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_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (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)
288 #define UNIPROTO(f,optional) { \
289 if (optional) PL_last_uni = PL_oldbufptr; \
293 #define UNIBRACK(f) { \
294 pl_yylval.ival = f; \
296 PL_last_uni = PL_oldbufptr; \
298 return REPORT( (int)FUNC1 ); \
300 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
303 /* grandfather return to old style */
306 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
307 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
308 pl_yylval.ival = (f); \
316 /* how to interpret the pl_yylval associated with the token */
320 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
326 static struct debug_tokens {
328 enum token_type type;
330 } const debug_tokens[] =
332 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
333 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
334 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
335 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
336 { ARROW, TOKENTYPE_NONE, "ARROW" },
337 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
338 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
339 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
340 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
341 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
342 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
343 { DO, TOKENTYPE_NONE, "DO" },
344 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
345 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
346 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
347 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
348 { ELSE, TOKENTYPE_NONE, "ELSE" },
349 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
350 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
351 { FOR, TOKENTYPE_IVAL, "FOR" },
352 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
353 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
354 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
355 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
356 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
357 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
358 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
359 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
360 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
361 { IF, TOKENTYPE_IVAL, "IF" },
362 { LABEL, TOKENTYPE_PVAL, "LABEL" },
363 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
364 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
365 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
366 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
367 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
368 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
369 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
370 { MY, TOKENTYPE_IVAL, "MY" },
371 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
372 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
373 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
374 { OROP, TOKENTYPE_IVAL, "OROP" },
375 { OROR, TOKENTYPE_NONE, "OROR" },
376 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
377 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
378 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
379 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
380 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
381 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
382 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
383 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
384 { PREINC, TOKENTYPE_NONE, "PREINC" },
385 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
386 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
387 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
388 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
389 { SUB, TOKENTYPE_NONE, "SUB" },
390 { THING, TOKENTYPE_OPVAL, "THING" },
391 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
392 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
393 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
394 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
395 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
396 { USE, TOKENTYPE_IVAL, "USE" },
397 { WHEN, TOKENTYPE_IVAL, "WHEN" },
398 { WHILE, TOKENTYPE_IVAL, "WHILE" },
399 { WORD, TOKENTYPE_OPVAL, "WORD" },
400 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
401 { 0, TOKENTYPE_NONE, NULL }
404 /* dump the returned token in rv, plus any optional arg in pl_yylval */
407 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
411 PERL_ARGS_ASSERT_TOKEREPORT;
414 const char *name = NULL;
415 enum token_type type = TOKENTYPE_NONE;
416 const struct debug_tokens *p;
417 SV* const report = newSVpvs("<== ");
419 for (p = debug_tokens; p->token; p++) {
420 if (p->token == (int)rv) {
427 Perl_sv_catpv(aTHX_ report, name);
428 else if ((char)rv > ' ' && (char)rv < '~')
429 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
431 sv_catpvs(report, "EOF");
433 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
436 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
439 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
441 case TOKENTYPE_OPNUM:
442 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
443 PL_op_name[lvalp->ival]);
446 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
448 case TOKENTYPE_OPVAL:
450 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
451 PL_op_name[lvalp->opval->op_type]);
452 if (lvalp->opval->op_type == OP_CONST) {
453 Perl_sv_catpvf(aTHX_ report, " %s",
454 SvPEEK(cSVOPx_sv(lvalp->opval)));
459 sv_catpvs(report, "(opval=null)");
462 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
468 /* print the buffer with suitable escapes */
471 S_printbuf(pTHX_ const char *const fmt, const char *const s)
473 SV* const tmp = newSVpvs("");
475 PERL_ARGS_ASSERT_PRINTBUF;
477 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
484 S_deprecate_commaless_var_list(pTHX) {
486 deprecate("comma-less variable list");
487 return REPORT(','); /* grandfather non-comma-format format */
493 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
494 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
498 S_ao(pTHX_ int toketype)
501 if (*PL_bufptr == '=') {
503 if (toketype == ANDAND)
504 pl_yylval.ival = OP_ANDASSIGN;
505 else if (toketype == OROR)
506 pl_yylval.ival = OP_ORASSIGN;
507 else if (toketype == DORDOR)
508 pl_yylval.ival = OP_DORASSIGN;
516 * When Perl expects an operator and finds something else, no_op
517 * prints the warning. It always prints "<something> found where
518 * operator expected. It prints "Missing semicolon on previous line?"
519 * if the surprise occurs at the start of the line. "do you need to
520 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
521 * where the compiler doesn't know if foo is a method call or a function.
522 * It prints "Missing operator before end of line" if there's nothing
523 * after the missing operator, or "... before <...>" if there is something
524 * after the missing operator.
528 S_no_op(pTHX_ const char *const what, char *s)
531 char * const oldbp = PL_bufptr;
532 const bool is_first = (PL_oldbufptr == PL_linestart);
534 PERL_ARGS_ASSERT_NO_OP;
540 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
541 if (ckWARN_d(WARN_SYNTAX)) {
543 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
544 "\t(Missing semicolon on previous line?)\n");
545 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
547 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
549 if (t < PL_bufptr && isSPACE(*t))
550 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
551 "\t(Do you need to predeclare %.*s?)\n",
552 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
556 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
557 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
565 * Complain about missing quote/regexp/heredoc terminator.
566 * If it's called with NULL then it cauterizes the line buffer.
567 * If we're in a delimited string and the delimiter is a control
568 * character, it's reformatted into a two-char sequence like ^C.
573 S_missingterm(pTHX_ char *s)
579 char * const nl = strrchr(s,'\n');
583 else if (isCNTRL(PL_multi_close)) {
585 tmpbuf[1] = (char)toCTRL(PL_multi_close);
590 *tmpbuf = (char)PL_multi_close;
594 q = strchr(s,'"') ? '\'' : '"';
595 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
601 * Check whether the named feature is enabled.
604 Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
607 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
609 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
611 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
613 if (namelen > MAX_FEATURE_LEN)
615 memcpy(&he_name[8], name, namelen);
619 PL_curcop, he_name, 8 + namelen, 0,
625 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
626 * utf16-to-utf8-reversed.
629 #ifdef PERL_CR_FILTER
633 register const char *s = SvPVX_const(sv);
634 register const char * const e = s + SvCUR(sv);
636 PERL_ARGS_ASSERT_STRIP_RETURN;
638 /* outer loop optimized to do nothing if there are no CR-LFs */
640 if (*s++ == '\r' && *s == '\n') {
641 /* hit a CR-LF, need to copy the rest */
642 register char *d = s - 1;
645 if (*s == '\r' && s[1] == '\n')
656 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
658 const I32 count = FILTER_READ(idx+1, sv, maxlen);
659 if (count > 0 && !maxlen)
666 =for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
668 Creates and initialises a new lexer/parser state object, supplying
669 a context in which to lex and parse from a new source of Perl code.
670 A pointer to the new state object is placed in L</PL_parser>. An entry
671 is made on the save stack so that upon unwinding the new state object
672 will be destroyed and the former value of L</PL_parser> will be restored.
673 Nothing else need be done to clean up the parsing context.
675 The code to be parsed comes from I<line> and I<rsfp>. I<line>, if
676 non-null, provides a string (in SV form) containing code to be parsed.
677 A copy of the string is made, so subsequent modification of I<line>
678 does not affect parsing. I<rsfp>, if non-null, provides an input stream
679 from which code will be read to be parsed. If both are non-null, the
680 code in I<line> comes first and must consist of complete lines of input,
681 and I<rsfp> supplies the remainder of the source.
683 The I<flags> parameter is reserved for future use. Currently it is only
684 used by perl internally, so extensions should always pass zero.
689 /* LEX_START_SAME_FILTER indicates that this is not a new file, so it
690 can share filters with the current parser. */
693 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
696 const char *s = NULL;
697 yy_parser *parser, *oparser;
698 if (flags && flags & ~LEX_START_FLAGS)
699 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
701 /* create and initialise a parser */
703 Newxz(parser, 1, yy_parser);
704 parser->old_parser = oparser = PL_parser;
707 parser->stack = NULL;
709 parser->stack_size = 0;
711 /* on scope exit, free this parser and restore any outer one */
713 parser->saved_curcop = PL_curcop;
715 /* initialise lexer state */
718 parser->curforce = -1;
720 parser->nexttoke = 0;
722 parser->error_count = oparser ? oparser->error_count : 0;
723 parser->copline = NOLINE;
724 parser->lex_state = LEX_NORMAL;
725 parser->expect = XSTATE;
727 parser->rsfp_filters =
728 !(flags & LEX_START_SAME_FILTER) || !oparser
730 : MUTABLE_AV(SvREFCNT_inc(
731 oparser->rsfp_filters
732 ? oparser->rsfp_filters
733 : (oparser->rsfp_filters = newAV())
736 Newx(parser->lex_brackstack, 120, char);
737 Newx(parser->lex_casestack, 12, char);
738 *parser->lex_casestack = '\0';
742 s = SvPV_const(line, len);
743 parser->linestr = flags & LEX_START_COPIED
744 ? SvREFCNT_inc_simple_NN(line)
745 : newSVpvn_flags(s, len, SvUTF8(line));
746 if (!len || s[len-1] != ';')
747 sv_catpvs(parser->linestr, "\n;");
749 parser->linestr = newSVpvs("\n;");
751 parser->oldoldbufptr =
754 parser->linestart = SvPVX(parser->linestr);
755 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
756 parser->last_lop = parser->last_uni = NULL;
757 parser->lex_flags = flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES);
759 parser->in_pod = parser->filtered = 0;
763 /* delete a parser object */
766 Perl_parser_free(pTHX_ const yy_parser *parser)
768 PERL_ARGS_ASSERT_PARSER_FREE;
770 PL_curcop = parser->saved_curcop;
771 SvREFCNT_dec(parser->linestr);
773 if (parser->rsfp == PerlIO_stdin())
774 PerlIO_clearerr(parser->rsfp);
775 else if (parser->rsfp && (!parser->old_parser ||
776 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
777 PerlIO_close(parser->rsfp);
778 SvREFCNT_dec(parser->rsfp_filters);
780 Safefree(parser->lex_brackstack);
781 Safefree(parser->lex_casestack);
782 PL_parser = parser->old_parser;
788 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
790 Buffer scalar containing the chunk currently under consideration of the
791 text currently being lexed. This is always a plain string scalar (for
792 which C<SvPOK> is true). It is not intended to be used as a scalar by
793 normal scalar means; instead refer to the buffer directly by the pointer
794 variables described below.
796 The lexer maintains various C<char*> pointers to things in the
797 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
798 reallocated, all of these pointers must be updated. Don't attempt to
799 do this manually, but rather use L</lex_grow_linestr> if you need to
800 reallocate the buffer.
802 The content of the text chunk in the buffer is commonly exactly one
803 complete line of input, up to and including a newline terminator,
804 but there are situations where it is otherwise. The octets of the
805 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
806 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
807 flag on this scalar, which may disagree with it.
809 For direct examination of the buffer, the variable
810 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
811 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
812 of these pointers is usually preferable to examination of the scalar
813 through normal scalar means.
815 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
817 Direct pointer to the end of the chunk of text currently being lexed, the
818 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
819 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
820 always located at the end of the buffer, and does not count as part of
821 the buffer's contents.
823 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
825 Points to the current position of lexing inside the lexer buffer.
826 Characters around this point may be freely examined, within
827 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
828 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
829 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
831 Lexing code (whether in the Perl core or not) moves this pointer past
832 the characters that it consumes. It is also expected to perform some
833 bookkeeping whenever a newline character is consumed. This movement
834 can be more conveniently performed by the function L</lex_read_to>,
835 which handles newlines appropriately.
837 Interpretation of the buffer's octets can be abstracted out by
838 using the slightly higher-level functions L</lex_peek_unichar> and
839 L</lex_read_unichar>.
841 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
843 Points to the start of the current line inside the lexer buffer.
844 This is useful for indicating at which column an error occurred, and
845 not much else. This must be updated by any lexing code that consumes
846 a newline; the function L</lex_read_to> handles this detail.
852 =for apidoc Amx|bool|lex_bufutf8
854 Indicates whether the octets in the lexer buffer
855 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
856 of Unicode characters. If not, they should be interpreted as Latin-1
857 characters. This is analogous to the C<SvUTF8> flag for scalars.
859 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
860 contains valid UTF-8. Lexing code must be robust in the face of invalid
863 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
864 is significant, but not the whole story regarding the input character
865 encoding. Normally, when a file is being read, the scalar contains octets
866 and its C<SvUTF8> flag is off, but the octets should be interpreted as
867 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
868 however, the scalar may have the C<SvUTF8> flag on, and in this case its
869 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
870 is in effect. This logic may change in the future; use this function
871 instead of implementing the logic yourself.
877 Perl_lex_bufutf8(pTHX)
883 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
885 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
886 at least I<len> octets (including terminating NUL). Returns a
887 pointer to the reallocated buffer. This is necessary before making
888 any direct modification of the buffer that would increase its length.
889 L</lex_stuff_pvn> provides a more convenient way to insert text into
892 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
893 this function updates all of the lexer's variables that point directly
900 Perl_lex_grow_linestr(pTHX_ STRLEN len)
904 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
905 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
906 linestr = PL_parser->linestr;
907 buf = SvPVX(linestr);
908 if (len <= SvLEN(linestr))
910 bufend_pos = PL_parser->bufend - buf;
911 bufptr_pos = PL_parser->bufptr - buf;
912 oldbufptr_pos = PL_parser->oldbufptr - buf;
913 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
914 linestart_pos = PL_parser->linestart - buf;
915 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
916 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
917 buf = sv_grow(linestr, len);
918 PL_parser->bufend = buf + bufend_pos;
919 PL_parser->bufptr = buf + bufptr_pos;
920 PL_parser->oldbufptr = buf + oldbufptr_pos;
921 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
922 PL_parser->linestart = buf + linestart_pos;
923 if (PL_parser->last_uni)
924 PL_parser->last_uni = buf + last_uni_pos;
925 if (PL_parser->last_lop)
926 PL_parser->last_lop = buf + last_lop_pos;
931 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
933 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
934 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
935 reallocating the buffer if necessary. This means that lexing code that
936 runs later will see the characters as if they had appeared in the input.
937 It is not recommended to do this as part of normal parsing, and most
938 uses of this facility run the risk of the inserted characters being
939 interpreted in an unintended manner.
941 The string to be inserted is represented by I<len> octets starting
942 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
943 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
944 The characters are recoded for the lexer buffer, according to how the
945 buffer is currently being interpreted (L</lex_bufutf8>). If a string
946 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
947 function is more convenient.
953 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
957 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
958 if (flags & ~(LEX_STUFF_UTF8))
959 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
961 if (flags & LEX_STUFF_UTF8) {
965 const char *p, *e = pv+len;
966 for (p = pv; p != e; p++)
967 highhalf += !!(((U8)*p) & 0x80);
970 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
971 bufptr = PL_parser->bufptr;
972 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
973 SvCUR_set(PL_parser->linestr,
974 SvCUR(PL_parser->linestr) + len+highhalf);
975 PL_parser->bufend += len+highhalf;
976 for (p = pv; p != e; p++) {
979 *bufptr++ = (char)(0xc0 | (c >> 6));
980 *bufptr++ = (char)(0x80 | (c & 0x3f));
987 if (flags & LEX_STUFF_UTF8) {
989 const char *p, *e = pv+len;
990 for (p = pv; p != e; p++) {
993 Perl_croak(aTHX_ "Lexing code attempted to stuff "
994 "non-Latin-1 character into Latin-1 input");
995 } else if (c >= 0xc2 && p+1 != e &&
996 (((U8)p[1]) & 0xc0) == 0x80) {
999 } else if (c >= 0x80) {
1000 /* malformed UTF-8 */
1002 SAVESPTR(PL_warnhook);
1003 PL_warnhook = PERL_WARNHOOK_FATAL;
1004 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
1010 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1011 bufptr = PL_parser->bufptr;
1012 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1013 SvCUR_set(PL_parser->linestr,
1014 SvCUR(PL_parser->linestr) + len-highhalf);
1015 PL_parser->bufend += len-highhalf;
1016 for (p = pv; p != e; p++) {
1019 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1022 *bufptr++ = (char)c;
1027 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1028 bufptr = PL_parser->bufptr;
1029 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1030 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1031 PL_parser->bufend += len;
1032 Copy(pv, bufptr, len, char);
1038 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1040 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1041 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1042 reallocating the buffer if necessary. This means that lexing code that
1043 runs later will see the characters as if they had appeared in the input.
1044 It is not recommended to do this as part of normal parsing, and most
1045 uses of this facility run the risk of the inserted characters being
1046 interpreted in an unintended manner.
1048 The string to be inserted is represented by octets starting at I<pv>
1049 and continuing to the first nul. These octets are interpreted as either
1050 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1051 in I<flags>. The characters are recoded for the lexer buffer, according
1052 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1053 If it is not convenient to nul-terminate a string to be inserted, the
1054 L</lex_stuff_pvn> function is more appropriate.
1060 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1062 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1063 lex_stuff_pvn(pv, strlen(pv), flags);
1067 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1069 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1070 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1071 reallocating the buffer if necessary. This means that lexing code that
1072 runs later will see the characters as if they had appeared in the input.
1073 It is not recommended to do this as part of normal parsing, and most
1074 uses of this facility run the risk of the inserted characters being
1075 interpreted in an unintended manner.
1077 The string to be inserted is the string value of I<sv>. The characters
1078 are recoded for the lexer buffer, according to how the buffer is currently
1079 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1080 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1081 need to construct a scalar.
1087 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1091 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1093 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1095 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1099 =for apidoc Amx|void|lex_unstuff|char *ptr
1101 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1102 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1103 This hides the discarded text from any lexing code that runs later,
1104 as if the text had never appeared.
1106 This is not the normal way to consume lexed text. For that, use
1113 Perl_lex_unstuff(pTHX_ char *ptr)
1117 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1118 buf = PL_parser->bufptr;
1120 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1123 bufend = PL_parser->bufend;
1125 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1126 unstuff_len = ptr - buf;
1127 Move(ptr, buf, bufend+1-ptr, char);
1128 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1129 PL_parser->bufend = bufend - unstuff_len;
1133 =for apidoc Amx|void|lex_read_to|char *ptr
1135 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1136 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1137 performing the correct bookkeeping whenever a newline character is passed.
1138 This is the normal way to consume lexed text.
1140 Interpretation of the buffer's octets can be abstracted out by
1141 using the slightly higher-level functions L</lex_peek_unichar> and
1142 L</lex_read_unichar>.
1148 Perl_lex_read_to(pTHX_ char *ptr)
1151 PERL_ARGS_ASSERT_LEX_READ_TO;
1152 s = PL_parser->bufptr;
1153 if (ptr < s || ptr > PL_parser->bufend)
1154 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1155 for (; s != ptr; s++)
1157 CopLINE_inc(PL_curcop);
1158 PL_parser->linestart = s+1;
1160 PL_parser->bufptr = ptr;
1164 =for apidoc Amx|void|lex_discard_to|char *ptr
1166 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1167 up to I<ptr>. The remaining content of the buffer will be moved, and
1168 all pointers into the buffer updated appropriately. I<ptr> must not
1169 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1170 it is not permitted to discard text that has yet to be lexed.
1172 Normally it is not necessarily to do this directly, because it suffices to
1173 use the implicit discarding behaviour of L</lex_next_chunk> and things
1174 based on it. However, if a token stretches across multiple lines,
1175 and the lexing code has kept multiple lines of text in the buffer for
1176 that purpose, then after completion of the token it would be wise to
1177 explicitly discard the now-unneeded earlier lines, to avoid future
1178 multi-line tokens growing the buffer without bound.
1184 Perl_lex_discard_to(pTHX_ char *ptr)
1188 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1189 buf = SvPVX(PL_parser->linestr);
1191 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1194 if (ptr > PL_parser->bufptr)
1195 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1196 discard_len = ptr - buf;
1197 if (PL_parser->oldbufptr < ptr)
1198 PL_parser->oldbufptr = ptr;
1199 if (PL_parser->oldoldbufptr < ptr)
1200 PL_parser->oldoldbufptr = ptr;
1201 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1202 PL_parser->last_uni = NULL;
1203 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1204 PL_parser->last_lop = NULL;
1205 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1206 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1207 PL_parser->bufend -= discard_len;
1208 PL_parser->bufptr -= discard_len;
1209 PL_parser->oldbufptr -= discard_len;
1210 PL_parser->oldoldbufptr -= discard_len;
1211 if (PL_parser->last_uni)
1212 PL_parser->last_uni -= discard_len;
1213 if (PL_parser->last_lop)
1214 PL_parser->last_lop -= discard_len;
1218 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1220 Reads in the next chunk of text to be lexed, appending it to
1221 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1222 looked to the end of the current chunk and wants to know more. It is
1223 usual, but not necessary, for lexing to have consumed the entirety of
1224 the current chunk at this time.
1226 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1227 chunk (i.e., the current chunk has been entirely consumed), normally the
1228 current chunk will be discarded at the same time that the new chunk is
1229 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1230 will not be discarded. If the current chunk has not been entirely
1231 consumed, then it will not be discarded regardless of the flag.
1233 Returns true if some new text was added to the buffer, or false if the
1234 buffer has reached the end of the input text.
1239 #define LEX_FAKE_EOF 0x80000000
1242 Perl_lex_next_chunk(pTHX_ U32 flags)
1246 STRLEN old_bufend_pos, new_bufend_pos;
1247 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1248 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1249 bool got_some_for_debugger = 0;
1251 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1252 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1253 linestr = PL_parser->linestr;
1254 buf = SvPVX(linestr);
1255 if (!(flags & LEX_KEEP_PREVIOUS) &&
1256 PL_parser->bufptr == PL_parser->bufend) {
1257 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1259 if (PL_parser->last_uni != PL_parser->bufend)
1260 PL_parser->last_uni = NULL;
1261 if (PL_parser->last_lop != PL_parser->bufend)
1262 PL_parser->last_lop = NULL;
1263 last_uni_pos = last_lop_pos = 0;
1267 old_bufend_pos = PL_parser->bufend - buf;
1268 bufptr_pos = PL_parser->bufptr - buf;
1269 oldbufptr_pos = PL_parser->oldbufptr - buf;
1270 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1271 linestart_pos = PL_parser->linestart - buf;
1272 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1273 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1275 if (flags & LEX_FAKE_EOF) {
1277 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
1279 } else if (filter_gets(linestr, old_bufend_pos)) {
1281 got_some_for_debugger = 1;
1283 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1284 sv_setpvs(linestr, "");
1286 /* End of real input. Close filehandle (unless it was STDIN),
1287 * then add implicit termination.
1289 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1290 PerlIO_clearerr(PL_parser->rsfp);
1291 else if (PL_parser->rsfp)
1292 (void)PerlIO_close(PL_parser->rsfp);
1293 PL_parser->rsfp = NULL;
1294 PL_parser->in_pod = PL_parser->filtered = 0;
1296 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1299 if (!PL_in_eval && PL_minus_p) {
1301 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1302 PL_minus_n = PL_minus_p = 0;
1303 } else if (!PL_in_eval && PL_minus_n) {
1304 sv_catpvs(linestr, /*{*/";}");
1307 sv_catpvs(linestr, ";");
1310 buf = SvPVX(linestr);
1311 new_bufend_pos = SvCUR(linestr);
1312 PL_parser->bufend = buf + new_bufend_pos;
1313 PL_parser->bufptr = buf + bufptr_pos;
1314 PL_parser->oldbufptr = buf + oldbufptr_pos;
1315 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1316 PL_parser->linestart = buf + linestart_pos;
1317 if (PL_parser->last_uni)
1318 PL_parser->last_uni = buf + last_uni_pos;
1319 if (PL_parser->last_lop)
1320 PL_parser->last_lop = buf + last_lop_pos;
1321 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1322 PL_curstash != PL_debstash) {
1323 /* debugger active and we're not compiling the debugger code,
1324 * so store the line into the debugger's array of lines
1326 update_debugger_info(NULL, buf+old_bufend_pos,
1327 new_bufend_pos-old_bufend_pos);
1333 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1335 Looks ahead one (Unicode) character in the text currently being lexed.
1336 Returns the codepoint (unsigned integer value) of the next character,
1337 or -1 if lexing has reached the end of the input text. To consume the
1338 peeked character, use L</lex_read_unichar>.
1340 If the next character is in (or extends into) the next chunk of input
1341 text, the next chunk will be read in. Normally the current chunk will be
1342 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1343 then the current chunk will not be discarded.
1345 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1346 is encountered, an exception is generated.
1352 Perl_lex_peek_unichar(pTHX_ U32 flags)
1356 if (flags & ~(LEX_KEEP_PREVIOUS))
1357 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1358 s = PL_parser->bufptr;
1359 bufend = PL_parser->bufend;
1365 if (!lex_next_chunk(flags))
1367 s = PL_parser->bufptr;
1368 bufend = PL_parser->bufend;
1374 len = PL_utf8skip[head];
1375 while ((STRLEN)(bufend-s) < len) {
1376 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1378 s = PL_parser->bufptr;
1379 bufend = PL_parser->bufend;
1382 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1383 if (retlen == (STRLEN)-1) {
1384 /* malformed UTF-8 */
1386 SAVESPTR(PL_warnhook);
1387 PL_warnhook = PERL_WARNHOOK_FATAL;
1388 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1394 if (!lex_next_chunk(flags))
1396 s = PL_parser->bufptr;
1403 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1405 Reads the next (Unicode) character in the text currently being lexed.
1406 Returns the codepoint (unsigned integer value) of the character read,
1407 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1408 if lexing has reached the end of the input text. To non-destructively
1409 examine the next character, use L</lex_peek_unichar> instead.
1411 If the next character is in (or extends into) the next chunk of input
1412 text, the next chunk will be read in. Normally the current chunk will be
1413 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1414 then the current chunk will not be discarded.
1416 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1417 is encountered, an exception is generated.
1423 Perl_lex_read_unichar(pTHX_ U32 flags)
1426 if (flags & ~(LEX_KEEP_PREVIOUS))
1427 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1428 c = lex_peek_unichar(flags);
1431 CopLINE_inc(PL_curcop);
1433 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1435 ++(PL_parser->bufptr);
1441 =for apidoc Amx|void|lex_read_space|U32 flags
1443 Reads optional spaces, in Perl style, in the text currently being
1444 lexed. The spaces may include ordinary whitespace characters and
1445 Perl-style comments. C<#line> directives are processed if encountered.
1446 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1447 at a non-space character (or the end of the input text).
1449 If spaces extend into the next chunk of input text, the next chunk will
1450 be read in. Normally the current chunk will be discarded at the same
1451 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1452 chunk will not be discarded.
1457 #define LEX_NO_NEXT_CHUNK 0x80000000
1460 Perl_lex_read_space(pTHX_ U32 flags)
1463 bool need_incline = 0;
1464 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1465 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1468 sv_free(PL_skipwhite);
1469 PL_skipwhite = NULL;
1472 PL_skipwhite = newSVpvs("");
1473 #endif /* PERL_MAD */
1474 s = PL_parser->bufptr;
1475 bufend = PL_parser->bufend;
1481 } while (!(c == '\n' || (c == 0 && s == bufend)));
1482 } else if (c == '\n') {
1484 PL_parser->linestart = s;
1489 } else if (isSPACE(c)) {
1491 } else if (c == 0 && s == bufend) {
1495 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1496 #endif /* PERL_MAD */
1497 if (flags & LEX_NO_NEXT_CHUNK)
1499 PL_parser->bufptr = s;
1500 CopLINE_inc(PL_curcop);
1501 got_more = lex_next_chunk(flags);
1502 CopLINE_dec(PL_curcop);
1503 s = PL_parser->bufptr;
1504 bufend = PL_parser->bufend;
1507 if (need_incline && PL_parser->rsfp) {
1517 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1518 #endif /* PERL_MAD */
1519 PL_parser->bufptr = s;
1524 * This subroutine has nothing to do with tilting, whether at windmills
1525 * or pinball tables. Its name is short for "increment line". It
1526 * increments the current line number in CopLINE(PL_curcop) and checks
1527 * to see whether the line starts with a comment of the form
1528 * # line 500 "foo.pm"
1529 * If so, it sets the current line number and file to the values in the comment.
1533 S_incline(pTHX_ const char *s)
1541 PERL_ARGS_ASSERT_INCLINE;
1543 CopLINE_inc(PL_curcop);
1546 while (SPACE_OR_TAB(*s))
1548 if (strnEQ(s, "line", 4))
1552 if (SPACE_OR_TAB(*s))
1556 while (SPACE_OR_TAB(*s))
1564 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1566 while (SPACE_OR_TAB(*s))
1568 if (*s == '"' && (t = strchr(s+1, '"'))) {
1574 while (!isSPACE(*t))
1578 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1580 if (*e != '\n' && *e != '\0')
1581 return; /* false alarm */
1583 line_num = atoi(n)-1;
1586 const STRLEN len = t - s;
1587 SV *const temp_sv = CopFILESV(PL_curcop);
1592 cf = SvPVX(temp_sv);
1593 tmplen = SvCUR(temp_sv);
1599 if (!PL_rsfp && !PL_parser->filtered) {
1600 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1601 * to *{"::_<newfilename"} */
1602 /* However, the long form of evals is only turned on by the
1603 debugger - usually they're "(eval %lu)" */
1607 STRLEN tmplen2 = len;
1608 if (tmplen + 2 <= sizeof smallbuf)
1611 Newx(tmpbuf, tmplen + 2, char);
1614 memcpy(tmpbuf + 2, cf, tmplen);
1616 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1621 if (tmplen2 + 2 <= sizeof smallbuf)
1624 Newx(tmpbuf2, tmplen2 + 2, char);
1626 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1627 /* Either they malloc'd it, or we malloc'd it,
1628 so no prefix is present in ours. */
1633 memcpy(tmpbuf2 + 2, s, tmplen2);
1636 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1638 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1639 /* adjust ${"::_<newfilename"} to store the new file name */
1640 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1641 /* The line number may differ. If that is the case,
1642 alias the saved lines that are in the array.
1643 Otherwise alias the whole array. */
1644 if (CopLINE(PL_curcop) == line_num) {
1645 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1646 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1648 else if (GvAV(*gvp)) {
1649 AV * const av = GvAV(*gvp);
1650 const I32 start = CopLINE(PL_curcop)+1;
1651 I32 items = AvFILLp(av) - start;
1653 AV * const av2 = GvAVn(gv2);
1654 SV **svp = AvARRAY(av) + start;
1655 I32 l = (I32)line_num+1;
1657 av_store(av2, l++, SvREFCNT_inc(*svp++));
1662 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1664 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1666 CopFILE_free(PL_curcop);
1667 CopFILE_setn(PL_curcop, s, len);
1669 CopLINE_set(PL_curcop, line_num);
1673 /* skip space before PL_thistoken */
1676 S_skipspace0(pTHX_ register char *s)
1678 PERL_ARGS_ASSERT_SKIPSPACE0;
1685 PL_thiswhite = newSVpvs("");
1686 sv_catsv(PL_thiswhite, PL_skipwhite);
1687 sv_free(PL_skipwhite);
1690 PL_realtokenstart = s - SvPVX(PL_linestr);
1694 /* skip space after PL_thistoken */
1697 S_skipspace1(pTHX_ register char *s)
1699 const char *start = s;
1700 I32 startoff = start - SvPVX(PL_linestr);
1702 PERL_ARGS_ASSERT_SKIPSPACE1;
1707 start = SvPVX(PL_linestr) + startoff;
1708 if (!PL_thistoken && PL_realtokenstart >= 0) {
1709 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1710 PL_thistoken = newSVpvn(tstart, start - tstart);
1712 PL_realtokenstart = -1;
1715 PL_nextwhite = newSVpvs("");
1716 sv_catsv(PL_nextwhite, PL_skipwhite);
1717 sv_free(PL_skipwhite);
1724 S_skipspace2(pTHX_ register char *s, SV **svp)
1727 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1728 const I32 startoff = s - SvPVX(PL_linestr);
1730 PERL_ARGS_ASSERT_SKIPSPACE2;
1733 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1734 if (!PL_madskills || !svp)
1736 start = SvPVX(PL_linestr) + startoff;
1737 if (!PL_thistoken && PL_realtokenstart >= 0) {
1738 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1739 PL_thistoken = newSVpvn(tstart, start - tstart);
1740 PL_realtokenstart = -1;
1744 *svp = newSVpvs("");
1745 sv_setsv(*svp, PL_skipwhite);
1746 sv_free(PL_skipwhite);
1755 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1757 AV *av = CopFILEAVx(PL_curcop);
1759 SV * const sv = newSV_type(SVt_PVMG);
1761 sv_setsv(sv, orig_sv);
1763 sv_setpvn(sv, buf, len);
1766 av_store(av, (I32)CopLINE(PL_curcop), sv);
1772 * Called to gobble the appropriate amount and type of whitespace.
1773 * Skips comments as well.
1777 S_skipspace(pTHX_ register char *s)
1781 #endif /* PERL_MAD */
1782 PERL_ARGS_ASSERT_SKIPSPACE;
1785 sv_free(PL_skipwhite);
1786 PL_skipwhite = NULL;
1788 #endif /* PERL_MAD */
1789 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1790 while (s < PL_bufend && SPACE_OR_TAB(*s))
1793 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1795 lex_read_space(LEX_KEEP_PREVIOUS |
1796 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1797 LEX_NO_NEXT_CHUNK : 0));
1799 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1800 if (PL_linestart > PL_bufptr)
1801 PL_bufptr = PL_linestart;
1806 PL_skipwhite = newSVpvn(start, s-start);
1807 #endif /* PERL_MAD */
1813 * Check the unary operators to ensure there's no ambiguity in how they're
1814 * used. An ambiguous piece of code would be:
1816 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1817 * the +5 is its argument.
1827 if (PL_oldoldbufptr != PL_last_uni)
1829 while (isSPACE(*PL_last_uni))
1832 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1834 if ((t = strchr(s, '(')) && t < PL_bufptr)
1837 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1838 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1839 (int)(s - PL_last_uni), PL_last_uni);
1843 * LOP : macro to build a list operator. Its behaviour has been replaced
1844 * with a subroutine, S_lop() for which LOP is just another name.
1847 #define LOP(f,x) return lop(f,x,s)
1851 * Build a list operator (or something that might be one). The rules:
1852 * - if we have a next token, then it's a list operator [why?]
1853 * - if the next thing is an opening paren, then it's a function
1854 * - else it's a list operator
1858 S_lop(pTHX_ I32 f, int x, char *s)
1862 PERL_ARGS_ASSERT_LOP;
1868 PL_last_lop = PL_oldbufptr;
1869 PL_last_lop_op = (OPCODE)f;
1878 return REPORT(FUNC);
1881 return REPORT(FUNC);
1884 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1885 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1886 return REPORT(LSTOP);
1893 * Sets up for an eventual force_next(). start_force(0) basically does
1894 * an unshift, while start_force(-1) does a push. yylex removes items
1899 S_start_force(pTHX_ int where)
1903 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1904 where = PL_lasttoke;
1905 assert(PL_curforce < 0 || PL_curforce == where);
1906 if (PL_curforce != where) {
1907 for (i = PL_lasttoke; i > where; --i) {
1908 PL_nexttoke[i] = PL_nexttoke[i-1];
1912 if (PL_curforce < 0) /* in case of duplicate start_force() */
1913 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1914 PL_curforce = where;
1917 curmad('^', newSVpvs(""));
1918 CURMAD('_', PL_nextwhite);
1923 S_curmad(pTHX_ char slot, SV *sv)
1929 if (PL_curforce < 0)
1930 where = &PL_thismad;
1932 where = &PL_nexttoke[PL_curforce].next_mad;
1938 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1940 else if (PL_encoding) {
1941 sv_recode_to_utf8(sv, PL_encoding);
1946 /* keep a slot open for the head of the list? */
1947 if (slot != '_' && *where && (*where)->mad_key == '^') {
1948 (*where)->mad_key = slot;
1949 sv_free(MUTABLE_SV(((*where)->mad_val)));
1950 (*where)->mad_val = (void*)sv;
1953 addmad(newMADsv(slot, sv), where, 0);
1956 # define start_force(where) NOOP
1957 # define curmad(slot, sv) NOOP
1962 * When the lexer realizes it knows the next token (for instance,
1963 * it is reordering tokens for the parser) then it can call S_force_next
1964 * to know what token to return the next time the lexer is called. Caller
1965 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1966 * and possibly PL_expect to ensure the lexer handles the token correctly.
1970 S_force_next(pTHX_ I32 type)
1975 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1976 tokereport(type, &NEXTVAL_NEXTTOKE);
1980 if (PL_curforce < 0)
1981 start_force(PL_lasttoke);
1982 PL_nexttoke[PL_curforce].next_type = type;
1983 if (PL_lex_state != LEX_KNOWNEXT)
1984 PL_lex_defer = PL_lex_state;
1985 PL_lex_state = LEX_KNOWNEXT;
1986 PL_lex_expect = PL_expect;
1989 PL_nexttype[PL_nexttoke] = type;
1991 if (PL_lex_state != LEX_KNOWNEXT) {
1992 PL_lex_defer = PL_lex_state;
1993 PL_lex_expect = PL_expect;
1994 PL_lex_state = LEX_KNOWNEXT;
2002 int yyc = PL_parser->yychar;
2003 if (yyc != YYEMPTY) {
2006 NEXTVAL_NEXTTOKE = PL_parser->yylval;
2007 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
2008 PL_lex_allbrackets--;
2010 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
2011 } else if (yyc == '('/*)*/) {
2012 PL_lex_allbrackets--;
2017 PL_parser->yychar = YYEMPTY;
2022 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
2025 SV * const sv = newSVpvn_utf8(start, len,
2028 && !is_ascii_string((const U8*)start, len)
2029 && is_utf8_string((const U8*)start, len));
2035 * When the lexer knows the next thing is a word (for instance, it has
2036 * just seen -> and it knows that the next char is a word char, then
2037 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2041 * char *start : buffer position (must be within PL_linestr)
2042 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2043 * int check_keyword : if true, Perl checks to make sure the word isn't
2044 * a keyword (do this if the word is a label, e.g. goto FOO)
2045 * int allow_pack : if true, : characters will also be allowed (require,
2046 * use, etc. do this)
2047 * int allow_initial_tick : used by the "sub" lexer only.
2051 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
2057 PERL_ARGS_ASSERT_FORCE_WORD;
2059 start = SKIPSPACE1(start);
2061 if (isIDFIRST_lazy_if(s,UTF) ||
2062 (allow_pack && *s == ':') ||
2063 (allow_initial_tick && *s == '\'') )
2065 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2066 if (check_keyword && keyword(PL_tokenbuf, len, 0))
2068 start_force(PL_curforce);
2070 curmad('X', newSVpvn(start,s-start));
2071 if (token == METHOD) {
2076 PL_expect = XOPERATOR;
2080 curmad('g', newSVpvs( "forced" ));
2081 NEXTVAL_NEXTTOKE.opval
2082 = (OP*)newSVOP(OP_CONST,0,
2083 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2084 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2092 * Called when the lexer wants $foo *foo &foo etc, but the program
2093 * text only contains the "foo" portion. The first argument is a pointer
2094 * to the "foo", and the second argument is the type symbol to prefix.
2095 * Forces the next token to be a "WORD".
2096 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2100 S_force_ident(pTHX_ register const char *s, int kind)
2104 PERL_ARGS_ASSERT_FORCE_IDENT;
2107 const STRLEN len = strlen(s);
2108 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2109 UTF ? SVf_UTF8 : 0));
2110 start_force(PL_curforce);
2111 NEXTVAL_NEXTTOKE.opval = o;
2114 o->op_private = OPpCONST_ENTERED;
2115 /* XXX see note in pp_entereval() for why we forgo typo
2116 warnings if the symbol must be introduced in an eval.
2118 gv_fetchpvn_flags(s, len,
2119 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2120 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2121 kind == '$' ? SVt_PV :
2122 kind == '@' ? SVt_PVAV :
2123 kind == '%' ? SVt_PVHV :
2131 Perl_str_to_version(pTHX_ SV *sv)
2136 const char *start = SvPV_const(sv,len);
2137 const char * const end = start + len;
2138 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2140 PERL_ARGS_ASSERT_STR_TO_VERSION;
2142 while (start < end) {
2146 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2151 retval += ((NV)n)/nshift;
2160 * Forces the next token to be a version number.
2161 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2162 * and if "guessing" is TRUE, then no new token is created (and the caller
2163 * must use an alternative parsing method).
2167 S_force_version(pTHX_ char *s, int guessing)
2173 I32 startoff = s - SvPVX(PL_linestr);
2176 PERL_ARGS_ASSERT_FORCE_VERSION;
2184 while (isDIGIT(*d) || *d == '_' || *d == '.')
2188 start_force(PL_curforce);
2189 curmad('X', newSVpvn(s,d-s));
2192 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2194 #ifdef USE_LOCALE_NUMERIC
2195 char *loc = savepv(setlocale(LC_NUMERIC, NULL));
2196 setlocale(LC_NUMERIC, "C");
2198 s = scan_num(s, &pl_yylval);
2199 #ifdef USE_LOCALE_NUMERIC
2200 setlocale(LC_NUMERIC, loc);
2203 version = pl_yylval.opval;
2204 ver = cSVOPx(version)->op_sv;
2205 if (SvPOK(ver) && !SvNIOK(ver)) {
2206 SvUPGRADE(ver, SVt_PVNV);
2207 SvNV_set(ver, str_to_version(ver));
2208 SvNOK_on(ver); /* hint that it is a version */
2211 else if (guessing) {
2214 sv_free(PL_nextwhite); /* let next token collect whitespace */
2216 s = SvPVX(PL_linestr) + startoff;
2224 if (PL_madskills && !version) {
2225 sv_free(PL_nextwhite); /* let next token collect whitespace */
2227 s = SvPVX(PL_linestr) + startoff;
2230 /* NOTE: The parser sees the package name and the VERSION swapped */
2231 start_force(PL_curforce);
2232 NEXTVAL_NEXTTOKE.opval = version;
2239 * S_force_strict_version
2240 * Forces the next token to be a version number using strict syntax rules.
2244 S_force_strict_version(pTHX_ char *s)
2249 I32 startoff = s - SvPVX(PL_linestr);
2251 const char *errstr = NULL;
2253 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2255 while (isSPACE(*s)) /* leading whitespace */
2258 if (is_STRICT_VERSION(s,&errstr)) {
2260 s = (char *)scan_version(s, ver, 0);
2261 version = newSVOP(OP_CONST, 0, ver);
2263 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2264 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2268 yyerror(errstr); /* version required */
2273 if (PL_madskills && !version) {
2274 sv_free(PL_nextwhite); /* let next token collect whitespace */
2276 s = SvPVX(PL_linestr) + startoff;
2279 /* NOTE: The parser sees the package name and the VERSION swapped */
2280 start_force(PL_curforce);
2281 NEXTVAL_NEXTTOKE.opval = version;
2289 * Tokenize a quoted string passed in as an SV. It finds the next
2290 * chunk, up to end of string or a backslash. It may make a new
2291 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2296 S_tokeq(pTHX_ SV *sv)
2300 register char *send;
2305 PERL_ARGS_ASSERT_TOKEQ;
2310 s = SvPV_force(sv, len);
2311 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2314 /* This is relying on the SV being "well formed" with a trailing '\0' */
2315 while (s < send && !(*s == '\\' && s[1] == '\\'))
2320 if ( PL_hints & HINT_NEW_STRING ) {
2321 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2325 if (s + 1 < send && (s[1] == '\\'))
2326 s++; /* all that, just for this */
2331 SvCUR_set(sv, d - SvPVX_const(sv));
2333 if ( PL_hints & HINT_NEW_STRING )
2334 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2339 * Now come three functions related to double-quote context,
2340 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2341 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2342 * interact with PL_lex_state, and create fake ( ... ) argument lists
2343 * to handle functions and concatenation.
2344 * They assume that whoever calls them will be setting up a fake
2345 * join call, because each subthing puts a ',' after it. This lets
2348 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2350 * (I'm not sure whether the spurious commas at the end of lcfirst's
2351 * arguments and join's arguments are created or not).
2356 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2358 * Pattern matching will set PL_lex_op to the pattern-matching op to
2359 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2361 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2363 * Everything else becomes a FUNC.
2365 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2366 * had an OP_CONST or OP_READLINE). This just sets us up for a
2367 * call to S_sublex_push().
2371 S_sublex_start(pTHX)
2374 register const I32 op_type = pl_yylval.ival;
2376 if (op_type == OP_NULL) {
2377 pl_yylval.opval = PL_lex_op;
2381 if (op_type == OP_CONST || op_type == OP_READLINE) {
2382 SV *sv = tokeq(PL_lex_stuff);
2384 if (SvTYPE(sv) == SVt_PVIV) {
2385 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2387 const char * const p = SvPV_const(sv, len);
2388 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2392 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2393 PL_lex_stuff = NULL;
2394 /* Allow <FH> // "foo" */
2395 if (op_type == OP_READLINE)
2396 PL_expect = XTERMORDORDOR;
2399 else if (op_type == OP_BACKTICK && PL_lex_op) {
2400 /* readpipe() vas overriden */
2401 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2402 pl_yylval.opval = PL_lex_op;
2404 PL_lex_stuff = NULL;
2408 PL_sublex_info.super_state = PL_lex_state;
2409 PL_sublex_info.sub_inwhat = (U16)op_type;
2410 PL_sublex_info.sub_op = PL_lex_op;
2411 PL_lex_state = LEX_INTERPPUSH;
2415 pl_yylval.opval = PL_lex_op;
2425 * Create a new scope to save the lexing state. The scope will be
2426 * ended in S_sublex_done. Returns a '(', starting the function arguments
2427 * to the uc, lc, etc. found before.
2428 * Sets PL_lex_state to LEX_INTERPCONCAT.
2437 PL_lex_state = PL_sublex_info.super_state;
2438 SAVEBOOL(PL_lex_dojoin);
2439 SAVEI32(PL_lex_brackets);
2440 SAVEI32(PL_lex_allbrackets);
2441 SAVEI8(PL_lex_fakeeof);
2442 SAVEI32(PL_lex_casemods);
2443 SAVEI32(PL_lex_starts);
2444 SAVEI8(PL_lex_state);
2445 SAVEVPTR(PL_lex_inpat);
2446 SAVEI16(PL_lex_inwhat);
2447 SAVECOPLINE(PL_curcop);
2448 SAVEPPTR(PL_bufptr);
2449 SAVEPPTR(PL_bufend);
2450 SAVEPPTR(PL_oldbufptr);
2451 SAVEPPTR(PL_oldoldbufptr);
2452 SAVEPPTR(PL_last_lop);
2453 SAVEPPTR(PL_last_uni);
2454 SAVEPPTR(PL_linestart);
2455 SAVESPTR(PL_linestr);
2456 SAVEGENERICPV(PL_lex_brackstack);
2457 SAVEGENERICPV(PL_lex_casestack);
2459 PL_linestr = PL_lex_stuff;
2460 PL_lex_stuff = NULL;
2462 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2463 = SvPVX(PL_linestr);
2464 PL_bufend += SvCUR(PL_linestr);
2465 PL_last_lop = PL_last_uni = NULL;
2466 SAVEFREESV(PL_linestr);
2468 PL_lex_dojoin = FALSE;
2469 PL_lex_brackets = 0;
2470 PL_lex_allbrackets = 0;
2471 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2472 Newx(PL_lex_brackstack, 120, char);
2473 Newx(PL_lex_casestack, 12, char);
2474 PL_lex_casemods = 0;
2475 *PL_lex_casestack = '\0';
2477 PL_lex_state = LEX_INTERPCONCAT;
2478 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2480 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2481 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2482 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2483 PL_lex_inpat = PL_sublex_info.sub_op;
2485 PL_lex_inpat = NULL;
2492 * Restores lexer state after a S_sublex_push.
2499 if (!PL_lex_starts++) {
2500 SV * const sv = newSVpvs("");
2501 if (SvUTF8(PL_linestr))
2503 PL_expect = XOPERATOR;
2504 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2508 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2509 PL_lex_state = LEX_INTERPCASEMOD;
2513 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2514 assert(PL_lex_inwhat != OP_TRANSR);
2515 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2516 PL_linestr = PL_lex_repl;
2518 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2519 PL_bufend += SvCUR(PL_linestr);
2520 PL_last_lop = PL_last_uni = NULL;
2521 SAVEFREESV(PL_linestr);
2522 PL_lex_dojoin = FALSE;
2523 PL_lex_brackets = 0;
2524 PL_lex_allbrackets = 0;
2525 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2526 PL_lex_casemods = 0;
2527 *PL_lex_casestack = '\0';
2529 if (SvEVALED(PL_lex_repl)) {
2530 PL_lex_state = LEX_INTERPNORMAL;
2532 /* we don't clear PL_lex_repl here, so that we can check later
2533 whether this is an evalled subst; that means we rely on the
2534 logic to ensure sublex_done() is called again only via the
2535 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2538 PL_lex_state = LEX_INTERPCONCAT;
2548 PL_endwhite = newSVpvs("");
2549 sv_catsv(PL_endwhite, PL_thiswhite);
2553 sv_setpvs(PL_thistoken,"");
2555 PL_realtokenstart = -1;
2559 PL_bufend = SvPVX(PL_linestr);
2560 PL_bufend += SvCUR(PL_linestr);
2561 PL_expect = XOPERATOR;
2562 PL_sublex_info.sub_inwhat = 0;
2570 Extracts a pattern, double-quoted string, or transliteration. This
2573 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2574 processing a pattern (PL_lex_inpat is true), a transliteration
2575 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2577 Returns a pointer to the character scanned up to. If this is
2578 advanced from the start pointer supplied (i.e. if anything was
2579 successfully parsed), will leave an OP for the substring scanned
2580 in pl_yylval. Caller must intuit reason for not parsing further
2581 by looking at the next characters herself.
2585 constants: \N{NAME} only
2586 case and quoting: \U \Q \E
2587 stops on @ and $, but not for $ as tail anchor
2589 In transliterations:
2590 characters are VERY literal, except for - not at the start or end
2591 of the string, which indicates a range. If the range is in bytes,
2592 scan_const expands the range to the full set of intermediate
2593 characters. If the range is in utf8, the hyphen is replaced with
2594 a certain range mark which will be handled by pmtrans() in op.c.
2596 In double-quoted strings:
2598 double-quoted style: \r and \n
2599 constants: \x31, etc.
2600 deprecated backrefs: \1 (in substitution replacements)
2601 case and quoting: \U \Q \E
2604 scan_const does *not* construct ops to handle interpolated strings.
2605 It stops processing as soon as it finds an embedded $ or @ variable
2606 and leaves it to the caller to work out what's going on.
2608 embedded arrays (whether in pattern or not) could be:
2609 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2611 $ in double-quoted strings must be the symbol of an embedded scalar.
2613 $ in pattern could be $foo or could be tail anchor. Assumption:
2614 it's a tail anchor if $ is the last thing in the string, or if it's
2615 followed by one of "()| \r\n\t"
2617 \1 (backreferences) are turned into $1
2619 The structure of the code is
2620 while (there's a character to process) {
2621 handle transliteration ranges
2622 skip regexp comments /(?#comment)/ and codes /(?{code})/
2623 skip #-initiated comments in //x patterns
2624 check for embedded arrays
2625 check for embedded scalars
2627 deprecate \1 in substitution replacements
2628 handle string-changing backslashes \l \U \Q \E, etc.
2629 switch (what was escaped) {
2630 handle \- in a transliteration (becomes a literal -)
2631 if a pattern and not \N{, go treat as regular character
2632 handle \132 (octal characters)
2633 handle \x15 and \x{1234} (hex characters)
2634 handle \N{name} (named characters, also \N{3,5} in a pattern)
2635 handle \cV (control characters)
2636 handle printf-style backslashes (\f, \r, \n, etc)
2639 } (end if backslash)
2640 handle regular character
2641 } (end while character to read)
2646 S_scan_const(pTHX_ char *start)
2649 register char *send = PL_bufend; /* end of the constant */
2650 SV *sv = newSV(send - start); /* sv for the constant. See
2651 note below on sizing. */
2652 register char *s = start; /* start of the constant */
2653 register char *d = SvPVX(sv); /* destination for copies */
2654 bool dorange = FALSE; /* are we in a translit range? */
2655 bool didrange = FALSE; /* did we just finish a range? */
2656 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2657 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
2658 to be UTF8? But, this can
2659 show as true when the source
2660 isn't utf8, as for example
2661 when it is entirely composed
2664 /* Note on sizing: The scanned constant is placed into sv, which is
2665 * initialized by newSV() assuming one byte of output for every byte of
2666 * input. This routine expects newSV() to allocate an extra byte for a
2667 * trailing NUL, which this routine will append if it gets to the end of
2668 * the input. There may be more bytes of input than output (eg., \N{LATIN
2669 * CAPITAL LETTER A}), or more output than input if the constant ends up
2670 * recoded to utf8, but each time a construct is found that might increase
2671 * the needed size, SvGROW() is called. Its size parameter each time is
2672 * based on the best guess estimate at the time, namely the length used so
2673 * far, plus the length the current construct will occupy, plus room for
2674 * the trailing NUL, plus one byte for every input byte still unscanned */
2678 UV literal_endpoint = 0;
2679 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2682 PERL_ARGS_ASSERT_SCAN_CONST;
2684 assert(PL_lex_inwhat != OP_TRANSR);
2685 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2686 /* If we are doing a trans and we know we want UTF8 set expectation */
2687 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2688 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2692 while (s < send || dorange) {
2694 /* get transliterations out of the way (they're most literal) */
2695 if (PL_lex_inwhat == OP_TRANS) {
2696 /* expand a range A-Z to the full set of characters. AIE! */
2698 I32 i; /* current expanded character */
2699 I32 min; /* first character in range */
2700 I32 max; /* last character in range */
2711 char * const c = (char*)utf8_hop((U8*)d, -1);
2715 *c = (char)UTF_TO_NATIVE(0xff);
2716 /* mark the range as done, and continue */
2722 i = d - SvPVX_const(sv); /* remember current offset */
2725 SvLEN(sv) + (has_utf8 ?
2726 (512 - UTF_CONTINUATION_MARK +
2729 /* How many two-byte within 0..255: 128 in UTF-8,
2730 * 96 in UTF-8-mod. */
2732 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2734 d = SvPVX(sv) + i; /* refresh d after realloc */
2738 for (j = 0; j <= 1; j++) {
2739 char * const c = (char*)utf8_hop((U8*)d, -1);
2740 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2746 max = (U8)0xff; /* only to \xff */
2747 uvmax = uv; /* \x{100} to uvmax */
2749 d = c; /* eat endpoint chars */
2754 d -= 2; /* eat the first char and the - */
2755 min = (U8)*d; /* first char in range */
2756 max = (U8)d[1]; /* last char in range */
2763 "Invalid range \"%c-%c\" in transliteration operator",
2764 (char)min, (char)max);
2768 if (literal_endpoint == 2 &&
2769 ((isLOWER(min) && isLOWER(max)) ||
2770 (isUPPER(min) && isUPPER(max)))) {
2772 for (i = min; i <= max; i++)
2774 *d++ = NATIVE_TO_NEED(has_utf8,i);
2776 for (i = min; i <= max; i++)
2778 *d++ = NATIVE_TO_NEED(has_utf8,i);
2783 for (i = min; i <= max; i++)
2786 const U8 ch = (U8)NATIVE_TO_UTF(i);
2787 if (UNI_IS_INVARIANT(ch))
2790 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2791 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2800 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2802 *d++ = (char)UTF_TO_NATIVE(0xff);
2804 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2808 /* mark the range as done, and continue */
2812 literal_endpoint = 0;
2817 /* range begins (ignore - as first or last char) */
2818 else if (*s == '-' && s+1 < send && s != start) {
2820 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2827 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2837 literal_endpoint = 0;
2838 native_range = TRUE;
2843 /* if we get here, we're not doing a transliteration */
2845 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2846 except for the last char, which will be done separately. */
2847 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2849 while (s+1 < send && *s != ')')
2850 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2852 else if (s[2] == '{' /* This should match regcomp.c */
2853 || (s[2] == '?' && s[3] == '{'))
2856 char *regparse = s + (s[2] == '{' ? 3 : 4);
2859 while (count && (c = *regparse)) {
2860 if (c == '\\' && regparse[1])
2868 if (*regparse != ')')
2869 regparse--; /* Leave one char for continuation. */
2870 while (s < regparse)
2871 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2875 /* likewise skip #-initiated comments in //x patterns */
2876 else if (*s == '#' && PL_lex_inpat &&
2877 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
2878 while (s+1 < send && *s != '\n')
2879 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2882 /* check for embedded arrays
2883 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2885 else if (*s == '@' && s[1]) {
2886 if (isALNUM_lazy_if(s+1,UTF))
2888 if (strchr(":'{$", s[1]))
2890 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2891 break; /* in regexp, neither @+ nor @- are interpolated */
2894 /* check for embedded scalars. only stop if we're sure it's a
2897 else if (*s == '$') {
2898 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2900 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2902 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2903 "Possible unintended interpolation of $\\ in regex");
2905 break; /* in regexp, $ might be tail anchor */
2909 /* End of else if chain - OP_TRANS rejoin rest */
2912 if (*s == '\\' && s+1 < send) {
2913 char* e; /* Can be used for ending '}', etc. */
2917 /* warn on \1 - \9 in substitution replacements, but note that \11
2918 * is an octal; and \19 is \1 followed by '9' */
2919 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2920 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2922 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2927 /* string-change backslash escapes */
2928 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2932 /* In a pattern, process \N, but skip any other backslash escapes.
2933 * This is because we don't want to translate an escape sequence
2934 * into a meta symbol and have the regex compiler use the meta
2935 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2936 * in spite of this, we do have to process \N here while the proper
2937 * charnames handler is in scope. See bugs #56444 and #62056.
2938 * There is a complication because \N in a pattern may also stand
2939 * for 'match a non-nl', and not mean a charname, in which case its
2940 * processing should be deferred to the regex compiler. To be a
2941 * charname it must be followed immediately by a '{', and not look
2942 * like \N followed by a curly quantifier, i.e., not something like
2943 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2945 else if (PL_lex_inpat
2948 || regcurly(s + 1)))
2950 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2951 goto default_action;
2956 /* quoted - in transliterations */
2958 if (PL_lex_inwhat == OP_TRANS) {
2965 if ((isALPHA(*s) || isDIGIT(*s)))
2966 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2967 "Unrecognized escape \\%c passed through",
2969 /* default action is to copy the quoted character */
2970 goto default_action;
2973 /* eg. \132 indicates the octal constant 0132 */
2974 case '0': case '1': case '2': case '3':
2975 case '4': case '5': case '6': case '7':
2979 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2982 goto NUM_ESCAPE_INSERT;
2984 /* eg. \o{24} indicates the octal constant \024 */
2990 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
2996 goto NUM_ESCAPE_INSERT;
2999 /* eg. \x24 indicates the hex constant 0x24 */
3003 char* const e = strchr(s, '}');
3004 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
3005 PERL_SCAN_DISALLOW_PREFIX;
3010 yyerror("Missing right brace on \\x{}");
3014 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
3020 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
3021 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
3027 /* Insert oct or hex escaped character. There will always be
3028 * enough room in sv since such escapes will be longer than any
3029 * UTF-8 sequence they can end up as, except if they force us
3030 * to recode the rest of the string into utf8 */
3032 /* Here uv is the ordinal of the next character being added in
3033 * unicode (converted from native). */
3034 if (!UNI_IS_INVARIANT(uv)) {
3035 if (!has_utf8 && uv > 255) {
3036 /* Might need to recode whatever we have accumulated so
3037 * far if it contains any chars variant in utf8 or
3040 SvCUR_set(sv, d - SvPVX_const(sv));
3043 /* See Note on sizing above. */
3044 sv_utf8_upgrade_flags_grow(sv,
3045 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3046 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3047 d = SvPVX(sv) + SvCUR(sv);
3052 d = (char*)uvuni_to_utf8((U8*)d, uv);
3053 if (PL_lex_inwhat == OP_TRANS &&
3054 PL_sublex_info.sub_op) {
3055 PL_sublex_info.sub_op->op_private |=
3056 (PL_lex_repl ? OPpTRANS_FROM_UTF
3060 if (uv > 255 && !dorange)
3061 native_range = FALSE;
3074 /* In a non-pattern \N must be a named character, like \N{LATIN
3075 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3076 * mean to match a non-newline. For non-patterns, named
3077 * characters are converted to their string equivalents. In
3078 * patterns, named characters are not converted to their
3079 * ultimate forms for the same reasons that other escapes
3080 * aren't. Instead, they are converted to the \N{U+...} form
3081 * to get the value from the charnames that is in effect right
3082 * now, while preserving the fact that it was a named character
3083 * so that the regex compiler knows this */
3085 /* This section of code doesn't generally use the
3086 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3087 * a close examination of this macro and determined it is a
3088 * no-op except on utfebcdic variant characters. Every
3089 * character generated by this that would normally need to be
3090 * enclosed by this macro is invariant, so the macro is not
3091 * needed, and would complicate use of copy(). XXX There are
3092 * other parts of this file where the macro is used
3093 * inconsistently, but are saved by it being a no-op */
3095 /* The structure of this section of code (besides checking for
3096 * errors and upgrading to utf8) is:
3097 * Further disambiguate between the two meanings of \N, and if
3098 * not a charname, go process it elsewhere
3099 * If of form \N{U+...}, pass it through if a pattern;
3100 * otherwise convert to utf8
3101 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3102 * pattern; otherwise convert to utf8 */
3104 /* Here, s points to the 'N'; the test below is guaranteed to
3105 * succeed if we are being called on a pattern as we already
3106 * know from a test above that the next character is a '{'.
3107 * On a non-pattern \N must mean 'named sequence, which
3108 * requires braces */
3111 yyerror("Missing braces on \\N{}");
3116 /* If there is no matching '}', it is an error. */
3117 if (! (e = strchr(s, '}'))) {
3118 if (! PL_lex_inpat) {
3119 yyerror("Missing right brace on \\N{}");
3121 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3126 /* Here it looks like a named character */
3130 /* XXX This block is temporary code. \N{} implies that the
3131 * pattern is to have Unicode semantics, and therefore
3132 * currently has to be encoded in utf8. By putting it in
3133 * utf8 now, we save a whole pass in the regular expression
3134 * compiler. Once that code is changed so Unicode
3135 * semantics doesn't necessarily have to be in utf8, this
3136 * block should be removed. However, the code that parses
3137 * the output of this would have to be changed to not
3138 * necessarily expect utf8 */
3140 SvCUR_set(sv, d - SvPVX_const(sv));
3143 /* See Note on sizing above. */
3144 sv_utf8_upgrade_flags_grow(sv,
3145 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3146 /* 5 = '\N{' + cur char + NUL */
3147 (STRLEN)(send - s) + 5);
3148 d = SvPVX(sv) + SvCUR(sv);
3153 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3154 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3155 | PERL_SCAN_DISALLOW_PREFIX;
3158 /* For \N{U+...}, the '...' is a unicode value even on
3159 * EBCDIC machines */
3160 s += 2; /* Skip to next char after the 'U+' */
3162 uv = grok_hex(s, &len, &flags, NULL);
3163 if (len == 0 || len != (STRLEN)(e - s)) {
3164 yyerror("Invalid hexadecimal number in \\N{U+...}");
3171 /* On non-EBCDIC platforms, pass through to the regex
3172 * compiler unchanged. The reason we evaluated the
3173 * number above is to make sure there wasn't a syntax
3174 * error. But on EBCDIC we convert to native so
3175 * downstream code can continue to assume it's native
3177 s -= 5; /* Include the '\N{U+' */
3179 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3182 (unsigned int) UNI_TO_NATIVE(uv));
3184 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3188 else { /* Not a pattern: convert the hex to string */
3190 /* If destination is not in utf8, unconditionally
3191 * recode it to be so. This is because \N{} implies
3192 * Unicode semantics, and scalars have to be in utf8
3193 * to guarantee those semantics */
3195 SvCUR_set(sv, d - SvPVX_const(sv));
3198 /* See Note on sizing above. */
3199 sv_utf8_upgrade_flags_grow(
3201 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3202 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3203 d = SvPVX(sv) + SvCUR(sv);
3207 /* Add the string to the output */
3208 if (UNI_IS_INVARIANT(uv)) {
3211 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3214 else { /* Here is \N{NAME} but not \N{U+...}. */
3216 SV *res; /* result from charnames */
3217 const char *str; /* the string in 'res' */
3218 STRLEN len; /* its length */
3220 /* Get the value for NAME */
3221 res = newSVpvn(s, e - s);
3222 res = new_constant( NULL, 0, "charnames",
3223 /* includes all of: \N{...} */
3224 res, NULL, s - 3, e - s + 4 );
3226 /* Most likely res will be in utf8 already since the
3227 * standard charnames uses pack U, but a custom translator
3228 * can leave it otherwise, so make sure. XXX This can be
3229 * revisited to not have charnames use utf8 for characters
3230 * that don't need it when regexes don't have to be in utf8
3231 * for Unicode semantics. If doing so, remember EBCDIC */
3232 sv_utf8_upgrade(res);
3233 str = SvPV_const(res, len);
3235 /* Don't accept malformed input */
3236 if (! is_utf8_string((U8 *) str, len)) {
3237 yyerror("Malformed UTF-8 returned by \\N");
3239 else if (PL_lex_inpat) {
3241 if (! len) { /* The name resolved to an empty string */
3242 Copy("\\N{}", d, 4, char);
3246 /* In order to not lose information for the regex
3247 * compiler, pass the result in the specially made
3248 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3249 * the code points in hex of each character
3250 * returned by charnames */
3252 const char *str_end = str + len;
3253 STRLEN char_length; /* cur char's byte length */
3254 STRLEN output_length; /* and the number of bytes
3255 after this is translated
3257 const STRLEN off = d - SvPVX_const(sv);
3259 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3260 * max('U+', '.'); and 1 for NUL */
3261 char hex_string[2 * UTF8_MAXBYTES + 5];
3263 /* Get the first character of the result. */
3264 U32 uv = utf8n_to_uvuni((U8 *) str,
3269 /* The call to is_utf8_string() above hopefully
3270 * guarantees that there won't be an error. But
3271 * it's easy here to make sure. The function just
3272 * above warns and returns 0 if invalid utf8, but
3273 * it can also return 0 if the input is validly a
3274 * NUL. Disambiguate */
3275 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3276 uv = UNICODE_REPLACEMENT;
3279 /* Convert first code point to hex, including the
3280 * boiler plate before it. For all these, we
3281 * convert to native format so that downstream code
3282 * can continue to assume the input is native */
3284 my_snprintf(hex_string, sizeof(hex_string),
3286 (unsigned int) UNI_TO_NATIVE(uv));
3288 /* Make sure there is enough space to hold it */
3289 d = off + SvGROW(sv, off
3291 + (STRLEN)(send - e)
3292 + 2); /* '}' + NUL */
3294 Copy(hex_string, d, output_length, char);
3297 /* For each subsequent character, append dot and
3298 * its ordinal in hex */
3299 while ((str += char_length) < str_end) {
3300 const STRLEN off = d - SvPVX_const(sv);
3301 U32 uv = utf8n_to_uvuni((U8 *) str,
3305 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3306 uv = UNICODE_REPLACEMENT;
3310 my_snprintf(hex_string, sizeof(hex_string),
3312 (unsigned int) UNI_TO_NATIVE(uv));
3314 d = off + SvGROW(sv, off
3316 + (STRLEN)(send - e)
3317 + 2); /* '}' + NUL */
3318 Copy(hex_string, d, output_length, char);
3322 *d++ = '}'; /* Done. Add the trailing brace */
3325 else { /* Here, not in a pattern. Convert the name to a
3328 /* If destination is not in utf8, unconditionally
3329 * recode it to be so. This is because \N{} implies
3330 * Unicode semantics, and scalars have to be in utf8
3331 * to guarantee those semantics */
3333 SvCUR_set(sv, d - SvPVX_const(sv));
3336 /* See Note on sizing above. */
3337 sv_utf8_upgrade_flags_grow(sv,
3338 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3339 len + (STRLEN)(send - s) + 1);
3340 d = SvPVX(sv) + SvCUR(sv);
3342 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3344 /* See Note on sizing above. (NOTE: SvCUR() is not
3345 * set correctly here). */
3346 const STRLEN off = d - SvPVX_const(sv);
3347 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3349 Copy(str, d, len, char);
3354 /* Deprecate non-approved name syntax */
3355 if (ckWARN_d(WARN_DEPRECATED)) {
3356 bool problematic = FALSE;
3359 /* For non-ut8 input, look to see that the first
3360 * character is an alpha, then loop through the rest
3361 * checking that each is a continuation */
3363 if (! isALPHAU(*i)) problematic = TRUE;
3364 else for (i = s + 1; i < e; i++) {
3365 if (isCHARNAME_CONT(*i)) continue;
3371 /* Similarly for utf8. For invariants can check
3372 * directly. We accept anything above the latin1
3373 * range because it is immaterial to Perl if it is
3374 * correct or not, and is expensive to check. But
3375 * it is fairly easy in the latin1 range to convert
3376 * the variants into a single character and check
3378 if (UTF8_IS_INVARIANT(*i)) {
3379 if (! isALPHAU(*i)) problematic = TRUE;
3380 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3381 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*i,
3387 if (! problematic) for (i = s + UTF8SKIP(s);
3391 if (UTF8_IS_INVARIANT(*i)) {
3392 if (isCHARNAME_CONT(*i)) continue;
3393 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3395 } else if (isCHARNAME_CONT(
3397 TWO_BYTE_UTF8_TO_UNI(*i, *(i+1)))))
3406 /* The e-i passed to the final %.*s makes sure that
3407 * should the trailing NUL be missing that this
3408 * print won't run off the end of the string */
3409 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3410 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s",
3411 (int)(i - s + 1), s, (int)(e - i), i + 1);
3414 } /* End \N{NAME} */
3417 native_range = FALSE; /* \N{} is defined to be Unicode */
3419 s = e + 1; /* Point to just after the '}' */
3422 /* \c is a control character */
3426 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3429 yyerror("Missing control char name in \\c");
3433 /* printf-style backslashes, formfeeds, newlines, etc */
3435 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3438 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3441 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3444 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3447 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3450 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3453 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3459 } /* end if (backslash) */
3466 /* If we started with encoded form, or already know we want it,
3467 then encode the next character */
3468 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3472 /* One might think that it is wasted effort in the case of the
3473 * source being utf8 (this_utf8 == TRUE) to take the next character
3474 * in the source, convert it to an unsigned value, and then convert
3475 * it back again. But the source has not been validated here. The
3476 * routine that does the conversion checks for errors like
3479 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3480 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3482 SvCUR_set(sv, d - SvPVX_const(sv));
3485 /* See Note on sizing above. */
3486 sv_utf8_upgrade_flags_grow(sv,
3487 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3488 need + (STRLEN)(send - s) + 1);
3489 d = SvPVX(sv) + SvCUR(sv);
3491 } else if (need > len) {
3492 /* encoded value larger than old, may need extra space (NOTE:
3493 * SvCUR() is not set correctly here). See Note on sizing
3495 const STRLEN off = d - SvPVX_const(sv);
3496 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3500 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3502 if (uv > 255 && !dorange)
3503 native_range = FALSE;
3507 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3509 } /* while loop to process each character */
3511 /* terminate the string and set up the sv */
3513 SvCUR_set(sv, d - SvPVX_const(sv));
3514 if (SvCUR(sv) >= SvLEN(sv))
3515 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3518 if (PL_encoding && !has_utf8) {
3519 sv_recode_to_utf8(sv, PL_encoding);
3525 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3526 PL_sublex_info.sub_op->op_private |=
3527 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3531 /* shrink the sv if we allocated more than we used */
3532 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3533 SvPV_shrink_to_cur(sv);
3536 /* return the substring (via pl_yylval) only if we parsed anything */
3537 if (s > PL_bufptr) {
3538 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3539 const char *const key = PL_lex_inpat ? "qr" : "q";
3540 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3544 if (PL_lex_inwhat == OP_TRANS) {
3547 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3555 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3558 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3565 * Returns TRUE if there's more to the expression (e.g., a subscript),
3568 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3570 * ->[ and ->{ return TRUE
3571 * { and [ outside a pattern are always subscripts, so return TRUE
3572 * if we're outside a pattern and it's not { or [, then return FALSE
3573 * if we're in a pattern and the first char is a {
3574 * {4,5} (any digits around the comma) returns FALSE
3575 * if we're in a pattern and the first char is a [
3577 * [SOMETHING] has a funky algorithm to decide whether it's a
3578 * character class or not. It has to deal with things like
3579 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3580 * anything else returns TRUE
3583 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3586 S_intuit_more(pTHX_ register char *s)
3590 PERL_ARGS_ASSERT_INTUIT_MORE;
3592 if (PL_lex_brackets)
3594 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3596 if (*s != '{' && *s != '[')
3601 /* In a pattern, so maybe we have {n,m}. */
3609 /* On the other hand, maybe we have a character class */
3612 if (*s == ']' || *s == '^')
3615 /* this is terrifying, and it works */
3616 int weight = 2; /* let's weigh the evidence */
3618 unsigned char un_char = 255, last_un_char;
3619 const char * const send = strchr(s,']');
3620 char tmpbuf[sizeof PL_tokenbuf * 4];
3622 if (!send) /* has to be an expression */
3625 Zero(seen,256,char);
3628 else if (isDIGIT(*s)) {
3630 if (isDIGIT(s[1]) && s[2] == ']')
3636 for (; s < send; s++) {
3637 last_un_char = un_char;
3638 un_char = (unsigned char)*s;
3643 weight -= seen[un_char] * 10;
3644 if (isALNUM_lazy_if(s+1,UTF)) {
3646 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3647 len = (int)strlen(tmpbuf);
3648 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3649 UTF ? SVf_UTF8 : 0, SVt_PV))
3654 else if (*s == '$' && s[1] &&
3655 strchr("[#!%*<>()-=",s[1])) {
3656 if (/*{*/ strchr("])} =",s[2]))
3665 if (strchr("wds]",s[1]))
3667 else if (seen[(U8)'\''] || seen[(U8)'"'])
3669 else if (strchr("rnftbxcav",s[1]))
3671 else if (isDIGIT(s[1])) {
3673 while (s[1] && isDIGIT(s[1]))
3683 if (strchr("aA01! ",last_un_char))
3685 if (strchr("zZ79~",s[1]))
3687 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3688 weight -= 5; /* cope with negative subscript */
3691 if (!isALNUM(last_un_char)
3692 && !(last_un_char == '$' || last_un_char == '@'
3693 || last_un_char == '&')
3694 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3699 if (keyword(tmpbuf, d - tmpbuf, 0))
3702 if (un_char == last_un_char + 1)
3704 weight -= seen[un_char];
3709 if (weight >= 0) /* probably a character class */
3719 * Does all the checking to disambiguate
3721 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3722 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3724 * First argument is the stuff after the first token, e.g. "bar".
3726 * Not a method if bar is a filehandle.
3727 * Not a method if foo is a subroutine prototyped to take a filehandle.
3728 * Not a method if it's really "Foo $bar"
3729 * Method if it's "foo $bar"
3730 * Not a method if it's really "print foo $bar"
3731 * Method if it's really "foo package::" (interpreted as package->foo)
3732 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3733 * Not a method if bar is a filehandle or package, but is quoted with
3738 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3741 char *s = start + (*start == '$');
3742 char tmpbuf[sizeof PL_tokenbuf];
3749 PERL_ARGS_ASSERT_INTUIT_METHOD;
3752 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3756 const char *proto = CvPROTO(cv);
3767 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3768 /* start is the beginning of the possible filehandle/object,
3769 * and s is the end of it
3770 * tmpbuf is a copy of it
3773 if (*start == '$') {
3774 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3775 isUPPER(*PL_tokenbuf))
3778 len = start - SvPVX(PL_linestr);
3782 start = SvPVX(PL_linestr) + len;
3786 return *s == '(' ? FUNCMETH : METHOD;
3788 if (!keyword(tmpbuf, len, 0)) {
3789 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3793 soff = s - SvPVX(PL_linestr);
3797 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
3798 if (indirgv && GvCVu(indirgv))
3800 /* filehandle or package name makes it a method */
3801 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
3803 soff = s - SvPVX(PL_linestr);
3806 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3807 return 0; /* no assumptions -- "=>" quotes bareword */
3809 start_force(PL_curforce);
3810 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3811 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3812 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3814 curmad('X', newSVpvn_flags(start,SvPVX(PL_linestr) + soff - start,
3815 ( UTF ? SVf_UTF8 : 0 )));
3820 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3822 return *s == '(' ? FUNCMETH : METHOD;
3828 /* Encoded script support. filter_add() effectively inserts a
3829 * 'pre-processing' function into the current source input stream.
3830 * Note that the filter function only applies to the current source file
3831 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3833 * The datasv parameter (which may be NULL) can be used to pass
3834 * private data to this instance of the filter. The filter function
3835 * can recover the SV using the FILTER_DATA macro and use it to
3836 * store private buffers and state information.
3838 * The supplied datasv parameter is upgraded to a PVIO type
3839 * and the IoDIRP/IoANY field is used to store the function pointer,
3840 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3841 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3842 * private use must be set using malloc'd pointers.
3846 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3855 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
3856 Perl_croak(aTHX_ "Source filters apply only to byte streams");
3858 if (!PL_rsfp_filters)
3859 PL_rsfp_filters = newAV();
3862 SvUPGRADE(datasv, SVt_PVIO);
3863 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3864 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3865 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3866 FPTR2DPTR(void *, IoANY(datasv)),
3867 SvPV_nolen(datasv)));
3868 av_unshift(PL_rsfp_filters, 1);
3869 av_store(PL_rsfp_filters, 0, datasv) ;
3871 !PL_parser->filtered
3872 && PL_parser->lex_flags & LEX_EVALBYTES
3873 && PL_bufptr < PL_bufend
3875 const char *s = PL_bufptr;
3876 while (s < PL_bufend) {
3878 SV *linestr = PL_parser->linestr;
3879 char *buf = SvPVX(linestr);
3880 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
3881 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
3882 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
3883 STRLEN const linestart_pos = PL_parser->linestart - buf;
3884 STRLEN const last_uni_pos =
3885 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
3886 STRLEN const last_lop_pos =
3887 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
3888 av_push(PL_rsfp_filters, linestr);
3889 PL_parser->linestr =
3890 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
3891 buf = SvPVX(PL_parser->linestr);
3892 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
3893 PL_parser->bufptr = buf + bufptr_pos;
3894 PL_parser->oldbufptr = buf + oldbufptr_pos;
3895 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
3896 PL_parser->linestart = buf + linestart_pos;
3897 if (PL_parser->last_uni)
3898 PL_parser->last_uni = buf + last_uni_pos;
3899 if (PL_parser->last_lop)
3900 PL_parser->last_lop = buf + last_lop_pos;
3901 SvLEN(linestr) = SvCUR(linestr);
3902 SvCUR(linestr) = s-SvPVX(linestr);
3903 PL_parser->filtered = 1;
3913 /* Delete most recently added instance of this filter function. */
3915 Perl_filter_del(pTHX_ filter_t funcp)
3920 PERL_ARGS_ASSERT_FILTER_DEL;
3923 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3924 FPTR2DPTR(void*, funcp)));
3926 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3928 /* if filter is on top of stack (usual case) just pop it off */
3929 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3930 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3931 sv_free(av_pop(PL_rsfp_filters));
3935 /* we need to search for the correct entry and clear it */
3936 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3940 /* Invoke the idxth filter function for the current rsfp. */
3941 /* maxlen 0 = read one text line */
3943 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3948 /* This API is bad. It should have been using unsigned int for maxlen.
3949 Not sure if we want to change the API, but if not we should sanity
3950 check the value here. */
3951 unsigned int correct_length
3960 PERL_ARGS_ASSERT_FILTER_READ;
3962 if (!PL_parser || !PL_rsfp_filters)
3964 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3965 /* Provide a default input filter to make life easy. */
3966 /* Note that we append to the line. This is handy. */
3967 DEBUG_P(PerlIO_printf(Perl_debug_log,
3968 "filter_read %d: from rsfp\n", idx));
3969 if (correct_length) {
3972 const int old_len = SvCUR(buf_sv);
3974 /* ensure buf_sv is large enough */
3975 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3976 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3977 correct_length)) <= 0) {
3978 if (PerlIO_error(PL_rsfp))
3979 return -1; /* error */
3981 return 0 ; /* end of file */
3983 SvCUR_set(buf_sv, old_len + len) ;
3984 SvPVX(buf_sv)[old_len + len] = '\0';
3987 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3988 if (PerlIO_error(PL_rsfp))
3989 return -1; /* error */
3991 return 0 ; /* end of file */
3994 return SvCUR(buf_sv);
3996 /* Skip this filter slot if filter has been deleted */
3997 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3998 DEBUG_P(PerlIO_printf(Perl_debug_log,
3999 "filter_read %d: skipped (filter deleted)\n",
4001 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4003 if (SvTYPE(datasv) != SVt_PVIO) {
4004 if (correct_length) {
4006 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4007 if (!remainder) return 0; /* eof */
4008 if (correct_length > remainder) correct_length = remainder;
4009 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4010 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4013 const char *s = SvEND(datasv);
4014 const char *send = SvPVX(datasv) + SvLEN(datasv);
4022 if (s == send) return 0; /* eof */
4023 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4024 SvCUR_set(datasv, s-SvPVX(datasv));
4026 return SvCUR(buf_sv);
4028 /* Get function pointer hidden within datasv */
4029 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4030 DEBUG_P(PerlIO_printf(Perl_debug_log,
4031 "filter_read %d: via function %p (%s)\n",
4032 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4033 /* Call function. The function is expected to */
4034 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4035 /* Return: <0:error, =0:eof, >0:not eof */
4036 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4040 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
4044 PERL_ARGS_ASSERT_FILTER_GETS;
4046 #ifdef PERL_CR_FILTER
4047 if (!PL_rsfp_filters) {
4048 filter_add(S_cr_textfilter,NULL);
4051 if (PL_rsfp_filters) {
4053 SvCUR_set(sv, 0); /* start with empty line */
4054 if (FILTER_READ(0, sv, 0) > 0)
4055 return ( SvPVX(sv) ) ;
4060 return (sv_gets(sv, PL_rsfp, append));
4064 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4069 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4071 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4075 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4076 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4078 return GvHV(gv); /* Foo:: */
4081 /* use constant CLASS => 'MyClass' */
4082 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4083 if (gv && GvCV(gv)) {
4084 SV * const sv = cv_const_sv(GvCV(gv));
4086 pkgname = SvPV_const(sv, len);
4089 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4093 * S_readpipe_override
4094 * Check whether readpipe() is overridden, and generates the appropriate
4095 * optree, provided sublex_start() is called afterwards.
4098 S_readpipe_override(pTHX)
4101 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4102 pl_yylval.ival = OP_BACKTICK;
4104 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4106 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4107 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4108 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4110 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4111 op_append_elem(OP_LIST,
4112 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4113 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4120 * The intent of this yylex wrapper is to minimize the changes to the
4121 * tokener when we aren't interested in collecting madprops. It remains
4122 * to be seen how successful this strategy will be...
4129 char *s = PL_bufptr;
4131 /* make sure PL_thiswhite is initialized */
4135 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
4136 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4137 return S_pending_ident(aTHX);
4139 /* previous token ate up our whitespace? */
4140 if (!PL_lasttoke && PL_nextwhite) {
4141 PL_thiswhite = PL_nextwhite;
4145 /* isolate the token, and figure out where it is without whitespace */
4146 PL_realtokenstart = -1;
4150 assert(PL_curforce < 0);
4152 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4153 if (!PL_thistoken) {
4154 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4155 PL_thistoken = newSVpvs("");
4157 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4158 PL_thistoken = newSVpvn(tstart, s - tstart);
4161 if (PL_thismad) /* install head */
4162 CURMAD('X', PL_thistoken);
4165 /* last whitespace of a sublex? */
4166 if (optype == ')' && PL_endwhite) {
4167 CURMAD('X', PL_endwhite);
4172 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4173 if (!PL_thiswhite && !PL_endwhite && !optype) {
4174 sv_free(PL_thistoken);
4179 /* put off final whitespace till peg */
4180 if (optype == ';' && !PL_rsfp && !PL_parser->filtered) {
4181 PL_nextwhite = PL_thiswhite;
4184 else if (PL_thisopen) {
4185 CURMAD('q', PL_thisopen);
4187 sv_free(PL_thistoken);
4191 /* Store actual token text as madprop X */
4192 CURMAD('X', PL_thistoken);
4196 /* add preceding whitespace as madprop _ */
4197 CURMAD('_', PL_thiswhite);
4201 /* add quoted material as madprop = */
4202 CURMAD('=', PL_thisstuff);
4206 /* add terminating quote as madprop Q */
4207 CURMAD('Q', PL_thisclose);
4211 /* special processing based on optype */
4215 /* opval doesn't need a TOKEN since it can already store mp */
4225 if (pl_yylval.opval)
4226 append_madprops(PL_thismad, pl_yylval.opval, 0);
4234 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4243 /* remember any fake bracket that lexer is about to discard */
4244 if (PL_lex_brackets == 1 &&
4245 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4248 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4251 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4252 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4255 break; /* don't bother looking for trailing comment */
4264 /* attach a trailing comment to its statement instead of next token */
4268 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4270 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4272 if (*s == '\n' || *s == '#') {
4273 while (s < PL_bufend && *s != '\n')
4277 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4278 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4295 /* Create new token struct. Note: opvals return early above. */
4296 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4303 S_tokenize_use(pTHX_ int is_use, char *s) {
4306 PERL_ARGS_ASSERT_TOKENIZE_USE;
4308 if (PL_expect != XSTATE)
4309 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4310 is_use ? "use" : "no"));
4312 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4313 s = force_version(s, TRUE);
4314 if (*s == ';' || *s == '}'
4315 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4316 start_force(PL_curforce);
4317 NEXTVAL_NEXTTOKE.opval = NULL;
4320 else if (*s == 'v') {
4321 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4322 s = force_version(s, FALSE);
4326 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4327 s = force_version(s, FALSE);
4329 pl_yylval.ival = is_use;
4333 static const char* const exp_name[] =
4334 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4335 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4339 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4341 S_word_takes_any_delimeter(char *p, STRLEN len)
4343 return (len == 1 && strchr("msyq", p[0])) ||
4345 (p[0] == 't' && p[1] == 'r') ||
4346 (p[0] == 'q' && strchr("qwxr", p[1]))));
4352 Works out what to call the token just pulled out of the input
4353 stream. The yacc parser takes care of taking the ops we return and
4354 stitching them into a tree.
4360 if read an identifier
4361 if we're in a my declaration
4362 croak if they tried to say my($foo::bar)
4363 build the ops for a my() declaration
4364 if it's an access to a my() variable
4365 are we in a sort block?
4366 croak if my($a); $a <=> $b
4367 build ops for access to a my() variable
4368 if in a dq string, and they've said @foo and we can't find @foo
4370 build ops for a bareword
4371 if we already built the token before, use it.
4376 #pragma segment Perl_yylex
4382 register char *s = PL_bufptr;
4388 /* orig_keyword, gvp, and gv are initialized here because
4389 * jump to the label just_a_word_zero can bypass their
4390 * initialization later. */
4391 I32 orig_keyword = 0;
4396 SV* tmp = newSVpvs("");
4397 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4398 (IV)CopLINE(PL_curcop),
4399 lex_state_names[PL_lex_state],
4400 exp_name[PL_expect],
4401 pv_display(tmp, s, strlen(s), 0, 60));
4404 /* check if there's an identifier for us to look at */
4405 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4406 return REPORT(S_pending_ident(aTHX));
4408 /* no identifier pending identification */
4410 switch (PL_lex_state) {
4412 case LEX_NORMAL: /* Some compilers will produce faster */
4413 case LEX_INTERPNORMAL: /* code if we comment these out. */
4417 /* when we've already built the next token, just pull it out of the queue */
4421 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4423 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4424 PL_nexttoke[PL_lasttoke].next_mad = 0;
4425 if (PL_thismad && PL_thismad->mad_key == '_') {
4426 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4427 PL_thismad->mad_val = 0;
4428 mad_free(PL_thismad);
4433 PL_lex_state = PL_lex_defer;
4434 PL_expect = PL_lex_expect;
4435 PL_lex_defer = LEX_NORMAL;
4436 if (!PL_nexttoke[PL_lasttoke].next_type)
4441 pl_yylval = PL_nextval[PL_nexttoke];
4443 PL_lex_state = PL_lex_defer;
4444 PL_expect = PL_lex_expect;
4445 PL_lex_defer = LEX_NORMAL;
4451 next_type = PL_nexttoke[PL_lasttoke].next_type;
4453 next_type = PL_nexttype[PL_nexttoke];
4455 if (next_type & (7<<24)) {
4456 if (next_type & (1<<24)) {
4457 if (PL_lex_brackets > 100)
4458 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4459 PL_lex_brackstack[PL_lex_brackets++] =
4460 (char) ((next_type >> 16) & 0xff);
4462 if (next_type & (2<<24))
4463 PL_lex_allbrackets++;
4464 if (next_type & (4<<24))
4465 PL_lex_allbrackets--;
4466 next_type &= 0xffff;
4469 /* FIXME - can these be merged? */
4472 return REPORT(next_type);
4476 /* interpolated case modifiers like \L \U, including \Q and \E.
4477 when we get here, PL_bufptr is at the \
4479 case LEX_INTERPCASEMOD:
4481 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4482 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
4484 /* handle \E or end of string */
4485 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4487 if (PL_lex_casemods) {
4488 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4489 PL_lex_casestack[PL_lex_casemods] = '\0';
4491 if (PL_bufptr != PL_bufend
4492 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
4494 PL_lex_state = LEX_INTERPCONCAT;
4497 PL_thistoken = newSVpvs("\\E");
4500 PL_lex_allbrackets--;
4503 else if ( PL_bufptr != PL_bufend && PL_bufptr[1] == 'E' ) {
4504 /* Got an unpaired \E */
4505 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4506 "Useless use of \\E");
4509 while (PL_bufptr != PL_bufend &&
4510 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4512 PL_thiswhite = newSVpvs("");
4513 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4517 if (PL_bufptr != PL_bufend)
4520 PL_lex_state = LEX_INTERPCONCAT;
4524 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4525 "### Saw case modifier\n"); });
4527 if (s[1] == '\\' && s[2] == 'E') {
4530 PL_thiswhite = newSVpvs("");
4531 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4534 PL_lex_state = LEX_INTERPCONCAT;
4539 if (!PL_madskills) /* when just compiling don't need correct */
4540 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4541 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4542 if ((*s == 'L' || *s == 'U') &&
4543 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4544 PL_lex_casestack[--PL_lex_casemods] = '\0';
4545 PL_lex_allbrackets--;
4548 if (PL_lex_casemods > 10)
4549 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4550 PL_lex_casestack[PL_lex_casemods++] = *s;
4551 PL_lex_casestack[PL_lex_casemods] = '\0';
4552 PL_lex_state = LEX_INTERPCONCAT;
4553 start_force(PL_curforce);
4554 NEXTVAL_NEXTTOKE.ival = 0;
4555 force_next((2<<24)|'(');
4556 start_force(PL_curforce);
4558 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4560 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4562 NEXTVAL_NEXTTOKE.ival = OP_LC;
4564 NEXTVAL_NEXTTOKE.ival = OP_UC;
4566 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4568 Perl_croak(aTHX_ "panic: yylex");
4570 SV* const tmpsv = newSVpvs("\\ ");
4571 /* replace the space with the character we want to escape
4573 SvPVX(tmpsv)[1] = *s;
4579 if (PL_lex_starts) {
4585 sv_free(PL_thistoken);
4586 PL_thistoken = newSVpvs("");
4589 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4590 if (PL_lex_casemods == 1 && PL_lex_inpat)
4599 case LEX_INTERPPUSH:
4600 return REPORT(sublex_push());
4602 case LEX_INTERPSTART:
4603 if (PL_bufptr == PL_bufend)
4604 return REPORT(sublex_done());
4605 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4606 "### Interpolated variable\n"); });
4608 PL_lex_dojoin = (*PL_bufptr == '@');
4609 PL_lex_state = LEX_INTERPNORMAL;
4610 if (PL_lex_dojoin) {
4611 start_force(PL_curforce);
4612 NEXTVAL_NEXTTOKE.ival = 0;
4614 start_force(PL_curforce);
4615 force_ident("\"", '$');
4616 start_force(PL_curforce);
4617 NEXTVAL_NEXTTOKE.ival = 0;
4619 start_force(PL_curforce);
4620 NEXTVAL_NEXTTOKE.ival = 0;
4621 force_next((2<<24)|'(');
4622 start_force(PL_curforce);
4623 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4626 if (PL_lex_starts++) {
4631 sv_free(PL_thistoken);
4632 PL_thistoken = newSVpvs("");
4635 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4636 if (!PL_lex_casemods && PL_lex_inpat)
4643 case LEX_INTERPENDMAYBE:
4644 if (intuit_more(PL_bufptr)) {
4645 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4651 if (PL_lex_dojoin) {
4652 PL_lex_dojoin = FALSE;
4653 PL_lex_state = LEX_INTERPCONCAT;
4657 sv_free(PL_thistoken);
4658 PL_thistoken = newSVpvs("");
4661 PL_lex_allbrackets--;
4664 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4665 && SvEVALED(PL_lex_repl))
4667 if (PL_bufptr != PL_bufend)
4668 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4672 case LEX_INTERPCONCAT:
4674 if (PL_lex_brackets)
4675 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4677 if (PL_bufptr == PL_bufend)
4678 return REPORT(sublex_done());
4680 if (SvIVX(PL_linestr) == '\'') {
4681 SV *sv = newSVsv(PL_linestr);
4684 else if ( PL_hints & HINT_NEW_RE )
4685 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4686 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4690 s = scan_const(PL_bufptr);
4692 PL_lex_state = LEX_INTERPCASEMOD;
4694 PL_lex_state = LEX_INTERPSTART;
4697 if (s != PL_bufptr) {
4698 start_force(PL_curforce);
4700 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4702 NEXTVAL_NEXTTOKE = pl_yylval;
4705 if (PL_lex_starts++) {
4709 sv_free(PL_thistoken);
4710 PL_thistoken = newSVpvs("");
4713 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4714 if (!PL_lex_casemods && PL_lex_inpat)
4727 PL_lex_state = LEX_NORMAL;
4728 s = scan_formline(PL_bufptr);
4729 if (!PL_lex_formbrack)
4735 PL_oldoldbufptr = PL_oldbufptr;
4741 sv_free(PL_thistoken);
4744 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4748 if (isIDFIRST_lazy_if(s,UTF))
4751 unsigned char c = *s;
4752 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4753 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4754 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4759 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4763 goto fake_eof; /* emulate EOF on ^D or ^Z */
4769 if (!PL_rsfp && (!PL_parser->filtered || s+1 < PL_bufend)) {
4772 if (PL_lex_brackets &&
4773 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
4774 yyerror((const char *)
4776 ? "Format not terminated"
4777 : "Missing right curly or square bracket"));
4779 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4780 "### Tokener got EOF\n");
4784 if (s++ < PL_bufend)
4785 goto retry; /* ignore stray nulls */
4788 if (!PL_in_eval && !PL_preambled) {
4789 PL_preambled = TRUE;
4795 /* Generate a string of Perl code to load the debugger.
4796 * If PERL5DB is set, it will return the contents of that,
4797 * otherwise a compile-time require of perl5db.pl. */
4799 const char * const pdb = PerlEnv_getenv("PERL5DB");
4802 sv_setpv(PL_linestr, pdb);
4803 sv_catpvs(PL_linestr,";");
4805 SETERRNO(0,SS_NORMAL);
4806 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4809 sv_setpvs(PL_linestr,"");
4810 if (PL_preambleav) {
4811 SV **svp = AvARRAY(PL_preambleav);
4812 SV **const end = svp + AvFILLp(PL_preambleav);
4814 sv_catsv(PL_linestr, *svp);
4816 sv_catpvs(PL_linestr, ";");
4818 sv_free(MUTABLE_SV(PL_preambleav));
4819 PL_preambleav = NULL;
4822 sv_catpvs(PL_linestr,
4823 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4824 if (PL_minus_n || PL_minus_p) {
4825 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4827 sv_catpvs(PL_linestr,"chomp;");
4830 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4831 || *PL_splitstr == '"')
4832 && strchr(PL_splitstr + 1, *PL_splitstr))
4833 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4835 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4836 bytes can be used as quoting characters. :-) */
4837 const char *splits = PL_splitstr;
4838 sv_catpvs(PL_linestr, "our @F=split(q\0");
4841 if (*splits == '\\')
4842 sv_catpvn(PL_linestr, splits, 1);
4843 sv_catpvn(PL_linestr, splits, 1);
4844 } while (*splits++);
4845 /* This loop will embed the trailing NUL of
4846 PL_linestr as the last thing it does before
4848 sv_catpvs(PL_linestr, ");");
4852 sv_catpvs(PL_linestr,"our @F=split(' ');");
4855 sv_catpvs(PL_linestr, "\n");
4856 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4857 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4858 PL_last_lop = PL_last_uni = NULL;
4859 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4860 update_debugger_info(PL_linestr, NULL, 0);
4865 bof = PL_rsfp ? TRUE : FALSE;
4868 fake_eof = LEX_FAKE_EOF;
4870 PL_bufptr = PL_bufend;
4871 CopLINE_inc(PL_curcop);
4872 if (!lex_next_chunk(fake_eof)) {
4873 CopLINE_dec(PL_curcop);
4875 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4877 CopLINE_dec(PL_curcop);
4880 PL_realtokenstart = -1;
4883 /* If it looks like the start of a BOM or raw UTF-16,
4884 * check if it in fact is. */
4885 if (bof && PL_rsfp &&
4890 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
4891 bof = (offset == (Off_t)SvCUR(PL_linestr));
4892 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
4893 /* offset may include swallowed CR */
4895 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
4898 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4899 s = swallow_bom((U8*)s);
4902 if (PL_parser->in_pod) {
4903 /* Incest with pod. */
4906 sv_catsv(PL_thiswhite, PL_linestr);
4908 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4909 sv_setpvs(PL_linestr, "");
4910 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4911 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4912 PL_last_lop = PL_last_uni = NULL;
4913 PL_parser->in_pod = 0;
4916 if (PL_rsfp || PL_parser->filtered)
4918 } while (PL_parser->in_pod);
4919 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4920 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4921 PL_last_lop = PL_last_uni = NULL;
4922 if (CopLINE(PL_curcop) == 1) {
4923 while (s < PL_bufend && isSPACE(*s))
4925 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4929 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4933 if (*s == '#' && *(s+1) == '!')
4935 #ifdef ALTERNATE_SHEBANG
4937 static char const as[] = ALTERNATE_SHEBANG;
4938 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4939 d = s + (sizeof(as) - 1);
4941 #endif /* ALTERNATE_SHEBANG */
4950 while (*d && !isSPACE(*d))
4954 #ifdef ARG_ZERO_IS_SCRIPT
4955 if (ipathend > ipath) {
4957 * HP-UX (at least) sets argv[0] to the script name,
4958 * which makes $^X incorrect. And Digital UNIX and Linux,
4959 * at least, set argv[0] to the basename of the Perl
4960 * interpreter. So, having found "#!", we'll set it right.
4962 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4964 assert(SvPOK(x) || SvGMAGICAL(x));
4965 if (sv_eq(x, CopFILESV(PL_curcop))) {
4966 sv_setpvn(x, ipath, ipathend - ipath);
4972 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4973 const char * const lstart = SvPV_const(x,llen);
4975 bstart += blen - llen;
4976 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4977 sv_setpvn(x, ipath, ipathend - ipath);
4982 TAINT_NOT; /* $^X is always tainted, but that's OK */
4984 #endif /* ARG_ZERO_IS_SCRIPT */
4989 d = instr(s,"perl -");
4991 d = instr(s,"perl");
4993 /* avoid getting into infinite loops when shebang
4994 * line contains "Perl" rather than "perl" */
4996 for (d = ipathend-4; d >= ipath; --d) {
4997 if ((*d == 'p' || *d == 'P')
4998 && !ibcmp(d, "perl", 4))
5008 #ifdef ALTERNATE_SHEBANG
5010 * If the ALTERNATE_SHEBANG on this system starts with a
5011 * character that can be part of a Perl expression, then if
5012 * we see it but not "perl", we're probably looking at the
5013 * start of Perl code, not a request to hand off to some
5014 * other interpreter. Similarly, if "perl" is there, but
5015 * not in the first 'word' of the line, we assume the line
5016 * contains the start of the Perl program.
5018 if (d && *s != '#') {
5019 const char *c = ipath;
5020 while (*c && !strchr("; \t\r\n\f\v#", *c))
5023 d = NULL; /* "perl" not in first word; ignore */
5025 *s = '#'; /* Don't try to parse shebang line */
5027 #endif /* ALTERNATE_SHEBANG */
5032 !instr(s,"indir") &&
5033 instr(PL_origargv[0],"perl"))
5040 while (s < PL_bufend && isSPACE(*s))
5042 if (s < PL_bufend) {
5043 Newx(newargv,PL_origargc+3,char*);
5045 while (s < PL_bufend && !isSPACE(*s))
5048 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
5051 newargv = PL_origargv;
5054 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
5056 Perl_croak(aTHX_ "Can't exec %s", ipath);
5059 while (*d && !isSPACE(*d))
5061 while (SPACE_OR_TAB(*d))
5065 const bool switches_done = PL_doswitches;
5066 const U32 oldpdb = PL_perldb;
5067 const bool oldn = PL_minus_n;
5068 const bool oldp = PL_minus_p;
5072 bool baduni = FALSE;
5074 const char *d2 = d1 + 1;
5075 if (parse_unicode_opts((const char **)&d2)
5079 if (baduni || *d1 == 'M' || *d1 == 'm') {
5080 const char * const m = d1;
5081 while (*d1 && !isSPACE(*d1))
5083 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
5086 d1 = moreswitches(d1);
5088 if (PL_doswitches && !switches_done) {
5089 int argc = PL_origargc;
5090 char **argv = PL_origargv;
5093 } while (argc && argv[0][0] == '-' && argv[0][1]);
5094 init_argv_symbols(argc,argv);
5096 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
5097 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
5098 /* if we have already added "LINE: while (<>) {",
5099 we must not do it again */
5101 sv_setpvs(PL_linestr, "");
5102 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5103 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5104 PL_last_lop = PL_last_uni = NULL;
5105 PL_preambled = FALSE;
5106 if (PERLDB_LINE || PERLDB_SAVESRC)
5107 (void)gv_fetchfile(PL_origfilename);
5114 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5116 PL_lex_state = LEX_FORMLINE;
5121 #ifdef PERL_STRICT_CR
5122 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5124 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5126 case ' ': case '\t': case '\f': case 013:
5128 PL_realtokenstart = -1;
5130 PL_thiswhite = newSVpvs("");
5131 sv_catpvn(PL_thiswhite, s, 1);
5138 PL_realtokenstart = -1;
5142 if (PL_lex_state != LEX_NORMAL ||
5143 (PL_in_eval && !PL_rsfp && !PL_parser->filtered)) {
5144 if (*s == '#' && s == PL_linestart && PL_in_eval
5145 && !PL_rsfp && !PL_parser->filtered) {
5146 /* handle eval qq[#line 1 "foo"\n ...] */
5147 CopLINE_dec(PL_curcop);
5150 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
5152 if (!PL_in_eval || PL_rsfp || PL_parser->filtered)
5157 while (d < PL_bufend && *d != '\n')
5161 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5162 Perl_croak(aTHX_ "panic: input overflow");
5165 PL_thiswhite = newSVpvn(s, d - s);
5170 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5172 PL_lex_state = LEX_FORMLINE;
5178 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
5179 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
5182 TOKEN(PEG); /* make sure any #! line is accessible */
5187 /* if (PL_madskills && PL_lex_formbrack) { */
5189 while (d < PL_bufend && *d != '\n')
5193 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5194 Perl_croak(aTHX_ "panic: input overflow");
5195 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
5197 PL_thiswhite = newSVpvs("");
5198 if (CopLINE(PL_curcop) == 1) {
5199 sv_setpvs(PL_thiswhite, "");
5202 sv_catpvn(PL_thiswhite, s, d - s);
5216 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
5224 while (s < PL_bufend && SPACE_OR_TAB(*s))
5227 if (strnEQ(s,"=>",2)) {
5228 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
5229 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5230 OPERATOR('-'); /* unary minus */
5232 PL_last_uni = PL_oldbufptr;
5234 case 'r': ftst = OP_FTEREAD; break;
5235 case 'w': ftst = OP_FTEWRITE; break;
5236 case 'x': ftst = OP_FTEEXEC; break;
5237 case 'o': ftst = OP_FTEOWNED; break;
5238 case 'R': ftst = OP_FTRREAD; break;
5239 case 'W': ftst = OP_FTRWRITE; break;
5240 case 'X': ftst = OP_FTREXEC; break;
5241 case 'O': ftst = OP_FTROWNED; break;
5242 case 'e': ftst = OP_FTIS; break;
5243 case 'z': ftst = OP_FTZERO; break;
5244 case 's': ftst = OP_FTSIZE; break;
5245 case 'f': ftst = OP_FTFILE; break;
5246 case 'd': ftst = OP_FTDIR; break;
5247 case 'l': ftst = OP_FTLINK; break;
5248 case 'p': ftst = OP_FTPIPE; break;
5249 case 'S': ftst = OP_FTSOCK; break;
5250 case 'u': ftst = OP_FTSUID; break;
5251 case 'g': ftst = OP_FTSGID; break;
5252 case 'k': ftst = OP_FTSVTX; break;
5253 case 'b': ftst = OP_FTBLK; break;
5254 case 'c': ftst = OP_FTCHR; break;
5255 case 't': ftst = OP_FTTTY; break;
5256 case 'T': ftst = OP_FTTEXT; break;
5257 case 'B': ftst = OP_FTBINARY; break;
5258 case 'M': case 'A': case 'C':
5259 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5261 case 'M': ftst = OP_FTMTIME; break;
5262 case 'A': ftst = OP_FTATIME; break;
5263 case 'C': ftst = OP_FTCTIME; break;
5271 PL_last_lop_op = (OPCODE)ftst;
5272 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5273 "### Saw file test %c\n", (int)tmp);
5278 /* Assume it was a minus followed by a one-letter named
5279 * subroutine call (or a -bareword), then. */
5280 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5281 "### '-%c' looked like a file test but was not\n",
5288 const char tmp = *s++;
5291 if (PL_expect == XOPERATOR)
5296 else if (*s == '>') {
5299 if (isIDFIRST_lazy_if(s,UTF)) {
5300 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5308 if (PL_expect == XOPERATOR) {
5309 if (*s == '=' && !PL_lex_allbrackets &&
5310 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5317 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5319 OPERATOR('-'); /* unary minus */
5325 const char tmp = *s++;
5328 if (PL_expect == XOPERATOR)
5333 if (PL_expect == XOPERATOR) {
5334 if (*s == '=' && !PL_lex_allbrackets &&
5335 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5342 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5349 if (PL_expect != XOPERATOR) {
5350 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5351 PL_expect = XOPERATOR;
5352 force_ident(PL_tokenbuf, '*');
5360 if (*s == '=' && !PL_lex_allbrackets &&
5361 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5367 if (*s == '=' && !PL_lex_allbrackets &&
5368 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5375 if (PL_expect == XOPERATOR) {
5376 if (s[1] == '=' && !PL_lex_allbrackets &&
5377 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5382 PL_tokenbuf[0] = '%';
5383 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5384 sizeof PL_tokenbuf - 1, FALSE);
5385 if (!PL_tokenbuf[1]) {
5388 PL_pending_ident = '%';
5392 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5393 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5398 if (PL_lex_brackets > 100)
5399 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5400 PL_lex_brackstack[PL_lex_brackets++] = 0;
5401 PL_lex_allbrackets++;
5403 const char tmp = *s++;
5408 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5410 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5418 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5425 goto just_a_word_zero_gv;
5428 switch (PL_expect) {
5434 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5436 PL_bufptr = s; /* update in case we back off */
5439 "Use of := for an empty attribute list is not allowed");
5446 PL_expect = XTERMBLOCK;
5449 stuffstart = s - SvPVX(PL_linestr) - 1;
5453 while (isIDFIRST_lazy_if(s,UTF)) {
5456 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5457 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5458 if (tmp < 0) tmp = -tmp;
5473 sv = newSVpvn_flags(s, len, UTF ? SVf_UTF8 : 0);
5475 d = scan_str(d,TRUE,TRUE);
5477 /* MUST advance bufptr here to avoid bogus
5478 "at end of line" context messages from yyerror().
5480 PL_bufptr = s + len;
5481 yyerror("Unterminated attribute parameter in attribute list");
5485 return REPORT(0); /* EOF indicator */
5489 sv_catsv(sv, PL_lex_stuff);
5490 attrs = op_append_elem(OP_LIST, attrs,
5491 newSVOP(OP_CONST, 0, sv));
5492 SvREFCNT_dec(PL_lex_stuff);
5493 PL_lex_stuff = NULL;
5496 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5498 if (PL_in_my == KEY_our) {
5499 deprecate(":unique");
5502 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5505 /* NOTE: any CV attrs applied here need to be part of
5506 the CVf_BUILTIN_ATTRS define in cv.h! */
5507 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5509 CvLVALUE_on(PL_compcv);
5511 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5513 deprecate(":locked");
5515 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5517 CvMETHOD_on(PL_compcv);
5519 /* After we've set the flags, it could be argued that
5520 we don't need to do the attributes.pm-based setting
5521 process, and shouldn't bother appending recognized
5522 flags. To experiment with that, uncomment the
5523 following "else". (Note that's already been
5524 uncommented. That keeps the above-applied built-in
5525 attributes from being intercepted (and possibly
5526 rejected) by a package's attribute routines, but is
5527 justified by the performance win for the common case
5528 of applying only built-in attributes.) */
5530 attrs = op_append_elem(OP_LIST, attrs,
5531 newSVOP(OP_CONST, 0,
5535 if (*s == ':' && s[1] != ':')
5538 break; /* require real whitespace or :'s */
5539 /* XXX losing whitespace on sequential attributes here */
5543 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5544 if (*s != ';' && *s != '}' && *s != tmp
5545 && (tmp != '=' || *s != ')')) {
5546 const char q = ((*s == '\'') ? '"' : '\'');
5547 /* If here for an expression, and parsed no attrs, back
5549 if (tmp == '=' && !attrs) {
5553 /* MUST advance bufptr here to avoid bogus "at end of line"
5554 context messages from yyerror().
5557 yyerror( (const char *)
5559 ? Perl_form(aTHX_ "Invalid separator character "
5560 "%c%c%c in attribute list", q, *s, q)
5561 : "Unterminated attribute list" ) );
5569 start_force(PL_curforce);
5570 NEXTVAL_NEXTTOKE.opval = attrs;
5571 CURMAD('_', PL_nextwhite);
5576 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5577 (s - SvPVX(PL_linestr)) - stuffstart);
5582 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5586 PL_lex_allbrackets--;
5590 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5591 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5595 PL_lex_allbrackets++;
5598 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5604 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5607 PL_lex_allbrackets--;
5613 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5616 if (PL_lex_brackets <= 0)
5617 yyerror("Unmatched right square bracket");
5620 PL_lex_allbrackets--;
5621 if (PL_lex_state == LEX_INTERPNORMAL) {
5622 if (PL_lex_brackets == 0) {
5623 if (*s == '-' && s[1] == '>')
5624 PL_lex_state = LEX_INTERPENDMAYBE;
5625 else if (*s != '[' && *s != '{')
5626 PL_lex_state = LEX_INTERPEND;
5633 if (PL_lex_brackets > 100) {
5634 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5636 switch (PL_expect) {
5638 if (PL_lex_formbrack) {
5642 if (PL_oldoldbufptr == PL_last_lop)
5643 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5645 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5646 PL_lex_allbrackets++;
5647 OPERATOR(HASHBRACK);
5649 while (s < PL_bufend && SPACE_OR_TAB(*s))
5652 PL_tokenbuf[0] = '\0';
5653 if (d < PL_bufend && *d == '-') {
5654 PL_tokenbuf[0] = '-';
5656 while (d < PL_bufend && SPACE_OR_TAB(*d))
5659 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5660 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5662 while (d < PL_bufend && SPACE_OR_TAB(*d))
5665 const char minus = (PL_tokenbuf[0] == '-');
5666 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5674 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5675 PL_lex_allbrackets++;
5680 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5681 PL_lex_allbrackets++;
5686 if (PL_oldoldbufptr == PL_last_lop)
5687 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5689 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5690 PL_lex_allbrackets++;
5693 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5695 /* This hack is to get the ${} in the message. */
5697 yyerror("syntax error");
5700 OPERATOR(HASHBRACK);
5702 /* This hack serves to disambiguate a pair of curlies
5703 * as being a block or an anon hash. Normally, expectation
5704 * determines that, but in cases where we're not in a
5705 * position to expect anything in particular (like inside
5706 * eval"") we have to resolve the ambiguity. This code
5707 * covers the case where the first term in the curlies is a
5708 * quoted string. Most other cases need to be explicitly
5709 * disambiguated by prepending a "+" before the opening
5710 * curly in order to force resolution as an anon hash.
5712 * XXX should probably propagate the outer expectation
5713 * into eval"" to rely less on this hack, but that could
5714 * potentially break current behavior of eval"".
5718 if (*s == '\'' || *s == '"' || *s == '`') {
5719 /* common case: get past first string, handling escapes */
5720 for (t++; t < PL_bufend && *t != *s;)
5721 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5725 else if (*s == 'q') {
5728 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5731 /* skip q//-like construct */
5733 char open, close, term;
5736 while (t < PL_bufend && isSPACE(*t))
5738 /* check for q => */
5739 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5740 OPERATOR(HASHBRACK);
5744 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5748 for (t++; t < PL_bufend; t++) {
5749 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5751 else if (*t == open)
5755 for (t++; t < PL_bufend; t++) {
5756 if (*t == '\\' && t+1 < PL_bufend)
5758 else if (*t == close && --brackets <= 0)
5760 else if (*t == open)
5767 /* skip plain q word */
5768 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5771 else if (isALNUM_lazy_if(t,UTF)) {
5773 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5776 while (t < PL_bufend && isSPACE(*t))
5778 /* if comma follows first term, call it an anon hash */
5779 /* XXX it could be a comma expression with loop modifiers */
5780 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5781 || (*t == '=' && t[1] == '>')))
5782 OPERATOR(HASHBRACK);
5783 if (PL_expect == XREF)
5786 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5792 pl_yylval.ival = CopLINE(PL_curcop);
5793 if (isSPACE(*s) || *s == '#')
5794 PL_copline = NOLINE; /* invalidate current command line number */
5797 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5801 if (PL_lex_brackets <= 0)
5802 yyerror("Unmatched right curly bracket");
5804 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5805 PL_lex_allbrackets--;
5806 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5807 PL_lex_formbrack = 0;
5808 if (PL_lex_state == LEX_INTERPNORMAL) {
5809 if (PL_lex_brackets == 0) {
5810 if (PL_expect & XFAKEBRACK) {
5811 PL_expect &= XENUMMASK;
5812 PL_lex_state = LEX_INTERPEND;
5817 PL_thiswhite = newSVpvs("");
5818 sv_catpvs(PL_thiswhite,"}");
5821 return yylex(); /* ignore fake brackets */
5823 if (*s == '-' && s[1] == '>')
5824 PL_lex_state = LEX_INTERPENDMAYBE;
5825 else if (*s != '[' && *s != '{')
5826 PL_lex_state = LEX_INTERPEND;
5829 if (PL_expect & XFAKEBRACK) {
5830 PL_expect &= XENUMMASK;
5832 return yylex(); /* ignore fake brackets */
5834 start_force(PL_curforce);
5836 curmad('X', newSVpvn(s-1,1));
5837 CURMAD('_', PL_thiswhite);
5842 PL_thistoken = newSVpvs("");
5848 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5849 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5856 if (PL_expect == XOPERATOR) {
5857 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5858 && isIDFIRST_lazy_if(s,UTF))
5860 CopLINE_dec(PL_curcop);
5861 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5862 CopLINE_inc(PL_curcop);
5864 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5865 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5872 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5874 PL_expect = XOPERATOR;
5875 force_ident(PL_tokenbuf, '&');
5879 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5885 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5886 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5893 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5894 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5902 const char tmp = *s++;
5904 if (!PL_lex_allbrackets &&
5905 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5912 if (!PL_lex_allbrackets &&
5913 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
5921 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5922 && strchr("+-*/%.^&|<",tmp))
5923 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5924 "Reversed %c= operator",(int)tmp);
5926 if (PL_expect == XSTATE && isALPHA(tmp) &&
5927 (s == PL_linestart+1 || s[-2] == '\n') )
5929 if (PL_in_eval && !PL_rsfp && !PL_parser->filtered) {
5934 if (strnEQ(s,"=cut",4)) {
5950 PL_thiswhite = newSVpvs("");
5951 sv_catpvn(PL_thiswhite, PL_linestart,
5952 PL_bufend - PL_linestart);
5956 PL_parser->in_pod = 1;
5960 if (PL_lex_brackets < PL_lex_formbrack) {
5962 #ifdef PERL_STRICT_CR
5963 while (SPACE_OR_TAB(*t))
5965 while (SPACE_OR_TAB(*t) || *t == '\r')
5968 if (*t == '\n' || *t == '#') {
5974 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5983 const char tmp = *s++;
5985 /* was this !=~ where !~ was meant?
5986 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5988 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5989 const char *t = s+1;
5991 while (t < PL_bufend && isSPACE(*t))
5994 if (*t == '/' || *t == '?' ||
5995 ((*t == 'm' || *t == 's' || *t == 'y')
5996 && !isALNUM(t[1])) ||
5997 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5998 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5999 "!=~ should be !~");
6001 if (!PL_lex_allbrackets &&
6002 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6014 if (PL_expect != XOPERATOR) {
6015 if (s[1] != '<' && !strchr(s,'>'))
6018 s = scan_heredoc(s);
6020 s = scan_inputsymbol(s);
6021 TERM(sublex_start());
6027 if (*s == '=' && !PL_lex_allbrackets &&
6028 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6032 SHop(OP_LEFT_SHIFT);
6037 if (!PL_lex_allbrackets &&
6038 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6045 if (!PL_lex_allbrackets &&
6046 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6054 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6062 const char tmp = *s++;
6064 if (*s == '=' && !PL_lex_allbrackets &&
6065 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6069 SHop(OP_RIGHT_SHIFT);
6071 else if (tmp == '=') {
6072 if (!PL_lex_allbrackets &&
6073 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6081 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6090 if (PL_expect == XOPERATOR) {
6091 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6092 return deprecate_commaless_var_list();
6096 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
6097 PL_tokenbuf[0] = '@';
6098 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
6099 sizeof PL_tokenbuf - 1, FALSE);
6100 if (PL_expect == XOPERATOR)
6101 no_op("Array length", s);
6102 if (!PL_tokenbuf[1])
6104 PL_expect = XOPERATOR;
6105 PL_pending_ident = '#';
6109 PL_tokenbuf[0] = '$';
6110 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
6111 sizeof PL_tokenbuf - 1, FALSE);
6112 if (PL_expect == XOPERATOR)
6114 if (!PL_tokenbuf[1]) {
6116 yyerror("Final $ should be \\$ or $name");
6122 const char tmp = *s;
6123 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6126 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6127 && intuit_more(s)) {
6129 PL_tokenbuf[0] = '@';
6130 if (ckWARN(WARN_SYNTAX)) {
6133 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
6136 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6137 while (t < PL_bufend && *t != ']')
6139 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6140 "Multidimensional syntax %.*s not supported",
6141 (int)((t - PL_bufptr) + 1), PL_bufptr);
6145 else if (*s == '{') {
6147 PL_tokenbuf[0] = '%';
6148 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6149 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6151 char tmpbuf[sizeof PL_tokenbuf];
6154 } while (isSPACE(*t));
6155 if (isIDFIRST_lazy_if(t,UTF)) {
6157 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6161 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
6162 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6163 "You need to quote \"%s\"",
6170 PL_expect = XOPERATOR;
6171 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6172 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6173 if (!islop || PL_last_lop_op == OP_GREPSTART)
6174 PL_expect = XOPERATOR;
6175 else if (strchr("$@\"'`q", *s))
6176 PL_expect = XTERM; /* e.g. print $fh "foo" */
6177 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6178 PL_expect = XTERM; /* e.g. print $fh &sub */
6179 else if (isIDFIRST_lazy_if(s,UTF)) {
6180 char tmpbuf[sizeof PL_tokenbuf];
6182 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6183 if ((t2 = keyword(tmpbuf, len, 0))) {
6184 /* binary operators exclude handle interpretations */
6196 PL_expect = XTERM; /* e.g. print $fh length() */
6201 PL_expect = XTERM; /* e.g. print $fh subr() */
6204 else if (isDIGIT(*s))
6205 PL_expect = XTERM; /* e.g. print $fh 3 */
6206 else if (*s == '.' && isDIGIT(s[1]))
6207 PL_expect = XTERM; /* e.g. print $fh .3 */
6208 else if ((*s == '?' || *s == '-' || *s == '+')
6209 && !isSPACE(s[1]) && s[1] != '=')
6210 PL_expect = XTERM; /* e.g. print $fh -1 */
6211 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6213 PL_expect = XTERM; /* e.g. print $fh /.../
6214 XXX except DORDOR operator
6216 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6218 PL_expect = XTERM; /* print $fh <<"EOF" */
6221 PL_pending_ident = '$';
6225 if (PL_expect == XOPERATOR)
6227 PL_tokenbuf[0] = '@';
6228 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6229 if (!PL_tokenbuf[1]) {
6232 if (PL_lex_state == LEX_NORMAL)
6234 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6236 PL_tokenbuf[0] = '%';
6238 /* Warn about @ where they meant $. */
6239 if (*s == '[' || *s == '{') {
6240 if (ckWARN(WARN_SYNTAX)) {
6241 const char *t = s + 1;
6242 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
6244 if (*t == '}' || *t == ']') {
6246 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6247 /* diag_listed_as: Scalar value @%s[%s] better written as $%s[%s] */
6248 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6249 "Scalar value %.*s better written as $%.*s",
6250 (int)(t-PL_bufptr), PL_bufptr,
6251 (int)(t-PL_bufptr-1), PL_bufptr+1);
6256 PL_pending_ident = '@';
6259 case '/': /* may be division, defined-or, or pattern */
6260 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
6261 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6262 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6267 case '?': /* may either be conditional or pattern */
6268 if (PL_expect == XOPERATOR) {
6271 if (!PL_lex_allbrackets &&
6272 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6276 PL_lex_allbrackets++;
6282 /* A // operator. */
6283 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6284 (*s == '=' ? LEX_FAKEEOF_ASSIGN :
6285 LEX_FAKEEOF_LOGIC)) {
6293 if (*s == '=' && !PL_lex_allbrackets &&
6294 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6303 /* Disable warning on "study /blah/" */
6304 if (PL_oldoldbufptr == PL_last_uni
6305 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6306 || memNE(PL_last_uni, "study", 5)
6307 || isALNUM_lazy_if(PL_last_uni+5,UTF)
6311 deprecate("?PATTERN? without explicit operator");
6312 s = scan_pat(s,OP_MATCH);
6313 TERM(sublex_start());
6317 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6318 #ifdef PERL_STRICT_CR
6321 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6323 && (s == PL_linestart || s[-1] == '\n') )
6325 PL_lex_formbrack = 0;
6329 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6333 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6336 if (!PL_lex_allbrackets &&
6337 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6344 pl_yylval.ival = OPf_SPECIAL;
6350 if (*s == '=' && !PL_lex_allbrackets &&
6351 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6358 case '0': case '1': case '2': case '3': case '4':
6359 case '5': case '6': case '7': case '8': case '9':
6360 s = scan_num(s, &pl_yylval);
6361 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6362 if (PL_expect == XOPERATOR)
6367 s = scan_str(s,!!PL_madskills,FALSE);
6368 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6369 if (PL_expect == XOPERATOR) {
6370 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6371 return deprecate_commaless_var_list();
6378 pl_yylval.ival = OP_CONST;
6379 TERM(sublex_start());
6382 s = scan_str(s,!!PL_madskills,FALSE);
6383 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6384 if (PL_expect == XOPERATOR) {
6385 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6386 return deprecate_commaless_var_list();
6393 pl_yylval.ival = OP_CONST;
6394 /* FIXME. I think that this can be const if char *d is replaced by
6395 more localised variables. */
6396 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6397 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6398 pl_yylval.ival = OP_STRINGIFY;
6402 TERM(sublex_start());
6405 s = scan_str(s,!!PL_madskills,FALSE);
6406 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6407 if (PL_expect == XOPERATOR)
6408 no_op("Backticks",s);
6411 readpipe_override();
6412 TERM(sublex_start());
6416 if (PL_lex_inwhat && isDIGIT(*s))
6417 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6419 if (PL_expect == XOPERATOR)
6420 no_op("Backslash",s);
6424 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6425 char *start = s + 2;
6426 while (isDIGIT(*start) || *start == '_')
6428 if (*start == '.' && isDIGIT(start[1])) {
6429 s = scan_num(s, &pl_yylval);
6432 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6433 else if (!isALPHA(*start) && (PL_expect == XTERM
6434 || PL_expect == XREF || PL_expect == XSTATE
6435 || PL_expect == XTERMORDORDOR)) {
6436 GV *const gv = gv_fetchpvn_flags(s, start - s,
6437 UTF ? SVf_UTF8 : 0, SVt_PVCV);
6439 s = scan_num(s, &pl_yylval);
6446 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6489 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6491 /* Some keywords can be followed by any delimiter, including ':' */
6492 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6494 /* x::* is just a word, unless x is "CORE" */
6495 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6499 while (d < PL_bufend && isSPACE(*d))
6500 d++; /* no comments skipped here, or s### is misparsed */
6502 /* Is this a word before a => operator? */
6503 if (*d == '=' && d[1] == '>') {
6506 = (OP*)newSVOP(OP_CONST, 0,
6507 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6508 pl_yylval.opval->op_private = OPpCONST_BARE;
6512 /* Check for plugged-in keyword */
6516 char *saved_bufptr = PL_bufptr;
6518 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6520 if (result == KEYWORD_PLUGIN_DECLINE) {
6521 /* not a plugged-in keyword */
6522 PL_bufptr = saved_bufptr;
6523 } else if (result == KEYWORD_PLUGIN_STMT) {
6524 pl_yylval.opval = o;
6527 return REPORT(PLUGSTMT);
6528 } else if (result == KEYWORD_PLUGIN_EXPR) {
6529 pl_yylval.opval = o;
6531 PL_expect = XOPERATOR;
6532 return REPORT(PLUGEXPR);
6534 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6539 /* Check for built-in keyword */
6540 tmp = keyword(PL_tokenbuf, len, 0);
6542 /* Is this a label? */
6543 if (!anydelim && PL_expect == XSTATE
6544 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6546 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
6551 if (tmp < 0) { /* second-class keyword? */
6552 GV *ogv = NULL; /* override (winner) */
6553 GV *hgv = NULL; /* hidden (loser) */
6554 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6556 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6557 UTF ? SVf_UTF8 : 0, SVt_PVCV)) &&
6560 if (GvIMPORTED_CV(gv))
6562 else if (! CvMETHOD(cv))
6566 (gvp = (GV**)hv_fetch(PL_globalstash, PL_tokenbuf,
6567 UTF ? -(I32)len : (I32)len, FALSE)) &&
6568 (gv = *gvp) && isGV_with_GP(gv) &&
6569 GvCVu(gv) && GvIMPORTED_CV(gv))
6576 tmp = 0; /* overridden by import or by GLOBAL */
6579 && -tmp==KEY_lock /* XXX generalizable kludge */
6582 tmp = 0; /* any sub overrides "weak" keyword */
6584 else { /* no override */
6586 if (tmp == KEY_dump) {
6587 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6588 "dump() better written as CORE::dump()");
6592 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6593 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6594 "Ambiguous call resolved as CORE::%s(), "
6595 "qualify as such or use &",
6603 default: /* not a keyword */
6604 /* Trade off - by using this evil construction we can pull the
6605 variable gv into the block labelled keylookup. If not, then
6606 we have to give it function scope so that the goto from the
6607 earlier ':' case doesn't bypass the initialisation. */
6609 just_a_word_zero_gv:
6617 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6621 SV *nextPL_nextwhite = 0;
6625 /* Get the rest if it looks like a package qualifier */
6627 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6629 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6632 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
6633 *s == '\'' ? "'" : "::");
6638 if (PL_expect == XOPERATOR) {
6639 if (PL_bufptr == PL_linestart) {
6640 CopLINE_dec(PL_curcop);
6641 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6642 CopLINE_inc(PL_curcop);
6645 no_op("Bareword",s);
6648 /* Look for a subroutine with this name in current package,
6649 unless name is "Foo::", in which case Foo is a bareword
6650 (and a package name). */
6652 if (len > 2 && !PL_madskills &&
6653 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6655 if (ckWARN(WARN_BAREWORD)
6656 && ! gv_fetchpvn_flags(PL_tokenbuf, len, UTF ? SVf_UTF8 : 0, SVt_PVHV))
6657 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6658 "Bareword \"%s\" refers to nonexistent package",
6661 PL_tokenbuf[len] = '\0';
6667 /* Mustn't actually add anything to a symbol table.
6668 But also don't want to "initialise" any placeholder
6669 constants that might already be there into full
6670 blown PVGVs with attached PVCV. */
6671 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6672 GV_NOADD_NOINIT | ( UTF ? SVf_UTF8 : 0 ),
6678 /* if we saw a global override before, get the right name */
6680 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
6681 len ? len : strlen(PL_tokenbuf));
6683 SV * const tmp_sv = sv;
6684 sv = newSVpvs("CORE::GLOBAL::");
6685 sv_catsv(sv, tmp_sv);
6686 SvREFCNT_dec(tmp_sv);
6690 if (PL_madskills && !PL_thistoken) {
6691 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6692 PL_thistoken = newSVpvn(start,s - start);
6693 PL_realtokenstart = s - SvPVX(PL_linestr);
6697 /* Presume this is going to be a bareword of some sort. */
6699 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6700 pl_yylval.opval->op_private = OPpCONST_BARE;
6702 /* And if "Foo::", then that's what it certainly is. */
6707 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
6708 const_op->op_private = OPpCONST_BARE;
6709 rv2cv_op = newCVREF(0, const_op);
6711 cv = rv2cv_op_cv(rv2cv_op, 0);
6713 /* See if it's the indirect object for a list operator. */
6715 if (PL_oldoldbufptr &&
6716 PL_oldoldbufptr < PL_bufptr &&
6717 (PL_oldoldbufptr == PL_last_lop
6718 || PL_oldoldbufptr == PL_last_uni) &&
6719 /* NO SKIPSPACE BEFORE HERE! */
6720 (PL_expect == XREF ||
6721 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6723 bool immediate_paren = *s == '(';
6725 /* (Now we can afford to cross potential line boundary.) */
6726 s = SKIPSPACE2(s,nextPL_nextwhite);
6728 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6731 /* Two barewords in a row may indicate method call. */
6733 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6734 (tmp = intuit_method(s, gv, cv))) {
6736 if (tmp == METHOD && !PL_lex_allbrackets &&
6737 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6738 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6742 /* If not a declared subroutine, it's an indirect object. */
6743 /* (But it's an indir obj regardless for sort.) */
6744 /* Also, if "_" follows a filetest operator, it's a bareword */
6747 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6749 (PL_last_lop_op != OP_MAPSTART &&
6750 PL_last_lop_op != OP_GREPSTART))))
6751 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6752 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6755 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6760 PL_expect = XOPERATOR;
6763 s = SKIPSPACE2(s,nextPL_nextwhite);
6764 PL_nextwhite = nextPL_nextwhite;
6769 /* Is this a word before a => operator? */
6770 if (*s == '=' && s[1] == '>' && !pkgname) {
6773 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6774 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6775 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6779 /* If followed by a paren, it's certainly a subroutine. */
6784 while (SPACE_OR_TAB(*d))
6786 if (*d == ')' && (sv = cv_const_sv(cv))) {
6793 PL_nextwhite = PL_thiswhite;
6796 start_force(PL_curforce);
6798 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6799 PL_expect = XOPERATOR;
6802 PL_nextwhite = nextPL_nextwhite;
6803 curmad('X', PL_thistoken);
6804 PL_thistoken = newSVpvs("");
6813 /* If followed by var or block, call it a method (unless sub) */
6815 if ((*s == '$' || *s == '{') && !cv) {
6817 PL_last_lop = PL_oldbufptr;
6818 PL_last_lop_op = OP_METHOD;
6819 if (!PL_lex_allbrackets &&
6820 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6821 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6825 /* If followed by a bareword, see if it looks like indir obj. */
6828 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6829 && (tmp = intuit_method(s, gv, cv))) {
6831 if (tmp == METHOD && !PL_lex_allbrackets &&
6832 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6833 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6837 /* Not a method, so call it a subroutine (if defined) */
6840 if (lastchar == '-')
6841 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6842 "Ambiguous use of -%s resolved as -&%s()",
6843 PL_tokenbuf, PL_tokenbuf);
6844 /* Check for a constant sub */
6845 if ((sv = cv_const_sv(cv))) {
6848 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6849 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6850 pl_yylval.opval->op_private = 0;
6851 pl_yylval.opval->op_flags |= OPf_SPECIAL;
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 /* Is there a prototype? */
6867 STRLEN protolen = CvPROTOLEN(cv);
6868 const char *proto = CvPROTO(cv);
6872 if ((optional = *proto == ';'))
6875 while (*proto == ';');
6879 *proto == '$' || *proto == '_'
6880 || *proto == '*' || *proto == '+'
6885 *proto == '\\' && proto[1] && proto[2] == '\0'
6888 UNIPROTO(UNIOPSUB,optional);
6889 if (*proto == '\\' && proto[1] == '[') {
6890 const char *p = proto + 2;
6891 while(*p && *p != ']')
6893 if(*p == ']' && !p[1])
6894 UNIPROTO(UNIOPSUB,optional);
6896 if (*proto == '&' && *s == '{') {
6898 sv_setpvs(PL_subname, "__ANON__");
6900 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6901 if (!PL_lex_allbrackets &&
6902 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6903 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6910 PL_nextwhite = PL_thiswhite;
6913 start_force(PL_curforce);
6914 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6917 PL_nextwhite = nextPL_nextwhite;
6918 curmad('X', PL_thistoken);
6919 PL_thistoken = newSVpvs("");
6922 if (!PL_lex_allbrackets &&
6923 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6924 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6929 /* Guess harder when madskills require "best effort". */
6930 if (PL_madskills && (!gv || !GvCVu(gv))) {
6931 int probable_sub = 0;
6932 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6934 else if (isALPHA(*s)) {
6938 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6939 if (!keyword(tmpbuf, tmplen, 0))
6942 while (d < PL_bufend && isSPACE(*d))
6944 if (*d == '=' && d[1] == '>')
6949 gv = gv_fetchpv(PL_tokenbuf, GV_ADD | ( UTF ? SVf_UTF8 : 0 ),
6951 op_free(pl_yylval.opval);
6952 pl_yylval.opval = rv2cv_op;
6953 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6954 PL_last_lop = PL_oldbufptr;
6955 PL_last_lop_op = OP_ENTERSUB;
6956 PL_nextwhite = PL_thiswhite;
6958 start_force(PL_curforce);
6959 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6961 PL_nextwhite = nextPL_nextwhite;
6962 curmad('X', PL_thistoken);
6963 PL_thistoken = newSVpvs("");
6965 if (!PL_lex_allbrackets &&
6966 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6967 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6971 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6974 if (!PL_lex_allbrackets &&
6975 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6976 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6981 /* Call it a bare word */
6983 if (PL_hints & HINT_STRICT_SUBS)
6984 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6987 /* after "print" and similar functions (corresponding to
6988 * "F? L" in opcode.pl), whatever wasn't already parsed as
6989 * a filehandle should be subject to "strict subs".
6990 * Likewise for the optional indirect-object argument to system
6991 * or exec, which can't be a bareword */
6992 if ((PL_last_lop_op == OP_PRINT
6993 || PL_last_lop_op == OP_PRTF
6994 || PL_last_lop_op == OP_SAY
6995 || PL_last_lop_op == OP_SYSTEM
6996 || PL_last_lop_op == OP_EXEC)
6997 && (PL_hints & HINT_STRICT_SUBS))
6998 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6999 if (lastchar != '-') {
7000 if (ckWARN(WARN_RESERVED)) {
7004 if (!*d && !gv_stashpv(PL_tokenbuf, UTF ? SVf_UTF8 : 0))
7005 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
7013 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
7014 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7015 "Operator or semicolon missing before %c%s",
7016 lastchar, PL_tokenbuf);
7017 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7018 "Ambiguous use of %c resolved as operator %c",
7019 lastchar, lastchar);
7026 (OP*)newSVOP(OP_CONST, 0, newSVpv(CopFILE(PL_curcop),0))
7031 (OP*)newSVOP(OP_CONST, 0,
7032 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
7035 case KEY___PACKAGE__:
7037 (OP*)newSVOP(OP_CONST, 0,
7039 ? newSVhek(HvNAME_HEK(PL_curstash))
7046 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
7047 const char *pname = "main";
7050 if (PL_tokenbuf[2] == 'D')
7053 PL_curstash ? PL_curstash : PL_defstash;
7054 pname = HvNAME_get(stash);
7055 plen = HvNAMELEN (stash);
7056 if(HvNAMEUTF8(stash)) putf8 = SVf_UTF8;
7058 gv = gv_fetchpvn_flags(
7059 Perl_form(aTHX_ "%*s::DATA", (int)plen, pname),
7060 plen+6, GV_ADD|putf8, SVt_PVIO
7064 GvIOp(gv) = newIO();
7065 IoIFP(GvIOp(gv)) = PL_rsfp;
7066 #if defined(HAS_FCNTL) && defined(F_SETFD)
7068 const int fd = PerlIO_fileno(PL_rsfp);
7069 fcntl(fd,F_SETFD,fd >= 3);
7072 /* Mark this internal pseudo-handle as clean */
7073 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
7074 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
7075 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
7077 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
7078 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
7079 /* if the script was opened in binmode, we need to revert
7080 * it to text mode for compatibility; but only iff it has CRs
7081 * XXX this is a questionable hack at best. */
7082 if (PL_bufend-PL_bufptr > 2
7083 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
7086 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
7087 loc = PerlIO_tell(PL_rsfp);
7088 (void)PerlIO_seek(PL_rsfp, 0L, 0);
7091 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
7093 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
7094 #endif /* NETWARE */
7096 PerlIO_seek(PL_rsfp, loc, 0);
7100 #ifdef PERLIO_LAYERS
7103 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
7104 else if (PL_encoding) {
7111 XPUSHs(PL_encoding);
7113 call_method("name", G_SCALAR);
7117 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7118 Perl_form(aTHX_ ":encoding(%"SVf")",
7127 if (PL_realtokenstart >= 0) {
7128 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7130 PL_endwhite = newSVpvs("");
7131 sv_catsv(PL_endwhite, PL_thiswhite);
7133 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
7134 PL_realtokenstart = -1;
7136 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
7146 FUN0OP(newPVOP(OP_RUNCV,0,NULL));
7155 if (PL_expect == XSTATE) {
7162 if (*s == ':' && s[1] == ':') {
7165 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7166 if (!(tmp = keyword(PL_tokenbuf, len, 1)))
7167 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
7170 else if (tmp == KEY_require || tmp == KEY_do
7172 /* that's a way to remember we saw "CORE::" */
7185 LOP(OP_ACCEPT,XTERM);
7188 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7193 LOP(OP_ATAN2,XTERM);
7199 LOP(OP_BINMODE,XTERM);
7202 LOP(OP_BLESS,XTERM);
7211 /* We have to disambiguate the two senses of
7212 "continue". If the next token is a '{' then
7213 treat it as the start of a continue block;
7214 otherwise treat it as a control operator.
7224 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7234 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7243 if (!PL_cryptseen) {
7244 PL_cryptseen = TRUE;
7248 LOP(OP_CRYPT,XTERM);
7251 LOP(OP_CHMOD,XTERM);
7254 LOP(OP_CHOWN,XTERM);
7257 LOP(OP_CONNECT,XTERM);
7276 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7277 if (orig_keyword == KEY_do) {
7286 PL_hints |= HINT_BLOCK_SCOPE;
7296 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7297 STR_WITH_LEN("NDBM_File::"),
7298 STR_WITH_LEN("DB_File::"),
7299 STR_WITH_LEN("GDBM_File::"),
7300 STR_WITH_LEN("SDBM_File::"),
7301 STR_WITH_LEN("ODBM_File::"),
7303 LOP(OP_DBMOPEN,XTERM);
7309 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7316 pl_yylval.ival = CopLINE(PL_curcop);
7320 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7334 if (*s == '{') { /* block eval */
7335 PL_expect = XTERMBLOCK;
7336 UNIBRACK(OP_ENTERTRY);
7338 else { /* string eval */
7340 UNIBRACK(OP_ENTEREVAL);
7345 UNIBRACK(-OP_ENTEREVAL);
7359 case KEY_endhostent:
7365 case KEY_endservent:
7368 case KEY_endprotoent:
7379 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7381 pl_yylval.ival = CopLINE(PL_curcop);
7383 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7386 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
7389 if ((PL_bufend - p) >= 3 &&
7390 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7392 else if ((PL_bufend - p) >= 4 &&
7393 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7396 if (isIDFIRST_lazy_if(p,UTF)) {
7397 p = scan_ident(p, PL_bufend,
7398 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
7402 Perl_croak(aTHX_ "Missing $ on loop variable");
7404 s = SvPVX(PL_linestr) + soff;
7410 LOP(OP_FORMLINE,XTERM);
7416 LOP(OP_FCNTL,XTERM);
7422 LOP(OP_FLOCK,XTERM);
7425 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7430 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7435 LOP(OP_GREPSTART, XREF);
7438 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7453 case KEY_getpriority:
7454 LOP(OP_GETPRIORITY,XTERM);
7456 case KEY_getprotobyname:
7459 case KEY_getprotobynumber:
7460 LOP(OP_GPBYNUMBER,XTERM);
7462 case KEY_getprotoent:
7474 case KEY_getpeername:
7475 UNI(OP_GETPEERNAME);
7477 case KEY_gethostbyname:
7480 case KEY_gethostbyaddr:
7481 LOP(OP_GHBYADDR,XTERM);
7483 case KEY_gethostent:
7486 case KEY_getnetbyname:
7489 case KEY_getnetbyaddr:
7490 LOP(OP_GNBYADDR,XTERM);
7495 case KEY_getservbyname:
7496 LOP(OP_GSBYNAME,XTERM);
7498 case KEY_getservbyport:
7499 LOP(OP_GSBYPORT,XTERM);
7501 case KEY_getservent:
7504 case KEY_getsockname:
7505 UNI(OP_GETSOCKNAME);
7507 case KEY_getsockopt:
7508 LOP(OP_GSOCKOPT,XTERM);
7523 pl_yylval.ival = CopLINE(PL_curcop);
7528 orig_keyword==KEY_glob ? (orig_keyword=0, -OP_GLOB) : OP_GLOB,
7536 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7538 pl_yylval.ival = CopLINE(PL_curcop);
7542 LOP(OP_INDEX,XTERM);
7548 LOP(OP_IOCTL,XTERM);
7560 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7577 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7582 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7596 LOP(OP_LISTEN,XTERM);
7605 s = scan_pat(s,OP_MATCH);
7606 TERM(sublex_start());
7609 LOP(OP_MAPSTART, XREF);
7612 LOP(OP_MKDIR,XTERM);
7615 LOP(OP_MSGCTL,XTERM);
7618 LOP(OP_MSGGET,XTERM);
7621 LOP(OP_MSGRCV,XTERM);
7624 LOP(OP_MSGSND,XTERM);
7629 PL_in_my = (U16)tmp;
7631 if (isIDFIRST_lazy_if(s,UTF)) {
7635 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7636 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7638 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7639 if (!PL_in_my_stash) {
7642 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7646 if (PL_madskills) { /* just add type to declarator token */
7647 sv_catsv(PL_thistoken, PL_nextwhite);
7649 sv_catpvn(PL_thistoken, start, s - start);
7657 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7661 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7666 s = tokenize_use(0, s);
7670 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7673 if (!PL_lex_allbrackets &&
7674 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7675 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7681 if (isIDFIRST_lazy_if(s,UTF)) {
7683 for (d = s; isALNUM_lazy_if(d,UTF);)
7685 for (t=d; isSPACE(*t);)
7687 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7689 && !(t[0] == '=' && t[1] == '>')
7690 && !(t[0] == ':' && t[1] == ':')
7691 && !keyword(s, d-s, 0)
7693 int parms_len = (int)(d-s);
7694 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7695 "Precedence problem: open %.*s should be open(%.*s)",
7696 parms_len, s, parms_len, s);
7702 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7704 pl_yylval.ival = OP_OR;
7714 LOP(OP_OPEN_DIR,XTERM);
7717 checkcomma(s,PL_tokenbuf,"filehandle");
7721 checkcomma(s,PL_tokenbuf,"filehandle");
7740 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7742 s = force_strict_version(s);
7743 PL_lex_expect = XBLOCK;
7747 LOP(OP_PIPE_OP,XTERM);
7750 s = scan_str(s,!!PL_madskills,FALSE);
7753 pl_yylval.ival = OP_CONST;
7754 TERM(sublex_start());
7761 s = scan_str(s,!!PL_madskills,FALSE);
7764 PL_expect = XOPERATOR;
7765 if (SvCUR(PL_lex_stuff)) {
7766 int warned_comma = !ckWARN(WARN_QW);
7767 int warned_comment = warned_comma;
7768 d = SvPV_force(PL_lex_stuff, len);
7770 for (; isSPACE(*d) && len; --len, ++d)
7775 if (!warned_comma || !warned_comment) {
7776 for (; !isSPACE(*d) && len; --len, ++d) {
7777 if (!warned_comma && *d == ',') {
7778 Perl_warner(aTHX_ packWARN(WARN_QW),
7779 "Possible attempt to separate words with commas");
7782 else if (!warned_comment && *d == '#') {
7783 Perl_warner(aTHX_ packWARN(WARN_QW),
7784 "Possible attempt to put comments in qw() list");
7790 for (; !isSPACE(*d) && len; --len, ++d)
7793 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7794 words = op_append_elem(OP_LIST, words,
7795 newSVOP(OP_CONST, 0, tokeq(sv)));
7800 words = newNULLLIST();
7802 SvREFCNT_dec(PL_lex_stuff);
7803 PL_lex_stuff = NULL;
7805 PL_expect = XOPERATOR;
7806 pl_yylval.opval = sawparens(words);
7811 s = scan_str(s,!!PL_madskills,FALSE);
7814 pl_yylval.ival = OP_STRINGIFY;
7815 if (SvIVX(PL_lex_stuff) == '\'')
7816 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
7817 TERM(sublex_start());
7820 s = scan_pat(s,OP_QR);
7821 TERM(sublex_start());
7824 s = scan_str(s,!!PL_madskills,FALSE);
7827 readpipe_override();
7828 TERM(sublex_start());
7836 s = force_version(s, FALSE);
7838 else if (*s != 'v' || !isDIGIT(s[1])
7839 || (s = force_version(s, TRUE), *s == 'v'))
7841 *PL_tokenbuf = '\0';
7842 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7843 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7844 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf),
7845 GV_ADD | (UTF ? SVf_UTF8 : 0));
7847 yyerror("<> should be quotes");
7849 if (orig_keyword == KEY_require) {
7857 PL_last_uni = PL_oldbufptr;
7858 PL_last_lop_op = OP_REQUIRE;
7860 return REPORT( (int)REQUIRE );
7866 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7870 LOP(OP_RENAME,XTERM);
7879 LOP(OP_RINDEX,XTERM);
7888 UNIDOR(OP_READLINE);
7891 UNIDOR(OP_BACKTICK);
7900 LOP(OP_REVERSE,XTERM);
7903 UNIDOR(OP_READLINK);
7910 if (pl_yylval.opval)
7911 TERM(sublex_start());
7913 TOKEN(1); /* force error */
7916 checkcomma(s,PL_tokenbuf,"filehandle");
7926 LOP(OP_SELECT,XTERM);
7932 LOP(OP_SEMCTL,XTERM);
7935 LOP(OP_SEMGET,XTERM);
7938 LOP(OP_SEMOP,XTERM);
7944 LOP(OP_SETPGRP,XTERM);
7946 case KEY_setpriority:
7947 LOP(OP_SETPRIORITY,XTERM);
7949 case KEY_sethostent:
7955 case KEY_setservent:
7958 case KEY_setprotoent:
7968 LOP(OP_SEEKDIR,XTERM);
7970 case KEY_setsockopt:
7971 LOP(OP_SSOCKOPT,XTERM);
7977 LOP(OP_SHMCTL,XTERM);
7980 LOP(OP_SHMGET,XTERM);
7983 LOP(OP_SHMREAD,XTERM);
7986 LOP(OP_SHMWRITE,XTERM);
7989 LOP(OP_SHUTDOWN,XTERM);
7998 LOP(OP_SOCKET,XTERM);
8000 case KEY_socketpair:
8001 LOP(OP_SOCKPAIR,XTERM);
8004 checkcomma(s,PL_tokenbuf,"subroutine name");
8007 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8011 LOP(OP_SPLIT,XTERM);
8014 LOP(OP_SPRINTF,XTERM);
8017 LOP(OP_SPLICE,XTERM);
8032 LOP(OP_SUBSTR,XTERM);
8038 char tmpbuf[sizeof PL_tokenbuf];
8039 SSize_t tboffset = 0;
8040 expectation attrful;
8041 bool have_name, have_proto;
8042 const int key = tmp;
8047 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
8048 SV *subtoken = newSVpvn_flags(tstart, s - tstart, SvUTF8(PL_linestr));
8052 s = SKIPSPACE2(s,tmpwhite);
8057 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
8058 (*s == ':' && s[1] == ':'))
8061 SV *nametoke = NULL;
8065 attrful = XATTRBLOCK;
8066 /* remember buffer pos'n for later force_word */
8067 tboffset = s - PL_oldbufptr;
8068 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
8071 nametoke = newSVpvn_flags(s, d - s, SvUTF8(PL_linestr));
8073 if (memchr(tmpbuf, ':', len))
8074 sv_setpvn(PL_subname, tmpbuf, len);
8076 sv_setsv(PL_subname,PL_curstname);
8077 sv_catpvs(PL_subname,"::");
8078 sv_catpvn(PL_subname,tmpbuf,len);
8080 if (SvUTF8(PL_linestr))
8081 SvUTF8_on(PL_subname);
8087 CURMAD('X', nametoke);
8088 CURMAD('_', tmpwhite);
8089 (void) force_word(PL_oldbufptr + tboffset, WORD,
8092 s = SKIPSPACE2(d,tmpwhite);
8099 Perl_croak(aTHX_ "Missing name in \"my sub\"");
8100 PL_expect = XTERMBLOCK;
8101 attrful = XATTRTERM;
8102 sv_setpvs(PL_subname,"?");
8106 if (key == KEY_format) {
8108 PL_lex_formbrack = PL_lex_brackets + 1;
8110 PL_thistoken = subtoken;
8114 (void) force_word(PL_oldbufptr + tboffset, WORD,
8120 /* Look for a prototype */
8123 bool bad_proto = FALSE;
8124 bool in_brackets = FALSE;
8125 char greedy_proto = ' ';
8126 bool proto_after_greedy_proto = FALSE;
8127 bool must_be_last = FALSE;
8128 bool underscore = FALSE;
8129 bool seen_underscore = FALSE;
8130 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
8133 s = scan_str(s,!!PL_madskills,FALSE);
8135 Perl_croak(aTHX_ "Prototype not terminated");
8136 /* strip spaces and check for bad characters */
8137 d = SvPV(PL_lex_stuff, tmplen);
8139 for (p = d; tmplen; tmplen--, ++p) {
8143 if (warnillegalproto) {
8145 proto_after_greedy_proto = TRUE;
8146 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
8158 else if ( *p == ']' ) {
8159 in_brackets = FALSE;
8161 else if ( (*p == '@' || *p == '%') &&
8162 ( tmp < 2 || d[tmp-2] != '\\' ) &&
8164 must_be_last = TRUE;
8167 else if ( *p == '_' ) {
8168 underscore = seen_underscore = TRUE;
8175 if (proto_after_greedy_proto)
8176 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8177 "Prototype after '%c' for %"SVf" : %s",
8178 greedy_proto, SVfARG(PL_subname), d);
8180 SV *dsv = newSVpvs_flags("", SVs_TEMP);
8181 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8182 "Illegal character %sin prototype for %"SVf" : %s",
8183 seen_underscore ? "after '_' " : "",
8186 newSVpvn_flags(d, tmp, SVs_TEMP | SvUTF8(PL_lex_stuff)),
8187 tmp, UNI_DISPLAY_ISPRINT));
8189 SvCUR_set(PL_lex_stuff, tmp);
8194 CURMAD('q', PL_thisopen);
8195 CURMAD('_', tmpwhite);
8196 CURMAD('=', PL_thisstuff);
8197 CURMAD('Q', PL_thisclose);
8198 NEXTVAL_NEXTTOKE.opval =
8199 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8200 PL_lex_stuff = NULL;
8203 s = SKIPSPACE2(s,tmpwhite);
8211 if (*s == ':' && s[1] != ':')
8212 PL_expect = attrful;
8213 else if (*s != '{' && key == KEY_sub) {
8215 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8216 else if (*s != ';' && *s != '}')
8217 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8224 curmad('^', newSVpvs(""));
8225 CURMAD('_', tmpwhite);
8229 PL_thistoken = subtoken;
8232 NEXTVAL_NEXTTOKE.opval =
8233 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8234 PL_lex_stuff = NULL;
8240 sv_setpvs(PL_subname, "__ANON__");
8242 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8246 (void) force_word(PL_oldbufptr + tboffset, WORD,
8255 LOP(OP_SYSTEM,XREF);
8258 LOP(OP_SYMLINK,XTERM);
8261 LOP(OP_SYSCALL,XTERM);
8264 LOP(OP_SYSOPEN,XTERM);
8267 LOP(OP_SYSSEEK,XTERM);
8270 LOP(OP_SYSREAD,XTERM);
8273 LOP(OP_SYSWRITE,XTERM);
8277 TERM(sublex_start());
8298 LOP(OP_TRUNCATE,XTERM);
8310 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8312 pl_yylval.ival = CopLINE(PL_curcop);
8316 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8318 pl_yylval.ival = CopLINE(PL_curcop);
8322 LOP(OP_UNLINK,XTERM);
8328 LOP(OP_UNPACK,XTERM);
8331 LOP(OP_UTIME,XTERM);
8337 LOP(OP_UNSHIFT,XTERM);
8340 s = tokenize_use(1, s);
8350 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8352 pl_yylval.ival = CopLINE(PL_curcop);
8356 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8358 pl_yylval.ival = CopLINE(PL_curcop);
8362 PL_hints |= HINT_BLOCK_SCOPE;
8369 LOP(OP_WAITPID,XTERM);
8378 ctl_l[0] = toCTRL('L');
8380 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
8383 /* Make sure $^L is defined */
8384 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
8389 if (PL_expect == XOPERATOR) {
8390 if (*s == '=' && !PL_lex_allbrackets &&
8391 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8399 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8401 pl_yylval.ival = OP_XOR;
8406 TERM(sublex_start());
8411 #pragma segment Main
8415 S_pending_ident(pTHX)
8420 /* pit holds the identifier we read and pending_ident is reset */
8421 char pit = PL_pending_ident;
8422 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8423 /* All routes through this function want to know if there is a colon. */
8424 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8425 PL_pending_ident = 0;
8427 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
8428 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8429 "### Pending identifier '%s'\n", PL_tokenbuf); });
8431 /* if we're in a my(), we can't allow dynamics here.
8432 $foo'bar has already been turned into $foo::bar, so
8433 just check for colons.
8435 if it's a legal name, the OP is a PADANY.
8438 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8440 yyerror(Perl_form(aTHX_ "No package name allowed for "
8441 "variable %s in \"our\"",
8443 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8447 yyerror(Perl_form(aTHX_ PL_no_myglob,
8448 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
8450 pl_yylval.opval = newOP(OP_PADANY, 0);
8451 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8452 UTF ? SVf_UTF8 : 0);
8458 build the ops for accesses to a my() variable.
8460 Deny my($a) or my($b) in a sort block, *if* $a or $b is
8461 then used in a comparison. This catches most, but not
8462 all cases. For instance, it catches
8463 sort { my($a); $a <=> $b }
8465 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
8466 (although why you'd do that is anyone's guess).
8471 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8472 UTF ? SVf_UTF8 : 0);
8473 if (tmp != NOT_IN_PAD) {
8474 /* might be an "our" variable" */
8475 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8476 /* build ops for a bareword */
8477 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8478 HEK * const stashname = HvNAME_HEK(stash);
8479 SV * const sym = newSVhek(stashname);
8480 sv_catpvs(sym, "::");
8481 sv_catpvn_flags(sym, PL_tokenbuf+1, tokenbuf_len - 1, (UTF ? SV_CATUTF8 : SV_CATBYTES ));
8482 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8483 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8486 ? (GV_ADDMULTI | GV_ADDINEVAL)
8489 ((PL_tokenbuf[0] == '$') ? SVt_PV
8490 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8495 /* if it's a sort block and they're naming $a or $b */
8496 if (PL_last_lop_op == OP_SORT &&
8497 PL_tokenbuf[0] == '$' &&
8498 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
8501 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
8502 d < PL_bufend && *d != '\n';
8505 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
8506 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
8512 pl_yylval.opval = newOP(OP_PADANY, 0);
8513 pl_yylval.opval->op_targ = tmp;
8519 Whine if they've said @foo in a doublequoted string,
8520 and @foo isn't a variable we can find in the symbol
8523 if (ckWARN(WARN_AMBIGUOUS) &&
8524 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8525 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1,
8526 ( UTF ? SVf_UTF8 : 0 ), SVt_PVAV);
8527 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8528 /* DO NOT warn for @- and @+ */
8529 && !( PL_tokenbuf[2] == '\0' &&
8530 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8533 /* Downgraded from fatal to warning 20000522 mjd */
8534 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8535 "Possible unintended interpolation of %s in string",
8540 /* build ops for a bareword */
8541 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(PL_tokenbuf + 1,
8543 UTF ? SVf_UTF8 : 0 ));
8544 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8545 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8546 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD)
8547 | ( UTF ? SVf_UTF8 : 0 ),
8548 ((PL_tokenbuf[0] == '$') ? SVt_PV
8549 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8555 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8559 PERL_ARGS_ASSERT_CHECKCOMMA;
8561 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8562 if (ckWARN(WARN_SYNTAX)) {
8565 for (w = s+2; *w && level; w++) {
8573 /* the list of chars below is for end of statements or
8574 * block / parens, boolean operators (&&, ||, //) and branch
8575 * constructs (or, and, if, until, unless, while, err, for).
8576 * Not a very solid hack... */
8577 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8578 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8579 "%s (...) interpreted as function",name);
8582 while (s < PL_bufend && isSPACE(*s))
8586 while (s < PL_bufend && isSPACE(*s))
8588 if (isIDFIRST_lazy_if(s,UTF)) {
8589 const char * const w = s++;
8590 while (isALNUM_lazy_if(s,UTF))
8592 while (s < PL_bufend && isSPACE(*s))
8596 if (keyword(w, s - w, 0))
8599 gv = gv_fetchpvn_flags(w, s - w, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
8600 if (gv && GvCVu(gv))
8602 Perl_croak(aTHX_ "No comma allowed after %s", what);
8607 /* Either returns sv, or mortalizes sv and returns a new SV*.
8608 Best used as sv=new_constant(..., sv, ...).
8609 If s, pv are NULL, calls subroutine with one argument,
8610 and type is used with error messages only. */
8613 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
8614 SV *sv, SV *pv, const char *type, STRLEN typelen)
8617 HV * table = GvHV(PL_hintgv); /* ^H */
8621 const char *why1 = "", *why2 = "", *why3 = "";
8623 PERL_ARGS_ASSERT_NEW_CONSTANT;
8625 /* charnames doesn't work well if there have been errors found */
8626 if (PL_error_count > 0 && strEQ(key,"charnames"))
8627 return &PL_sv_undef;
8630 || ! (PL_hints & HINT_LOCALIZE_HH)
8631 || ! (cvp = hv_fetch(table, key, keylen, FALSE))
8636 /* Here haven't found what we're looking for. If it is charnames,
8637 * perhaps it needs to be loaded. Try doing that before giving up */
8638 if (strEQ(key,"charnames")) {
8639 Perl_load_module(aTHX_
8641 newSVpvs("_charnames"),
8642 /* version parameter; no need to specify it, as if
8643 * we get too early a version, will fail anyway,
8644 * not being able to find '_charnames' */
8650 table = GvHV(PL_hintgv);
8652 && (PL_hints & HINT_LOCALIZE_HH)
8653 && (cvp = hv_fetch(table, key, keylen, FALSE))
8659 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8660 msg = Perl_newSVpvf(aTHX_
8661 "Constant(%s) unknown", (type ? type: "undef"));
8666 why3 = "} is not defined";
8668 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
8669 (type ? type: "undef"), why1, why2, why3);
8671 yyerror(SvPVX_const(msg));
8676 sv_2mortal(sv); /* Parent created it permanently */
8679 pv = newSVpvn_flags(s, len, SVs_TEMP);
8681 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
8683 typesv = &PL_sv_undef;
8685 PUSHSTACKi(PERLSI_OVERLOAD);
8697 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
8701 /* Check the eval first */
8702 if (!PL_in_eval && SvTRUE(ERRSV)) {
8703 sv_catpvs(ERRSV, "Propagated");
8704 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
8706 res = SvREFCNT_inc_simple(sv);
8710 SvREFCNT_inc_simple_void(res);
8719 why1 = "Call to &{$^H{";
8721 why3 = "}} did not return a defined value";
8729 /* Returns a NUL terminated string, with the length of the string written to
8733 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
8736 register char *d = dest;
8737 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
8739 PERL_ARGS_ASSERT_SCAN_WORD;
8743 Perl_croak(aTHX_ ident_too_long);
8744 if (isALNUM(*s)) /* UTF handled below */
8746 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
8751 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
8755 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
8756 char *t = s + UTF8SKIP(s);
8758 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
8762 Perl_croak(aTHX_ ident_too_long);
8763 Copy(s, d, len, char);
8776 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
8779 char *bracket = NULL;
8781 register char *d = dest;
8782 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
8784 PERL_ARGS_ASSERT_SCAN_IDENT;
8789 while (isDIGIT(*s)) {
8791 Perl_croak(aTHX_ ident_too_long);
8798 Perl_croak(aTHX_ ident_too_long);
8799 if (isALNUM(*s)) /* UTF handled below */
8801 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
8806 else if (*s == ':' && s[1] == ':') {
8810 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
8811 char *t = s + UTF8SKIP(s);
8812 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
8814 if (d + (t - s) > e)
8815 Perl_croak(aTHX_ ident_too_long);
8816 Copy(s, d, t - s, char);
8827 if (PL_lex_state != LEX_NORMAL)
8828 PL_lex_state = LEX_INTERPENDMAYBE;
8831 if (*s == '$' && s[1] &&
8832 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
8844 const STRLEN skip = UTF8SKIP(s);
8847 for ( i = 0; i < skip; i++ )
8855 if (*d == '^' && *s && isCONTROLVAR(*s)) {
8860 if (isSPACE(s[-1])) {
8862 const char ch = *s++;
8863 if (!SPACE_OR_TAB(ch)) {
8869 if (isIDFIRST_lazy_if(d,UTF)) {
8873 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
8874 end += UTF8SKIP(end);
8875 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
8876 end += UTF8SKIP(end);
8878 Copy(s, d, end - s, char);
8883 while ((isALNUM(*s) || *s == ':') && d < e)
8886 Perl_croak(aTHX_ ident_too_long);
8889 while (s < send && SPACE_OR_TAB(*s))
8891 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
8892 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
8893 const char * const brack =
8895 ((*s == '[') ? "[...]" : "{...}");
8896 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
8897 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8898 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
8899 funny, dest, brack, funny, dest, brack);
8902 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
8903 PL_lex_allbrackets++;
8907 /* Handle extended ${^Foo} variables
8908 * 1999-02-27 mjd-perl-patch@plover.com */
8909 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
8913 while (isALNUM(*s) && d < e) {
8917 Perl_croak(aTHX_ ident_too_long);
8922 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
8923 PL_lex_state = LEX_INTERPEND;
8926 if (PL_lex_state == LEX_NORMAL) {
8927 if (ckWARN(WARN_AMBIGUOUS) &&
8928 (keyword(dest, d - dest, 0)
8929 || get_cvn_flags(dest, d - dest, 0)))
8933 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8934 "Ambiguous use of %c{%s} resolved to %c%s",
8935 funny, dest, funny, dest);
8940 s = bracket; /* let the parser handle it */
8944 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
8945 PL_lex_state = LEX_INTERPEND;
8950 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
8952 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
8953 * the parse starting at 's', based on the subset that are valid in this
8954 * context input to this routine in 'valid_flags'. Advances s. Returns
8955 * TRUE if the input was a valid flag, so the next char may be as well;
8956 * otherwise FALSE. 'charset' should point to a NUL upon first call on the
8957 * current regex. This routine will set it to any charset modifier found.
8958 * The caller shouldn't change it. This way, another charset modifier
8959 * encountered in the parse can be detected as an error, as we have decided
8964 if (! strchr(valid_flags, c)) {
8973 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
8974 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
8975 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
8976 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
8977 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
8978 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
8979 case LOCALE_PAT_MOD:
8981 /* In 5.14, qr//lt is legal but deprecated; the 't' means they
8982 * can't be regex modifiers.
8983 * In 5.14, s///le is legal and ambiguous. Try to disambiguate as
8984 * much as easily done. s///lei, for example, has to mean regex
8985 * modifiers if it's not an error (as does any word character
8986 * following the 'e'). Otherwise, we resolve to the backwards-
8987 * compatible, but less likely 's/// le ...', i.e. as meaning
8988 * less-than-or-equal. The reason it's not likely is that s//
8989 * returns a number for code in the field (/r returns a string, but
8990 * that wasn't added until the 5.13 series), and so '<=' should be
8991 * used for comparing, not 'le'. */
8992 if (*((*s) + 1) == 't') {
8995 else if (*((*s) + 1) == 'e' && ! isALNUM(*((*s) + 2))) {
8997 /* 'e' is valid only for substitutes, s///e. If it is not
8998 * valid in the current context, then 'm//le' must mean the
8999 * comparison operator, so use the regular deprecation message.
9001 if (! strchr(valid_flags, 'e')) {
9004 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
9005 "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 5.18, it will be resolved the other way");
9009 goto multiple_charsets;
9011 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
9014 case UNICODE_PAT_MOD:
9015 /* In 5.14, qr//unless and qr//until are legal but deprecated; the
9016 * 'n' means they can't be regex modifiers */
9017 if (*((*s) + 1) == 'n') {
9021 goto multiple_charsets;
9023 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
9026 case ASCII_RESTRICT_PAT_MOD:
9027 /* In 5.14, qr//and is legal but deprecated; the 'n' means they
9028 * can't be regex modifiers */
9029 if (*((*s) + 1) == 'n') {
9034 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
9038 /* Error if previous modifier wasn't an 'a', but if it was, see
9039 * if, and accept, a second occurrence (only) */
9041 || get_regex_charset(*pmfl)
9042 != REGEX_ASCII_RESTRICTED_CHARSET)
9044 goto multiple_charsets;
9046 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
9050 case DEPENDS_PAT_MOD:
9052 goto multiple_charsets;
9054 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
9063 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX),
9064 "Having no space between pattern and following word is deprecated");
9068 if (*charset != c) {
9069 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
9071 else if (c == 'a') {
9072 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
9075 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
9078 /* Pretend that it worked, so will continue processing before dieing */
9084 S_scan_pat(pTHX_ char *start, I32 type)
9088 char *s = scan_str(start,!!PL_madskills,FALSE);
9089 const char * const valid_flags =
9090 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
9091 char charset = '\0'; /* character set modifier */
9096 PERL_ARGS_ASSERT_SCAN_PAT;
9099 const char * const delimiter = skipspace(start);
9103 ? "Search pattern not terminated or ternary operator parsed as search pattern"
9104 : "Search pattern not terminated" ));
9107 pm = (PMOP*)newPMOP(type, 0);
9108 if (PL_multi_open == '?') {
9109 /* This is the only point in the code that sets PMf_ONCE: */
9110 pm->op_pmflags |= PMf_ONCE;
9112 /* Hence it's safe to do this bit of PMOP book-keeping here, which
9113 allows us to restrict the list needed by reset to just the ??
9115 assert(type != OP_TRANS);
9117 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
9120 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
9123 elements = mg->mg_len / sizeof(PMOP**);
9124 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
9125 ((PMOP**)mg->mg_ptr) [elements++] = pm;
9126 mg->mg_len = elements * sizeof(PMOP**);
9127 PmopSTASH_set(pm,PL_curstash);
9133 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
9135 if (PL_madskills && modstart != s) {
9136 SV* tmptoken = newSVpvn(modstart, s - modstart);
9137 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
9140 /* issue a warning if /c is specified,but /g is not */
9141 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
9143 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
9144 "Use of /c modifier is meaningless without /g" );
9147 PL_lex_op = (OP*)pm;
9148 pl_yylval.ival = OP_MATCH;
9153 S_scan_subst(pTHX_ char *start)
9160 char charset = '\0'; /* character set modifier */
9165 PERL_ARGS_ASSERT_SCAN_SUBST;
9167 pl_yylval.ival = OP_NULL;
9169 s = scan_str(start,!!PL_madskills,FALSE);
9172 Perl_croak(aTHX_ "Substitution pattern not terminated");
9174 if (s[-1] == PL_multi_open)
9178 CURMAD('q', PL_thisopen);
9179 CURMAD('_', PL_thiswhite);
9180 CURMAD('E', PL_thisstuff);
9181 CURMAD('Q', PL_thisclose);
9182 PL_realtokenstart = s - SvPVX(PL_linestr);
9186 first_start = PL_multi_start;
9187 s = scan_str(s,!!PL_madskills,FALSE);
9190 SvREFCNT_dec(PL_lex_stuff);
9191 PL_lex_stuff = NULL;
9193 Perl_croak(aTHX_ "Substitution replacement not terminated");
9195 PL_multi_start = first_start; /* so whole substitution is taken together */
9197 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9201 CURMAD('z', PL_thisopen);
9202 CURMAD('R', PL_thisstuff);
9203 CURMAD('Z', PL_thisclose);
9209 if (*s == EXEC_PAT_MOD) {
9213 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
9222 curmad('m', newSVpvn(modstart, s - modstart));
9223 append_madprops(PL_thismad, (OP*)pm, 0);
9227 if ((pm->op_pmflags & PMf_CONTINUE)) {
9228 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9232 SV * const repl = newSVpvs("");
9234 PL_sublex_info.super_bufptr = s;
9235 PL_sublex_info.super_bufend = PL_bufend;
9237 pm->op_pmflags |= PMf_EVAL;
9240 sv_catpvs(repl, "eval ");
9242 sv_catpvs(repl, "do ");
9244 sv_catpvs(repl, "{");
9245 sv_catsv(repl, PL_lex_repl);
9246 if (strchr(SvPVX(PL_lex_repl), '#'))
9247 sv_catpvs(repl, "\n");
9248 sv_catpvs(repl, "}");
9250 SvREFCNT_dec(PL_lex_repl);
9254 PL_lex_op = (OP*)pm;
9255 pl_yylval.ival = OP_SUBST;
9260 S_scan_trans(pTHX_ char *start)
9269 bool nondestruct = 0;
9274 PERL_ARGS_ASSERT_SCAN_TRANS;
9276 pl_yylval.ival = OP_NULL;
9278 s = scan_str(start,!!PL_madskills,FALSE);
9280 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9282 if (s[-1] == PL_multi_open)
9286 CURMAD('q', PL_thisopen);
9287 CURMAD('_', PL_thiswhite);
9288 CURMAD('E', PL_thisstuff);
9289 CURMAD('Q', PL_thisclose);
9290 PL_realtokenstart = s - SvPVX(PL_linestr);
9294 s = scan_str(s,!!PL_madskills,FALSE);
9297 SvREFCNT_dec(PL_lex_stuff);
9298 PL_lex_stuff = NULL;
9300 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9303 CURMAD('z', PL_thisopen);
9304 CURMAD('R', PL_thisstuff);
9305 CURMAD('Z', PL_thisclose);
9308 complement = del = squash = 0;
9315 complement = OPpTRANS_COMPLEMENT;
9318 del = OPpTRANS_DELETE;
9321 squash = OPpTRANS_SQUASH;
9333 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
9334 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)tbl);
9335 o->op_private &= ~OPpTRANS_ALL;
9336 o->op_private |= del|squash|complement|
9337 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9338 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
9341 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9346 curmad('m', newSVpvn(modstart, s - modstart));
9347 append_madprops(PL_thismad, o, 0);
9356 S_scan_heredoc(pTHX_ register char *s)
9360 I32 op_type = OP_SCALAR;
9364 const char *found_newline;
9368 const int outer = (PL_rsfp || PL_parser->filtered)
9369 && !(PL_lex_inwhat == OP_SCALAR);
9371 I32 stuffstart = s - SvPVX(PL_linestr);
9374 PL_realtokenstart = -1;
9377 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9381 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9385 while (SPACE_OR_TAB(*peek))
9387 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9390 s = delimcpy(d, e, s, PL_bufend, term, &len);
9400 if (!isALNUM_lazy_if(s,UTF))
9401 deprecate("bare << to mean <<\"\"");
9402 for (; isALNUM_lazy_if(s,UTF); s++) {
9407 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9408 Perl_croak(aTHX_ "Delimiter for here document is too long");
9411 len = d - PL_tokenbuf;
9415 tstart = PL_tokenbuf + !outer;
9416 PL_thisclose = newSVpvn(tstart, len - !outer);
9417 tstart = SvPVX(PL_linestr) + stuffstart;
9418 PL_thisopen = newSVpvn(tstart, s - tstart);
9419 stuffstart = s - SvPVX(PL_linestr);
9422 #ifndef PERL_STRICT_CR
9423 d = strchr(s, '\r');
9425 char * const olds = s;
9427 while (s < PL_bufend) {
9433 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9442 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9449 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
9450 herewas = newSVpvn(s,PL_bufend-s);
9454 herewas = newSVpvn(s-1,found_newline-s+1);
9457 herewas = newSVpvn(s,found_newline-s);
9462 tstart = SvPVX(PL_linestr) + stuffstart;
9464 sv_catpvn(PL_thisstuff, tstart, s - tstart);
9466 PL_thisstuff = newSVpvn(tstart, s - tstart);
9469 s += SvCUR(herewas);
9472 stuffstart = s - SvPVX(PL_linestr);
9478 tmpstr = newSV_type(SVt_PVIV);
9482 SvIV_set(tmpstr, -1);
9484 else if (term == '`') {
9485 op_type = OP_BACKTICK;
9486 SvIV_set(tmpstr, '\\');
9490 PL_multi_start = CopLINE(PL_curcop);
9491 PL_multi_open = PL_multi_close = '<';
9492 term = *PL_tokenbuf;
9493 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp
9494 && !PL_parser->filtered) {
9495 char * const bufptr = PL_sublex_info.super_bufptr;
9496 char * const bufend = PL_sublex_info.super_bufend;
9497 char * const olds = s - SvCUR(herewas);
9498 s = strchr(bufptr, '\n');
9502 while (s < bufend &&
9503 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9505 CopLINE_inc(PL_curcop);
9508 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9509 missingterm(PL_tokenbuf);
9511 sv_setpvn(herewas,bufptr,d-bufptr+1);
9512 sv_setpvn(tmpstr,d+1,s-d);
9514 sv_catpvn(herewas,s,bufend-s);
9515 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
9522 while (s < PL_bufend &&
9523 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9525 CopLINE_inc(PL_curcop);
9527 if (s >= PL_bufend) {
9528 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9529 missingterm(PL_tokenbuf);
9531 sv_setpvn(tmpstr,d+1,s-d);
9535 sv_catpvn(PL_thisstuff, d + 1, s - d);
9537 PL_thisstuff = newSVpvn(d + 1, s - d);
9538 stuffstart = s - SvPVX(PL_linestr);
9542 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
9544 sv_catpvn(herewas,s,PL_bufend-s);
9545 sv_setsv(PL_linestr,herewas);
9546 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
9547 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9548 PL_last_lop = PL_last_uni = NULL;
9551 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9552 while (s >= PL_bufend) { /* multiple line string? */
9555 tstart = SvPVX(PL_linestr) + stuffstart;
9557 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
9559 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
9563 CopLINE_inc(PL_curcop);
9564 if (!outer || !lex_next_chunk(0)) {
9565 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9566 missingterm(PL_tokenbuf);
9568 CopLINE_dec(PL_curcop);
9571 stuffstart = s - SvPVX(PL_linestr);
9573 CopLINE_inc(PL_curcop);
9574 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9575 PL_last_lop = PL_last_uni = NULL;
9576 #ifndef PERL_STRICT_CR
9577 if (PL_bufend - PL_linestart >= 2) {
9578 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
9579 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9581 PL_bufend[-2] = '\n';
9583 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9585 else if (PL_bufend[-1] == '\r')
9586 PL_bufend[-1] = '\n';
9588 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9589 PL_bufend[-1] = '\n';
9591 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
9592 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
9593 *(SvPVX(PL_linestr) + off ) = ' ';
9594 lex_grow_linestr(SvCUR(PL_linestr) + SvCUR(herewas) + 1);
9595 sv_catsv(PL_linestr,herewas);
9596 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9597 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
9601 sv_catsv(tmpstr,PL_linestr);
9606 PL_multi_end = CopLINE(PL_curcop);
9607 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9608 SvPV_shrink_to_cur(tmpstr);
9610 SvREFCNT_dec(herewas);
9612 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9614 else if (PL_encoding)
9615 sv_recode_to_utf8(tmpstr, PL_encoding);
9617 PL_lex_stuff = tmpstr;
9618 pl_yylval.ival = op_type;
9623 takes: current position in input buffer
9624 returns: new position in input buffer
9625 side-effects: pl_yylval and lex_op are set.
9630 <FH> read from filehandle
9631 <pkg::FH> read from package qualified filehandle
9632 <pkg'FH> read from package qualified filehandle
9633 <$fh> read from filehandle in $fh
9639 S_scan_inputsymbol(pTHX_ char *start)
9642 register char *s = start; /* current position in buffer */
9645 char *d = PL_tokenbuf; /* start of temp holding space */
9646 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9648 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
9650 end = strchr(s, '\n');
9653 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9655 /* die if we didn't have space for the contents of the <>,
9656 or if it didn't end, or if we see a newline
9659 if (len >= (I32)sizeof PL_tokenbuf)
9660 Perl_croak(aTHX_ "Excessively long <> operator");
9662 Perl_croak(aTHX_ "Unterminated <> operator");
9667 Remember, only scalar variables are interpreted as filehandles by
9668 this code. Anything more complex (e.g., <$fh{$num}>) will be
9669 treated as a glob() call.
9670 This code makes use of the fact that except for the $ at the front,
9671 a scalar variable and a filehandle look the same.
9673 if (*d == '$' && d[1]) d++;
9675 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9676 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9677 d += UTF ? UTF8SKIP(d) : 1;
9679 /* If we've tried to read what we allow filehandles to look like, and
9680 there's still text left, then it must be a glob() and not a getline.
9681 Use scan_str to pull out the stuff between the <> and treat it
9682 as nothing more than a string.
9685 if (d - PL_tokenbuf != len) {
9686 pl_yylval.ival = OP_GLOB;
9687 s = scan_str(start,!!PL_madskills,FALSE);
9689 Perl_croak(aTHX_ "Glob not terminated");
9693 bool readline_overriden = FALSE;
9696 /* we're in a filehandle read situation */
9699 /* turn <> into <ARGV> */
9701 Copy("ARGV",d,5,char);
9703 /* Check whether readline() is overriden */
9704 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
9706 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
9708 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
9709 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
9710 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
9711 readline_overriden = TRUE;
9713 /* if <$fh>, create the ops to turn the variable into a
9717 /* try to find it in the pad for this block, otherwise find
9718 add symbol table ops
9720 const PADOFFSET tmp = pad_findmy_pvn(d, len, UTF ? SVf_UTF8 : 0);
9721 if (tmp != NOT_IN_PAD) {
9722 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
9723 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
9724 HEK * const stashname = HvNAME_HEK(stash);
9725 SV * const sym = sv_2mortal(newSVhek(stashname));
9726 sv_catpvs(sym, "::");
9732 OP * const o = newOP(OP_PADSV, 0);
9734 PL_lex_op = readline_overriden
9735 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9736 op_append_elem(OP_LIST, o,
9737 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
9738 : (OP*)newUNOP(OP_READLINE, 0, o);
9747 ? (GV_ADDMULTI | GV_ADDINEVAL)
9748 : GV_ADDMULTI) | ( UTF ? SVf_UTF8 : 0 ),
9750 PL_lex_op = readline_overriden
9751 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9752 op_append_elem(OP_LIST,
9753 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
9754 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9755 : (OP*)newUNOP(OP_READLINE, 0,
9756 newUNOP(OP_RV2SV, 0,
9757 newGVOP(OP_GV, 0, gv)));
9759 if (!readline_overriden)
9760 PL_lex_op->op_flags |= OPf_SPECIAL;
9761 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
9762 pl_yylval.ival = OP_NULL;
9765 /* If it's none of the above, it must be a literal filehandle
9766 (<Foo::BAR> or <FOO>) so build a simple readline OP */
9768 GV * const gv = gv_fetchpv(d, GV_ADD | ( UTF ? SVf_UTF8 : 0 ), SVt_PVIO);
9769 PL_lex_op = readline_overriden
9770 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9771 op_append_elem(OP_LIST,
9772 newGVOP(OP_GV, 0, gv),
9773 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9774 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
9775 pl_yylval.ival = OP_NULL;
9784 takes: start position in buffer
9785 keep_quoted preserve \ on the embedded delimiter(s)
9786 keep_delims preserve the delimiters around the string
9787 returns: position to continue reading from buffer
9788 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
9789 updates the read buffer.
9791 This subroutine pulls a string out of the input. It is called for:
9792 q single quotes q(literal text)
9793 ' single quotes 'literal text'
9794 qq double quotes qq(interpolate $here please)
9795 " double quotes "interpolate $here please"
9796 qx backticks qx(/bin/ls -l)
9797 ` backticks `/bin/ls -l`
9798 qw quote words @EXPORT_OK = qw( func() $spam )
9799 m// regexp match m/this/
9800 s/// regexp substitute s/this/that/
9801 tr/// string transliterate tr/this/that/
9802 y/// string transliterate y/this/that/
9803 ($*@) sub prototypes sub foo ($)
9804 (stuff) sub attr parameters sub foo : attr(stuff)
9805 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
9807 In most of these cases (all but <>, patterns and transliterate)
9808 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
9809 calls scan_str(). s/// makes yylex() call scan_subst() which calls
9810 scan_str(). tr/// and y/// make yylex() call scan_trans() which
9813 It skips whitespace before the string starts, and treats the first
9814 character as the delimiter. If the delimiter is one of ([{< then
9815 the corresponding "close" character )]}> is used as the closing
9816 delimiter. It allows quoting of delimiters, and if the string has
9817 balanced delimiters ([{<>}]) it allows nesting.
9819 On success, the SV with the resulting string is put into lex_stuff or,
9820 if that is already non-NULL, into lex_repl. The second case occurs only
9821 when parsing the RHS of the special constructs s/// and tr/// (y///).
9822 For convenience, the terminating delimiter character is stuffed into
9827 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
9830 SV *sv; /* scalar value: string */
9831 const char *tmps; /* temp string, used for delimiter matching */
9832 register char *s = start; /* current position in the buffer */
9833 register char term; /* terminating character */
9834 register char *to; /* current position in the sv's data */
9835 I32 brackets = 1; /* bracket nesting level */
9836 bool has_utf8 = FALSE; /* is there any utf8 content? */
9837 I32 termcode; /* terminating char. code */
9838 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
9839 STRLEN termlen; /* length of terminating string */
9840 int last_off = 0; /* last position for nesting bracket */
9846 PERL_ARGS_ASSERT_SCAN_STR;
9848 /* skip space before the delimiter */
9854 if (PL_realtokenstart >= 0) {
9855 stuffstart = PL_realtokenstart;
9856 PL_realtokenstart = -1;
9859 stuffstart = start - SvPVX(PL_linestr);
9861 /* mark where we are, in case we need to report errors */
9864 /* after skipping whitespace, the next character is the terminator */
9867 termcode = termstr[0] = term;
9871 termcode = utf8_to_uvchr((U8*)s, &termlen);
9872 Copy(s, termstr, termlen, U8);
9873 if (!UTF8_IS_INVARIANT(term))
9877 /* mark where we are */
9878 PL_multi_start = CopLINE(PL_curcop);
9879 PL_multi_open = term;
9881 /* find corresponding closing delimiter */
9882 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
9883 termcode = termstr[0] = term = tmps[5];
9885 PL_multi_close = term;
9887 /* create a new SV to hold the contents. 79 is the SV's initial length.
9888 What a random number. */
9889 sv = newSV_type(SVt_PVIV);
9891 SvIV_set(sv, termcode);
9892 (void)SvPOK_only(sv); /* validate pointer */
9894 /* move past delimiter and try to read a complete string */
9896 sv_catpvn(sv, s, termlen);
9899 tstart = SvPVX(PL_linestr) + stuffstart;
9900 if (!PL_thisopen && !keep_delims) {
9901 PL_thisopen = newSVpvn(tstart, s - tstart);
9902 stuffstart = s - SvPVX(PL_linestr);
9906 if (PL_encoding && !UTF) {
9910 int offset = s - SvPVX_const(PL_linestr);
9911 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
9912 &offset, (char*)termstr, termlen);
9913 const char * const ns = SvPVX_const(PL_linestr) + offset;
9914 char * const svlast = SvEND(sv) - 1;
9916 for (; s < ns; s++) {
9917 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9918 CopLINE_inc(PL_curcop);
9921 goto read_more_line;
9923 /* handle quoted delimiters */
9924 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
9926 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
9928 if ((svlast-1 - t) % 2) {
9932 SvCUR_set(sv, SvCUR(sv) - 1);
9937 if (PL_multi_open == PL_multi_close) {
9943 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
9944 /* At here, all closes are "was quoted" one,
9945 so we don't check PL_multi_close. */
9947 if (!keep_quoted && *(t+1) == PL_multi_open)
9952 else if (*t == PL_multi_open)
9960 SvCUR_set(sv, w - SvPVX_const(sv));
9962 last_off = w - SvPVX(sv);
9963 if (--brackets <= 0)
9969 SvCUR_set(sv, SvCUR(sv) - 1);
9975 /* extend sv if need be */
9976 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
9977 /* set 'to' to the next character in the sv's string */
9978 to = SvPVX(sv)+SvCUR(sv);
9980 /* if open delimiter is the close delimiter read unbridle */
9981 if (PL_multi_open == PL_multi_close) {
9982 for (; s < PL_bufend; s++,to++) {
9983 /* embedded newlines increment the current line number */
9984 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9985 CopLINE_inc(PL_curcop);
9986 /* handle quoted delimiters */
9987 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
9988 if (!keep_quoted && s[1] == term)
9990 /* any other quotes are simply copied straight through */
9994 /* terminate when run out of buffer (the for() condition), or
9995 have found the terminator */
9996 else if (*s == term) {
9999 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
10002 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10008 /* if the terminator isn't the same as the start character (e.g.,
10009 matched brackets), we have to allow more in the quoting, and
10010 be prepared for nested brackets.
10013 /* read until we run out of string, or we find the terminator */
10014 for (; s < PL_bufend; s++,to++) {
10015 /* embedded newlines increment the line count */
10016 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10017 CopLINE_inc(PL_curcop);
10018 /* backslashes can escape the open or closing characters */
10019 if (*s == '\\' && s+1 < PL_bufend) {
10020 if (!keep_quoted &&
10021 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
10026 /* allow nested opens and closes */
10027 else if (*s == PL_multi_close && --brackets <= 0)
10029 else if (*s == PL_multi_open)
10031 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10036 /* terminate the copied string and update the sv's end-of-string */
10038 SvCUR_set(sv, to - SvPVX_const(sv));
10041 * this next chunk reads more into the buffer if we're not done yet
10045 break; /* handle case where we are done yet :-) */
10047 #ifndef PERL_STRICT_CR
10048 if (to - SvPVX_const(sv) >= 2) {
10049 if ((to[-2] == '\r' && to[-1] == '\n') ||
10050 (to[-2] == '\n' && to[-1] == '\r'))
10054 SvCUR_set(sv, to - SvPVX_const(sv));
10056 else if (to[-1] == '\r')
10059 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
10064 /* if we're out of file, or a read fails, bail and reset the current
10065 line marker so we can report where the unterminated string began
10068 if (PL_madskills) {
10069 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10071 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10073 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10076 CopLINE_inc(PL_curcop);
10077 PL_bufptr = PL_bufend;
10078 if (!lex_next_chunk(0)) {
10080 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10089 /* at this point, we have successfully read the delimited string */
10091 if (!PL_encoding || UTF) {
10093 if (PL_madskills) {
10094 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10095 const int len = s - tstart;
10097 sv_catpvn(PL_thisstuff, tstart, len);
10099 PL_thisstuff = newSVpvn(tstart, len);
10100 if (!PL_thisclose && !keep_delims)
10101 PL_thisclose = newSVpvn(s,termlen);
10106 sv_catpvn(sv, s, termlen);
10111 if (PL_madskills) {
10112 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10113 const int len = s - tstart - termlen;
10115 sv_catpvn(PL_thisstuff, tstart, len);
10117 PL_thisstuff = newSVpvn(tstart, len);
10118 if (!PL_thisclose && !keep_delims)
10119 PL_thisclose = newSVpvn(s - termlen,termlen);
10123 if (has_utf8 || PL_encoding)
10126 PL_multi_end = CopLINE(PL_curcop);
10128 /* if we allocated too much space, give some back */
10129 if (SvCUR(sv) + 5 < SvLEN(sv)) {
10130 SvLEN_set(sv, SvCUR(sv) + 1);
10131 SvPV_renew(sv, SvLEN(sv));
10134 /* decide whether this is the first or second quoted string we've read
10147 takes: pointer to position in buffer
10148 returns: pointer to new position in buffer
10149 side-effects: builds ops for the constant in pl_yylval.op
10151 Read a number in any of the formats that Perl accepts:
10153 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
10154 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
10157 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
10159 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
10162 If it reads a number without a decimal point or an exponent, it will
10163 try converting the number to an integer and see if it can do so
10164 without loss of precision.
10168 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10171 register const char *s = start; /* current position in buffer */
10172 register char *d; /* destination in temp buffer */
10173 register char *e; /* end of temp buffer */
10174 NV nv; /* number read, as a double */
10175 SV *sv = NULL; /* place to put the converted number */
10176 bool floatit; /* boolean: int or float? */
10177 const char *lastub = NULL; /* position of last underbar */
10178 static char const number_too_long[] = "Number too long";
10180 PERL_ARGS_ASSERT_SCAN_NUM;
10182 /* We use the first character to decide what type of number this is */
10186 Perl_croak(aTHX_ "panic: scan_num");
10188 /* if it starts with a 0, it could be an octal number, a decimal in
10189 0.13 disguise, or a hexadecimal number, or a binary number. */
10193 u holds the "number so far"
10194 shift the power of 2 of the base
10195 (hex == 4, octal == 3, binary == 1)
10196 overflowed was the number more than we can hold?
10198 Shift is used when we add a digit. It also serves as an "are
10199 we in octal/hex/binary?" indicator to disallow hex characters
10200 when in octal mode.
10205 bool overflowed = FALSE;
10206 bool just_zero = TRUE; /* just plain 0 or binary number? */
10207 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10208 static const char* const bases[5] =
10209 { "", "binary", "", "octal", "hexadecimal" };
10210 static const char* const Bases[5] =
10211 { "", "Binary", "", "Octal", "Hexadecimal" };
10212 static const char* const maxima[5] =
10214 "0b11111111111111111111111111111111",
10218 const char *base, *Base, *max;
10220 /* check for hex */
10221 if (s[1] == 'x' || s[1] == 'X') {
10225 } else if (s[1] == 'b' || s[1] == 'B') {
10230 /* check for a decimal in disguise */
10231 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
10233 /* so it must be octal */
10240 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10241 "Misplaced _ in number");
10245 base = bases[shift];
10246 Base = Bases[shift];
10247 max = maxima[shift];
10249 /* read the rest of the number */
10251 /* x is used in the overflow test,
10252 b is the digit we're adding on. */
10257 /* if we don't mention it, we're done */
10261 /* _ are ignored -- but warned about if consecutive */
10263 if (lastub && s == lastub + 1)
10264 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10265 "Misplaced _ in number");
10269 /* 8 and 9 are not octal */
10270 case '8': case '9':
10272 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10276 case '2': case '3': case '4':
10277 case '5': case '6': case '7':
10279 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10282 case '0': case '1':
10283 b = *s++ & 15; /* ASCII digit -> value of digit */
10287 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10288 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10289 /* make sure they said 0x */
10292 b = (*s++ & 7) + 9;
10294 /* Prepare to put the digit we have onto the end
10295 of the number so far. We check for overflows.
10301 x = u << shift; /* make room for the digit */
10303 if ((x >> shift) != u
10304 && !(PL_hints & HINT_NEW_BINARY)) {
10307 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10308 "Integer overflow in %s number",
10311 u = x | b; /* add the digit to the end */
10314 n *= nvshift[shift];
10315 /* If an NV has not enough bits in its
10316 * mantissa to represent an UV this summing of
10317 * small low-order numbers is a waste of time
10318 * (because the NV cannot preserve the
10319 * low-order bits anyway): we could just
10320 * remember when did we overflow and in the
10321 * end just multiply n by the right
10329 /* if we get here, we had success: make a scalar value from
10334 /* final misplaced underbar check */
10335 if (s[-1] == '_') {
10336 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10340 if (n > 4294967295.0)
10341 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10342 "%s number > %s non-portable",
10348 if (u > 0xffffffff)
10349 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10350 "%s number > %s non-portable",
10355 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10356 sv = new_constant(start, s - start, "integer",
10357 sv, NULL, NULL, 0);
10358 else if (PL_hints & HINT_NEW_BINARY)
10359 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10364 handle decimal numbers.
10365 we're also sent here when we read a 0 as the first digit
10367 case '1': case '2': case '3': case '4': case '5':
10368 case '6': case '7': case '8': case '9': case '.':
10371 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10374 /* read next group of digits and _ and copy into d */
10375 while (isDIGIT(*s) || *s == '_') {
10376 /* skip underscores, checking for misplaced ones
10380 if (lastub && s == lastub + 1)
10381 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10382 "Misplaced _ in number");
10386 /* check for end of fixed-length buffer */
10388 Perl_croak(aTHX_ number_too_long);
10389 /* if we're ok, copy the character */
10394 /* final misplaced underbar check */
10395 if (lastub && s == lastub + 1) {
10396 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10399 /* read a decimal portion if there is one. avoid
10400 3..5 being interpreted as the number 3. followed
10403 if (*s == '.' && s[1] != '.') {
10408 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10409 "Misplaced _ in number");
10413 /* copy, ignoring underbars, until we run out of digits.
10415 for (; isDIGIT(*s) || *s == '_'; s++) {
10416 /* fixed length buffer check */
10418 Perl_croak(aTHX_ number_too_long);
10420 if (lastub && s == lastub + 1)
10421 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10422 "Misplaced _ in number");
10428 /* fractional part ending in underbar? */
10429 if (s[-1] == '_') {
10430 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10431 "Misplaced _ in number");
10433 if (*s == '.' && isDIGIT(s[1])) {
10434 /* oops, it's really a v-string, but without the "v" */
10440 /* read exponent part, if present */
10441 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10445 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10446 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10448 /* stray preinitial _ */
10450 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10451 "Misplaced _ in number");
10455 /* allow positive or negative exponent */
10456 if (*s == '+' || *s == '-')
10459 /* stray initial _ */
10461 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10462 "Misplaced _ in number");
10466 /* read digits of exponent */
10467 while (isDIGIT(*s) || *s == '_') {
10470 Perl_croak(aTHX_ number_too_long);
10474 if (((lastub && s == lastub + 1) ||
10475 (!isDIGIT(s[1]) && s[1] != '_')))
10476 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10477 "Misplaced _ in number");
10485 We try to do an integer conversion first if no characters
10486 indicating "float" have been found.
10491 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10493 if (flags == IS_NUMBER_IN_UV) {
10495 sv = newSViv(uv); /* Prefer IVs over UVs. */
10498 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10499 if (uv <= (UV) IV_MIN)
10500 sv = newSViv(-(IV)uv);
10507 /* terminate the string */
10509 nv = Atof(PL_tokenbuf);
10514 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10515 const char *const key = floatit ? "float" : "integer";
10516 const STRLEN keylen = floatit ? 5 : 7;
10517 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10518 key, keylen, sv, NULL, NULL, 0);
10522 /* if it starts with a v, it could be a v-string */
10525 sv = newSV(5); /* preallocate storage space */
10526 s = scan_vstring(s, PL_bufend, sv);
10530 /* make the op for the constant and return */
10533 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10535 lvalp->opval = NULL;
10541 S_scan_formline(pTHX_ register char *s)
10544 register char *eol;
10546 SV * const stuff = newSVpvs("");
10547 bool needargs = FALSE;
10548 bool eofmt = FALSE;
10550 char *tokenstart = s;
10551 SV* savewhite = NULL;
10553 if (PL_madskills) {
10554 savewhite = PL_thiswhite;
10559 PERL_ARGS_ASSERT_SCAN_FORMLINE;
10561 while (!needargs) {
10564 #ifdef PERL_STRICT_CR
10565 while (SPACE_OR_TAB(*t))
10568 while (SPACE_OR_TAB(*t) || *t == '\r')
10571 if (*t == '\n' || t == PL_bufend) {
10576 if (PL_in_eval && !PL_rsfp && !PL_parser->filtered) {
10577 eol = (char *) memchr(s,'\n',PL_bufend-s);
10582 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10584 for (t = s; t < eol; t++) {
10585 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10587 goto enough; /* ~~ must be first line in formline */
10589 if (*t == '@' || *t == '^')
10593 sv_catpvn(stuff, s, eol-s);
10594 #ifndef PERL_STRICT_CR
10595 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10596 char *end = SvPVX(stuff) + SvCUR(stuff);
10599 SvCUR_set(stuff, SvCUR(stuff) - 1);
10607 if (PL_rsfp || PL_parser->filtered) {
10610 if (PL_madskills) {
10612 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
10614 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
10617 PL_bufptr = PL_bufend;
10618 CopLINE_inc(PL_curcop);
10619 got_some = lex_next_chunk(0);
10620 CopLINE_dec(PL_curcop);
10623 tokenstart = PL_bufptr;
10631 if (SvCUR(stuff)) {
10634 PL_lex_state = LEX_NORMAL;
10635 start_force(PL_curforce);
10636 NEXTVAL_NEXTTOKE.ival = 0;
10640 PL_lex_state = LEX_FORMLINE;
10642 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10644 else if (PL_encoding)
10645 sv_recode_to_utf8(stuff, PL_encoding);
10647 start_force(PL_curforce);
10648 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10650 start_force(PL_curforce);
10651 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
10655 SvREFCNT_dec(stuff);
10657 PL_lex_formbrack = 0;
10661 if (PL_madskills) {
10663 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
10665 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
10666 PL_thiswhite = savewhite;
10673 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10676 const I32 oldsavestack_ix = PL_savestack_ix;
10677 CV* const outsidecv = PL_compcv;
10680 assert(SvTYPE(PL_compcv) == SVt_PVCV);
10682 SAVEI32(PL_subline);
10683 save_item(PL_subname);
10684 SAVESPTR(PL_compcv);
10686 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
10687 CvFLAGS(PL_compcv) |= flags;
10689 PL_subline = CopLINE(PL_curcop);
10690 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10691 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
10692 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10694 return oldsavestack_ix;
10698 #pragma segment Perl_yylex
10701 S_yywarn(pTHX_ const char *const s)
10705 PERL_ARGS_ASSERT_YYWARN;
10707 PL_in_eval |= EVAL_WARNONLY;
10709 PL_in_eval &= ~EVAL_WARNONLY;
10714 Perl_yyerror(pTHX_ const char *const s)
10717 const char *where = NULL;
10718 const char *context = NULL;
10721 int yychar = PL_parser->yychar;
10723 PERL_ARGS_ASSERT_YYERROR;
10725 if (!yychar || (yychar == ';' && !PL_rsfp))
10727 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
10728 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
10729 PL_oldbufptr != PL_bufptr) {
10732 The code below is removed for NetWare because it abends/crashes on NetWare
10733 when the script has error such as not having the closing quotes like:
10734 if ($var eq "value)
10735 Checking of white spaces is anyway done in NetWare code.
10738 while (isSPACE(*PL_oldoldbufptr))
10741 context = PL_oldoldbufptr;
10742 contlen = PL_bufptr - PL_oldoldbufptr;
10744 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
10745 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
10748 The code below is removed for NetWare because it abends/crashes on NetWare
10749 when the script has error such as not having the closing quotes like:
10750 if ($var eq "value)
10751 Checking of white spaces is anyway done in NetWare code.
10754 while (isSPACE(*PL_oldbufptr))
10757 context = PL_oldbufptr;
10758 contlen = PL_bufptr - PL_oldbufptr;
10760 else if (yychar > 255)
10761 where = "next token ???";
10762 else if (yychar == -2) { /* YYEMPTY */
10763 if (PL_lex_state == LEX_NORMAL ||
10764 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
10765 where = "at end of line";
10766 else if (PL_lex_inpat)
10767 where = "within pattern";
10769 where = "within string";
10772 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
10774 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
10775 else if (isPRINT_LC(yychar)) {
10776 const char string = yychar;
10777 sv_catpvn(where_sv, &string, 1);
10780 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
10781 where = SvPVX_const(where_sv);
10783 msg = sv_2mortal(newSVpv(s, 0));
10784 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
10785 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
10787 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
10789 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
10790 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
10791 Perl_sv_catpvf(aTHX_ msg,
10792 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
10793 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
10796 if (PL_in_eval & EVAL_WARNONLY) {
10797 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
10801 if (PL_error_count >= 10) {
10802 if (PL_in_eval && SvCUR(ERRSV))
10803 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
10804 SVfARG(ERRSV), OutCopFILE(PL_curcop));
10806 Perl_croak(aTHX_ "%s has too many errors.\n",
10807 OutCopFILE(PL_curcop));
10810 PL_in_my_stash = NULL;
10814 #pragma segment Main
10818 S_swallow_bom(pTHX_ U8 *s)
10821 const STRLEN slen = SvCUR(PL_linestr);
10823 PERL_ARGS_ASSERT_SWALLOW_BOM;
10827 if (s[1] == 0xFE) {
10828 /* UTF-16 little-endian? (or UTF-32LE?) */
10829 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
10830 /* diag_listed_as: Unsupported script encoding %s */
10831 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
10832 #ifndef PERL_NO_UTF16_FILTER
10833 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
10835 if (PL_bufend > (char*)s) {
10836 s = add_utf16_textfilter(s, TRUE);
10839 /* diag_listed_as: Unsupported script encoding %s */
10840 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10845 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
10846 #ifndef PERL_NO_UTF16_FILTER
10847 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
10849 if (PL_bufend > (char *)s) {
10850 s = add_utf16_textfilter(s, FALSE);
10853 /* diag_listed_as: Unsupported script encoding %s */
10854 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10859 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
10860 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10861 s += 3; /* UTF-8 */
10867 if (s[2] == 0xFE && s[3] == 0xFF) {
10868 /* UTF-32 big-endian */
10869 /* diag_listed_as: Unsupported script encoding %s */
10870 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
10873 else if (s[2] == 0 && s[3] != 0) {
10876 * are a good indicator of UTF-16BE. */
10877 #ifndef PERL_NO_UTF16_FILTER
10878 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
10879 s = add_utf16_textfilter(s, FALSE);
10881 /* diag_listed_as: Unsupported script encoding %s */
10882 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10888 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
10889 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10890 s += 4; /* UTF-8 */
10896 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
10899 * are a good indicator of UTF-16LE. */
10900 #ifndef PERL_NO_UTF16_FILTER
10901 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
10902 s = add_utf16_textfilter(s, TRUE);
10904 /* diag_listed_as: Unsupported script encoding %s */
10905 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10913 #ifndef PERL_NO_UTF16_FILTER
10915 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
10918 SV *const filter = FILTER_DATA(idx);
10919 /* We re-use this each time round, throwing the contents away before we
10921 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
10922 SV *const utf8_buffer = filter;
10923 IV status = IoPAGE(filter);
10924 const bool reverse = cBOOL(IoLINES(filter));
10927 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
10929 /* As we're automatically added, at the lowest level, and hence only called
10930 from this file, we can be sure that we're not called in block mode. Hence
10931 don't bother writing code to deal with block mode. */
10933 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
10936 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
10938 DEBUG_P(PerlIO_printf(Perl_debug_log,
10939 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10940 FPTR2DPTR(void *, S_utf16_textfilter),
10941 reverse ? 'l' : 'b', idx, maxlen, status,
10942 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10949 /* First, look in our buffer of existing UTF-8 data: */
10950 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
10954 } else if (status == 0) {
10956 IoPAGE(filter) = 0;
10957 nl = SvEND(utf8_buffer);
10960 STRLEN got = nl - SvPVX(utf8_buffer);
10961 /* Did we have anything to append? */
10963 sv_catpvn(sv, SvPVX(utf8_buffer), got);
10964 /* Everything else in this code works just fine if SVp_POK isn't
10965 set. This, however, needs it, and we need it to work, else
10966 we loop infinitely because the buffer is never consumed. */
10967 sv_chop(utf8_buffer, nl);
10971 /* OK, not a complete line there, so need to read some more UTF-16.
10972 Read an extra octect if the buffer currently has an odd number. */
10976 if (SvCUR(utf16_buffer) >= 2) {
10977 /* Location of the high octet of the last complete code point.
10978 Gosh, UTF-16 is a pain. All the benefits of variable length,
10979 *coupled* with all the benefits of partial reads and
10981 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
10982 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
10984 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
10988 /* We have the first half of a surrogate. Read more. */
10989 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
10992 status = FILTER_READ(idx + 1, utf16_buffer,
10993 160 + (SvCUR(utf16_buffer) & 1));
10994 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
10995 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
10998 IoPAGE(filter) = status;
11003 chars = SvCUR(utf16_buffer) >> 1;
11004 have = SvCUR(utf8_buffer);
11005 SvGROW(utf8_buffer, have + chars * 3 + 1);
11008 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
11009 (U8*)SvPVX_const(utf8_buffer) + have,
11010 chars * 2, &newlen);
11012 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
11013 (U8*)SvPVX_const(utf8_buffer) + have,
11014 chars * 2, &newlen);
11016 SvCUR_set(utf8_buffer, have + newlen);
11019 /* No need to keep this SV "well-formed" with a '\0' after the end, as
11020 it's private to us, and utf16_to_utf8{,reversed} take a
11021 (pointer,length) pair, rather than a NUL-terminated string. */
11022 if(SvCUR(utf16_buffer) & 1) {
11023 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
11024 SvCUR_set(utf16_buffer, 1);
11026 SvCUR_set(utf16_buffer, 0);
11029 DEBUG_P(PerlIO_printf(Perl_debug_log,
11030 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11032 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11033 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
11038 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
11040 SV *filter = filter_add(S_utf16_textfilter, NULL);
11042 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
11044 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
11045 sv_setpvs(filter, "");
11046 IoLINES(filter) = reversed;
11047 IoPAGE(filter) = 1; /* Not EOF */
11049 /* Sadly, we have to return a valid pointer, come what may, so we have to
11050 ignore any error return from this. */
11051 SvCUR_set(PL_linestr, 0);
11052 if (FILTER_READ(0, PL_linestr, 0)) {
11053 SvUTF8_on(PL_linestr);
11055 SvUTF8_on(PL_linestr);
11057 PL_bufend = SvEND(PL_linestr);
11058 return (U8*)SvPVX(PL_linestr);
11063 Returns a pointer to the next character after the parsed
11064 vstring, as well as updating the passed in sv.
11066 Function must be called like
11069 s = scan_vstring(s,e,sv);
11071 where s and e are the start and end of the string.
11072 The sv should already be large enough to store the vstring
11073 passed in, for performance reasons.
11078 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
11081 const char *pos = s;
11082 const char *start = s;
11084 PERL_ARGS_ASSERT_SCAN_VSTRING;
11086 if (*pos == 'v') pos++; /* get past 'v' */
11087 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11089 if ( *pos != '.') {
11090 /* this may not be a v-string if followed by => */
11091 const char *next = pos;
11092 while (next < e && isSPACE(*next))
11094 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
11095 /* return string not v-string */
11096 sv_setpvn(sv,(char *)s,pos-s);
11097 return (char *)pos;
11101 if (!isALPHA(*pos)) {
11102 U8 tmpbuf[UTF8_MAXBYTES+1];
11105 s++; /* get past 'v' */
11110 /* this is atoi() that tolerates underscores */
11113 const char *end = pos;
11115 while (--end >= s) {
11117 const UV orev = rev;
11118 rev += (*end - '0') * mult;
11121 /* diag_listed_as: Integer overflow in %s number */
11122 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
11123 "Integer overflow in decimal number");
11127 if (rev > 0x7FFFFFFF)
11128 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
11130 /* Append native character for the rev point */
11131 tmpend = uvchr_to_utf8(tmpbuf, rev);
11132 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
11133 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
11135 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
11141 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11145 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
11152 Perl_keyword_plugin_standard(pTHX_
11153 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
11155 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
11156 PERL_UNUSED_CONTEXT;
11157 PERL_UNUSED_ARG(keyword_ptr);
11158 PERL_UNUSED_ARG(keyword_len);
11159 PERL_UNUSED_ARG(op_ptr);
11160 return KEYWORD_PLUGIN_DECLINE;
11163 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
11165 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
11167 SAVEI32(PL_lex_brackets);
11168 if (PL_lex_brackets > 100)
11169 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
11170 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
11171 SAVEI32(PL_lex_allbrackets);
11172 PL_lex_allbrackets = 0;
11173 SAVEI8(PL_lex_fakeeof);
11174 PL_lex_fakeeof = (U8)fakeeof;
11175 if(yyparse(gramtype) && !PL_parser->error_count)
11176 qerror(Perl_mess(aTHX_ "Parse error"));
11179 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11181 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11185 SAVEVPTR(PL_eval_root);
11186 PL_eval_root = NULL;
11187 parse_recdescent(gramtype, fakeeof);
11193 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11195 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11198 if (flags & ~PARSE_OPTIONAL)
11199 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11200 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11201 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11202 if (!PL_parser->error_count)
11203 qerror(Perl_mess(aTHX_ "Parse error"));
11204 exprop = newOP(OP_NULL, 0);
11210 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11212 Parse a Perl arithmetic expression. This may contain operators of precedence
11213 down to the bit shift operators. The expression must be followed (and thus
11214 terminated) either by a comparison or lower-precedence operator or by
11215 something that would normally terminate an expression such as semicolon.
11216 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11217 otherwise it is mandatory. It is up to the caller to ensure that the
11218 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11219 the source of the code to be parsed and the lexical context for the
11222 The op tree representing the expression is returned. If an optional
11223 expression is absent, a null pointer is returned, otherwise the pointer
11226 If an error occurs in parsing or compilation, in most cases a valid op
11227 tree is returned anyway. The error is reflected in the parser state,
11228 normally resulting in a single exception at the top level of parsing
11229 which covers all the compilation errors that occurred. Some compilation
11230 errors, however, will throw an exception immediately.
11236 Perl_parse_arithexpr(pTHX_ U32 flags)
11238 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11242 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11244 Parse a Perl term expression. This may contain operators of precedence
11245 down to the assignment operators. The expression must be followed (and thus
11246 terminated) either by a comma or lower-precedence operator or by
11247 something that would normally terminate an expression such as semicolon.
11248 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11249 otherwise it is mandatory. It is up to the caller to ensure that the
11250 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11251 the source of the code to be parsed and the lexical context for the
11254 The op tree representing the expression is returned. If an optional
11255 expression is absent, a null pointer is returned, otherwise the pointer
11258 If an error occurs in parsing or compilation, in most cases a valid op
11259 tree is returned anyway. The error is reflected in the parser state,
11260 normally resulting in a single exception at the top level of parsing
11261 which covers all the compilation errors that occurred. Some compilation
11262 errors, however, will throw an exception immediately.
11268 Perl_parse_termexpr(pTHX_ U32 flags)
11270 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11274 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11276 Parse a Perl list expression. This may contain operators of precedence
11277 down to the comma operator. The expression must be followed (and thus
11278 terminated) either by a low-precedence logic operator such as C<or> or by
11279 something that would normally terminate an expression such as semicolon.
11280 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11281 otherwise it is mandatory. It is up to the caller to ensure that the
11282 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11283 the source of the code to be parsed and the lexical context for the
11286 The op tree representing the expression is returned. If an optional
11287 expression is absent, a null pointer is returned, otherwise the pointer
11290 If an error occurs in parsing or compilation, in most cases a valid op
11291 tree is returned anyway. The error is reflected in the parser state,
11292 normally resulting in a single exception at the top level of parsing
11293 which covers all the compilation errors that occurred. Some compilation
11294 errors, however, will throw an exception immediately.
11300 Perl_parse_listexpr(pTHX_ U32 flags)
11302 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11306 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11308 Parse a single complete Perl expression. This allows the full
11309 expression grammar, including the lowest-precedence operators such
11310 as C<or>. The expression must be followed (and thus terminated) by a
11311 token that an expression would normally be terminated by: end-of-file,
11312 closing bracketing punctuation, semicolon, or one of the keywords that
11313 signals a postfix expression-statement modifier. If I<flags> includes
11314 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11315 mandatory. It is up to the caller to ensure that the dynamic parser
11316 state (L</PL_parser> et al) is correctly set to reflect the source of
11317 the code to be parsed and the lexical context for the expression.
11319 The op tree representing the expression is returned. If an optional
11320 expression is absent, a null pointer is returned, otherwise the pointer
11323 If an error occurs in parsing or compilation, in most cases a valid op
11324 tree is returned anyway. The error is reflected in the parser state,
11325 normally resulting in a single exception at the top level of parsing
11326 which covers all the compilation errors that occurred. Some compilation
11327 errors, however, will throw an exception immediately.
11333 Perl_parse_fullexpr(pTHX_ U32 flags)
11335 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11339 =for apidoc Amx|OP *|parse_block|U32 flags
11341 Parse a single complete Perl code block. This consists of an opening
11342 brace, a sequence of statements, and a closing brace. The block
11343 constitutes a lexical scope, so C<my> variables and various compile-time
11344 effects can be contained within it. It is up to the caller to ensure
11345 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11346 reflect the source of the code to be parsed and the lexical context for
11349 The op tree representing the code block is returned. This is always a
11350 real op, never a null pointer. It will normally be a C<lineseq> list,
11351 including C<nextstate> or equivalent ops. No ops to construct any kind
11352 of runtime scope are included by virtue of it being a block.
11354 If an error occurs in parsing or compilation, in most cases a valid op
11355 tree (most likely null) is returned anyway. The error is reflected in
11356 the parser state, normally resulting in a single exception at the top
11357 level of parsing which covers all the compilation errors that occurred.
11358 Some compilation errors, however, will throw an exception immediately.
11360 The I<flags> parameter is reserved for future use, and must always
11367 Perl_parse_block(pTHX_ U32 flags)
11370 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11371 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11375 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11377 Parse a single unadorned Perl statement. This may be a normal imperative
11378 statement or a declaration that has compile-time effect. It does not
11379 include any label or other affixture. It is up to the caller to ensure
11380 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11381 reflect the source of the code to be parsed and the lexical context for
11384 The op tree representing the statement is returned. This may be a
11385 null pointer if the statement is null, for example if it was actually
11386 a subroutine definition (which has compile-time side effects). If not
11387 null, it will be ops directly implementing the statement, suitable to
11388 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11389 equivalent op (except for those embedded in a scope contained entirely
11390 within the statement).
11392 If an error occurs in parsing or compilation, in most cases a valid op
11393 tree (most likely null) is returned anyway. The error is reflected in
11394 the parser state, normally resulting in a single exception at the top
11395 level of parsing which covers all the compilation errors that occurred.
11396 Some compilation errors, however, will throw an exception immediately.
11398 The I<flags> parameter is reserved for future use, and must always
11405 Perl_parse_barestmt(pTHX_ U32 flags)
11408 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11409 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11413 =for apidoc Amx|SV *|parse_label|U32 flags
11415 Parse a single label, possibly optional, of the type that may prefix a
11416 Perl statement. It is up to the caller to ensure that the dynamic parser
11417 state (L</PL_parser> et al) is correctly set to reflect the source of
11418 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11419 label is optional, otherwise it is mandatory.
11421 The name of the label is returned in the form of a fresh scalar. If an
11422 optional label is absent, a null pointer is returned.
11424 If an error occurs in parsing, which can only occur if the label is
11425 mandatory, a valid label is returned anyway. The error is reflected in
11426 the parser state, normally resulting in a single exception at the top
11427 level of parsing which covers all the compilation errors that occurred.
11433 Perl_parse_label(pTHX_ U32 flags)
11435 if (flags & ~PARSE_OPTIONAL)
11436 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11437 if (PL_lex_state == LEX_KNOWNEXT) {
11438 PL_parser->yychar = yylex();
11439 if (PL_parser->yychar == LABEL) {
11440 char *lpv = pl_yylval.pval;
11441 STRLEN llen = strlen(lpv);
11443 PL_parser->yychar = YYEMPTY;
11444 lsv = newSV_type(SVt_PV);
11445 SvPV_set(lsv, lpv);
11446 SvCUR_set(lsv, llen);
11447 SvLEN_set(lsv, llen+1);
11457 STRLEN wlen, bufptr_pos;
11461 if (!isIDFIRST_A(c))
11465 } while(isWORDCHAR_A(c));
11467 if (word_takes_any_delimeter(s, wlen))
11469 bufptr_pos = s - SvPVX(PL_linestr);
11471 lex_read_space(LEX_KEEP_PREVIOUS);
11473 s = SvPVX(PL_linestr) + bufptr_pos;
11474 if (t[0] == ':' && t[1] != ':') {
11475 PL_oldoldbufptr = PL_oldbufptr;
11478 return newSVpvn(s, wlen);
11482 if (flags & PARSE_OPTIONAL) {
11485 qerror(Perl_mess(aTHX_ "Parse error"));
11486 return newSVpvs("x");
11493 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11495 Parse a single complete Perl statement. This may be a normal imperative
11496 statement or a declaration that has compile-time effect, and may include
11497 optional labels. It is up to the caller to ensure that the dynamic
11498 parser state (L</PL_parser> et al) is correctly set to reflect the source
11499 of the code to be parsed and the lexical context for the statement.
11501 The op tree representing the statement is returned. This may be a
11502 null pointer if the statement is null, for example if it was actually
11503 a subroutine definition (which has compile-time side effects). If not
11504 null, it will be the result of a L</newSTATEOP> call, normally including
11505 a C<nextstate> or equivalent op.
11507 If an error occurs in parsing or compilation, in most cases a valid op
11508 tree (most likely null) is returned anyway. The error is reflected in
11509 the parser state, normally resulting in a single exception at the top
11510 level of parsing which covers all the compilation errors that occurred.
11511 Some compilation errors, however, will throw an exception immediately.
11513 The I<flags> parameter is reserved for future use, and must always
11520 Perl_parse_fullstmt(pTHX_ U32 flags)
11523 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11524 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11528 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11530 Parse a sequence of zero or more Perl statements. These may be normal
11531 imperative statements, including optional labels, or declarations
11532 that have compile-time effect, or any mixture thereof. The statement
11533 sequence ends when a closing brace or end-of-file is encountered in a
11534 place where a new statement could have validly started. It is up to
11535 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11536 is correctly set to reflect the source of the code to be parsed and the
11537 lexical context for the statements.
11539 The op tree representing the statement sequence is returned. This may
11540 be a null pointer if the statements were all null, for example if there
11541 were no statements or if there were only subroutine definitions (which
11542 have compile-time side effects). If not null, it will be a C<lineseq>
11543 list, normally including C<nextstate> or equivalent ops.
11545 If an error occurs in parsing or compilation, in most cases a valid op
11546 tree is returned anyway. The error is reflected in the parser state,
11547 normally resulting in a single exception at the top level of parsing
11548 which covers all the compilation errors that occurred. Some compilation
11549 errors, however, will throw an exception immediately.
11551 The I<flags> parameter is reserved for future use, and must always
11558 Perl_parse_stmtseq(pTHX_ U32 flags)
11563 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
11564 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
11565 c = lex_peek_unichar(0);
11566 if (c != -1 && c != /*{*/'}')
11567 qerror(Perl_mess(aTHX_ "Parse error"));
11572 Perl_munge_qwlist_to_paren_list(pTHX_ OP *qwlist)
11574 PERL_ARGS_ASSERT_MUNGE_QWLIST_TO_PAREN_LIST;
11575 deprecate("qw(...) as parentheses");
11576 force_next((4<<24)|')');
11577 if (qwlist->op_type == OP_STUB) {
11581 start_force(PL_curforce);
11582 NEXTVAL_NEXTTOKE.opval = qwlist;
11585 force_next((2<<24)|'(');
11590 * c-indentation-style: bsd
11591 * c-basic-offset: 4
11592 * indent-tabs-mode: t
11595 * ex: set ts=8 sts=4 sw=4 noet: