563 by looking at the next characters herself.
2567 constants: \N{NAME} only
2568 case and quoting: \U \Q \E
2569 stops on @ and $, but not for $ as tail anchor
2571 In transliterations:
2572 characters are VERY literal, except for - not at the start or end
2573 of the string, which indicates a range. If the range is in bytes,
2574 scan_const expands the range to the full set of intermediate
2575 characters. If the range is in utf8, the hyphen is replaced with
2576 a certain range mark which will be handled by pmtrans() in op.c.
2578 In double-quoted strings:
2580 double-quoted style: \r and \n
2581 constants: \x31, etc.
2582 deprecated backrefs: \1 (in substitution replacements)
2583 case and quoting: \U \Q \E
2586 scan_const does *not* construct ops to handle interpolated strings.
2587 It stops processing as soon as it finds an embedded $ or @ variable
2588 and leaves it to the caller to work out what's going on.
2590 embedded arrays (whether in pattern or not) could be:
2591 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2593 $ in double-quoted strings must be the symbol of an embedded scalar.
2595 $ in pattern could be $foo or could be tail anchor. Assumption:
2596 it's a tail anchor if $ is the last thing in the string, or if it's
2597 followed by one of "()| \r\n\t"
2599 \1 (backreferences) are turned into $1
2601 The structure of the code is
2602 while (there's a character to process) {
2603 handle transliteration ranges
2604 skip regexp comments /(?#comment)/ and codes /(?{code})/
2605 skip #-initiated comments in //x patterns
2606 check for embedded arrays
2607 check for embedded scalars
2609 deprecate \1 in substitution replacements
2610 handle string-changing backslashes \l \U \Q \E, etc.
2611 switch (what was escaped) {
2612 handle \- in a transliteration (becomes a literal -)
2613 if a pattern and not \N{, go treat as regular character
2614 handle \132 (octal characters)
2615 handle \x15 and \x{1234} (hex characters)
2616 handle \N{name} (named characters, also \N{3,5} in a pattern)
2617 handle \cV (control characters)
2618 handle printf-style backslashes (\f, \r, \n, etc)
2621 } (end if backslash)
2622 handle regular character
2623 } (end while character to read)
2628 S_scan_const(pTHX_ char *start)
2631 register char *send = PL_bufend; /* end of the constant */
2632 SV *sv = newSV(send - start); /* sv for the constant. See
2633 note below on sizing. */
2634 register char *s = start; /* start of the constant */
2635 register char *d = SvPVX(sv); /* destination for copies */
2636 bool dorange = FALSE; /* are we in a translit range? */
2637 bool didrange = FALSE; /* did we just finish a range? */
2638 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2639 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
2640 to be UTF8? But, this can
2641 show as true when the source
2642 isn't utf8, as for example
2643 when it is entirely composed
2646 /* Note on sizing: The scanned constant is placed into sv, which is
2647 * initialized by newSV() assuming one byte of output for every byte of
2648 * input. This routine expects newSV() to allocate an extra byte for a
2649 * trailing NUL, which this routine will append if it gets to the end of
2650 * the input. There may be more bytes of input than output (eg., \N{LATIN
2651 * CAPITAL LETTER A}), or more output than input if the constant ends up
2652 * recoded to utf8, but each time a construct is found that might increase
2653 * the needed size, SvGROW() is called. Its size parameter each time is
2654 * based on the best guess estimate at the time, namely the length used so
2655 * far, plus the length the current construct will occupy, plus room for
2656 * the trailing NUL, plus one byte for every input byte still unscanned */
2660 UV literal_endpoint = 0;
2661 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2664 PERL_ARGS_ASSERT_SCAN_CONST;
2666 assert(PL_lex_inwhat != OP_TRANSR);
2667 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2668 /* If we are doing a trans and we know we want UTF8 set expectation */
2669 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2670 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2674 while (s < send || dorange) {
2676 /* get transliterations out of the way (they're most literal) */
2677 if (PL_lex_inwhat == OP_TRANS) {
2678 /* expand a range A-Z to the full set of characters. AIE! */
2680 I32 i; /* current expanded character */
2681 I32 min; /* first character in range */
2682 I32 max; /* last character in range */
2693 char * const c = (char*)utf8_hop((U8*)d, -1);
2697 *c = (char)UTF_TO_NATIVE(0xff);
2698 /* mark the range as done, and continue */
2704 i = d - SvPVX_const(sv); /* remember current offset */
2707 SvLEN(sv) + (has_utf8 ?
2708 (512 - UTF_CONTINUATION_MARK +
2711 /* How many two-byte within 0..255: 128 in UTF-8,
2712 * 96 in UTF-8-mod. */
2714 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2716 d = SvPVX(sv) + i; /* refresh d after realloc */
2720 for (j = 0; j <= 1; j++) {
2721 char * const c = (char*)utf8_hop((U8*)d, -1);
2722 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2728 max = (U8)0xff; /* only to \xff */
2729 uvmax = uv; /* \x{100} to uvmax */
2731 d = c; /* eat endpoint chars */
2736 d -= 2; /* eat the first char and the - */
2737 min = (U8)*d; /* first char in range */
2738 max = (U8)d[1]; /* last char in range */
2745 "Invalid range \"%c-%c\" in transliteration operator",
2746 (char)min, (char)max);
2750 if (literal_endpoint == 2 &&
2751 ((isLOWER(min) && isLOWER(max)) ||
2752 (isUPPER(min) && isUPPER(max)))) {
2754 for (i = min; i <= max; i++)
2756 *d++ = NATIVE_TO_NEED(has_utf8,i);
2758 for (i = min; i <= max; i++)
2760 *d++ = NATIVE_TO_NEED(has_utf8,i);
2765 for (i = min; i <= max; i++)
2768 const U8 ch = (U8)NATIVE_TO_UTF(i);
2769 if (UNI_IS_INVARIANT(ch))
2772 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2773 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2782 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2784 *d++ = (char)UTF_TO_NATIVE(0xff);
2786 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2790 /* mark the range as done, and continue */
2794 literal_endpoint = 0;
2799 /* range begins (ignore - as first or last char) */
2800 else if (*s == '-' && s+1 < send && s != start) {
2802 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2809 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2819 literal_endpoint = 0;
2820 native_range = TRUE;
2825 /* if we get here, we're not doing a transliteration */
2827 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2828 except for the last char, which will be done separately. */
2829 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2831 while (s+1 < send && *s != ')')
2832 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2834 else if (s[2] == '{' /* This should match regcomp.c */
2835 || (s[2] == '?' && s[3] == '{'))
2838 char *regparse = s + (s[2] == '{' ? 3 : 4);
2841 while (count && (c = *regparse)) {
2842 if (c == '\\' && regparse[1])
2850 if (*regparse != ')')
2851 regparse--; /* Leave one char for continuation. */
2852 while (s < regparse)
2853 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2857 /* likewise skip #-initiated comments in //x patterns */
2858 else if (*s == '#' && PL_lex_inpat &&
2859 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
2860 while (s+1 < send && *s != '\n')
2861 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2864 /* check for embedded arrays
2865 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2867 else if (*s == '@' && s[1]) {
2868 if (isALNUM_lazy_if(s+1,UTF))
2870 if (strchr(":'{$", s[1]))
2872 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2873 break; /* in regexp, neither @+ nor @- are interpolated */
2876 /* check for embedded scalars. only stop if we're sure it's a
2879 else if (*s == '$') {
2880 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2882 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2884 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2885 "Possible unintended interpolation of $\\ in regex");
2887 break; /* in regexp, $ might be tail anchor */
2891 /* End of else if chain - OP_TRANS rejoin rest */
2894 if (*s == '\\' && s+1 < send) {
2895 char* e; /* Can be used for ending '}', etc. */
2899 /* warn on \1 - \9 in substitution replacements, but note that \11
2900 * is an octal; and \19 is \1 followed by '9' */
2901 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2902 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2904 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2909 /* string-change backslash escapes */
2910 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2914 /* In a pattern, process \N, but skip any other backslash escapes.
2915 * This is because we don't want to translate an escape sequence
2916 * into a meta symbol and have the regex compiler use the meta
2917 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2918 * in spite of this, we do have to process \N here while the proper
2919 * charnames handler is in scope. See bugs #56444 and #62056.
2920 * There is a complication because \N in a pattern may also stand
2921 * for 'match a non-nl', and not mean a charname, in which case its
2922 * processing should be deferred to the regex compiler. To be a
2923 * charname it must be followed immediately by a '{', and not look
2924 * like \N followed by a curly quantifier, i.e., not something like
2925 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2927 else if (PL_lex_inpat
2930 || regcurly(s + 1)))
2932 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2933 goto default_action;
2938 /* quoted - in transliterations */
2940 if (PL_lex_inwhat == OP_TRANS) {
2947 if ((isALPHA(*s) || isDIGIT(*s)))
2948 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2949 "Unrecognized escape \\%c passed through",
2951 /* default action is to copy the quoted character */
2952 goto default_action;
2955 /* eg. \132 indicates the octal constant 0132 */
2956 case '0': case '1': case '2': case '3':
2957 case '4': case '5': case '6': case '7':
2961 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2964 goto NUM_ESCAPE_INSERT;
2966 /* eg. \o{24} indicates the octal constant \024 */
2972 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
2978 goto NUM_ESCAPE_INSERT;
2981 /* eg. \x24 indicates the hex constant 0x24 */
2985 char* const e = strchr(s, '}');
2986 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2987 PERL_SCAN_DISALLOW_PREFIX;
2992 yyerror("Missing right brace on \\x{}");
2996 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
3002 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
3003 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
3009 /* Insert oct or hex escaped character. There will always be
3010 * enough room in sv since such escapes will be longer than any
3011 * UTF-8 sequence they can end up as, except if they force us
3012 * to recode the rest of the string into utf8 */
3014 /* Here uv is the ordinal of the next character being added in
3015 * unicode (converted from native). */
3016 if (!UNI_IS_INVARIANT(uv)) {
3017 if (!has_utf8 && uv > 255) {
3018 /* Might need to recode whatever we have accumulated so
3019 * far if it contains any chars variant in utf8 or
3022 SvCUR_set(sv, d - SvPVX_const(sv));
3025 /* See Note on sizing above. */
3026 sv_utf8_upgrade_flags_grow(sv,
3027 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3028 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3029 d = SvPVX(sv) + SvCUR(sv);
3034 d = (char*)uvuni_to_utf8((U8*)d, uv);
3035 if (PL_lex_inwhat == OP_TRANS &&
3036 PL_sublex_info.sub_op) {
3037 PL_sublex_info.sub_op->op_private |=
3038 (PL_lex_repl ? OPpTRANS_FROM_UTF
3042 if (uv > 255 && !dorange)
3043 native_range = FALSE;
3056 /* In a non-pattern \N must be a named character, like \N{LATIN
3057 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3058 * mean to match a non-newline. For non-patterns, named
3059 * characters are converted to their string equivalents. In
3060 * patterns, named characters are not converted to their
3061 * ultimate forms for the same reasons that other escapes
3062 * aren't. Instead, they are converted to the \N{U+...} form
3063 * to get the value from the charnames that is in effect right
3064 * now, while preserving the fact that it was a named character
3065 * so that the regex compiler knows this */
3067 /* This section of code doesn't generally use the
3068 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3069 * a close examination of this macro and determined it is a
3070 * no-op except on utfebcdic variant characters. Every
3071 * character generated by this that would normally need to be
3072 * enclosed by this macro is invariant, so the macro is not
3073 * needed, and would complicate use of copy(). XXX There are
3074 * other parts of this file where the macro is used
3075 * inconsistently, but are saved by it being a no-op */
3077 /* The structure of this section of code (besides checking for
3078 * errors and upgrading to utf8) is:
3079 * Further disambiguate between the two meanings of \N, and if
3080 * not a charname, go process it elsewhere
3081 * If of form \N{U+...}, pass it through if a pattern;
3082 * otherwise convert to utf8
3083 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3084 * pattern; otherwise convert to utf8 */
3086 /* Here, s points to the 'N'; the test below is guaranteed to
3087 * succeed if we are being called on a pattern as we already
3088 * know from a test above that the next character is a '{'.
3089 * On a non-pattern \N must mean 'named sequence, which
3090 * requires braces */
3093 yyerror("Missing braces on \\N{}");
3098 /* If there is no matching '}', it is an error. */
3099 if (! (e = strchr(s, '}'))) {
3100 if (! PL_lex_inpat) {
3101 yyerror("Missing right brace on \\N{}");
3103 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3108 /* Here it looks like a named character */
3112 /* XXX This block is temporary code. \N{} implies that the
3113 * pattern is to have Unicode semantics, and therefore
3114 * currently has to be encoded in utf8. By putting it in
3115 * utf8 now, we save a whole pass in the regular expression
3116 * compiler. Once that code is changed so Unicode
3117 * semantics doesn't necessarily have to be in utf8, this
3118 * block should be removed. However, the code that parses
3119 * the output of this would have to be changed to not
3120 * necessarily expect utf8 */
3122 SvCUR_set(sv, d - SvPVX_const(sv));
3125 /* See Note on sizing above. */
3126 sv_utf8_upgrade_flags_grow(sv,
3127 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3128 /* 5 = '\N{' + cur char + NUL */
3129 (STRLEN)(send - s) + 5);
3130 d = SvPVX(sv) + SvCUR(sv);
3135 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3136 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3137 | PERL_SCAN_DISALLOW_PREFIX;
3140 /* For \N{U+...}, the '...' is a unicode value even on
3141 * EBCDIC machines */
3142 s += 2; /* Skip to next char after the 'U+' */
3144 uv = grok_hex(s, &len, &flags, NULL);
3145 if (len == 0 || len != (STRLEN)(e - s)) {
3146 yyerror("Invalid hexadecimal number in \\N{U+...}");
3153 /* On non-EBCDIC platforms, pass through to the regex
3154 * compiler unchanged. The reason we evaluated the
3155 * number above is to make sure there wasn't a syntax
3156 * error. But on EBCDIC we convert to native so
3157 * downstream code can continue to assume it's native
3159 s -= 5; /* Include the '\N{U+' */
3161 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3164 (unsigned int) UNI_TO_NATIVE(uv));
3166 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3170 else { /* Not a pattern: convert the hex to string */
3172 /* If destination is not in utf8, unconditionally
3173 * recode it to be so. This is because \N{} implies
3174 * Unicode semantics, and scalars have to be in utf8
3175 * to guarantee those semantics */
3177 SvCUR_set(sv, d - SvPVX_const(sv));
3180 /* See Note on sizing above. */
3181 sv_utf8_upgrade_flags_grow(
3183 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3184 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3185 d = SvPVX(sv) + SvCUR(sv);
3189 /* Add the string to the output */
3190 if (UNI_IS_INVARIANT(uv)) {
3193 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3196 else { /* Here is \N{NAME} but not \N{U+...}. */
3198 SV *res; /* result from charnames */
3199 const char *str; /* the string in 'res' */
3200 STRLEN len; /* its length */
3202 /* Get the value for NAME */
3203 res = newSVpvn(s, e - s);
3204 res = new_constant( NULL, 0, "charnames",
3205 /* includes all of: \N{...} */
3206 res, NULL, s - 3, e - s + 4 );
3208 /* Most likely res will be in utf8 already since the
3209 * standard charnames uses pack U, but a custom translator
3210 * can leave it otherwise, so make sure. XXX This can be
3211 * revisited to not have charnames use utf8 for characters
3212 * that don't need it when regexes don't have to be in utf8
3213 * for Unicode semantics. If doing so, remember EBCDIC */
3214 sv_utf8_upgrade(res);
3215 str = SvPV_const(res, len);
3217 /* Don't accept malformed input */
3218 if (! is_utf8_string((U8 *) str, len)) {
3219 yyerror("Malformed UTF-8 returned by \\N");
3221 else if (PL_lex_inpat) {
3223 if (! len) { /* The name resolved to an empty string */
3224 Copy("\\N{}", d, 4, char);
3228 /* In order to not lose information for the regex
3229 * compiler, pass the result in the specially made
3230 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3231 * the code points in hex of each character
3232 * returned by charnames */
3234 const char *str_end = str + len;
3235 STRLEN char_length; /* cur char's byte length */
3236 STRLEN output_length; /* and the number of bytes
3237 after this is translated
3239 const STRLEN off = d - SvPVX_const(sv);
3241 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3242 * max('U+', '.'); and 1 for NUL */
3243 char hex_string[2 * UTF8_MAXBYTES + 5];
3245 /* Get the first character of the result. */
3246 U32 uv = utf8n_to_uvuni((U8 *) str,
3251 /* The call to is_utf8_string() above hopefully
3252 * guarantees that there won't be an error. But
3253 * it's easy here to make sure. The function just
3254 * above warns and returns 0 if invalid utf8, but
3255 * it can also return 0 if the input is validly a
3256 * NUL. Disambiguate */
3257 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3258 uv = UNICODE_REPLACEMENT;
3261 /* Convert first code point to hex, including the
3262 * boiler plate before it. For all these, we
3263 * convert to native format so that downstream code
3264 * can continue to assume the input is native */
3266 my_snprintf(hex_string, sizeof(hex_string),
3268 (unsigned int) UNI_TO_NATIVE(uv));
3270 /* Make sure there is enough space to hold it */
3271 d = off + SvGROW(sv, off
3273 + (STRLEN)(send - e)
3274 + 2); /* '}' + NUL */
3276 Copy(hex_string, d, output_length, char);
3279 /* For each subsequent character, append dot and
3280 * its ordinal in hex */
3281 while ((str += char_length) < str_end) {
3282 const STRLEN off = d - SvPVX_const(sv);
3283 U32 uv = utf8n_to_uvuni((U8 *) str,
3287 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3288 uv = UNICODE_REPLACEMENT;
3292 my_snprintf(hex_string, sizeof(hex_string),
3294 (unsigned int) UNI_TO_NATIVE(uv));
3296 d = off + SvGROW(sv, off
3298 + (STRLEN)(send - e)
3299 + 2); /* '}' + NUL */
3300 Copy(hex_string, d, output_length, char);
3304 *d++ = '}'; /* Done. Add the trailing brace */
3307 else { /* Here, not in a pattern. Convert the name to a
3310 /* If destination is not in utf8, unconditionally
3311 * recode it to be so. This is because \N{} implies
3312 * Unicode semantics, and scalars have to be in utf8
3313 * to guarantee those semantics */
3315 SvCUR_set(sv, d - SvPVX_const(sv));
3318 /* See Note on sizing above. */
3319 sv_utf8_upgrade_flags_grow(sv,
3320 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3321 len + (STRLEN)(send - s) + 1);
3322 d = SvPVX(sv) + SvCUR(sv);
3324 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3326 /* See Note on sizing above. (NOTE: SvCUR() is not
3327 * set correctly here). */
3328 const STRLEN off = d - SvPVX_const(sv);
3329 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3331 Copy(str, d, len, char);
3336 /* Deprecate non-approved name syntax */
3337 if (ckWARN_d(WARN_DEPRECATED)) {
3338 bool problematic = FALSE;
3341 /* For non-ut8 input, look to see that the first
3342 * character is an alpha, then loop through the rest
3343 * checking that each is a continuation */
3345 if (! isALPHAU(*i)) problematic = TRUE;
3346 else for (i = s + 1; i < e; i++) {
3347 if (isCHARNAME_CONT(*i)) continue;
3353 /* Similarly for utf8. For invariants can check
3354 * directly. We accept anything above the latin1
3355 * range because it is immaterial to Perl if it is
3356 * correct or not, and is expensive to check. But
3357 * it is fairly easy in the latin1 range to convert
3358 * the variants into a single character and check
3360 if (UTF8_IS_INVARIANT(*i)) {
3361 if (! isALPHAU(*i)) problematic = TRUE;
3362 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3363 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*i,
3369 if (! problematic) for (i = s + UTF8SKIP(s);
3373 if (UTF8_IS_INVARIANT(*i)) {
3374 if (isCHARNAME_CONT(*i)) continue;
3375 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3377 } else if (isCHARNAME_CONT(
3379 TWO_BYTE_UTF8_TO_UNI(*i, *(i+1)))))
3388 /* The e-i passed to the final %.*s makes sure that
3389 * should the trailing NUL be missing that this
3390 * print won't run off the end of the string */
3391 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3392 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s",
3393 (int)(i - s + 1), s, (int)(e - i), i + 1);
3396 } /* End \N{NAME} */
3399 native_range = FALSE; /* \N{} is defined to be Unicode */
3401 s = e + 1; /* Point to just after the '}' */
3404 /* \c is a control character */
3408 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3411 yyerror("Missing control char name in \\c");
3415 /* printf-style backslashes, formfeeds, newlines, etc */
3417 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3420 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3423 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3426 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3429 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3432 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3435 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3441 } /* end if (backslash) */
3448 /* If we started with encoded form, or already know we want it,
3449 then encode the next character */
3450 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3454 /* One might think that it is wasted effort in the case of the
3455 * source being utf8 (this_utf8 == TRUE) to take the next character
3456 * in the source, convert it to an unsigned value, and then convert
3457 * it back again. But the source has not been validated here. The
3458 * routine that does the conversion checks for errors like
3461 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3462 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3464 SvCUR_set(sv, d - SvPVX_const(sv));
3467 /* See Note on sizing above. */
3468 sv_utf8_upgrade_flags_grow(sv,
3469 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3470 need + (STRLEN)(send - s) + 1);
3471 d = SvPVX(sv) + SvCUR(sv);
3473 } else if (need > len) {
3474 /* encoded value larger than old, may need extra space (NOTE:
3475 * SvCUR() is not set correctly here). See Note on sizing
3477 const STRLEN off = d - SvPVX_const(sv);
3478 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3482 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3484 if (uv > 255 && !dorange)
3485 native_range = FALSE;
3489 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3491 } /* while loop to process each character */
3493 /* terminate the string and set up the sv */
3495 SvCUR_set(sv, d - SvPVX_const(sv));
3496 if (SvCUR(sv) >= SvLEN(sv))
3497 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3500 if (PL_encoding && !has_utf8) {
3501 sv_recode_to_utf8(sv, PL_encoding);
3507 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3508 PL_sublex_info.sub_op->op_private |=
3509 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3513 /* shrink the sv if we allocated more than we used */
3514 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3515 SvPV_shrink_to_cur(sv);
3518 /* return the substring (via pl_yylval) only if we parsed anything */
3519 if (s > PL_bufptr) {
3520 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3521 const char *const key = PL_lex_inpat ? "qr" : "q";
3522 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3526 if (PL_lex_inwhat == OP_TRANS) {
3529 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3537 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3540 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3547 * Returns TRUE if there's more to the expression (e.g., a subscript),
3550 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3552 * ->[ and ->{ return TRUE
3553 * { and [ outside a pattern are always subscripts, so return TRUE
3554 * if we're outside a pattern and it's not { or [, then return FALSE
3555 * if we're in a pattern and the first char is a {
3556 * {4,5} (any digits around the comma) returns FALSE
3557 * if we're in a pattern and the first char is a [
3559 * [SOMETHING] has a funky algorithm to decide whether it's a
3560 * character class or not. It has to deal with things like
3561 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3562 * anything else returns TRUE
3565 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3568 S_intuit_more(pTHX_ register char *s)
3572 PERL_ARGS_ASSERT_INTUIT_MORE;
3574 if (PL_lex_brackets)
3576 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3578 if (*s != '{' && *s != '[')
3583 /* In a pattern, so maybe we have {n,m}. */
3591 /* On the other hand, maybe we have a character class */
3594 if (*s == ']' || *s == '^')
3597 /* this is terrifying, and it works */
3598 int weight = 2; /* let's weigh the evidence */
3600 unsigned char un_char = 255, last_un_char;
3601 const char * const send = strchr(s,']');
3602 char tmpbuf[sizeof PL_tokenbuf * 4];
3604 if (!send) /* has to be an expression */
3607 Zero(seen,256,char);
3610 else if (isDIGIT(*s)) {
3612 if (isDIGIT(s[1]) && s[2] == ']')
3618 for (; s < send; s++) {
3619 last_un_char = un_char;
3620 un_char = (unsigned char)*s;
3625 weight -= seen[un_char] * 10;
3626 if (isALNUM_lazy_if(s+1,UTF)) {
3628 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3629 len = (int)strlen(tmpbuf);
3630 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
3635 else if (*s == '$' && s[1] &&
3636 strchr("[#!%*<>()-=",s[1])) {
3637 if (/*{*/ strchr("])} =",s[2]))
3646 if (strchr("wds]",s[1]))
3648 else if (seen[(U8)'\''] || seen[(U8)'"'])
3650 else if (strchr("rnftbxcav",s[1]))
3652 else if (isDIGIT(s[1])) {
3654 while (s[1] && isDIGIT(s[1]))
3664 if (strchr("aA01! ",last_un_char))
3666 if (strchr("zZ79~",s[1]))
3668 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3669 weight -= 5; /* cope with negative subscript */
3672 if (!isALNUM(last_un_char)
3673 && !(last_un_char == '$' || last_un_char == '@'
3674 || last_un_char == '&')
3675 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3680 if (keyword(tmpbuf, d - tmpbuf, 0))
3683 if (un_char == last_un_char + 1)
3685 weight -= seen[un_char];
3690 if (weight >= 0) /* probably a character class */
3700 * Does all the checking to disambiguate
3702 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3703 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3705 * First argument is the stuff after the first token, e.g. "bar".
3707 * Not a method if bar is a filehandle.
3708 * Not a method if foo is a subroutine prototyped to take a filehandle.
3709 * Not a method if it's really "Foo $bar"
3710 * Method if it's "foo $bar"
3711 * Not a method if it's really "print foo $bar"
3712 * Method if it's really "foo package::" (interpreted as package->foo)
3713 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3714 * Not a method if bar is a filehandle or package, but is quoted with
3719 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3722 char *s = start + (*start == '$');
3723 char tmpbuf[sizeof PL_tokenbuf];
3730 PERL_ARGS_ASSERT_INTUIT_METHOD;
3733 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3737 const char *proto = SvPVX_const(cv);
3748 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3749 /* start is the beginning of the possible filehandle/object,
3750 * and s is the end of it
3751 * tmpbuf is a copy of it
3754 if (*start == '$') {
3755 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3756 isUPPER(*PL_tokenbuf))
3759 len = start - SvPVX(PL_linestr);
3763 start = SvPVX(PL_linestr) + len;
3767 return *s == '(' ? FUNCMETH : METHOD;
3769 if (!keyword(tmpbuf, len, 0)) {
3770 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3774 soff = s - SvPVX(PL_linestr);
3778 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
3779 if (indirgv && GvCVu(indirgv))
3781 /* filehandle or package name makes it a method */
3782 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
3784 soff = s - SvPVX(PL_linestr);
3787 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3788 return 0; /* no assumptions -- "=>" quotes bareword */
3790 start_force(PL_curforce);
3791 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3792 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3793 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3795 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3800 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3802 return *s == '(' ? FUNCMETH : METHOD;
3808 /* Encoded script support. filter_add() effectively inserts a
3809 * 'pre-processing' function into the current source input stream.
3810 * Note that the filter function only applies to the current source file
3811 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3813 * The datasv parameter (which may be NULL) can be used to pass
3814 * private data to this instance of the filter. The filter function
3815 * can recover the SV using the FILTER_DATA macro and use it to
3816 * store private buffers and state information.
3818 * The supplied datasv parameter is upgraded to a PVIO type
3819 * and the IoDIRP/IoANY field is used to store the function pointer,
3820 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3821 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3822 * private use must be set using malloc'd pointers.
3826 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3835 if (!PL_rsfp_filters)
3836 PL_rsfp_filters = newAV();
3839 SvUPGRADE(datasv, SVt_PVIO);
3840 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3841 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3842 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3843 FPTR2DPTR(void *, IoANY(datasv)),
3844 SvPV_nolen(datasv)));
3845 av_unshift(PL_rsfp_filters, 1);
3846 av_store(PL_rsfp_filters, 0, datasv) ;
3851 /* Delete most recently added instance of this filter function. */
3853 Perl_filter_del(pTHX_ filter_t funcp)
3858 PERL_ARGS_ASSERT_FILTER_DEL;
3861 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3862 FPTR2DPTR(void*, funcp)));
3864 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3866 /* if filter is on top of stack (usual case) just pop it off */
3867 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3868 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3869 sv_free(av_pop(PL_rsfp_filters));
3873 /* we need to search for the correct entry and clear it */
3874 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3878 /* Invoke the idxth filter function for the current rsfp. */
3879 /* maxlen 0 = read one text line */
3881 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3886 /* This API is bad. It should have been using unsigned int for maxlen.
3887 Not sure if we want to change the API, but if not we should sanity
3888 check the value here. */
3889 const unsigned int correct_length
3898 PERL_ARGS_ASSERT_FILTER_READ;
3900 if (!PL_parser || !PL_rsfp_filters)
3902 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3903 /* Provide a default input filter to make life easy. */
3904 /* Note that we append to the line. This is handy. */
3905 DEBUG_P(PerlIO_printf(Perl_debug_log,
3906 "filter_read %d: from rsfp\n", idx));
3907 if (correct_length) {
3910 const int old_len = SvCUR(buf_sv);
3912 /* ensure buf_sv is large enough */
3913 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3914 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3915 correct_length)) <= 0) {
3916 if (PerlIO_error(PL_rsfp))
3917 return -1; /* error */
3919 return 0 ; /* end of file */
3921 SvCUR_set(buf_sv, old_len + len) ;
3922 SvPVX(buf_sv)[old_len + len] = '\0';
3925 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3926 if (PerlIO_error(PL_rsfp))
3927 return -1; /* error */
3929 return 0 ; /* end of file */
3932 return SvCUR(buf_sv);
3934 /* Skip this filter slot if filter has been deleted */
3935 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3936 DEBUG_P(PerlIO_printf(Perl_debug_log,
3937 "filter_read %d: skipped (filter deleted)\n",
3939 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
3941 /* Get function pointer hidden within datasv */
3942 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
3943 DEBUG_P(PerlIO_printf(Perl_debug_log,
3944 "filter_read %d: via function %p (%s)\n",
3945 idx, (void*)datasv, SvPV_nolen_const(datasv)));
3946 /* Call function. The function is expected to */
3947 /* call "FILTER_READ(idx+1, buf_sv)" first. */
3948 /* Return: <0:error, =0:eof, >0:not eof */
3949 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
3953 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
3957 PERL_ARGS_ASSERT_FILTER_GETS;
3959 #ifdef PERL_CR_FILTER
3960 if (!PL_rsfp_filters) {
3961 filter_add(S_cr_textfilter,NULL);
3964 if (PL_rsfp_filters) {
3966 SvCUR_set(sv, 0); /* start with empty line */
3967 if (FILTER_READ(0, sv, 0) > 0)
3968 return ( SvPVX(sv) ) ;
3973 return (sv_gets(sv, PL_rsfp, append));
3977 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3982 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3984 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3988 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3989 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3991 return GvHV(gv); /* Foo:: */
3994 /* use constant CLASS => 'MyClass' */
3995 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3996 if (gv && GvCV(gv)) {
3997 SV * const sv = cv_const_sv(GvCV(gv));
3999 pkgname = SvPV_const(sv, len);
4002 return gv_stashpvn(pkgname, len, 0);
4006 * S_readpipe_override
4007 * Check whether readpipe() is overridden, and generates the appropriate
4008 * optree, provided sublex_start() is called afterwards.
4011 S_readpipe_override(pTHX)
4014 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4015 pl_yylval.ival = OP_BACKTICK;
4017 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4019 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4020 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4021 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4023 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4024 op_append_elem(OP_LIST,
4025 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4026 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4033 * The intent of this yylex wrapper is to minimize the changes to the
4034 * tokener when we aren't interested in collecting madprops. It remains
4035 * to be seen how successful this strategy will be...
4042 char *s = PL_bufptr;
4044 /* make sure PL_thiswhite is initialized */
4048 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
4049 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4050 return S_pending_ident(aTHX);
4052 /* previous token ate up our whitespace? */
4053 if (!PL_lasttoke && PL_nextwhite) {
4054 PL_thiswhite = PL_nextwhite;
4058 /* isolate the token, and figure out where it is without whitespace */
4059 PL_realtokenstart = -1;
4063 assert(PL_curforce < 0);
4065 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4066 if (!PL_thistoken) {
4067 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4068 PL_thistoken = newSVpvs("");
4070 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4071 PL_thistoken = newSVpvn(tstart, s - tstart);
4074 if (PL_thismad) /* install head */
4075 CURMAD('X', PL_thistoken);
4078 /* last whitespace of a sublex? */
4079 if (optype == ')' && PL_endwhite) {
4080 CURMAD('X', PL_endwhite);
4085 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4086 if (!PL_thiswhite && !PL_endwhite && !optype) {
4087 sv_free(PL_thistoken);
4092 /* put off final whitespace till peg */
4093 if (optype == ';' && !PL_rsfp) {
4094 PL_nextwhite = PL_thiswhite;
4097 else if (PL_thisopen) {
4098 CURMAD('q', PL_thisopen);
4100 sv_free(PL_thistoken);
4104 /* Store actual token text as madprop X */
4105 CURMAD('X', PL_thistoken);
4109 /* add preceding whitespace as madprop _ */
4110 CURMAD('_', PL_thiswhite);
4114 /* add quoted material as madprop = */
4115 CURMAD('=', PL_thisstuff);
4119 /* add terminating quote as madprop Q */
4120 CURMAD('Q', PL_thisclose);
4124 /* special processing based on optype */
4128 /* opval doesn't need a TOKEN since it can already store mp */
4138 if (pl_yylval.opval)
4139 append_madprops(PL_thismad, pl_yylval.opval, 0);
4147 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4156 /* remember any fake bracket that lexer is about to discard */
4157 if (PL_lex_brackets == 1 &&
4158 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4161 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4164 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4165 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4168 break; /* don't bother looking for trailing comment */
4177 /* attach a trailing comment to its statement instead of next token */
4181 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4183 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4185 if (*s == '\n' || *s == '#') {
4186 while (s < PL_bufend && *s != '\n')
4190 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4191 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4208 /* Create new token struct. Note: opvals return early above. */
4209 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4216 S_tokenize_use(pTHX_ int is_use, char *s) {
4219 PERL_ARGS_ASSERT_TOKENIZE_USE;
4221 if (PL_expect != XSTATE)
4222 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4223 is_use ? "use" : "no"));
4225 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4226 s = force_version(s, TRUE);
4227 if (*s == ';' || *s == '}'
4228 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4229 start_force(PL_curforce);
4230 NEXTVAL_NEXTTOKE.opval = NULL;
4233 else if (*s == 'v') {
4234 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4235 s = force_version(s, FALSE);
4239 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4240 s = force_version(s, FALSE);
4242 pl_yylval.ival = is_use;
4246 static const char* const exp_name[] =
4247 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4248 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4252 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4254 S_word_takes_any_delimeter(char *p, STRLEN len)
4256 return (len == 1 && strchr("msyq", p[0])) ||
4258 (p[0] == 't' && p[1] == 'r') ||
4259 (p[0] == 'q' && strchr("qwxr", p[1]))));
4265 Works out what to call the token just pulled out of the input
4266 stream. The yacc parser takes care of taking the ops we return and
4267 stitching them into a tree.
4273 if read an identifier
4274 if we're in a my declaration
4275 croak if they tried to say my($foo::bar)
4276 build the ops for a my() declaration
4277 if it's an access to a my() variable
4278 are we in a sort block?
4279 croak if my($a); $a <=> $b
4280 build ops for access to a my() variable
4281 if in a dq string, and they've said @foo and we can't find @foo
4283 build ops for a bareword
4284 if we already built the token before, use it.
4289 #pragma segment Perl_yylex
4295 register char *s = PL_bufptr;
4301 /* orig_keyword, gvp, and gv are initialized here because
4302 * jump to the label just_a_word_zero can bypass their
4303 * initialization later. */
4304 I32 orig_keyword = 0;
4309 SV* tmp = newSVpvs("");
4310 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4311 (IV)CopLINE(PL_curcop),
4312 lex_state_names[PL_lex_state],
4313 exp_name[PL_expect],
4314 pv_display(tmp, s, strlen(s), 0, 60));
4317 /* check if there's an identifier for us to look at */
4318 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4319 return REPORT(S_pending_ident(aTHX));
4321 /* no identifier pending identification */
4323 switch (PL_lex_state) {
4325 case LEX_NORMAL: /* Some compilers will produce faster */
4326 case LEX_INTERPNORMAL: /* code if we comment these out. */
4330 /* when we've already built the next token, just pull it out of the queue */
4334 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4336 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4337 PL_nexttoke[PL_lasttoke].next_mad = 0;
4338 if (PL_thismad && PL_thismad->mad_key == '_') {
4339 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4340 PL_thismad->mad_val = 0;
4341 mad_free(PL_thismad);
4346 PL_lex_state = PL_lex_defer;
4347 PL_expect = PL_lex_expect;
4348 PL_lex_defer = LEX_NORMAL;
4349 if (!PL_nexttoke[PL_lasttoke].next_type)
4354 pl_yylval = PL_nextval[PL_nexttoke];
4356 PL_lex_state = PL_lex_defer;
4357 PL_expect = PL_lex_expect;
4358 PL_lex_defer = LEX_NORMAL;
4364 next_type = PL_nexttoke[PL_lasttoke].next_type;
4366 next_type = PL_nexttype[PL_nexttoke];
4368 if (next_type & (7<<24)) {
4369 if (next_type & (1<<24)) {
4370 if (PL_lex_brackets > 100)
4371 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4372 PL_lex_brackstack[PL_lex_brackets++] =
4373 (char) ((next_type >> 16) & 0xff);
4375 if (next_type & (2<<24))
4376 PL_lex_allbrackets++;
4377 if (next_type & (4<<24))
4378 PL_lex_allbrackets--;
4379 next_type &= 0xffff;
4382 /* FIXME - can these be merged? */
4385 return REPORT(next_type);
4389 /* interpolated case modifiers like \L \U, including \Q and \E.
4390 when we get here, PL_bufptr is at the \
4392 case LEX_INTERPCASEMOD:
4394 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4395 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
4397 /* handle \E or end of string */
4398 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4400 if (PL_lex_casemods) {
4401 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4402 PL_lex_casestack[PL_lex_casemods] = '\0';
4404 if (PL_bufptr != PL_bufend
4405 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
4407 PL_lex_state = LEX_INTERPCONCAT;
4410 PL_thistoken = newSVpvs("\\E");
4413 PL_lex_allbrackets--;
4417 while (PL_bufptr != PL_bufend &&
4418 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4420 PL_thiswhite = newSVpvs("");
4421 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4425 if (PL_bufptr != PL_bufend)
4428 PL_lex_state = LEX_INTERPCONCAT;
4432 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4433 "### Saw case modifier\n"); });
4435 if (s[1] == '\\' && s[2] == 'E') {
4438 PL_thiswhite = newSVpvs("");
4439 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4442 PL_lex_state = LEX_INTERPCONCAT;
4447 if (!PL_madskills) /* when just compiling don't need correct */
4448 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4449 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4450 if ((*s == 'L' || *s == 'U') &&
4451 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4452 PL_lex_casestack[--PL_lex_casemods] = '\0';
4453 PL_lex_allbrackets--;
4456 if (PL_lex_casemods > 10)
4457 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4458 PL_lex_casestack[PL_lex_casemods++] = *s;
4459 PL_lex_casestack[PL_lex_casemods] = '\0';
4460 PL_lex_state = LEX_INTERPCONCAT;
4461 start_force(PL_curforce);
4462 NEXTVAL_NEXTTOKE.ival = 0;
4463 force_next((2<<24)|'(');
4464 start_force(PL_curforce);
4466 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4468 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4470 NEXTVAL_NEXTTOKE.ival = OP_LC;
4472 NEXTVAL_NEXTTOKE.ival = OP_UC;
4474 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4476 Perl_croak(aTHX_ "panic: yylex");
4478 SV* const tmpsv = newSVpvs("\\ ");
4479 /* replace the space with the character we want to escape
4481 SvPVX(tmpsv)[1] = *s;
4487 if (PL_lex_starts) {
4493 sv_free(PL_thistoken);
4494 PL_thistoken = newSVpvs("");
4497 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4498 if (PL_lex_casemods == 1 && PL_lex_inpat)
4507 case LEX_INTERPPUSH:
4508 return REPORT(sublex_push());
4510 case LEX_INTERPSTART:
4511 if (PL_bufptr == PL_bufend)
4512 return REPORT(sublex_done());
4513 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4514 "### Interpolated variable\n"); });
4516 PL_lex_dojoin = (*PL_bufptr == '@');
4517 PL_lex_state = LEX_INTERPNORMAL;
4518 if (PL_lex_dojoin) {
4519 start_force(PL_curforce);
4520 NEXTVAL_NEXTTOKE.ival = 0;
4522 start_force(PL_curforce);
4523 force_ident("\"", '$');
4524 start_force(PL_curforce);
4525 NEXTVAL_NEXTTOKE.ival = 0;
4527 start_force(PL_curforce);
4528 NEXTVAL_NEXTTOKE.ival = 0;
4529 force_next((2<<24)|'(');
4530 start_force(PL_curforce);
4531 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4534 if (PL_lex_starts++) {
4539 sv_free(PL_thistoken);
4540 PL_thistoken = newSVpvs("");
4543 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4544 if (!PL_lex_casemods && PL_lex_inpat)
4551 case LEX_INTERPENDMAYBE:
4552 if (intuit_more(PL_bufptr)) {
4553 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4559 if (PL_lex_dojoin) {
4560 PL_lex_dojoin = FALSE;
4561 PL_lex_state = LEX_INTERPCONCAT;
4565 sv_free(PL_thistoken);
4566 PL_thistoken = newSVpvs("");
4569 PL_lex_allbrackets--;
4572 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4573 && SvEVALED(PL_lex_repl))
4575 if (PL_bufptr != PL_bufend)
4576 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4580 case LEX_INTERPCONCAT:
4582 if (PL_lex_brackets)
4583 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4585 if (PL_bufptr == PL_bufend)
4586 return REPORT(sublex_done());
4588 if (SvIVX(PL_linestr) == '\'') {
4589 SV *sv = newSVsv(PL_linestr);
4592 else if ( PL_hints & HINT_NEW_RE )
4593 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4594 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4598 s = scan_const(PL_bufptr);
4600 PL_lex_state = LEX_INTERPCASEMOD;
4602 PL_lex_state = LEX_INTERPSTART;
4605 if (s != PL_bufptr) {
4606 start_force(PL_curforce);
4608 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4610 NEXTVAL_NEXTTOKE = pl_yylval;
4613 if (PL_lex_starts++) {
4617 sv_free(PL_thistoken);
4618 PL_thistoken = newSVpvs("");
4621 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4622 if (!PL_lex_casemods && PL_lex_inpat)
4635 PL_lex_state = LEX_NORMAL;
4636 s = scan_formline(PL_bufptr);
4637 if (!PL_lex_formbrack)
4643 PL_oldoldbufptr = PL_oldbufptr;
4649 sv_free(PL_thistoken);
4652 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4656 if (isIDFIRST_lazy_if(s,UTF))
4659 unsigned char c = *s;
4660 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4661 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4662 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4667 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4671 goto fake_eof; /* emulate EOF on ^D or ^Z */
4680 if (PL_lex_brackets &&
4681 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
4682 yyerror((const char *)
4684 ? "Format not terminated"
4685 : "Missing right curly or square bracket"));
4687 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4688 "### Tokener got EOF\n");
4692 if (s++ < PL_bufend)
4693 goto retry; /* ignore stray nulls */
4696 if (!PL_in_eval && !PL_preambled) {
4697 PL_preambled = TRUE;
4703 /* Generate a string of Perl code to load the debugger.
4704 * If PERL5DB is set, it will return the contents of that,
4705 * otherwise a compile-time require of perl5db.pl. */
4707 const char * const pdb = PerlEnv_getenv("PERL5DB");
4710 sv_setpv(PL_linestr, pdb);
4711 sv_catpvs(PL_linestr,";");
4713 SETERRNO(0,SS_NORMAL);
4714 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4717 sv_setpvs(PL_linestr,"");
4718 if (PL_preambleav) {
4719 SV **svp = AvARRAY(PL_preambleav);
4720 SV **const end = svp + AvFILLp(PL_preambleav);
4722 sv_catsv(PL_linestr, *svp);
4724 sv_catpvs(PL_linestr, ";");
4726 sv_free(MUTABLE_SV(PL_preambleav));
4727 PL_preambleav = NULL;
4730 sv_catpvs(PL_linestr,
4731 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4732 if (PL_minus_n || PL_minus_p) {
4733 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4735 sv_catpvs(PL_linestr,"chomp;");
4738 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4739 || *PL_splitstr == '"')
4740 && strchr(PL_splitstr + 1, *PL_splitstr))
4741 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4743 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4744 bytes can be used as quoting characters. :-) */
4745 const char *splits = PL_splitstr;
4746 sv_catpvs(PL_linestr, "our @F=split(q\0");
4749 if (*splits == '\\')
4750 sv_catpvn(PL_linestr, splits, 1);
4751 sv_catpvn(PL_linestr, splits, 1);
4752 } while (*splits++);
4753 /* This loop will embed the trailing NUL of
4754 PL_linestr as the last thing it does before
4756 sv_catpvs(PL_linestr, ");");
4760 sv_catpvs(PL_linestr,"our @F=split(' ');");
4763 sv_catpvs(PL_linestr, "\n");
4764 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4765 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4766 PL_last_lop = PL_last_uni = NULL;
4767 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4768 update_debugger_info(PL_linestr, NULL, 0);
4773 bof = PL_rsfp ? TRUE : FALSE;
4776 fake_eof = LEX_FAKE_EOF;
4778 PL_bufptr = PL_bufend;
4779 CopLINE_inc(PL_curcop);
4780 if (!lex_next_chunk(fake_eof)) {
4781 CopLINE_dec(PL_curcop);
4783 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4785 CopLINE_dec(PL_curcop);
4788 PL_realtokenstart = -1;
4791 /* If it looks like the start of a BOM or raw UTF-16,
4792 * check if it in fact is. */
4793 if (bof && PL_rsfp &&
4798 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
4799 bof = (offset == (Off_t)SvCUR(PL_linestr));
4800 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
4801 /* offset may include swallowed CR */
4803 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
4806 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4807 s = swallow_bom((U8*)s);
4810 if (PL_parser->in_pod) {
4811 /* Incest with pod. */
4814 sv_catsv(PL_thiswhite, PL_linestr);
4816 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4817 sv_setpvs(PL_linestr, "");
4818 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4819 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4820 PL_last_lop = PL_last_uni = NULL;
4821 PL_parser->in_pod = 0;
4826 } while (PL_parser->in_pod);
4827 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4828 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4829 PL_last_lop = PL_last_uni = NULL;
4830 if (CopLINE(PL_curcop) == 1) {
4831 while (s < PL_bufend && isSPACE(*s))
4833 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4837 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4841 if (*s == '#' && *(s+1) == '!')
4843 #ifdef ALTERNATE_SHEBANG
4845 static char const as[] = ALTERNATE_SHEBANG;
4846 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4847 d = s + (sizeof(as) - 1);
4849 #endif /* ALTERNATE_SHEBANG */
4858 while (*d && !isSPACE(*d))
4862 #ifdef ARG_ZERO_IS_SCRIPT
4863 if (ipathend > ipath) {
4865 * HP-UX (at least) sets argv[0] to the script name,
4866 * which makes $^X incorrect. And Digital UNIX and Linux,
4867 * at least, set argv[0] to the basename of the Perl
4868 * interpreter. So, having found "#!", we'll set it right.
4870 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4872 assert(SvPOK(x) || SvGMAGICAL(x));
4873 if (sv_eq(x, CopFILESV(PL_curcop))) {
4874 sv_setpvn(x, ipath, ipathend - ipath);
4880 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4881 const char * const lstart = SvPV_const(x,llen);
4883 bstart += blen - llen;
4884 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4885 sv_setpvn(x, ipath, ipathend - ipath);
4890 TAINT_NOT; /* $^X is always tainted, but that's OK */
4892 #endif /* ARG_ZERO_IS_SCRIPT */
4897 d = instr(s,"perl -");
4899 d = instr(s,"perl");
4901 /* avoid getting into infinite loops when shebang
4902 * line contains "Perl" rather than "perl" */
4904 for (d = ipathend-4; d >= ipath; --d) {
4905 if ((*d == 'p' || *d == 'P')
4906 && !ibcmp(d, "perl", 4))
4916 #ifdef ALTERNATE_SHEBANG
4918 * If the ALTERNATE_SHEBANG on this system starts with a
4919 * character that can be part of a Perl expression, then if
4920 * we see it but not "perl", we're probably looking at the
4921 * start of Perl code, not a request to hand off to some
4922 * other interpreter. Similarly, if "perl" is there, but
4923 * not in the first 'word' of the line, we assume the line
4924 * contains the start of the Perl program.
4926 if (d && *s != '#') {
4927 const char *c = ipath;
4928 while (*c && !strchr("; \t\r\n\f\v#", *c))
4931 d = NULL; /* "perl" not in first word; ignore */
4933 *s = '#'; /* Don't try to parse shebang line */
4935 #endif /* ALTERNATE_SHEBANG */
4940 !instr(s,"indir") &&
4941 instr(PL_origargv[0],"perl"))
4948 while (s < PL_bufend && isSPACE(*s))
4950 if (s < PL_bufend) {
4951 Newx(newargv,PL_origargc+3,char*);
4953 while (s < PL_bufend && !isSPACE(*s))
4956 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4959 newargv = PL_origargv;
4962 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4964 Perl_croak(aTHX_ "Can't exec %s", ipath);
4967 while (*d && !isSPACE(*d))
4969 while (SPACE_OR_TAB(*d))
4973 const bool switches_done = PL_doswitches;
4974 const U32 oldpdb = PL_perldb;
4975 const bool oldn = PL_minus_n;
4976 const bool oldp = PL_minus_p;
4980 bool baduni = FALSE;
4982 const char *d2 = d1 + 1;
4983 if (parse_unicode_opts((const char **)&d2)
4987 if (baduni || *d1 == 'M' || *d1 == 'm') {
4988 const char * const m = d1;
4989 while (*d1 && !isSPACE(*d1))
4991 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4994 d1 = moreswitches(d1);
4996 if (PL_doswitches && !switches_done) {
4997 int argc = PL_origargc;
4998 char **argv = PL_origargv;
5001 } while (argc && argv[0][0] == '-' && argv[0][1]);
5002 init_argv_symbols(argc,argv);
5004 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
5005 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
5006 /* if we have already added "LINE: while (<>) {",
5007 we must not do it again */
5009 sv_setpvs(PL_linestr, "");
5010 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5011 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5012 PL_last_lop = PL_last_uni = NULL;
5013 PL_preambled = FALSE;
5014 if (PERLDB_LINE || PERLDB_SAVESRC)
5015 (void)gv_fetchfile(PL_origfilename);
5022 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5024 PL_lex_state = LEX_FORMLINE;
5029 #ifdef PERL_STRICT_CR
5030 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5032 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5034 case ' ': case '\t': case '\f': case 013:
5036 PL_realtokenstart = -1;
5038 PL_thiswhite = newSVpvs("");
5039 sv_catpvn(PL_thiswhite, s, 1);
5046 PL_realtokenstart = -1;
5050 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
5051 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
5052 /* handle eval qq[#line 1 "foo"\n ...] */
5053 CopLINE_dec(PL_curcop);
5056 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
5058 if (!PL_in_eval || PL_rsfp)
5063 while (d < PL_bufend && *d != '\n')
5067 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5068 Perl_croak(aTHX_ "panic: input overflow");
5071 PL_thiswhite = newSVpvn(s, d - s);
5076 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5078 PL_lex_state = LEX_FORMLINE;
5084 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
5085 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
5088 TOKEN(PEG); /* make sure any #! line is accessible */
5093 /* if (PL_madskills && PL_lex_formbrack) { */
5095 while (d < PL_bufend && *d != '\n')
5099 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5100 Perl_croak(aTHX_ "panic: input overflow");
5101 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
5103 PL_thiswhite = newSVpvs("");
5104 if (CopLINE(PL_curcop) == 1) {
5105 sv_setpvs(PL_thiswhite, "");
5108 sv_catpvn(PL_thiswhite, s, d - s);
5122 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
5130 while (s < PL_bufend && SPACE_OR_TAB(*s))
5133 if (strnEQ(s,"=>",2)) {
5134 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
5135 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5136 OPERATOR('-'); /* unary minus */
5138 PL_last_uni = PL_oldbufptr;
5140 case 'r': ftst = OP_FTEREAD; break;
5141 case 'w': ftst = OP_FTEWRITE; break;
5142 case 'x': ftst = OP_FTEEXEC; break;
5143 case 'o': ftst = OP_FTEOWNED; break;
5144 case 'R': ftst = OP_FTRREAD; break;
5145 case 'W': ftst = OP_FTRWRITE; break;
5146 case 'X': ftst = OP_FTREXEC; break;
5147 case 'O': ftst = OP_FTROWNED; break;
5148 case 'e': ftst = OP_FTIS; break;
5149 case 'z': ftst = OP_FTZERO; break;
5150 case 's': ftst = OP_FTSIZE; break;
5151 case 'f': ftst = OP_FTFILE; break;
5152 case 'd': ftst = OP_FTDIR; break;
5153 case 'l': ftst = OP_FTLINK; break;
5154 case 'p': ftst = OP_FTPIPE; break;
5155 case 'S': ftst = OP_FTSOCK; break;
5156 case 'u': ftst = OP_FTSUID; break;
5157 case 'g': ftst = OP_FTSGID; break;
5158 case 'k': ftst = OP_FTSVTX; break;
5159 case 'b': ftst = OP_FTBLK; break;
5160 case 'c': ftst = OP_FTCHR; break;
5161 case 't': ftst = OP_FTTTY; break;
5162 case 'T': ftst = OP_FTTEXT; break;
5163 case 'B': ftst = OP_FTBINARY; break;
5164 case 'M': case 'A': case 'C':
5165 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5167 case 'M': ftst = OP_FTMTIME; break;
5168 case 'A': ftst = OP_FTATIME; break;
5169 case 'C': ftst = OP_FTCTIME; break;
5177 PL_last_lop_op = (OPCODE)ftst;
5178 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5179 "### Saw file test %c\n", (int)tmp);
5184 /* Assume it was a minus followed by a one-letter named
5185 * subroutine call (or a -bareword), then. */
5186 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5187 "### '-%c' looked like a file test but was not\n",
5194 const char tmp = *s++;
5197 if (PL_expect == XOPERATOR)
5202 else if (*s == '>') {
5205 if (isIDFIRST_lazy_if(s,UTF)) {
5206 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5214 if (PL_expect == XOPERATOR) {
5215 if (*s == '=' && !PL_lex_allbrackets &&
5216 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5223 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5225 OPERATOR('-'); /* unary minus */
5231 const char tmp = *s++;
5234 if (PL_expect == XOPERATOR)
5239 if (PL_expect == XOPERATOR) {
5240 if (*s == '=' && !PL_lex_allbrackets &&
5241 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5248 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5255 if (PL_expect != XOPERATOR) {
5256 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5257 PL_expect = XOPERATOR;
5258 force_ident(PL_tokenbuf, '*');
5266 if (*s == '=' && !PL_lex_allbrackets &&
5267 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5273 if (*s == '=' && !PL_lex_allbrackets &&
5274 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5281 if (PL_expect == XOPERATOR) {
5282 if (s[1] == '=' && !PL_lex_allbrackets &&
5283 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5288 PL_tokenbuf[0] = '%';
5289 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5290 sizeof PL_tokenbuf - 1, FALSE);
5291 if (!PL_tokenbuf[1]) {
5294 PL_pending_ident = '%';
5298 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5299 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5304 if (PL_lex_brackets > 100)
5305 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5306 PL_lex_brackstack[PL_lex_brackets++] = 0;
5307 PL_lex_allbrackets++;
5309 const char tmp = *s++;
5314 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5316 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5324 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5331 goto just_a_word_zero_gv;
5334 switch (PL_expect) {
5340 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5342 PL_bufptr = s; /* update in case we back off */
5345 "Use of := for an empty attribute list is not allowed");
5352 PL_expect = XTERMBLOCK;
5355 stuffstart = s - SvPVX(PL_linestr) - 1;
5359 while (isIDFIRST_lazy_if(s,UTF)) {
5362 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5363 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5364 if (tmp < 0) tmp = -tmp;
5379 sv = newSVpvn(s, len);
5381 d = scan_str(d,TRUE,TRUE);
5383 /* MUST advance bufptr here to avoid bogus
5384 "at end of line" context messages from yyerror().
5386 PL_bufptr = s + len;
5387 yyerror("Unterminated attribute parameter in attribute list");
5391 return REPORT(0); /* EOF indicator */
5395 sv_catsv(sv, PL_lex_stuff);
5396 attrs = op_append_elem(OP_LIST, attrs,
5397 newSVOP(OP_CONST, 0, sv));
5398 SvREFCNT_dec(PL_lex_stuff);
5399 PL_lex_stuff = NULL;
5402 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5404 if (PL_in_my == KEY_our) {
5405 deprecate(":unique");
5408 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5411 /* NOTE: any CV attrs applied here need to be part of
5412 the CVf_BUILTIN_ATTRS define in cv.h! */
5413 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5415 CvLVALUE_on(PL_compcv);
5417 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5419 deprecate(":locked");
5421 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5423 CvMETHOD_on(PL_compcv);
5425 /* After we've set the flags, it could be argued that
5426 we don't need to do the attributes.pm-based setting
5427 process, and shouldn't bother appending recognized
5428 flags. To experiment with that, uncomment the
5429 following "else". (Note that's already been
5430 uncommented. That keeps the above-applied built-in
5431 attributes from being intercepted (and possibly
5432 rejected) by a package's attribute routines, but is
5433 justified by the performance win for the common case
5434 of applying only built-in attributes.) */
5436 attrs = op_append_elem(OP_LIST, attrs,
5437 newSVOP(OP_CONST, 0,
5441 if (*s == ':' && s[1] != ':')
5444 break; /* require real whitespace or :'s */
5445 /* XXX losing whitespace on sequential attributes here */
5449 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5450 if (*s != ';' && *s != '}' && *s != tmp
5451 && (tmp != '=' || *s != ')')) {
5452 const char q = ((*s == '\'') ? '"' : '\'');
5453 /* If here for an expression, and parsed no attrs, back
5455 if (tmp == '=' && !attrs) {
5459 /* MUST advance bufptr here to avoid bogus "at end of line"
5460 context messages from yyerror().
5463 yyerror( (const char *)
5465 ? Perl_form(aTHX_ "Invalid separator character "
5466 "%c%c%c in attribute list", q, *s, q)
5467 : "Unterminated attribute list" ) );
5475 start_force(PL_curforce);
5476 NEXTVAL_NEXTTOKE.opval = attrs;
5477 CURMAD('_', PL_nextwhite);
5482 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5483 (s - SvPVX(PL_linestr)) - stuffstart);
5488 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5492 PL_lex_allbrackets--;
5496 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5497 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5501 PL_lex_allbrackets++;
5504 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5510 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5513 PL_lex_allbrackets--;
5519 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5522 if (PL_lex_brackets <= 0)
5523 yyerror("Unmatched right square bracket");
5526 PL_lex_allbrackets--;
5527 if (PL_lex_state == LEX_INTERPNORMAL) {
5528 if (PL_lex_brackets == 0) {
5529 if (*s == '-' && s[1] == '>')
5530 PL_lex_state = LEX_INTERPENDMAYBE;
5531 else if (*s != '[' && *s != '{')
5532 PL_lex_state = LEX_INTERPEND;
5539 if (PL_lex_brackets > 100) {
5540 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5542 switch (PL_expect) {
5544 if (PL_lex_formbrack) {
5548 if (PL_oldoldbufptr == PL_last_lop)
5549 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5551 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5552 PL_lex_allbrackets++;
5553 OPERATOR(HASHBRACK);
5555 while (s < PL_bufend && SPACE_OR_TAB(*s))
5558 PL_tokenbuf[0] = '\0';
5559 if (d < PL_bufend && *d == '-') {
5560 PL_tokenbuf[0] = '-';
5562 while (d < PL_bufend && SPACE_OR_TAB(*d))
5565 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5566 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5568 while (d < PL_bufend && SPACE_OR_TAB(*d))
5571 const char minus = (PL_tokenbuf[0] == '-');
5572 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5580 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5581 PL_lex_allbrackets++;
5586 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5587 PL_lex_allbrackets++;
5592 if (PL_oldoldbufptr == PL_last_lop)
5593 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5595 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5596 PL_lex_allbrackets++;
5599 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5601 /* This hack is to get the ${} in the message. */
5603 yyerror("syntax error");
5606 OPERATOR(HASHBRACK);
5608 /* This hack serves to disambiguate a pair of curlies
5609 * as being a block or an anon hash. Normally, expectation
5610 * determines that, but in cases where we're not in a
5611 * position to expect anything in particular (like inside
5612 * eval"") we have to resolve the ambiguity. This code
5613 * covers the case where the first term in the curlies is a
5614 * quoted string. Most other cases need to be explicitly
5615 * disambiguated by prepending a "+" before the opening
5616 * curly in order to force resolution as an anon hash.
5618 * XXX should probably propagate the outer expectation
5619 * into eval"" to rely less on this hack, but that could
5620 * potentially break current behavior of eval"".
5624 if (*s == '\'' || *s == '"' || *s == '`') {
5625 /* common case: get past first string, handling escapes */
5626 for (t++; t < PL_bufend && *t != *s;)
5627 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5631 else if (*s == 'q') {
5634 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5637 /* skip q//-like construct */
5639 char open, close, term;
5642 while (t < PL_bufend && isSPACE(*t))
5644 /* check for q => */
5645 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5646 OPERATOR(HASHBRACK);
5650 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5654 for (t++; t < PL_bufend; t++) {
5655 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5657 else if (*t == open)
5661 for (t++; t < PL_bufend; t++) {
5662 if (*t == '\\' && t+1 < PL_bufend)
5664 else if (*t == close && --brackets <= 0)
5666 else if (*t == open)
5673 /* skip plain q word */
5674 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5677 else if (isALNUM_lazy_if(t,UTF)) {
5679 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5682 while (t < PL_bufend && isSPACE(*t))
5684 /* if comma follows first term, call it an anon hash */
5685 /* XXX it could be a comma expression with loop modifiers */
5686 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5687 || (*t == '=' && t[1] == '>')))
5688 OPERATOR(HASHBRACK);
5689 if (PL_expect == XREF)
5692 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5698 pl_yylval.ival = CopLINE(PL_curcop);
5699 if (isSPACE(*s) || *s == '#')
5700 PL_copline = NOLINE; /* invalidate current command line number */
5703 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5707 if (PL_lex_brackets <= 0)
5708 yyerror("Unmatched right curly bracket");
5710 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5711 PL_lex_allbrackets--;
5712 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5713 PL_lex_formbrack = 0;
5714 if (PL_lex_state == LEX_INTERPNORMAL) {
5715 if (PL_lex_brackets == 0) {
5716 if (PL_expect & XFAKEBRACK) {
5717 PL_expect &= XENUMMASK;
5718 PL_lex_state = LEX_INTERPEND;
5723 PL_thiswhite = newSVpvs("");
5724 sv_catpvs(PL_thiswhite,"}");
5727 return yylex(); /* ignore fake brackets */
5729 if (*s == '-' && s[1] == '>')
5730 PL_lex_state = LEX_INTERPENDMAYBE;
5731 else if (*s != '[' && *s != '{')
5732 PL_lex_state = LEX_INTERPEND;
5735 if (PL_expect & XFAKEBRACK) {
5736 PL_expect &= XENUMMASK;
5738 return yylex(); /* ignore fake brackets */
5740 start_force(PL_curforce);
5742 curmad('X', newSVpvn(s-1,1));
5743 CURMAD('_', PL_thiswhite);
5748 PL_thistoken = newSVpvs("");
5754 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5755 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5762 if (PL_expect == XOPERATOR) {
5763 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5764 && isIDFIRST_lazy_if(s,UTF))
5766 CopLINE_dec(PL_curcop);
5767 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5768 CopLINE_inc(PL_curcop);
5770 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5771 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5778 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5780 PL_expect = XOPERATOR;
5781 force_ident(PL_tokenbuf, '&');
5785 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5791 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5792 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5799 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5800 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5808 const char tmp = *s++;
5810 if (!PL_lex_allbrackets &&
5811 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5818 if (!PL_lex_allbrackets &&
5819 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
5827 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5828 && strchr("+-*/%.^&|<",tmp))
5829 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5830 "Reversed %c= operator",(int)tmp);
5832 if (PL_expect == XSTATE && isALPHA(tmp) &&
5833 (s == PL_linestart+1 || s[-2] == '\n') )
5835 if (PL_in_eval && !PL_rsfp) {
5840 if (strnEQ(s,"=cut",4)) {
5856 PL_thiswhite = newSVpvs("");
5857 sv_catpvn(PL_thiswhite, PL_linestart,
5858 PL_bufend - PL_linestart);
5862 PL_parser->in_pod = 1;
5866 if (PL_lex_brackets < PL_lex_formbrack) {
5868 #ifdef PERL_STRICT_CR
5869 while (SPACE_OR_TAB(*t))
5871 while (SPACE_OR_TAB(*t) || *t == '\r')
5874 if (*t == '\n' || *t == '#') {
5880 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5889 const char tmp = *s++;
5891 /* was this !=~ where !~ was meant?
5892 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5894 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5895 const char *t = s+1;
5897 while (t < PL_bufend && isSPACE(*t))
5900 if (*t == '/' || *t == '?' ||
5901 ((*t == 'm' || *t == 's' || *t == 'y')
5902 && !isALNUM(t[1])) ||
5903 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5904 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5905 "!=~ should be !~");
5907 if (!PL_lex_allbrackets &&
5908 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5920 if (PL_expect != XOPERATOR) {
5921 if (s[1] != '<' && !strchr(s,'>'))
5924 s = scan_heredoc(s);
5926 s = scan_inputsymbol(s);
5927 TERM(sublex_start());
5933 if (*s == '=' && !PL_lex_allbrackets &&
5934 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5938 SHop(OP_LEFT_SHIFT);
5943 if (!PL_lex_allbrackets &&
5944 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5951 if (!PL_lex_allbrackets &&
5952 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5960 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5968 const char tmp = *s++;
5970 if (*s == '=' && !PL_lex_allbrackets &&
5971 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5975 SHop(OP_RIGHT_SHIFT);
5977 else if (tmp == '=') {
5978 if (!PL_lex_allbrackets &&
5979 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5987 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5996 if (PL_expect == XOPERATOR) {
5997 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5998 return deprecate_commaless_var_list();
6002 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
6003 PL_tokenbuf[0] = '@';
6004 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
6005 sizeof PL_tokenbuf - 1, FALSE);
6006 if (PL_expect == XOPERATOR)
6007 no_op("Array length", s);
6008 if (!PL_tokenbuf[1])
6010 PL_expect = XOPERATOR;
6011 PL_pending_ident = '#';
6015 PL_tokenbuf[0] = '$';
6016 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
6017 sizeof PL_tokenbuf - 1, FALSE);
6018 if (PL_expect == XOPERATOR)
6020 if (!PL_tokenbuf[1]) {
6022 yyerror("Final $ should be \\$ or $name");
6026 /* This kludge not intended to be bulletproof. */
6027 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
6028 pl_yylval.opval = newSVOP(OP_CONST, 0,
6029 newSViv(CopARYBASE_get(&PL_compiling)));
6030 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
6036 const char tmp = *s;
6037 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6040 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6041 && intuit_more(s)) {
6043 PL_tokenbuf[0] = '@';
6044 if (ckWARN(WARN_SYNTAX)) {
6047 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
6050 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6051 while (t < PL_bufend && *t != ']')
6053 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6054 "Multidimensional syntax %.*s not supported",
6055 (int)((t - PL_bufptr) + 1), PL_bufptr);
6059 else if (*s == '{') {
6061 PL_tokenbuf[0] = '%';
6062 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6063 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6065 char tmpbuf[sizeof PL_tokenbuf];
6068 } while (isSPACE(*t));
6069 if (isIDFIRST_lazy_if(t,UTF)) {
6071 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6075 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
6076 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6077 "You need to quote \"%s\"",
6084 PL_expect = XOPERATOR;
6085 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6086 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6087 if (!islop || PL_last_lop_op == OP_GREPSTART)
6088 PL_expect = XOPERATOR;
6089 else if (strchr("$@\"'`q", *s))
6090 PL_expect = XTERM; /* e.g. print $fh "foo" */
6091 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6092 PL_expect = XTERM; /* e.g. print $fh &sub */
6093 else if (isIDFIRST_lazy_if(s,UTF)) {
6094 char tmpbuf[sizeof PL_tokenbuf];
6096 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6097 if ((t2 = keyword(tmpbuf, len, 0))) {
6098 /* binary operators exclude handle interpretations */
6110 PL_expect = XTERM; /* e.g. print $fh length() */
6115 PL_expect = XTERM; /* e.g. print $fh subr() */
6118 else if (isDIGIT(*s))
6119 PL_expect = XTERM; /* e.g. print $fh 3 */
6120 else if (*s == '.' && isDIGIT(s[1]))
6121 PL_expect = XTERM; /* e.g. print $fh .3 */
6122 else if ((*s == '?' || *s == '-' || *s == '+')
6123 && !isSPACE(s[1]) && s[1] != '=')
6124 PL_expect = XTERM; /* e.g. print $fh -1 */
6125 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6127 PL_expect = XTERM; /* e.g. print $fh /.../
6128 XXX except DORDOR operator
6130 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6132 PL_expect = XTERM; /* print $fh <<"EOF" */
6135 PL_pending_ident = '$';
6139 if (PL_expect == XOPERATOR)
6141 PL_tokenbuf[0] = '@';
6142 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6143 if (!PL_tokenbuf[1]) {
6146 if (PL_lex_state == LEX_NORMAL)
6148 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6150 PL_tokenbuf[0] = '%';
6152 /* Warn about @ where they meant $. */
6153 if (*s == '[' || *s == '{') {
6154 if (ckWARN(WARN_SYNTAX)) {
6155 const char *t = s + 1;
6156 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
6158 if (*t == '}' || *t == ']') {
6160 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6161 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6162 "Scalar value %.*s better written as $%.*s",
6163 (int)(t-PL_bufptr), PL_bufptr,
6164 (int)(t-PL_bufptr-1), PL_bufptr+1);
6169 PL_pending_ident = '@';
6172 case '/': /* may be division, defined-or, or pattern */
6173 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
6174 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6175 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6180 case '?': /* may either be conditional or pattern */
6181 if (PL_expect == XOPERATOR) {
6184 if (!PL_lex_allbrackets &&
6185 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6189 PL_lex_allbrackets++;
6195 /* A // operator. */
6196 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6197 (*s == '=' ? LEX_FAKEEOF_ASSIGN :
6198 LEX_FAKEEOF_LOGIC)) {
6206 if (*s == '=' && !PL_lex_allbrackets &&
6207 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6216 /* Disable warning on "study /blah/" */
6217 if (PL_oldoldbufptr == PL_last_uni
6218 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6219 || memNE(PL_last_uni, "study", 5)
6220 || isALNUM_lazy_if(PL_last_uni+5,UTF)
6224 deprecate("?PATTERN? without explicit operator");
6225 s = scan_pat(s,OP_MATCH);
6226 TERM(sublex_start());
6230 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6231 #ifdef PERL_STRICT_CR
6234 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6236 && (s == PL_linestart || s[-1] == '\n') )
6238 PL_lex_formbrack = 0;
6242 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6246 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6249 if (!PL_lex_allbrackets &&
6250 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6257 pl_yylval.ival = OPf_SPECIAL;
6263 if (*s == '=' && !PL_lex_allbrackets &&
6264 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6271 case '0': case '1': case '2': case '3': case '4':
6272 case '5': case '6': case '7': case '8': case '9':
6273 s = scan_num(s, &pl_yylval);
6274 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6275 if (PL_expect == XOPERATOR)
6280 s = scan_str(s,!!PL_madskills,FALSE);
6281 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6282 if (PL_expect == XOPERATOR) {
6283 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6284 return deprecate_commaless_var_list();
6291 pl_yylval.ival = OP_CONST;
6292 TERM(sublex_start());
6295 s = scan_str(s,!!PL_madskills,FALSE);
6296 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6297 if (PL_expect == XOPERATOR) {
6298 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6299 return deprecate_commaless_var_list();
6306 pl_yylval.ival = OP_CONST;
6307 /* FIXME. I think that this can be const if char *d is replaced by
6308 more localised variables. */
6309 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6310 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6311 pl_yylval.ival = OP_STRINGIFY;
6315 TERM(sublex_start());
6318 s = scan_str(s,!!PL_madskills,FALSE);
6319 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6320 if (PL_expect == XOPERATOR)
6321 no_op("Backticks",s);
6324 readpipe_override();
6325 TERM(sublex_start());
6329 if (PL_lex_inwhat && isDIGIT(*s))
6330 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6332 if (PL_expect == XOPERATOR)
6333 no_op("Backslash",s);
6337 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6338 char *start = s + 2;
6339 while (isDIGIT(*start) || *start == '_')
6341 if (*start == '.' && isDIGIT(start[1])) {
6342 s = scan_num(s, &pl_yylval);
6345 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6346 else if (!isALPHA(*start) && (PL_expect == XTERM
6347 || PL_expect == XREF || PL_expect == XSTATE
6348 || PL_expect == XTERMORDORDOR)) {
6349 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
6351 s = scan_num(s, &pl_yylval);
6358 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6401 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6403 /* Some keywords can be followed by any delimiter, including ':' */
6404 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6406 /* x::* is just a word, unless x is "CORE" */
6407 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6411 while (d < PL_bufend && isSPACE(*d))
6412 d++; /* no comments skipped here, or s### is misparsed */
6414 /* Is this a word before a => operator? */
6415 if (*d == '=' && d[1] == '>') {
6418 = (OP*)newSVOP(OP_CONST, 0,
6419 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6420 pl_yylval.opval->op_private = OPpCONST_BARE;
6424 /* Check for plugged-in keyword */
6428 char *saved_bufptr = PL_bufptr;
6430 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6432 if (result == KEYWORD_PLUGIN_DECLINE) {
6433 /* not a plugged-in keyword */
6434 PL_bufptr = saved_bufptr;
6435 } else if (result == KEYWORD_PLUGIN_STMT) {
6436 pl_yylval.opval = o;
6439 return REPORT(PLUGSTMT);
6440 } else if (result == KEYWORD_PLUGIN_EXPR) {
6441 pl_yylval.opval = o;
6443 PL_expect = XOPERATOR;
6444 return REPORT(PLUGEXPR);
6446 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6451 /* Check for built-in keyword */
6452 tmp = keyword(PL_tokenbuf, len, 0);
6454 /* Is this a label? */
6455 if (!anydelim && PL_expect == XSTATE
6456 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6458 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
6463 if (tmp < 0) { /* second-class keyword? */
6464 GV *ogv = NULL; /* override (winner) */
6465 GV *hgv = NULL; /* hidden (loser) */
6466 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6468 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
6471 if (GvIMPORTED_CV(gv))
6473 else if (! CvMETHOD(cv))
6477 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
6478 (gv = *gvp) && isGV_with_GP(gv) &&
6479 GvCVu(gv) && GvIMPORTED_CV(gv))
6486 tmp = 0; /* overridden by import or by GLOBAL */
6489 && -tmp==KEY_lock /* XXX generalizable kludge */
6492 tmp = 0; /* any sub overrides "weak" keyword */
6494 else { /* no override */
6496 if (tmp == KEY_dump) {
6497 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6498 "dump() better written as CORE::dump()");
6502 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6503 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6504 "Ambiguous call resolved as CORE::%s(), "
6505 "qualify as such or use &",
6513 default: /* not a keyword */
6514 /* Trade off - by using this evil construction we can pull the
6515 variable gv into the block labelled keylookup. If not, then
6516 we have to give it function scope so that the goto from the
6517 earlier ':' case doesn't bypass the initialisation. */
6519 just_a_word_zero_gv:
6527 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6531 SV *nextPL_nextwhite = 0;
6535 /* Get the rest if it looks like a package qualifier */
6537 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6539 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6542 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
6543 *s == '\'' ? "'" : "::");
6548 if (PL_expect == XOPERATOR) {
6549 if (PL_bufptr == PL_linestart) {
6550 CopLINE_dec(PL_curcop);
6551 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6552 CopLINE_inc(PL_curcop);
6555 no_op("Bareword",s);
6558 /* Look for a subroutine with this name in current package,
6559 unless name is "Foo::", in which case Foo is a bareword
6560 (and a package name). */
6562 if (len > 2 && !PL_madskills &&
6563 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6565 if (ckWARN(WARN_BAREWORD)
6566 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
6567 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6568 "Bareword \"%s\" refers to nonexistent package",
6571 PL_tokenbuf[len] = '\0';
6577 /* Mustn't actually add anything to a symbol table.
6578 But also don't want to "initialise" any placeholder
6579 constants that might already be there into full
6580 blown PVGVs with attached PVCV. */
6581 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6582 GV_NOADD_NOINIT, SVt_PVCV);
6587 /* if we saw a global override before, get the right name */
6589 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
6590 len ? len : strlen(PL_tokenbuf));
6592 SV * const tmp_sv = sv;
6593 sv = newSVpvs("CORE::GLOBAL::");
6594 sv_catsv(sv, tmp_sv);
6595 SvREFCNT_dec(tmp_sv);
6599 if (PL_madskills && !PL_thistoken) {
6600 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6601 PL_thistoken = newSVpvn(start,s - start);
6602 PL_realtokenstart = s - SvPVX(PL_linestr);
6606 /* Presume this is going to be a bareword of some sort. */
6608 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6609 pl_yylval.opval->op_private = OPpCONST_BARE;
6611 /* And if "Foo::", then that's what it certainly is. */
6616 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
6617 const_op->op_private = OPpCONST_BARE;
6618 rv2cv_op = newCVREF(0, const_op);
6620 cv = rv2cv_op_cv(rv2cv_op, 0);
6622 /* See if it's the indirect object for a list operator. */
6624 if (PL_oldoldbufptr &&
6625 PL_oldoldbufptr < PL_bufptr &&
6626 (PL_oldoldbufptr == PL_last_lop
6627 || PL_oldoldbufptr == PL_last_uni) &&
6628 /* NO SKIPSPACE BEFORE HERE! */
6629 (PL_expect == XREF ||
6630 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6632 bool immediate_paren = *s == '(';
6634 /* (Now we can afford to cross potential line boundary.) */
6635 s = SKIPSPACE2(s,nextPL_nextwhite);
6637 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6640 /* Two barewords in a row may indicate method call. */
6642 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6643 (tmp = intuit_method(s, gv, cv))) {
6645 if (tmp == METHOD && !PL_lex_allbrackets &&
6646 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6647 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6651 /* If not a declared subroutine, it's an indirect object. */
6652 /* (But it's an indir obj regardless for sort.) */
6653 /* Also, if "_" follows a filetest operator, it's a bareword */
6656 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6658 (PL_last_lop_op != OP_MAPSTART &&
6659 PL_last_lop_op != OP_GREPSTART))))
6660 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6661 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6664 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6669 PL_expect = XOPERATOR;
6672 s = SKIPSPACE2(s,nextPL_nextwhite);
6673 PL_nextwhite = nextPL_nextwhite;
6678 /* Is this a word before a => operator? */
6679 if (*s == '=' && s[1] == '>' && !pkgname) {
6682 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6683 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6684 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6688 /* If followed by a paren, it's certainly a subroutine. */
6693 while (SPACE_OR_TAB(*d))
6695 if (*d == ')' && (sv = cv_const_sv(cv))) {
6702 PL_nextwhite = PL_thiswhite;
6705 start_force(PL_curforce);
6707 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6708 PL_expect = XOPERATOR;
6711 PL_nextwhite = nextPL_nextwhite;
6712 curmad('X', PL_thistoken);
6713 PL_thistoken = newSVpvs("");
6722 /* If followed by var or block, call it a method (unless sub) */
6724 if ((*s == '$' || *s == '{') && !cv) {
6726 PL_last_lop = PL_oldbufptr;
6727 PL_last_lop_op = OP_METHOD;
6728 if (!PL_lex_allbrackets &&
6729 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6730 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6734 /* If followed by a bareword, see if it looks like indir obj. */
6737 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6738 && (tmp = intuit_method(s, gv, cv))) {
6740 if (tmp == METHOD && !PL_lex_allbrackets &&
6741 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6742 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6746 /* Not a method, so call it a subroutine (if defined) */
6749 if (lastchar == '-')
6750 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6751 "Ambiguous use of -%s resolved as -&%s()",
6752 PL_tokenbuf, PL_tokenbuf);
6753 /* Check for a constant sub */
6754 if ((sv = cv_const_sv(cv))) {
6757 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6758 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6759 pl_yylval.opval->op_private = 0;
6760 pl_yylval.opval->op_flags |= OPf_SPECIAL;
6764 op_free(pl_yylval.opval);
6765 pl_yylval.opval = rv2cv_op;
6766 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6767 PL_last_lop = PL_oldbufptr;
6768 PL_last_lop_op = OP_ENTERSUB;
6769 /* Is there a prototype? */
6777 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
6780 while (*proto == ';')
6785 *proto == '$' || *proto == '_'
6786 || *proto == '*' || *proto == '+'
6791 *proto == '\\' && proto[1] && proto[2] == '\0'
6795 if (*proto == '\\' && proto[1] == '[') {
6796 const char *p = proto + 2;
6797 while(*p && *p != ']')
6799 if(*p == ']' && !p[1]) OPERATOR(UNIOPSUB);
6801 if (*proto == '&' && *s == '{') {
6803 sv_setpvs(PL_subname, "__ANON__");
6805 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6806 if (!PL_lex_allbrackets &&
6807 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6808 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6815 PL_nextwhite = PL_thiswhite;
6818 start_force(PL_curforce);
6819 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6822 PL_nextwhite = nextPL_nextwhite;
6823 curmad('X', PL_thistoken);
6824 PL_thistoken = newSVpvs("");
6827 if (!PL_lex_allbrackets &&
6828 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6829 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6834 /* Guess harder when madskills require "best effort". */
6835 if (PL_madskills && (!gv || !GvCVu(gv))) {
6836 int probable_sub = 0;
6837 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6839 else if (isALPHA(*s)) {
6843 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6844 if (!keyword(tmpbuf, tmplen, 0))
6847 while (d < PL_bufend && isSPACE(*d))
6849 if (*d == '=' && d[1] == '>')
6854 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
6855 op_free(pl_yylval.opval);
6856 pl_yylval.opval = rv2cv_op;
6857 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6858 PL_last_lop = PL_oldbufptr;
6859 PL_last_lop_op = OP_ENTERSUB;
6860 PL_nextwhite = PL_thiswhite;
6862 start_force(PL_curforce);
6863 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6865 PL_nextwhite = nextPL_nextwhite;
6866 curmad('X', PL_thistoken);
6867 PL_thistoken = newSVpvs("");
6869 if (!PL_lex_allbrackets &&
6870 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6871 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6875 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6878 if (!PL_lex_allbrackets &&
6879 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6880 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6885 /* Call it a bare word */
6887 if (PL_hints & HINT_STRICT_SUBS)
6888 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6891 /* after "print" and similar functions (corresponding to
6892 * "F? L" in opcode.pl), whatever wasn't already parsed as
6893 * a filehandle should be subject to "strict subs".
6894 * Likewise for the optional indirect-object argument to system
6895 * or exec, which can't be a bareword */
6896 if ((PL_last_lop_op == OP_PRINT
6897 || PL_last_lop_op == OP_PRTF
6898 || PL_last_lop_op == OP_SAY
6899 || PL_last_lop_op == OP_SYSTEM
6900 || PL_last_lop_op == OP_EXEC)
6901 && (PL_hints & HINT_STRICT_SUBS))
6902 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6903 if (lastchar != '-') {
6904 if (ckWARN(WARN_RESERVED)) {
6908 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
6909 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6917 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
6918 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6919 "Operator or semicolon missing before %c%s",
6920 lastchar, PL_tokenbuf);
6921 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6922 "Ambiguous use of %c resolved as operator %c",
6923 lastchar, lastchar);
6930 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6931 newSVpv(CopFILE(PL_curcop),0))
6936 (OP*)newSVOP(OP_CONST, 0,
6937 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
6940 case KEY___PACKAGE__:
6942 (OP*)newSVOP(OP_CONST, 0,
6944 ? newSVhek(HvNAME_HEK(PL_curstash))
6951 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6952 const char *pname = "main";
6953 if (PL_tokenbuf[2] == 'D')
6954 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
6955 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
6959 GvIOp(gv) = newIO();
6960 IoIFP(GvIOp(gv)) = PL_rsfp;
6961 #if defined(HAS_FCNTL) && defined(F_SETFD)
6963 const int fd = PerlIO_fileno(PL_rsfp);
6964 fcntl(fd,F_SETFD,fd >= 3);
6967 /* Mark this internal pseudo-handle as clean */
6968 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6969 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6970 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6972 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6973 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6974 /* if the script was opened in binmode, we need to revert
6975 * it to text mode for compatibility; but only iff it has CRs
6976 * XXX this is a questionable hack at best. */
6977 if (PL_bufend-PL_bufptr > 2
6978 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6981 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6982 loc = PerlIO_tell(PL_rsfp);
6983 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6986 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6988 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6989 #endif /* NETWARE */
6990 #ifdef PERLIO_IS_STDIO /* really? */
6991 # if defined(__BORLANDC__)
6992 /* XXX see note in do_binmode() */
6993 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
6997 PerlIO_seek(PL_rsfp, loc, 0);
7001 #ifdef PERLIO_LAYERS
7004 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
7005 else if (PL_encoding) {
7012 XPUSHs(PL_encoding);
7014 call_method("name", G_SCALAR);
7018 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7019 Perl_form(aTHX_ ":encoding(%"SVf")",
7028 if (PL_realtokenstart >= 0) {
7029 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7031 PL_endwhite = newSVpvs("");
7032 sv_catsv(PL_endwhite, PL_thiswhite);
7034 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
7035 PL_realtokenstart = -1;
7037 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
7053 if (PL_expect == XSTATE) {
7060 if (*s == ':' && s[1] == ':') {
7063 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7064 if (!(tmp = keyword(PL_tokenbuf, len, 1)))
7065 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
7068 else if (tmp == KEY_require || tmp == KEY_do)
7069 /* that's a way to remember we saw "CORE::" */
7082 LOP(OP_ACCEPT,XTERM);
7085 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7090 LOP(OP_ATAN2,XTERM);
7096 LOP(OP_BINMODE,XTERM);
7099 LOP(OP_BLESS,XTERM);
7108 /* We have to disambiguate the two senses of
7109 "continue". If the next token is a '{' then
7110 treat it as the start of a continue block;
7111 otherwise treat it as a control operator.
7121 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7131 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7140 if (!PL_cryptseen) {
7141 PL_cryptseen = TRUE;
7145 LOP(OP_CRYPT,XTERM);
7148 LOP(OP_CHMOD,XTERM);
7151 LOP(OP_CHOWN,XTERM);
7154 LOP(OP_CONNECT,XTERM);
7173 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7174 if (orig_keyword == KEY_do) {
7183 PL_hints |= HINT_BLOCK_SCOPE;
7193 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7194 STR_WITH_LEN("NDBM_File::"),
7195 STR_WITH_LEN("DB_File::"),
7196 STR_WITH_LEN("GDBM_File::"),
7197 STR_WITH_LEN("SDBM_File::"),
7198 STR_WITH_LEN("ODBM_File::"),
7200 LOP(OP_DBMOPEN,XTERM);
7206 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7213 pl_yylval.ival = CopLINE(PL_curcop);
7217 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7231 if (*s == '{') { /* block eval */
7232 PL_expect = XTERMBLOCK;
7233 UNIBRACK(OP_ENTERTRY);
7235 else { /* string eval */
7237 UNIBRACK(OP_ENTEREVAL);
7252 case KEY_endhostent:
7258 case KEY_endservent:
7261 case KEY_endprotoent:
7272 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7274 pl_yylval.ival = CopLINE(PL_curcop);
7276 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7279 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
7282 if ((PL_bufend - p) >= 3 &&
7283 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7285 else if ((PL_bufend - p) >= 4 &&
7286 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7289 if (isIDFIRST_lazy_if(p,UTF)) {
7290 p = scan_ident(p, PL_bufend,
7291 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
7295 Perl_croak(aTHX_ "Missing $ on loop variable");
7297 s = SvPVX(PL_linestr) + soff;
7303 LOP(OP_FORMLINE,XTERM);
7309 LOP(OP_FCNTL,XTERM);
7315 LOP(OP_FLOCK,XTERM);
7318 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7323 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7328 LOP(OP_GREPSTART, XREF);
7331 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7346 case KEY_getpriority:
7347 LOP(OP_GETPRIORITY,XTERM);
7349 case KEY_getprotobyname:
7352 case KEY_getprotobynumber:
7353 LOP(OP_GPBYNUMBER,XTERM);
7355 case KEY_getprotoent:
7367 case KEY_getpeername:
7368 UNI(OP_GETPEERNAME);
7370 case KEY_gethostbyname:
7373 case KEY_gethostbyaddr:
7374 LOP(OP_GHBYADDR,XTERM);
7376 case KEY_gethostent:
7379 case KEY_getnetbyname:
7382 case KEY_getnetbyaddr:
7383 LOP(OP_GNBYADDR,XTERM);
7388 case KEY_getservbyname:
7389 LOP(OP_GSBYNAME,XTERM);
7391 case KEY_getservbyport:
7392 LOP(OP_GSBYPORT,XTERM);
7394 case KEY_getservent:
7397 case KEY_getsockname:
7398 UNI(OP_GETSOCKNAME);
7400 case KEY_getsockopt:
7401 LOP(OP_GSOCKOPT,XTERM);
7416 pl_yylval.ival = CopLINE(PL_curcop);
7426 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7428 pl_yylval.ival = CopLINE(PL_curcop);
7432 LOP(OP_INDEX,XTERM);
7438 LOP(OP_IOCTL,XTERM);
7450 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7467 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7472 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7486 LOP(OP_LISTEN,XTERM);
7495 s = scan_pat(s,OP_MATCH);
7496 TERM(sublex_start());
7499 LOP(OP_MAPSTART, XREF);
7502 LOP(OP_MKDIR,XTERM);
7505 LOP(OP_MSGCTL,XTERM);
7508 LOP(OP_MSGGET,XTERM);
7511 LOP(OP_MSGRCV,XTERM);
7514 LOP(OP_MSGSND,XTERM);
7519 PL_in_my = (U16)tmp;
7521 if (isIDFIRST_lazy_if(s,UTF)) {
7525 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7526 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7528 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7529 if (!PL_in_my_stash) {
7532 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7536 if (PL_madskills) { /* just add type to declarator token */
7537 sv_catsv(PL_thistoken, PL_nextwhite);
7539 sv_catpvn(PL_thistoken, start, s - start);
7547 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7551 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7556 s = tokenize_use(0, s);
7560 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7563 if (!PL_lex_allbrackets &&
7564 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7565 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7571 if (isIDFIRST_lazy_if(s,UTF)) {
7573 for (d = s; isALNUM_lazy_if(d,UTF);)
7575 for (t=d; isSPACE(*t);)
7577 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7579 && !(t[0] == '=' && t[1] == '>')
7581 int parms_len = (int)(d-s);
7582 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7583 "Precedence problem: open %.*s should be open(%.*s)",
7584 parms_len, s, parms_len, s);
7590 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7592 pl_yylval.ival = OP_OR;
7602 LOP(OP_OPEN_DIR,XTERM);
7605 checkcomma(s,PL_tokenbuf,"filehandle");
7609 checkcomma(s,PL_tokenbuf,"filehandle");
7628 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7630 s = force_strict_version(s);
7631 PL_lex_expect = XBLOCK;
7635 LOP(OP_PIPE_OP,XTERM);
7638 s = scan_str(s,!!PL_madskills,FALSE);
7641 pl_yylval.ival = OP_CONST;
7642 TERM(sublex_start());
7649 s = scan_str(s,!!PL_madskills,FALSE);
7652 PL_expect = XOPERATOR;
7653 if (SvCUR(PL_lex_stuff)) {
7654 int warned_comma = !ckWARN(WARN_QW);
7655 int warned_comment = warned_comma;
7656 d = SvPV_force(PL_lex_stuff, len);
7658 for (; isSPACE(*d) && len; --len, ++d)
7663 if (!warned_comma || !warned_comment) {
7664 for (; !isSPACE(*d) && len; --len, ++d) {
7665 if (!warned_comma && *d == ',') {
7666 Perl_warner(aTHX_ packWARN(WARN_QW),
7667 "Possible attempt to separate words with commas");
7670 else if (!warned_comment && *d == '#') {
7671 Perl_warner(aTHX_ packWARN(WARN_QW),
7672 "Possible attempt to put comments in qw() list");
7678 for (; !isSPACE(*d) && len; --len, ++d)
7681 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7682 words = op_append_elem(OP_LIST, words,
7683 newSVOP(OP_CONST, 0, tokeq(sv)));
7688 words = newNULLLIST();
7690 SvREFCNT_dec(PL_lex_stuff);
7691 PL_lex_stuff = NULL;
7693 PL_expect = XOPERATOR;
7694 pl_yylval.opval = sawparens(words);
7699 s = scan_str(s,!!PL_madskills,FALSE);
7702 pl_yylval.ival = OP_STRINGIFY;
7703 if (SvIVX(PL_lex_stuff) == '\'')
7704 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
7705 TERM(sublex_start());
7708 s = scan_pat(s,OP_QR);
7709 TERM(sublex_start());
7712 s = scan_str(s,!!PL_madskills,FALSE);
7715 readpipe_override();
7716 TERM(sublex_start());
7724 s = force_version(s, FALSE);
7726 else if (*s != 'v' || !isDIGIT(s[1])
7727 || (s = force_version(s, TRUE), *s == 'v'))
7729 *PL_tokenbuf = '\0';
7730 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7731 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7732 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7734 yyerror("<> should be quotes");
7736 if (orig_keyword == KEY_require) {
7744 PL_last_uni = PL_oldbufptr;
7745 PL_last_lop_op = OP_REQUIRE;
7747 return REPORT( (int)REQUIRE );
7753 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7757 LOP(OP_RENAME,XTERM);
7766 LOP(OP_RINDEX,XTERM);
7775 UNIDOR(OP_READLINE);
7778 UNIDOR(OP_BACKTICK);
7787 LOP(OP_REVERSE,XTERM);
7790 UNIDOR(OP_READLINK);
7797 if (pl_yylval.opval)
7798 TERM(sublex_start());
7800 TOKEN(1); /* force error */
7803 checkcomma(s,PL_tokenbuf,"filehandle");
7813 LOP(OP_SELECT,XTERM);
7819 LOP(OP_SEMCTL,XTERM);
7822 LOP(OP_SEMGET,XTERM);
7825 LOP(OP_SEMOP,XTERM);
7831 LOP(OP_SETPGRP,XTERM);
7833 case KEY_setpriority:
7834 LOP(OP_SETPRIORITY,XTERM);
7836 case KEY_sethostent:
7842 case KEY_setservent:
7845 case KEY_setprotoent:
7855 LOP(OP_SEEKDIR,XTERM);
7857 case KEY_setsockopt:
7858 LOP(OP_SSOCKOPT,XTERM);
7864 LOP(OP_SHMCTL,XTERM);
7867 LOP(OP_SHMGET,XTERM);
7870 LOP(OP_SHMREAD,XTERM);
7873 LOP(OP_SHMWRITE,XTERM);
7876 LOP(OP_SHUTDOWN,XTERM);
7885 LOP(OP_SOCKET,XTERM);
7887 case KEY_socketpair:
7888 LOP(OP_SOCKPAIR,XTERM);
7891 checkcomma(s,PL_tokenbuf,"subroutine name");
7893 if (*s == ';' || *s == ')') /* probably a close */
7894 Perl_croak(aTHX_ "sort is now a reserved word");
7896 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7900 LOP(OP_SPLIT,XTERM);
7903 LOP(OP_SPRINTF,XTERM);
7906 LOP(OP_SPLICE,XTERM);
7921 LOP(OP_SUBSTR,XTERM);
7927 char tmpbuf[sizeof PL_tokenbuf];
7928 SSize_t tboffset = 0;
7929 expectation attrful;
7930 bool have_name, have_proto;
7931 const int key = tmp;
7936 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7937 SV *subtoken = newSVpvn(tstart, s - tstart);
7941 s = SKIPSPACE2(s,tmpwhite);
7946 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7947 (*s == ':' && s[1] == ':'))
7950 SV *nametoke = NULL;
7954 attrful = XATTRBLOCK;
7955 /* remember buffer pos'n for later force_word */
7956 tboffset = s - PL_oldbufptr;
7957 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7960 nametoke = newSVpvn(s, d - s);
7962 if (memchr(tmpbuf, ':', len))
7963 sv_setpvn(PL_subname, tmpbuf, len);
7965 sv_setsv(PL_subname,PL_curstname);
7966 sv_catpvs(PL_subname,"::");
7967 sv_catpvn(PL_subname,tmpbuf,len);
7974 CURMAD('X', nametoke);
7975 CURMAD('_', tmpwhite);
7976 (void) force_word(PL_oldbufptr + tboffset, WORD,
7979 s = SKIPSPACE2(d,tmpwhite);
7986 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7987 PL_expect = XTERMBLOCK;
7988 attrful = XATTRTERM;
7989 sv_setpvs(PL_subname,"?");
7993 if (key == KEY_format) {
7995 PL_lex_formbrack = PL_lex_brackets + 1;
7997 PL_thistoken = subtoken;
8001 (void) force_word(PL_oldbufptr + tboffset, WORD,
8007 /* Look for a prototype */
8010 bool bad_proto = FALSE;
8011 bool in_brackets = FALSE;
8012 char greedy_proto = ' ';
8013 bool proto_after_greedy_proto = FALSE;
8014 bool must_be_last = FALSE;
8015 bool underscore = FALSE;
8016 bool seen_underscore = FALSE;
8017 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
8019 s = scan_str(s,!!PL_madskills,FALSE);
8021 Perl_croak(aTHX_ "Prototype not terminated");
8022 /* strip spaces and check for bad characters */
8023 d = SvPVX(PL_lex_stuff);
8025 for (p = d; *p; ++p) {
8029 if (warnillegalproto) {
8031 proto_after_greedy_proto = TRUE;
8032 if (!strchr("$@%*;[]&\\_+", *p)) {
8044 else if ( *p == ']' ) {
8045 in_brackets = FALSE;
8047 else if ( (*p == '@' || *p == '%') &&
8048 ( tmp < 2 || d[tmp-2] != '\\' ) &&
8050 must_be_last = TRUE;
8053 else if ( *p == '_' ) {
8054 underscore = seen_underscore = TRUE;
8061 if (proto_after_greedy_proto)
8062 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8063 "Prototype after '%c' for %"SVf" : %s",
8064 greedy_proto, SVfARG(PL_subname), d);
8066 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8067 "Illegal character %sin prototype for %"SVf" : %s",
8068 seen_underscore ? "after '_' " : "",
8069 SVfARG(PL_subname), d);
8070 SvCUR_set(PL_lex_stuff, tmp);
8075 CURMAD('q', PL_thisopen);
8076 CURMAD('_', tmpwhite);
8077 CURMAD('=', PL_thisstuff);
8078 CURMAD('Q', PL_thisclose);
8079 NEXTVAL_NEXTTOKE.opval =
8080 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8081 PL_lex_stuff = NULL;
8084 s = SKIPSPACE2(s,tmpwhite);
8092 if (*s == ':' && s[1] != ':')
8093 PL_expect = attrful;
8094 else if (*s != '{' && key == KEY_sub) {
8096 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8097 else if (*s != ';' && *s != '}')
8098 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8105 curmad('^', newSVpvs(""));
8106 CURMAD('_', tmpwhite);
8110 PL_thistoken = subtoken;
8113 NEXTVAL_NEXTTOKE.opval =
8114 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8115 PL_lex_stuff = NULL;
8121 sv_setpvs(PL_subname, "__ANON__");
8123 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8127 (void) force_word(PL_oldbufptr + tboffset, WORD,
8136 LOP(OP_SYSTEM,XREF);
8139 LOP(OP_SYMLINK,XTERM);
8142 LOP(OP_SYSCALL,XTERM);
8145 LOP(OP_SYSOPEN,XTERM);
8148 LOP(OP_SYSSEEK,XTERM);
8151 LOP(OP_SYSREAD,XTERM);
8154 LOP(OP_SYSWRITE,XTERM);
8158 TERM(sublex_start());
8179 LOP(OP_TRUNCATE,XTERM);
8191 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8193 pl_yylval.ival = CopLINE(PL_curcop);
8197 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8199 pl_yylval.ival = CopLINE(PL_curcop);
8203 LOP(OP_UNLINK,XTERM);
8209 LOP(OP_UNPACK,XTERM);
8212 LOP(OP_UTIME,XTERM);
8218 LOP(OP_UNSHIFT,XTERM);
8221 s = tokenize_use(1, s);
8231 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8233 pl_yylval.ival = CopLINE(PL_curcop);
8237 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8239 pl_yylval.ival = CopLINE(PL_curcop);
8243 PL_hints |= HINT_BLOCK_SCOPE;
8250 LOP(OP_WAITPID,XTERM);
8259 ctl_l[0] = toCTRL('L');
8261 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
8264 /* Make sure $^L is defined */
8265 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
8270 if (PL_expect == XOPERATOR) {
8271 if (*s == '=' && !PL_lex_allbrackets &&
8272 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8280 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8282 pl_yylval.ival = OP_XOR;
8287 TERM(sublex_start());
8292 #pragma segment Main
8296 S_pending_ident(pTHX)
8301 /* pit holds the identifier we read and pending_ident is reset */
8302 char pit = PL_pending_ident;
8303 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8304 /* All routes through this function want to know if there is a colon. */
8305 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8306 PL_pending_ident = 0;
8308 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
8309 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8310 "### Pending identifier '%s'\n", PL_tokenbuf); });
8312 /* if we're in a my(), we can't allow dynamics here.
8313 $foo'bar has already been turned into $foo::bar, so
8314 just check for colons.
8316 if it's a legal name, the OP is a PADANY.
8319 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8321 yyerror(Perl_form(aTHX_ "No package name allowed for "
8322 "variable %s in \"our\"",
8324 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8328 yyerror(Perl_form(aTHX_ PL_no_myglob,
8329 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
8331 pl_yylval.opval = newOP(OP_PADANY, 0);
8332 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8333 UTF ? SVf_UTF8 : 0);
8339 build the ops for accesses to a my() variable.
8341 Deny my($a) or my($b) in a sort block, *if* $a or $b is
8342 then used in a comparison. This catches most, but not
8343 all cases. For instance, it catches
8344 sort { my($a); $a <=> $b }
8346 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
8347 (although why you'd do that is anyone's guess).
8352 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8353 UTF ? SVf_UTF8 : 0);
8354 if (tmp != NOT_IN_PAD) {
8355 /* might be an "our" variable" */
8356 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8357 /* build ops for a bareword */
8358 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8359 HEK * const stashname = HvNAME_HEK(stash);
8360 SV * const sym = newSVhek(stashname);
8361 sv_catpvs(sym, "::");
8362 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
8363 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8364 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8367 ? (GV_ADDMULTI | GV_ADDINEVAL)
8370 ((PL_tokenbuf[0] == '$') ? SVt_PV
8371 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8376 /* if it's a sort block and they're naming $a or $b */
8377 if (PL_last_lop_op == OP_SORT &&
8378 PL_tokenbuf[0] == '$' &&
8379 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
8382 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
8383 d < PL_bufend && *d != '\n';
8386 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
8387 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
8393 pl_yylval.opval = newOP(OP_PADANY, 0);
8394 pl_yylval.opval->op_targ = tmp;
8400 Whine if they've said @foo in a doublequoted string,
8401 and @foo isn't a variable we can find in the symbol
8404 if (ckWARN(WARN_AMBIGUOUS) &&
8405 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8406 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
8408 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8409 /* DO NOT warn for @- and @+ */
8410 && !( PL_tokenbuf[2] == '\0' &&
8411 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8414 /* Downgraded from fatal to warning 20000522 mjd */
8415 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8416 "Possible unintended interpolation of %s in string",
8421 /* build ops for a bareword */
8422 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
8424 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8425 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8426 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD,
8427 ((PL_tokenbuf[0] == '$') ? SVt_PV
8428 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8434 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8438 PERL_ARGS_ASSERT_CHECKCOMMA;
8440 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8441 if (ckWARN(WARN_SYNTAX)) {
8444 for (w = s+2; *w && level; w++) {
8452 /* the list of chars below is for end of statements or
8453 * block / parens, boolean operators (&&, ||, //) and branch
8454 * constructs (or, and, if, until, unless, while, err, for).
8455 * Not a very solid hack... */
8456 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8457 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8458 "%s (...) interpreted as function",name);
8461 while (s < PL_bufend && isSPACE(*s))
8465 while (s < PL_bufend && isSPACE(*s))
8467 if (isIDFIRST_lazy_if(s,UTF)) {
8468 const char * const w = s++;
8469 while (isALNUM_lazy_if(s,UTF))
8471 while (s < PL_bufend && isSPACE(*s))
8475 if (keyword(w, s - w, 0))
8478 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
8479 if (gv && GvCVu(gv))
8481 Perl_croak(aTHX_ "No comma allowed after %s", what);
8486 /* Either returns sv, or mortalizes sv and returns a new SV*.
8487 Best used as sv=new_constant(..., sv, ...).
8488 If s, pv are NULL, calls subroutine with one argument,
8489 and type is used with error messages only. */
8492 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
8493 SV *sv, SV *pv, const char *type, STRLEN typelen)
8496 HV * const table = GvHV(PL_hintgv); /* ^H */
8500 const char *why1 = "", *why2 = "", *why3 = "";
8502 PERL_ARGS_ASSERT_NEW_CONSTANT;
8504 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8507 why2 = (const char *)
8508 (strEQ(key,"charnames")
8509 ? "(possibly a missing \"use charnames ...\")"
8511 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
8512 (type ? type: "undef"), why2);
8514 /* This is convoluted and evil ("goto considered harmful")
8515 * but I do not understand the intricacies of all the different
8516 * failure modes of %^H in here. The goal here is to make
8517 * the most probable error message user-friendly. --jhi */
8522 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
8523 (type ? type: "undef"), why1, why2, why3);
8525 yyerror(SvPVX_const(msg));
8530 /* charnames doesn't work well if there have been errors found */
8531 if (PL_error_count > 0 && strEQ(key,"charnames"))
8532 return &PL_sv_undef;
8534 cvp = hv_fetch(table, key, keylen, FALSE);
8535 if (!cvp || !SvOK(*cvp)) {
8538 why3 = "} is not defined";
8541 sv_2mortal(sv); /* Parent created it permanently */
8544 pv = newSVpvn_flags(s, len, SVs_TEMP);
8546 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
8548 typesv = &PL_sv_undef;
8550 PUSHSTACKi(PERLSI_OVERLOAD);
8562 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
8566 /* Check the eval first */
8567 if (!PL_in_eval && SvTRUE(ERRSV)) {
8568 sv_catpvs(ERRSV, "Propagated");
8569 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
8571 res = SvREFCNT_inc_simple(sv);
8575 SvREFCNT_inc_simple_void(res);
8584 why1 = "Call to &{$^H{";
8586 why3 = "}} did not return a defined value";
8594 /* Returns a NUL terminated string, with the length of the string written to
8598 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
8601 register char *d = dest;
8602 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
8604 PERL_ARGS_ASSERT_SCAN_WORD;
8608 Perl_croak(aTHX_ ident_too_long);
8609 if (isALNUM(*s)) /* UTF handled below */
8611 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
8616 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
8620 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
8621 char *t = s + UTF8SKIP(s);
8623 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
8627 Perl_croak(aTHX_ ident_too_long);
8628 Copy(s, d, len, char);
8641 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
8644 char *bracket = NULL;
8646 register char *d = dest;
8647 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
8649 PERL_ARGS_ASSERT_SCAN_IDENT;
8654 while (isDIGIT(*s)) {
8656 Perl_croak(aTHX_ ident_too_long);
8663 Perl_croak(aTHX_ ident_too_long);
8664 if (isALNUM(*s)) /* UTF handled below */
8666 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
8671 else if (*s == ':' && s[1] == ':') {
8675 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
8676 char *t = s + UTF8SKIP(s);
8677 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
8679 if (d + (t - s) > e)
8680 Perl_croak(aTHX_ ident_too_long);
8681 Copy(s, d, t - s, char);
8692 if (PL_lex_state != LEX_NORMAL)
8693 PL_lex_state = LEX_INTERPENDMAYBE;
8696 if (*s == '$' && s[1] &&
8697 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
8710 if (*d == '^' && *s && isCONTROLVAR(*s)) {
8715 if (isSPACE(s[-1])) {
8717 const char ch = *s++;
8718 if (!SPACE_OR_TAB(ch)) {
8724 if (isIDFIRST_lazy_if(d,UTF)) {
8728 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
8729 end += UTF8SKIP(end);
8730 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
8731 end += UTF8SKIP(end);
8733 Copy(s, d, end - s, char);
8738 while ((isALNUM(*s) || *s == ':') && d < e)
8741 Perl_croak(aTHX_ ident_too_long);
8744 while (s < send && SPACE_OR_TAB(*s))
8746 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
8747 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
8748 const char * const brack =
8750 ((*s == '[') ? "[...]" : "{...}");
8751 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
8752 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8753 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
8754 funny, dest, brack, funny, dest, brack);
8757 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
8758 PL_lex_allbrackets++;
8762 /* Handle extended ${^Foo} variables
8763 * 1999-02-27 mjd-perl-patch@plover.com */
8764 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
8768 while (isALNUM(*s) && d < e) {
8772 Perl_croak(aTHX_ ident_too_long);
8777 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
8778 PL_lex_state = LEX_INTERPEND;
8781 if (PL_lex_state == LEX_NORMAL) {
8782 if (ckWARN(WARN_AMBIGUOUS) &&
8783 (keyword(dest, d - dest, 0)
8784 || get_cvn_flags(dest, d - dest, 0)))
8788 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8789 "Ambiguous use of %c{%s} resolved to %c%s",
8790 funny, dest, funny, dest);
8795 s = bracket; /* let the parser handle it */
8799 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
8800 PL_lex_state = LEX_INTERPEND;
8805 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
8807 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
8808 * the parse starting at 's', based on the subset that are valid in this
8809 * context input to this routine in 'valid_flags'. Advances s. Returns
8810 * TRUE if the input was a valid flag, so the next char may be as well;
8811 * otherwise FALSE. 'charset' should point to a NUL upon first call on the
8812 * current regex. This routine will set it to any charset modifier found.
8813 * The caller shouldn't change it. This way, another charset modifier
8814 * encountered in the parse can be detected as an error, as we have decided
8819 if (! strchr(valid_flags, c)) {
8828 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
8829 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
8830 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
8831 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
8832 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
8833 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
8834 case LOCALE_PAT_MOD:
8836 /* In 5.14, qr//lt is legal but deprecated; the 't' means they
8837 * can't be regex modifiers.
8838 * In 5.14, s///le is legal and ambiguous. Try to disambiguate as
8839 * much as easily done. s///lei, for example, has to mean regex
8840 * modifiers if it's not an error (as does any word character
8841 * following the 'e'). Otherwise, we resolve to the backwards-
8842 * compatible, but less likely 's/// le ...', i.e. as meaning
8843 * less-than-or-equal. The reason it's not likely is that s//
8844 * returns a number for code in the field (/r returns a string, but
8845 * that wasn't added until the 5.13 series), and so '<=' should be
8846 * used for comparing, not 'le'. */
8847 if (*((*s) + 1) == 't') {
8850 else if (*((*s) + 1) == 'e' && ! isALNUM(*((*s) + 2))) {
8852 /* 'e' is valid only for substitutes, s///e. If it is not
8853 * valid in the current context, then 'm//le' must mean the
8854 * comparison operator, so use the regular deprecation message.
8856 if (! strchr(valid_flags, 'e')) {
8859 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
8860 "Ambiguous use of 's//le...' resolved as 's// le...'; Rewrite as 's//el' if you meant 'use locale rules and evaluate rhs as an expression'. In Perl 5.16, it will be resolved the other way");
8864 goto multiple_charsets;
8866 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
8869 case UNICODE_PAT_MOD:
8870 /* In 5.14, qr//unless and qr//until are legal but deprecated; the
8871 * 'n' means they can't be regex modifiers */
8872 if (*((*s) + 1) == 'n') {
8876 goto multiple_charsets;
8878 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
8881 case ASCII_RESTRICT_PAT_MOD:
8882 /* In 5.14, qr//and is legal but deprecated; the 'n' means they
8883 * can't be regex modifiers */
8884 if (*((*s) + 1) == 'n') {
8889 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
8893 /* Error if previous modifier wasn't an 'a', but if it was, see
8894 * if, and accept, a second occurrence (only) */
8896 || get_regex_charset(*pmfl)
8897 != REGEX_ASCII_RESTRICTED_CHARSET)
8899 goto multiple_charsets;
8901 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
8905 case DEPENDS_PAT_MOD:
8907 goto multiple_charsets;
8909 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
8918 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX),
8919 "Having no space between pattern and following word is deprecated");
8923 if (*charset != c) {
8924 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
8926 else if (c == 'a') {
8927 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
8930 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
8933 /* Pretend that it worked, so will continue processing before dieing */
8939 S_scan_pat(pTHX_ char *start, I32 type)
8943 char *s = scan_str(start,!!PL_madskills,FALSE);
8944 const char * const valid_flags =
8945 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
8946 char charset = '\0'; /* character set modifier */
8951 PERL_ARGS_ASSERT_SCAN_PAT;
8954 const char * const delimiter = skipspace(start);
8958 ? "Search pattern not terminated or ternary operator parsed as search pattern"
8959 : "Search pattern not terminated" ));
8962 pm = (PMOP*)newPMOP(type, 0);
8963 if (PL_multi_open == '?') {
8964 /* This is the only point in the code that sets PMf_ONCE: */
8965 pm->op_pmflags |= PMf_ONCE;
8967 /* Hence it's safe to do this bit of PMOP book-keeping here, which
8968 allows us to restrict the list needed by reset to just the ??
8970 assert(type != OP_TRANS);
8972 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
8975 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
8978 elements = mg->mg_len / sizeof(PMOP**);
8979 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
8980 ((PMOP**)mg->mg_ptr) [elements++] = pm;
8981 mg->mg_len = elements * sizeof(PMOP**);
8982 PmopSTASH_set(pm,PL_curstash);
8988 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
8990 if (PL_madskills && modstart != s) {
8991 SV* tmptoken = newSVpvn(modstart, s - modstart);
8992 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
8995 /* issue a warning if /c is specified,but /g is not */
8996 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
8998 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
8999 "Use of /c modifier is meaningless without /g" );
9002 PL_lex_op = (OP*)pm;
9003 pl_yylval.ival = OP_MATCH;
9008 S_scan_subst(pTHX_ char *start)
9015 char charset = '\0'; /* character set modifier */
9020 PERL_ARGS_ASSERT_SCAN_SUBST;
9022 pl_yylval.ival = OP_NULL;
9024 s = scan_str(start,!!PL_madskills,FALSE);
9027 Perl_croak(aTHX_ "Substitution pattern not terminated");
9029 if (s[-1] == PL_multi_open)
9033 CURMAD('q', PL_thisopen);
9034 CURMAD('_', PL_thiswhite);
9035 CURMAD('E', PL_thisstuff);
9036 CURMAD('Q', PL_thisclose);
9037 PL_realtokenstart = s - SvPVX(PL_linestr);
9041 first_start = PL_multi_start;
9042 s = scan_str(s,!!PL_madskills,FALSE);
9045 SvREFCNT_dec(PL_lex_stuff);
9046 PL_lex_stuff = NULL;
9048 Perl_croak(aTHX_ "Substitution replacement not terminated");
9050 PL_multi_start = first_start; /* so whole substitution is taken together */
9052 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9056 CURMAD('z', PL_thisopen);
9057 CURMAD('R', PL_thisstuff);
9058 CURMAD('Z', PL_thisclose);
9064 if (*s == EXEC_PAT_MOD) {
9068 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
9077 curmad('m', newSVpvn(modstart, s - modstart));
9078 append_madprops(PL_thismad, (OP*)pm, 0);
9082 if ((pm->op_pmflags & PMf_CONTINUE)) {
9083 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9087 SV * const repl = newSVpvs("");
9089 PL_sublex_info.super_bufptr = s;
9090 PL_sublex_info.super_bufend = PL_bufend;
9092 pm->op_pmflags |= PMf_EVAL;
9095 sv_catpvs(repl, "eval ");
9097 sv_catpvs(repl, "do ");
9099 sv_catpvs(repl, "{");
9100 sv_catsv(repl, PL_lex_repl);
9101 if (strchr(SvPVX(PL_lex_repl), '#'))
9102 sv_catpvs(repl, "\n");
9103 sv_catpvs(repl, "}");
9105 SvREFCNT_dec(PL_lex_repl);
9109 PL_lex_op = (OP*)pm;
9110 pl_yylval.ival = OP_SUBST;
9115 S_scan_trans(pTHX_ char *start)
9124 bool nondestruct = 0;
9129 PERL_ARGS_ASSERT_SCAN_TRANS;
9131 pl_yylval.ival = OP_NULL;
9133 s = scan_str(start,!!PL_madskills,FALSE);
9135 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9137 if (s[-1] == PL_multi_open)
9141 CURMAD('q', PL_thisopen);
9142 CURMAD('_', PL_thiswhite);
9143 CURMAD('E', PL_thisstuff);
9144 CURMAD('Q', PL_thisclose);
9145 PL_realtokenstart = s - SvPVX(PL_linestr);
9149 s = scan_str(s,!!PL_madskills,FALSE);
9152 SvREFCNT_dec(PL_lex_stuff);
9153 PL_lex_stuff = NULL;
9155 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9158 CURMAD('z', PL_thisopen);
9159 CURMAD('R', PL_thisstuff);
9160 CURMAD('Z', PL_thisclose);
9163 complement = del = squash = 0;
9170 complement = OPpTRANS_COMPLEMENT;
9173 del = OPpTRANS_DELETE;
9176 squash = OPpTRANS_SQUASH;
9188 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
9189 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)tbl);
9190 o->op_private &= ~OPpTRANS_ALL;
9191 o->op_private |= del|squash|complement|
9192 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9193 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
9196 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9201 curmad('m', newSVpvn(modstart, s - modstart));
9202 append_madprops(PL_thismad, o, 0);
9211 S_scan_heredoc(pTHX_ register char *s)
9215 I32 op_type = OP_SCALAR;
9219 const char *found_newline;
9223 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
9225 I32 stuffstart = s - SvPVX(PL_linestr);
9228 PL_realtokenstart = -1;
9231 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9235 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9239 while (SPACE_OR_TAB(*peek))
9241 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9244 s = delimcpy(d, e, s, PL_bufend, term, &len);
9254 if (!isALNUM_lazy_if(s,UTF))
9255 deprecate("bare << to mean <<\"\"");
9256 for (; isALNUM_lazy_if(s,UTF); s++) {
9261 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9262 Perl_croak(aTHX_ "Delimiter for here document is too long");
9265 len = d - PL_tokenbuf;
9269 tstart = PL_tokenbuf + !outer;
9270 PL_thisclose = newSVpvn(tstart, len - !outer);
9271 tstart = SvPVX(PL_linestr) + stuffstart;
9272 PL_thisopen = newSVpvn(tstart, s - tstart);
9273 stuffstart = s - SvPVX(PL_linestr);
9276 #ifndef PERL_STRICT_CR
9277 d = strchr(s, '\r');
9279 char * const olds = s;
9281 while (s < PL_bufend) {
9287 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9296 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9303 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
9304 herewas = newSVpvn(s,PL_bufend-s);
9308 herewas = newSVpvn(s-1,found_newline-s+1);
9311 herewas = newSVpvn(s,found_newline-s);
9316 tstart = SvPVX(PL_linestr) + stuffstart;
9318 sv_catpvn(PL_thisstuff, tstart, s - tstart);
9320 PL_thisstuff = newSVpvn(tstart, s - tstart);
9323 s += SvCUR(herewas);
9326 stuffstart = s - SvPVX(PL_linestr);
9332 tmpstr = newSV_type(SVt_PVIV);
9336 SvIV_set(tmpstr, -1);
9338 else if (term == '`') {
9339 op_type = OP_BACKTICK;
9340 SvIV_set(tmpstr, '\\');
9344 PL_multi_start = CopLINE(PL_curcop);
9345 PL_multi_open = PL_multi_close = '<';
9346 term = *PL_tokenbuf;
9347 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
9348 char * const bufptr = PL_sublex_info.super_bufptr;
9349 char * const bufend = PL_sublex_info.super_bufend;
9350 char * const olds = s - SvCUR(herewas);
9351 s = strchr(bufptr, '\n');
9355 while (s < bufend &&
9356 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9358 CopLINE_inc(PL_curcop);
9361 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9362 missingterm(PL_tokenbuf);
9364 sv_setpvn(herewas,bufptr,d-bufptr+1);
9365 sv_setpvn(tmpstr,d+1,s-d);
9367 sv_catpvn(herewas,s,bufend-s);
9368 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
9375 while (s < PL_bufend &&
9376 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9378 CopLINE_inc(PL_curcop);
9380 if (s >= PL_bufend) {
9381 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9382 missingterm(PL_tokenbuf);
9384 sv_setpvn(tmpstr,d+1,s-d);
9388 sv_catpvn(PL_thisstuff, d + 1, s - d);
9390 PL_thisstuff = newSVpvn(d + 1, s - d);
9391 stuffstart = s - SvPVX(PL_linestr);
9395 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
9397 sv_catpvn(herewas,s,PL_bufend-s);
9398 sv_setsv(PL_linestr,herewas);
9399 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
9400 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9401 PL_last_lop = PL_last_uni = NULL;
9404 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9405 while (s >= PL_bufend) { /* multiple line string? */
9408 tstart = SvPVX(PL_linestr) + stuffstart;
9410 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
9412 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
9416 CopLINE_inc(PL_curcop);
9417 if (!outer || !lex_next_chunk(0)) {
9418 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9419 missingterm(PL_tokenbuf);
9421 CopLINE_dec(PL_curcop);
9424 stuffstart = s - SvPVX(PL_linestr);
9426 CopLINE_inc(PL_curcop);
9427 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9428 PL_last_lop = PL_last_uni = NULL;
9429 #ifndef PERL_STRICT_CR
9430 if (PL_bufend - PL_linestart >= 2) {
9431 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
9432 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9434 PL_bufend[-2] = '\n';
9436 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9438 else if (PL_bufend[-1] == '\r')
9439 PL_bufend[-1] = '\n';
9441 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9442 PL_bufend[-1] = '\n';
9444 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
9445 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
9446 *(SvPVX(PL_linestr) + off ) = ' ';
9447 lex_grow_linestr(SvCUR(PL_linestr) + SvCUR(herewas) + 1);
9448 sv_catsv(PL_linestr,herewas);
9449 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9450 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
9454 sv_catsv(tmpstr,PL_linestr);
9459 PL_multi_end = CopLINE(PL_curcop);
9460 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9461 SvPV_shrink_to_cur(tmpstr);
9463 SvREFCNT_dec(herewas);
9465 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9467 else if (PL_encoding)
9468 sv_recode_to_utf8(tmpstr, PL_encoding);
9470 PL_lex_stuff = tmpstr;
9471 pl_yylval.ival = op_type;
9476 takes: current position in input buffer
9477 returns: new position in input buffer
9478 side-effects: pl_yylval and lex_op are set.
9483 <FH> read from filehandle
9484 <pkg::FH> read from package qualified filehandle
9485 <pkg'FH> read from package qualified filehandle
9486 <$fh> read from filehandle in $fh
9492 S_scan_inputsymbol(pTHX_ char *start)
9495 register char *s = start; /* current position in buffer */
9498 char *d = PL_tokenbuf; /* start of temp holding space */
9499 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9501 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
9503 end = strchr(s, '\n');
9506 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9508 /* die if we didn't have space for the contents of the <>,
9509 or if it didn't end, or if we see a newline
9512 if (len >= (I32)sizeof PL_tokenbuf)
9513 Perl_croak(aTHX_ "Excessively long <> operator");
9515 Perl_croak(aTHX_ "Unterminated <> operator");
9520 Remember, only scalar variables are interpreted as filehandles by
9521 this code. Anything more complex (e.g., <$fh{$num}>) will be
9522 treated as a glob() call.
9523 This code makes use of the fact that except for the $ at the front,
9524 a scalar variable and a filehandle look the same.
9526 if (*d == '$' && d[1]) d++;
9528 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9529 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9532 /* If we've tried to read what we allow filehandles to look like, and
9533 there's still text left, then it must be a glob() and not a getline.
9534 Use scan_str to pull out the stuff between the <> and treat it
9535 as nothing more than a string.
9538 if (d - PL_tokenbuf != len) {
9539 pl_yylval.ival = OP_GLOB;
9540 s = scan_str(start,!!PL_madskills,FALSE);
9542 Perl_croak(aTHX_ "Glob not terminated");
9546 bool readline_overriden = FALSE;
9549 /* we're in a filehandle read situation */
9552 /* turn <> into <ARGV> */
9554 Copy("ARGV",d,5,char);
9556 /* Check whether readline() is overriden */
9557 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
9559 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
9561 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
9562 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
9563 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
9564 readline_overriden = TRUE;
9566 /* if <$fh>, create the ops to turn the variable into a
9570 /* try to find it in the pad for this block, otherwise find
9571 add symbol table ops
9573 const PADOFFSET tmp = pad_findmy_pvn(d, len, UTF ? SVf_UTF8 : 0);
9574 if (tmp != NOT_IN_PAD) {
9575 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
9576 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
9577 HEK * const stashname = HvNAME_HEK(stash);
9578 SV * const sym = sv_2mortal(newSVhek(stashname));
9579 sv_catpvs(sym, "::");
9585 OP * const o = newOP(OP_PADSV, 0);
9587 PL_lex_op = readline_overriden
9588 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9589 op_append_elem(OP_LIST, o,
9590 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
9591 : (OP*)newUNOP(OP_READLINE, 0, o);
9600 ? (GV_ADDMULTI | GV_ADDINEVAL)
9603 PL_lex_op = readline_overriden
9604 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9605 op_append_elem(OP_LIST,
9606 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
9607 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9608 : (OP*)newUNOP(OP_READLINE, 0,
9609 newUNOP(OP_RV2SV, 0,
9610 newGVOP(OP_GV, 0, gv)));
9612 if (!readline_overriden)
9613 PL_lex_op->op_flags |= OPf_SPECIAL;
9614 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
9615 pl_yylval.ival = OP_NULL;
9618 /* If it's none of the above, it must be a literal filehandle
9619 (<Foo::BAR> or <FOO>) so build a simple readline OP */
9621 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
9622 PL_lex_op = readline_overriden
9623 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9624 op_append_elem(OP_LIST,
9625 newGVOP(OP_GV, 0, gv),
9626 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9627 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
9628 pl_yylval.ival = OP_NULL;
9637 takes: start position in buffer
9638 keep_quoted preserve \ on the embedded delimiter(s)
9639 keep_delims preserve the delimiters around the string
9640 returns: position to continue reading from buffer
9641 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
9642 updates the read buffer.
9644 This subroutine pulls a string out of the input. It is called for:
9645 q single quotes q(literal text)
9646 ' single quotes 'literal text'
9647 qq double quotes qq(interpolate $here please)
9648 " double quotes "interpolate $here please"
9649 qx backticks qx(/bin/ls -l)
9650 ` backticks `/bin/ls -l`
9651 qw quote words @EXPORT_OK = qw( func() $spam )
9652 m// regexp match m/this/
9653 s/// regexp substitute s/this/that/
9654 tr/// string transliterate tr/this/that/
9655 y/// string transliterate y/this/that/
9656 ($*@) sub prototypes sub foo ($)
9657 (stuff) sub attr parameters sub foo : attr(stuff)
9658 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
9660 In most of these cases (all but <>, patterns and transliterate)
9661 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
9662 calls scan_str(). s/// makes yylex() call scan_subst() which calls
9663 scan_str(). tr/// and y/// make yylex() call scan_trans() which
9666 It skips whitespace before the string starts, and treats the first
9667 character as the delimiter. If the delimiter is one of ([{< then
9668 the corresponding "close" character )]}> is used as the closing
9669 delimiter. It allows quoting of delimiters, and if the string has
9670 balanced delimiters ([{<>}]) it allows nesting.
9672 On success, the SV with the resulting string is put into lex_stuff or,
9673 if that is already non-NULL, into lex_repl. The second case occurs only
9674 when parsing the RHS of the special constructs s/// and tr/// (y///).
9675 For convenience, the terminating delimiter character is stuffed into
9680 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
9683 SV *sv; /* scalar value: string */
9684 const char *tmps; /* temp string, used for delimiter matching */
9685 register char *s = start; /* current position in the buffer */
9686 register char term; /* terminating character */
9687 register char *to; /* current position in the sv's data */
9688 I32 brackets = 1; /* bracket nesting level */
9689 bool has_utf8 = FALSE; /* is there any utf8 content? */
9690 I32 termcode; /* terminating char. code */
9691 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
9692 STRLEN termlen; /* length of terminating string */
9693 int last_off = 0; /* last position for nesting bracket */
9699 PERL_ARGS_ASSERT_SCAN_STR;
9701 /* skip space before the delimiter */
9707 if (PL_realtokenstart >= 0) {
9708 stuffstart = PL_realtokenstart;
9709 PL_realtokenstart = -1;
9712 stuffstart = start - SvPVX(PL_linestr);
9714 /* mark where we are, in case we need to report errors */
9717 /* after skipping whitespace, the next character is the terminator */
9720 termcode = termstr[0] = term;
9724 termcode = utf8_to_uvchr((U8*)s, &termlen);
9725 Copy(s, termstr, termlen, U8);
9726 if (!UTF8_IS_INVARIANT(term))
9730 /* mark where we are */
9731 PL_multi_start = CopLINE(PL_curcop);
9732 PL_multi_open = term;
9734 /* find corresponding closing delimiter */
9735 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
9736 termcode = termstr[0] = term = tmps[5];
9738 PL_multi_close = term;
9740 /* create a new SV to hold the contents. 79 is the SV's initial length.
9741 What a random number. */
9742 sv = newSV_type(SVt_PVIV);
9744 SvIV_set(sv, termcode);
9745 (void)SvPOK_only(sv); /* validate pointer */
9747 /* move past delimiter and try to read a complete string */
9749 sv_catpvn(sv, s, termlen);
9752 tstart = SvPVX(PL_linestr) + stuffstart;
9753 if (!PL_thisopen && !keep_delims) {
9754 PL_thisopen = newSVpvn(tstart, s - tstart);
9755 stuffstart = s - SvPVX(PL_linestr);
9759 if (PL_encoding && !UTF) {
9763 int offset = s - SvPVX_const(PL_linestr);
9764 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
9765 &offset, (char*)termstr, termlen);
9766 const char * const ns = SvPVX_const(PL_linestr) + offset;
9767 char * const svlast = SvEND(sv) - 1;
9769 for (; s < ns; s++) {
9770 if (*s == '\n' && !PL_rsfp)
9771 CopLINE_inc(PL_curcop);
9774 goto read_more_line;
9776 /* handle quoted delimiters */
9777 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
9779 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
9781 if ((svlast-1 - t) % 2) {
9785 SvCUR_set(sv, SvCUR(sv) - 1);
9790 if (PL_multi_open == PL_multi_close) {
9796 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
9797 /* At here, all closes are "was quoted" one,
9798 so we don't check PL_multi_close. */
9800 if (!keep_quoted && *(t+1) == PL_multi_open)
9805 else if (*t == PL_multi_open)
9813 SvCUR_set(sv, w - SvPVX_const(sv));
9815 last_off = w - SvPVX(sv);
9816 if (--brackets <= 0)
9822 SvCUR_set(sv, SvCUR(sv) - 1);
9828 /* extend sv if need be */
9829 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
9830 /* set 'to' to the next character in the sv's string */
9831 to = SvPVX(sv)+SvCUR(sv);
9833 /* if open delimiter is the close delimiter read unbridle */
9834 if (PL_multi_open == PL_multi_close) {
9835 for (; s < PL_bufend; s++,to++) {
9836 /* embedded newlines increment the current line number */
9837 if (*s == '\n' && !PL_rsfp)
9838 CopLINE_inc(PL_curcop);
9839 /* handle quoted delimiters */
9840 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
9841 if (!keep_quoted && s[1] == term)
9843 /* any other quotes are simply copied straight through */
9847 /* terminate when run out of buffer (the for() condition), or
9848 have found the terminator */
9849 else if (*s == term) {
9852 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
9855 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9861 /* if the terminator isn't the same as the start character (e.g.,
9862 matched brackets), we have to allow more in the quoting, and
9863 be prepared for nested brackets.
9866 /* read until we run out of string, or we find the terminator */
9867 for (; s < PL_bufend; s++,to++) {
9868 /* embedded newlines increment the line count */
9869 if (*s == '\n' && !PL_rsfp)
9870 CopLINE_inc(PL_curcop);
9871 /* backslashes can escape the open or closing characters */
9872 if (*s == '\\' && s+1 < PL_bufend) {
9874 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
9879 /* allow nested opens and closes */
9880 else if (*s == PL_multi_close && --brackets <= 0)
9882 else if (*s == PL_multi_open)
9884 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9889 /* terminate the copied string and update the sv's end-of-string */
9891 SvCUR_set(sv, to - SvPVX_const(sv));
9894 * this next chunk reads more into the buffer if we're not done yet
9898 break; /* handle case where we are done yet :-) */
9900 #ifndef PERL_STRICT_CR
9901 if (to - SvPVX_const(sv) >= 2) {
9902 if ((to[-2] == '\r' && to[-1] == '\n') ||
9903 (to[-2] == '\n' && to[-1] == '\r'))
9907 SvCUR_set(sv, to - SvPVX_const(sv));
9909 else if (to[-1] == '\r')
9912 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
9917 /* if we're out of file, or a read fails, bail and reset the current
9918 line marker so we can report where the unterminated string began
9922 char * const tstart = SvPVX(PL_linestr) + stuffstart;
9924 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
9926 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
9929 CopLINE_inc(PL_curcop);
9930 PL_bufptr = PL_bufend;
9931 if (!lex_next_chunk(0)) {
9933 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9942 /* at this point, we have successfully read the delimited string */
9944 if (!PL_encoding || UTF) {
9947 char * const tstart = SvPVX(PL_linestr) + stuffstart;
9948 const int len = s - tstart;
9950 sv_catpvn(PL_thisstuff, tstart, len);
9952 PL_thisstuff = newSVpvn(tstart, len);
9953 if (!PL_thisclose && !keep_delims)
9954 PL_thisclose = newSVpvn(s,termlen);
9959 sv_catpvn(sv, s, termlen);
9965 char * const tstart = SvPVX(PL_linestr) + stuffstart;
9966 const int len = s - tstart - termlen;
9968 sv_catpvn(PL_thisstuff, tstart, len);
9970 PL_thisstuff = newSVpvn(tstart, len);
9971 if (!PL_thisclose && !keep_delims)
9972 PL_thisclose = newSVpvn(s - termlen,termlen);
9976 if (has_utf8 || PL_encoding)
9979 PL_multi_end = CopLINE(PL_curcop);
9981 /* if we allocated too much space, give some back */
9982 if (SvCUR(sv) + 5 < SvLEN(sv)) {
9983 SvLEN_set(sv, SvCUR(sv) + 1);
9984 SvPV_renew(sv, SvLEN(sv));
9987 /* decide whether this is the first or second quoted string we've read
10000 takes: pointer to position in buffer
10001 returns: pointer to new position in buffer
10002 side-effects: builds ops for the constant in pl_yylval.op
10004 Read a number in any of the formats that Perl accepts:
10006 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
10007 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
10010 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
10012 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
10015 If it reads a number without a decimal point or an exponent, it will
10016 try converting the number to an integer and see if it can do so
10017 without loss of precision.
10021 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10024 register const char *s = start; /* current position in buffer */
10025 register char *d; /* destination in temp buffer */
10026 register char *e; /* end of temp buffer */
10027 NV nv; /* number read, as a double */
10028 SV *sv = NULL; /* place to put the converted number */
10029 bool floatit; /* boolean: int or float? */
10030 const char *lastub = NULL; /* position of last underbar */
10031 static char const number_too_long[] = "Number too long";
10033 PERL_ARGS_ASSERT_SCAN_NUM;
10035 /* We use the first character to decide what type of number this is */
10039 Perl_croak(aTHX_ "panic: scan_num");
10041 /* if it starts with a 0, it could be an octal number, a decimal in
10042 0.13 disguise, or a hexadecimal number, or a binary number. */
10046 u holds the "number so far"
10047 shift the power of 2 of the base
10048 (hex == 4, octal == 3, binary == 1)
10049 overflowed was the number more than we can hold?
10051 Shift is used when we add a digit. It also serves as an "are
10052 we in octal/hex/binary?" indicator to disallow hex characters
10053 when in octal mode.
10058 bool overflowed = FALSE;
10059 bool just_zero = TRUE; /* just plain 0 or binary number? */
10060 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10061 static const char* const bases[5] =
10062 { "", "binary", "", "octal", "hexadecimal" };
10063 static const char* const Bases[5] =
10064 { "", "Binary", "", "Octal", "Hexadecimal" };
10065 static const char* const maxima[5] =
10067 "0b11111111111111111111111111111111",
10071 const char *base, *Base, *max;
10073 /* check for hex */
10074 if (s[1] == 'x' || s[1] == 'X') {
10078 } else if (s[1] == 'b' || s[1] == 'B') {
10083 /* check for a decimal in disguise */
10084 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
10086 /* so it must be octal */
10093 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10094 "Misplaced _ in number");
10098 base = bases[shift];
10099 Base = Bases[shift];
10100 max = maxima[shift];
10102 /* read the rest of the number */
10104 /* x is used in the overflow test,
10105 b is the digit we're adding on. */
10110 /* if we don't mention it, we're done */
10114 /* _ are ignored -- but warned about if consecutive */
10116 if (lastub && s == lastub + 1)
10117 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10118 "Misplaced _ in number");
10122 /* 8 and 9 are not octal */
10123 case '8': case '9':
10125 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10129 case '2': case '3': case '4':
10130 case '5': case '6': case '7':
10132 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10135 case '0': case '1':
10136 b = *s++ & 15; /* ASCII digit -> value of digit */
10140 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10141 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10142 /* make sure they said 0x */
10145 b = (*s++ & 7) + 9;
10147 /* Prepare to put the digit we have onto the end
10148 of the number so far. We check for overflows.
10154 x = u << shift; /* make room for the digit */
10156 if ((x >> shift) != u
10157 && !(PL_hints & HINT_NEW_BINARY)) {
10160 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10161 "Integer overflow in %s number",
10164 u = x | b; /* add the digit to the end */
10167 n *= nvshift[shift];
10168 /* If an NV has not enough bits in its
10169 * mantissa to represent an UV this summing of
10170 * small low-order numbers is a waste of time
10171 * (because the NV cannot preserve the
10172 * low-order bits anyway): we could just
10173 * remember when did we overflow and in the
10174 * end just multiply n by the right
10182 /* if we get here, we had success: make a scalar value from
10187 /* final misplaced underbar check */
10188 if (s[-1] == '_') {
10189 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10193 if (n > 4294967295.0)
10194 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10195 "%s number > %s non-portable",
10201 if (u > 0xffffffff)
10202 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10203 "%s number > %s non-portable",
10208 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10209 sv = new_constant(start, s - start, "integer",
10210 sv, NULL, NULL, 0);
10211 else if (PL_hints & HINT_NEW_BINARY)
10212 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10217 handle decimal numbers.
10218 we're also sent here when we read a 0 as the first digit
10220 case '1': case '2': case '3': case '4': case '5':
10221 case '6': case '7': case '8': case '9': case '.':
10224 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10227 /* read next group of digits and _ and copy into d */
10228 while (isDIGIT(*s) || *s == '_') {
10229 /* skip underscores, checking for misplaced ones
10233 if (lastub && s == lastub + 1)
10234 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10235 "Misplaced _ in number");
10239 /* check for end of fixed-length buffer */
10241 Perl_croak(aTHX_ number_too_long);
10242 /* if we're ok, copy the character */
10247 /* final misplaced underbar check */
10248 if (lastub && s == lastub + 1) {
10249 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10252 /* read a decimal portion if there is one. avoid
10253 3..5 being interpreted as the number 3. followed
10256 if (*s == '.' && s[1] != '.') {
10261 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10262 "Misplaced _ in number");
10266 /* copy, ignoring underbars, until we run out of digits.
10268 for (; isDIGIT(*s) || *s == '_'; s++) {
10269 /* fixed length buffer check */
10271 Perl_croak(aTHX_ number_too_long);
10273 if (lastub && s == lastub + 1)
10274 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10275 "Misplaced _ in number");
10281 /* fractional part ending in underbar? */
10282 if (s[-1] == '_') {
10283 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10284 "Misplaced _ in number");
10286 if (*s == '.' && isDIGIT(s[1])) {
10287 /* oops, it's really a v-string, but without the "v" */
10293 /* read exponent part, if present */
10294 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10298 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10299 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10301 /* stray preinitial _ */
10303 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10304 "Misplaced _ in number");
10308 /* allow positive or negative exponent */
10309 if (*s == '+' || *s == '-')
10312 /* stray initial _ */
10314 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10315 "Misplaced _ in number");
10319 /* read digits of exponent */
10320 while (isDIGIT(*s) || *s == '_') {
10323 Perl_croak(aTHX_ number_too_long);
10327 if (((lastub && s == lastub + 1) ||
10328 (!isDIGIT(s[1]) && s[1] != '_')))
10329 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10330 "Misplaced _ in number");
10338 We try to do an integer conversion first if no characters
10339 indicating "float" have been found.
10344 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10346 if (flags == IS_NUMBER_IN_UV) {
10348 sv = newSViv(uv); /* Prefer IVs over UVs. */
10351 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10352 if (uv <= (UV) IV_MIN)
10353 sv = newSViv(-(IV)uv);
10360 /* terminate the string */
10362 nv = Atof(PL_tokenbuf);
10367 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10368 const char *const key = floatit ? "float" : "integer";
10369 const STRLEN keylen = floatit ? 5 : 7;
10370 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10371 key, keylen, sv, NULL, NULL, 0);
10375 /* if it starts with a v, it could be a v-string */
10378 sv = newSV(5); /* preallocate storage space */
10379 s = scan_vstring(s, PL_bufend, sv);
10383 /* make the op for the constant and return */
10386 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10388 lvalp->opval = NULL;
10394 S_scan_formline(pTHX_ register char *s)
10397 register char *eol;
10399 SV * const stuff = newSVpvs("");
10400 bool needargs = FALSE;
10401 bool eofmt = FALSE;
10403 char *tokenstart = s;
10404 SV* savewhite = NULL;
10406 if (PL_madskills) {
10407 savewhite = PL_thiswhite;
10412 PERL_ARGS_ASSERT_SCAN_FORMLINE;
10414 while (!needargs) {
10417 #ifdef PERL_STRICT_CR
10418 while (SPACE_OR_TAB(*t))
10421 while (SPACE_OR_TAB(*t) || *t == '\r')
10424 if (*t == '\n' || t == PL_bufend) {
10429 if (PL_in_eval && !PL_rsfp) {
10430 eol = (char *) memchr(s,'\n',PL_bufend-s);
10435 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10437 for (t = s; t < eol; t++) {
10438 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10440 goto enough; /* ~~ must be first line in formline */
10442 if (*t == '@' || *t == '^')
10446 sv_catpvn(stuff, s, eol-s);
10447 #ifndef PERL_STRICT_CR
10448 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10449 char *end = SvPVX(stuff) + SvCUR(stuff);
10452 SvCUR_set(stuff, SvCUR(stuff) - 1);
10463 if (PL_madskills) {
10465 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
10467 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
10470 PL_bufptr = PL_bufend;
10471 CopLINE_inc(PL_curcop);
10472 got_some = lex_next_chunk(0);
10473 CopLINE_dec(PL_curcop);
10476 tokenstart = PL_bufptr;
10484 if (SvCUR(stuff)) {
10487 PL_lex_state = LEX_NORMAL;
10488 start_force(PL_curforce);
10489 NEXTVAL_NEXTTOKE.ival = 0;
10493 PL_lex_state = LEX_FORMLINE;
10495 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10497 else if (PL_encoding)
10498 sv_recode_to_utf8(stuff, PL_encoding);
10500 start_force(PL_curforce);
10501 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10503 start_force(PL_curforce);
10504 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
10508 SvREFCNT_dec(stuff);
10510 PL_lex_formbrack = 0;
10514 if (PL_madskills) {
10516 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
10518 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
10519 PL_thiswhite = savewhite;
10526 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10529 const I32 oldsavestack_ix = PL_savestack_ix;
10530 CV* const outsidecv = PL_compcv;
10533 assert(SvTYPE(PL_compcv) == SVt_PVCV);
10535 SAVEI32(PL_subline);
10536 save_item(PL_subname);
10537 SAVESPTR(PL_compcv);
10539 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
10540 CvFLAGS(PL_compcv) |= flags;
10542 PL_subline = CopLINE(PL_curcop);
10543 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10544 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
10545 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10547 return oldsavestack_ix;
10551 #pragma segment Perl_yylex
10554 S_yywarn(pTHX_ const char *const s)
10558 PERL_ARGS_ASSERT_YYWARN;
10560 PL_in_eval |= EVAL_WARNONLY;
10562 PL_in_eval &= ~EVAL_WARNONLY;
10567 Perl_yyerror(pTHX_ const char *const s)
10570 const char *where = NULL;
10571 const char *context = NULL;
10574 int yychar = PL_parser->yychar;
10576 PERL_ARGS_ASSERT_YYERROR;
10578 if (!yychar || (yychar == ';' && !PL_rsfp))
10580 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
10581 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
10582 PL_oldbufptr != PL_bufptr) {
10585 The code below is removed for NetWare because it abends/crashes on NetWare
10586 when the script has error such as not having the closing quotes like:
10587 if ($var eq "value)
10588 Checking of white spaces is anyway done in NetWare code.
10591 while (isSPACE(*PL_oldoldbufptr))
10594 context = PL_oldoldbufptr;
10595 contlen = PL_bufptr - PL_oldoldbufptr;
10597 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
10598 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
10601 The code below is removed for NetWare because it abends/crashes on NetWare
10602 when the script has error such as not having the closing quotes like:
10603 if ($var eq "value)
10604 Checking of white spaces is anyway done in NetWare code.
10607 while (isSPACE(*PL_oldbufptr))
10610 context = PL_oldbufptr;
10611 contlen = PL_bufptr - PL_oldbufptr;
10613 else if (yychar > 255)
10614 where = "next token ???";
10615 else if (yychar == -2) { /* YYEMPTY */
10616 if (PL_lex_state == LEX_NORMAL ||
10617 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
10618 where = "at end of line";
10619 else if (PL_lex_inpat)
10620 where = "within pattern";
10622 where = "within string";
10625 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
10627 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
10628 else if (isPRINT_LC(yychar)) {
10629 const char string = yychar;
10630 sv_catpvn(where_sv, &string, 1);
10633 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
10634 where = SvPVX_const(where_sv);
10636 msg = sv_2mortal(newSVpv(s, 0));
10637 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
10638 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
10640 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
10642 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
10643 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
10644 Perl_sv_catpvf(aTHX_ msg,
10645 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
10646 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
10649 if (PL_in_eval & EVAL_WARNONLY) {
10650 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
10654 if (PL_error_count >= 10) {
10655 if (PL_in_eval && SvCUR(ERRSV))
10656 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
10657 SVfARG(ERRSV), OutCopFILE(PL_curcop));
10659 Perl_croak(aTHX_ "%s has too many errors.\n",
10660 OutCopFILE(PL_curcop));
10663 PL_in_my_stash = NULL;
10667 #pragma segment Main
10671 S_swallow_bom(pTHX_ U8 *s)
10674 const STRLEN slen = SvCUR(PL_linestr);
10676 PERL_ARGS_ASSERT_SWALLOW_BOM;
10680 if (s[1] == 0xFE) {
10681 /* UTF-16 little-endian? (or UTF-32LE?) */
10682 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
10683 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
10684 #ifndef PERL_NO_UTF16_FILTER
10685 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
10687 if (PL_bufend > (char*)s) {
10688 s = add_utf16_textfilter(s, TRUE);
10691 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10696 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
10697 #ifndef PERL_NO_UTF16_FILTER
10698 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
10700 if (PL_bufend > (char *)s) {
10701 s = add_utf16_textfilter(s, FALSE);
10704 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10709 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
10710 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10711 s += 3; /* UTF-8 */
10717 if (s[2] == 0xFE && s[3] == 0xFF) {
10718 /* UTF-32 big-endian */
10719 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
10722 else if (s[2] == 0 && s[3] != 0) {
10725 * are a good indicator of UTF-16BE. */
10726 #ifndef PERL_NO_UTF16_FILTER
10727 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
10728 s = add_utf16_textfilter(s, FALSE);
10730 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10736 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
10737 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10738 s += 4; /* UTF-8 */
10744 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
10747 * are a good indicator of UTF-16LE. */
10748 #ifndef PERL_NO_UTF16_FILTER
10749 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
10750 s = add_utf16_textfilter(s, TRUE);
10752 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10760 #ifndef PERL_NO_UTF16_FILTER
10762 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
10765 SV *const filter = FILTER_DATA(idx);
10766 /* We re-use this each time round, throwing the contents away before we
10768 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
10769 SV *const utf8_buffer = filter;
10770 IV status = IoPAGE(filter);
10771 const bool reverse = cBOOL(IoLINES(filter));
10774 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
10776 /* As we're automatically added, at the lowest level, and hence only called
10777 from this file, we can be sure that we're not called in block mode. Hence
10778 don't bother writing code to deal with block mode. */
10780 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
10783 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
10785 DEBUG_P(PerlIO_printf(Perl_debug_log,
10786 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10787 FPTR2DPTR(void *, S_utf16_textfilter),
10788 reverse ? 'l' : 'b', idx, maxlen, status,
10789 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10796 /* First, look in our buffer of existing UTF-8 data: */
10797 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
10801 } else if (status == 0) {
10803 IoPAGE(filter) = 0;
10804 nl = SvEND(utf8_buffer);
10807 STRLEN got = nl - SvPVX(utf8_buffer);
10808 /* Did we have anything to append? */
10810 sv_catpvn(sv, SvPVX(utf8_buffer), got);
10811 /* Everything else in this code works just fine if SVp_POK isn't
10812 set. This, however, needs it, and we need it to work, else
10813 we loop infinitely because the buffer is never consumed. */
10814 sv_chop(utf8_buffer, nl);
10818 /* OK, not a complete line there, so need to read some more UTF-16.
10819 Read an extra octect if the buffer currently has an odd number. */
10823 if (SvCUR(utf16_buffer) >= 2) {
10824 /* Location of the high octet of the last complete code point.
10825 Gosh, UTF-16 is a pain. All the benefits of variable length,
10826 *coupled* with all the benefits of partial reads and
10828 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
10829 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
10831 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
10835 /* We have the first half of a surrogate. Read more. */
10836 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
10839 status = FILTER_READ(idx + 1, utf16_buffer,
10840 160 + (SvCUR(utf16_buffer) & 1));
10841 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
10842 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
10845 IoPAGE(filter) = status;
10850 chars = SvCUR(utf16_buffer) >> 1;
10851 have = SvCUR(utf8_buffer);
10852 SvGROW(utf8_buffer, have + chars * 3 + 1);
10855 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
10856 (U8*)SvPVX_const(utf8_buffer) + have,
10857 chars * 2, &newlen);
10859 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
10860 (U8*)SvPVX_const(utf8_buffer) + have,
10861 chars * 2, &newlen);
10863 SvCUR_set(utf8_buffer, have + newlen);
10866 /* No need to keep this SV "well-formed" with a '\0' after the end, as
10867 it's private to us, and utf16_to_utf8{,reversed} take a
10868 (pointer,length) pair, rather than a NUL-terminated string. */
10869 if(SvCUR(utf16_buffer) & 1) {
10870 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
10871 SvCUR_set(utf16_buffer, 1);
10873 SvCUR_set(utf16_buffer, 0);
10876 DEBUG_P(PerlIO_printf(Perl_debug_log,
10877 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10879 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10880 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
10885 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
10887 SV *filter = filter_add(S_utf16_textfilter, NULL);
10889 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
10891 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
10892 sv_setpvs(filter, "");
10893 IoLINES(filter) = reversed;
10894 IoPAGE(filter) = 1; /* Not EOF */
10896 /* Sadly, we have to return a valid pointer, come what may, so we have to
10897 ignore any error return from this. */
10898 SvCUR_set(PL_linestr, 0);
10899 if (FILTER_READ(0, PL_linestr, 0)) {
10900 SvUTF8_on(PL_linestr);
10902 SvUTF8_on(PL_linestr);
10904 PL_bufend = SvEND(PL_linestr);
10905 return (U8*)SvPVX(PL_linestr);
10910 Returns a pointer to the next character after the parsed
10911 vstring, as well as updating the passed in sv.
10913 Function must be called like
10916 s = scan_vstring(s,e,sv);
10918 where s and e are the start and end of the string.
10919 The sv should already be large enough to store the vstring
10920 passed in, for performance reasons.
10925 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
10928 const char *pos = s;
10929 const char *start = s;
10931 PERL_ARGS_ASSERT_SCAN_VSTRING;
10933 if (*pos == 'v') pos++; /* get past 'v' */
10934 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
10936 if ( *pos != '.') {
10937 /* this may not be a v-string if followed by => */
10938 const char *next = pos;
10939 while (next < e && isSPACE(*next))
10941 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
10942 /* return string not v-string */
10943 sv_setpvn(sv,(char *)s,pos-s);
10944 return (char *)pos;
10948 if (!isALPHA(*pos)) {
10949 U8 tmpbuf[UTF8_MAXBYTES+1];
10952 s++; /* get past 'v' */
10957 /* this is atoi() that tolerates underscores */
10960 const char *end = pos;
10962 while (--end >= s) {
10964 const UV orev = rev;
10965 rev += (*end - '0') * mult;
10968 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10969 "Integer overflow in decimal number");
10973 if (rev > 0x7FFFFFFF)
10974 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
10976 /* Append native character for the rev point */
10977 tmpend = uvchr_to_utf8(tmpbuf, rev);
10978 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
10979 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
10981 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
10987 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
10991 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
10998 Perl_keyword_plugin_standard(pTHX_
10999 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
11001 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
11002 PERL_UNUSED_CONTEXT;
11003 PERL_UNUSED_ARG(keyword_ptr);
11004 PERL_UNUSED_ARG(keyword_len);
11005 PERL_UNUSED_ARG(op_ptr);
11006 return KEYWORD_PLUGIN_DECLINE;
11009 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
11011 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
11013 SAVEI32(PL_lex_brackets);
11014 if (PL_lex_brackets > 100)
11015 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
11016 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
11017 SAVEI32(PL_lex_allbrackets);
11018 PL_lex_allbrackets = 0;
11019 SAVEI8(PL_lex_fakeeof);
11020 PL_lex_fakeeof = (U8)fakeeof;
11021 if(yyparse(gramtype) && !PL_parser->error_count)
11022 qerror(Perl_mess(aTHX_ "Parse error"));
11025 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11027 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11031 SAVEVPTR(PL_eval_root);
11032 PL_eval_root = NULL;
11033 parse_recdescent(gramtype, fakeeof);
11039 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11041 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11044 if (flags & ~PARSE_OPTIONAL)
11045 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11046 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11047 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11048 if (!PL_parser->error_count)
11049 qerror(Perl_mess(aTHX_ "Parse error"));
11050 exprop = newOP(OP_NULL, 0);
11056 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11058 Parse a Perl arithmetic expression. This may contain operators of precedence
11059 down to the bit shift operators. The expression must be followed (and thus
11060 terminated) either by a comparison or lower-precedence operator or by
11061 something that would normally terminate an expression such as semicolon.
11062 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11063 otherwise it is mandatory. It is up to the caller to ensure that the
11064 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11065 the source of the code to be parsed and the lexical context for the
11068 The op tree representing the expression is returned. If an optional
11069 expression is absent, a null pointer is returned, otherwise the pointer
11072 If an error occurs in parsing or compilation, in most cases a valid op
11073 tree is returned anyway. The error is reflected in the parser state,
11074 normally resulting in a single exception at the top level of parsing
11075 which covers all the compilation errors that occurred. Some compilation
11076 errors, however, will throw an exception immediately.
11082 Perl_parse_arithexpr(pTHX_ U32 flags)
11084 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11088 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11090 Parse a Perl term expression. This may contain operators of precedence
11091 down to the assignment operators. The expression must be followed (and thus
11092 terminated) either by a comma or lower-precedence operator or by
11093 something that would normally terminate an expression such as semicolon.
11094 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11095 otherwise it is mandatory. It is up to the caller to ensure that the
11096 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11097 the source of the code to be parsed and the lexical context for the
11100 The op tree representing the expression is returned. If an optional
11101 expression is absent, a null pointer is returned, otherwise the pointer
11104 If an error occurs in parsing or compilation, in most cases a valid op
11105 tree is returned anyway. The error is reflected in the parser state,
11106 normally resulting in a single exception at the top level of parsing
11107 which covers all the compilation errors that occurred. Some compilation
11108 errors, however, will throw an exception immediately.
11114 Perl_parse_termexpr(pTHX_ U32 flags)
11116 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11120 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11122 Parse a Perl list expression. This may contain operators of precedence
11123 down to the comma operator. The expression must be followed (and thus
11124 terminated) either by a low-precedence logic operator such as C<or> or by
11125 something that would normally terminate an expression such as semicolon.
11126 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11127 otherwise it is mandatory. It is up to the caller to ensure that the
11128 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11129 the source of the code to be parsed and the lexical context for the
11132 The op tree representing the expression is returned. If an optional
11133 expression is absent, a null pointer is returned, otherwise the pointer
11136 If an error occurs in parsing or compilation, in most cases a valid op
11137 tree is returned anyway. The error is reflected in the parser state,
11138 normally resulting in a single exception at the top level of parsing
11139 which covers all the compilation errors that occurred. Some compilation
11140 errors, however, will throw an exception immediately.
11146 Perl_parse_listexpr(pTHX_ U32 flags)
11148 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11152 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11154 Parse a single complete Perl expression. This allows the full
11155 expression grammar, including the lowest-precedence operators such
11156 as C<or>. The expression must be followed (and thus terminated) by a
11157 token that an expression would normally be terminated by: end-of-file,
11158 closing bracketing punctuation, semicolon, or one of the keywords that
11159 signals a postfix expression-statement modifier. If I<flags> includes
11160 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11161 mandatory. It is up to the caller to ensure that the dynamic parser
11162 state (L</PL_parser> et al) is correctly set to reflect the source of
11163 the code to be parsed and the lexical context for the expression.
11165 The op tree representing the expression is returned. If an optional
11166 expression is absent, a null pointer is returned, otherwise the pointer
11169 If an error occurs in parsing or compilation, in most cases a valid op
11170 tree is returned anyway. The error is reflected in the parser state,
11171 normally resulting in a single exception at the top level of parsing
11172 which covers all the compilation errors that occurred. Some compilation
11173 errors, however, will throw an exception immediately.
11179 Perl_parse_fullexpr(pTHX_ U32 flags)
11181 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11185 =for apidoc Amx|OP *|parse_block|U32 flags
11187 Parse a single complete Perl code block. This consists of an opening
11188 brace, a sequence of statements, and a closing brace. The block
11189 constitutes a lexical scope, so C<my> variables and various compile-time
11190 effects can be contained within it. It is up to the caller to ensure
11191 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11192 reflect the source of the code to be parsed and the lexical context for
11195 The op tree representing the code block is returned. This is always a
11196 real op, never a null pointer. It will normally be a C<lineseq> list,
11197 including C<nextstate> or equivalent ops. No ops to construct any kind
11198 of runtime scope are included by virtue of it being a block.
11200 If an error occurs in parsing or compilation, in most cases a valid op
11201 tree (most likely null) is returned anyway. The error is reflected in
11202 the parser state, normally resulting in a single exception at the top
11203 level of parsing which covers all the compilation errors that occurred.
11204 Some compilation errors, however, will throw an exception immediately.
11206 The I<flags> parameter is reserved for future use, and must always
11213 Perl_parse_block(pTHX_ U32 flags)
11216 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11217 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11221 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11223 Parse a single unadorned Perl statement. This may be a normal imperative
11224 statement or a declaration that has compile-time effect. It does not
11225 include any label or other affixture. It is up to the caller to ensure
11226 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11227 reflect the source of the code to be parsed and the lexical context for
11230 The op tree representing the statement is returned. This may be a
11231 null pointer if the statement is null, for example if it was actually
11232 a subroutine definition (which has compile-time side effects). If not
11233 null, it will be ops directly implementing the statement, suitable to
11234 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11235 equivalent op (except for those embedded in a scope contained entirely
11236 within the statement).
11238 If an error occurs in parsing or compilation, in most cases a valid op
11239 tree (most likely null) is returned anyway. The error is reflected in
11240 the parser state, normally resulting in a single exception at the top
11241 level of parsing which covers all the compilation errors that occurred.
11242 Some compilation errors, however, will throw an exception immediately.
11244 The I<flags> parameter is reserved for future use, and must always
11251 Perl_parse_barestmt(pTHX_ U32 flags)
11254 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11255 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11259 =for apidoc Amx|SV *|parse_label|U32 flags
11261 Parse a single label, possibly optional, of the type that may prefix a
11262 Perl statement. It is up to the caller to ensure that the dynamic parser
11263 state (L</PL_parser> et al) is correctly set to reflect the source of
11264 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11265 label is optional, otherwise it is mandatory.
11267 The name of the label is returned in the form of a fresh scalar. If an
11268 optional label is absent, a null pointer is returned.
11270 If an error occurs in parsing, which can only occur if the label is
11271 mandatory, a valid label is returned anyway. The error is reflected in
11272 the parser state, normally resulting in a single exception at the top
11273 level of parsing which covers all the compilation errors that occurred.
11279 Perl_parse_label(pTHX_ U32 flags)
11281 if (flags & ~PARSE_OPTIONAL)
11282 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11283 if (PL_lex_state == LEX_KNOWNEXT) {
11284 PL_parser->yychar = yylex();
11285 if (PL_parser->yychar == LABEL) {
11286 char *lpv = pl_yylval.pval;
11287 STRLEN llen = strlen(lpv);
11289 PL_parser->yychar = YYEMPTY;
11290 lsv = newSV_type(SVt_PV);
11291 SvPV_set(lsv, lpv);
11292 SvCUR_set(lsv, llen);
11293 SvLEN_set(lsv, llen+1);
11303 STRLEN wlen, bufptr_pos;
11307 if (!isIDFIRST_A(c))
11311 } while(isWORDCHAR_A(c));
11313 if (word_takes_any_delimeter(s, wlen))
11315 bufptr_pos = s - SvPVX(PL_linestr);
11317 lex_read_space(LEX_KEEP_PREVIOUS);
11319 s = SvPVX(PL_linestr) + bufptr_pos;
11320 if (t[0] == ':' && t[1] != ':') {
11321 PL_oldoldbufptr = PL_oldbufptr;
11324 return newSVpvn(s, wlen);
11328 if (flags & PARSE_OPTIONAL) {
11331 qerror(Perl_mess(aTHX_ "Parse error"));
11332 return newSVpvs("x");
11339 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11341 Parse a single complete Perl statement. This may be a normal imperative
11342 statement or a declaration that has compile-time effect, and may include
11343 optional labels. It is up to the caller to ensure that the dynamic
11344 parser state (L</PL_parser> et al) is correctly set to reflect the source
11345 of the code to be parsed and the lexical context for the statement.
11347 The op tree representing the statement is returned. This may be a
11348 null pointer if the statement is null, for example if it was actually
11349 a subroutine definition (which has compile-time side effects). If not
11350 null, it will be the result of a L</newSTATEOP> call, normally including
11351 a C<nextstate> or equivalent op.
11353 If an error occurs in parsing or compilation, in most cases a valid op
11354 tree (most likely null) is returned anyway. The error is reflected in
11355 the parser state, normally resulting in a single exception at the top
11356 level of parsing which covers all the compilation errors that occurred.
11357 Some compilation errors, however, will throw an exception immediately.
11359 The I<flags> parameter is reserved for future use, and must always
11366 Perl_parse_fullstmt(pTHX_ U32 flags)
11369 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11370 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11374 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11376 Parse a sequence of zero or more Perl statements. These may be normal
11377 imperative statements, including optional labels, or declarations
11378 that have compile-time effect, or any mixture thereof. The statement
11379 sequence ends when a closing brace or end-of-file is encountered in a
11380 place where a new statement could have validly started. It is up to
11381 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11382 is correctly set to reflect the source of the code to be parsed and the
11383 lexical context for the statements.
11385 The op tree representing the statement sequence is returned. This may
11386 be a null pointer if the statements were all null, for example if there
11387 were no statements or if there were only subroutine definitions (which
11388 have compile-time side effects). If not null, it will be a C<lineseq>
11389 list, normally including C<nextstate> or equivalent ops.
11391 If an error occurs in parsing or compilation, in most cases a valid op
11392 tree is returned anyway. The error is reflected in the parser state,
11393 normally resulting in a single exception at the top level of parsing
11394 which covers all the compilation errors that occurred. Some compilation
11395 errors, however, will throw an exception immediately.
11397 The I<flags> parameter is reserved for future use, and must always
11404 Perl_parse_stmtseq(pTHX_ U32 flags)
11409 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
11410 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
11411 c = lex_peek_unichar(0);
11412 if (c != -1 && c != /*{*/'}')
11413 qerror(Perl_mess(aTHX_ "Parse error"));
11418 Perl_munge_qwlist_to_paren_list(pTHX_ OP *qwlist)
11420 PERL_ARGS_ASSERT_MUNGE_QWLIST_TO_PAREN_LIST;
11421 deprecate("qw(...) as parentheses");
11422 force_next((4<<24)|')');
11423 if (qwlist->op_type == OP_STUB) {
11427 start_force(PL_curforce);
11428 NEXTVAL_NEXTTOKE.opval = qwlist;
11431 force_next((2<<24)|'(');
11436 * c-indentation-style: bsd
11437 * c-basic-offset: 4
11438 * indent-tabs-mode: t
11441 * ex: set ts=8 sts=4 sw=4 noet: