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);
617 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
618 REFCOUNTED_HE_EXISTS));
622 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
623 * utf16-to-utf8-reversed.
626 #ifdef PERL_CR_FILTER
630 register const char *s = SvPVX_const(sv);
631 register const char * const e = s + SvCUR(sv);
633 PERL_ARGS_ASSERT_STRIP_RETURN;
635 /* outer loop optimized to do nothing if there are no CR-LFs */
637 if (*s++ == '\r' && *s == '\n') {
638 /* hit a CR-LF, need to copy the rest */
639 register char *d = s - 1;
642 if (*s == '\r' && s[1] == '\n')
653 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
655 const I32 count = FILTER_READ(idx+1, sv, maxlen);
656 if (count > 0 && !maxlen)
663 =for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
665 Creates and initialises a new lexer/parser state object, supplying
666 a context in which to lex and parse from a new source of Perl code.
667 A pointer to the new state object is placed in L</PL_parser>. An entry
668 is made on the save stack so that upon unwinding the new state object
669 will be destroyed and the former value of L</PL_parser> will be restored.
670 Nothing else need be done to clean up the parsing context.
672 The code to be parsed comes from I<line> and I<rsfp>. I<line>, if
673 non-null, provides a string (in SV form) containing code to be parsed.
674 A copy of the string is made, so subsequent modification of I<line>
675 does not affect parsing. I<rsfp>, if non-null, provides an input stream
676 from which code will be read to be parsed. If both are non-null, the
677 code in I<line> comes first and must consist of complete lines of input,
678 and I<rsfp> supplies the remainder of the source.
680 The I<flags> parameter is reserved for future use. Currently it is only
681 used by perl internally, so extensions should always pass zero.
686 /* LEX_START_SAME_FILTER indicates that this is not a new file, so it
687 can share filters with the current parser. */
690 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
693 const char *s = NULL;
694 yy_parser *parser, *oparser;
695 if (flags && flags & ~LEX_START_FLAGS)
696 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
698 /* create and initialise a parser */
700 Newxz(parser, 1, yy_parser);
701 parser->old_parser = oparser = PL_parser;
704 parser->stack = NULL;
706 parser->stack_size = 0;
708 /* on scope exit, free this parser and restore any outer one */
710 parser->saved_curcop = PL_curcop;
712 /* initialise lexer state */
715 parser->curforce = -1;
717 parser->nexttoke = 0;
719 parser->error_count = oparser ? oparser->error_count : 0;
720 parser->copline = NOLINE;
721 parser->lex_state = LEX_NORMAL;
722 parser->expect = XSTATE;
724 parser->rsfp_filters =
725 !(flags & LEX_START_SAME_FILTER) || !oparser
727 : MUTABLE_AV(SvREFCNT_inc(
728 oparser->rsfp_filters
729 ? oparser->rsfp_filters
730 : (oparser->rsfp_filters = newAV())
733 Newx(parser->lex_brackstack, 120, char);
734 Newx(parser->lex_casestack, 12, char);
735 *parser->lex_casestack = '\0';
739 s = SvPV_const(line, len);
740 parser->linestr = flags & LEX_START_COPIED
741 ? SvREFCNT_inc_simple_NN(line)
742 : newSVpvn_flags(s, len, SvUTF8(line));
743 if (!len || s[len-1] != ';')
744 sv_catpvs(parser->linestr, "\n;");
746 parser->linestr = newSVpvs("\n;");
748 parser->oldoldbufptr =
751 parser->linestart = SvPVX(parser->linestr);
752 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
753 parser->last_lop = parser->last_uni = NULL;
754 parser->lex_flags = flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES);
756 parser->in_pod = parser->filtered = 0;
760 /* delete a parser object */
763 Perl_parser_free(pTHX_ const yy_parser *parser)
765 PERL_ARGS_ASSERT_PARSER_FREE;
767 PL_curcop = parser->saved_curcop;
768 SvREFCNT_dec(parser->linestr);
770 if (parser->rsfp == PerlIO_stdin())
771 PerlIO_clearerr(parser->rsfp);
772 else if (parser->rsfp && (!parser->old_parser ||
773 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
774 PerlIO_close(parser->rsfp);
775 SvREFCNT_dec(parser->rsfp_filters);
777 Safefree(parser->lex_brackstack);
778 Safefree(parser->lex_casestack);
779 PL_parser = parser->old_parser;
785 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
787 Buffer scalar containing the chunk currently under consideration of the
788 text currently being lexed. This is always a plain string scalar (for
789 which C<SvPOK> is true). It is not intended to be used as a scalar by
790 normal scalar means; instead refer to the buffer directly by the pointer
791 variables described below.
793 The lexer maintains various C<char*> pointers to things in the
794 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
795 reallocated, all of these pointers must be updated. Don't attempt to
796 do this manually, but rather use L</lex_grow_linestr> if you need to
797 reallocate the buffer.
799 The content of the text chunk in the buffer is commonly exactly one
800 complete line of input, up to and including a newline terminator,
801 but there are situations where it is otherwise. The octets of the
802 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
803 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
804 flag on this scalar, which may disagree with it.
806 For direct examination of the buffer, the variable
807 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
808 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
809 of these pointers is usually preferable to examination of the scalar
810 through normal scalar means.
812 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
814 Direct pointer to the end of the chunk of text currently being lexed, the
815 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
816 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
817 always located at the end of the buffer, and does not count as part of
818 the buffer's contents.
820 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
822 Points to the current position of lexing inside the lexer buffer.
823 Characters around this point may be freely examined, within
824 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
825 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
826 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
828 Lexing code (whether in the Perl core or not) moves this pointer past
829 the characters that it consumes. It is also expected to perform some
830 bookkeeping whenever a newline character is consumed. This movement
831 can be more conveniently performed by the function L</lex_read_to>,
832 which handles newlines appropriately.
834 Interpretation of the buffer's octets can be abstracted out by
835 using the slightly higher-level functions L</lex_peek_unichar> and
836 L</lex_read_unichar>.
838 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
840 Points to the start of the current line inside the lexer buffer.
841 This is useful for indicating at which column an error occurred, and
842 not much else. This must be updated by any lexing code that consumes
843 a newline; the function L</lex_read_to> handles this detail.
849 =for apidoc Amx|bool|lex_bufutf8
851 Indicates whether the octets in the lexer buffer
852 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
853 of Unicode characters. If not, they should be interpreted as Latin-1
854 characters. This is analogous to the C<SvUTF8> flag for scalars.
856 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
857 contains valid UTF-8. Lexing code must be robust in the face of invalid
860 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
861 is significant, but not the whole story regarding the input character
862 encoding. Normally, when a file is being read, the scalar contains octets
863 and its C<SvUTF8> flag is off, but the octets should be interpreted as
864 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
865 however, the scalar may have the C<SvUTF8> flag on, and in this case its
866 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
867 is in effect. This logic may change in the future; use this function
868 instead of implementing the logic yourself.
874 Perl_lex_bufutf8(pTHX)
880 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
882 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
883 at least I<len> octets (including terminating NUL). Returns a
884 pointer to the reallocated buffer. This is necessary before making
885 any direct modification of the buffer that would increase its length.
886 L</lex_stuff_pvn> provides a more convenient way to insert text into
889 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
890 this function updates all of the lexer's variables that point directly
897 Perl_lex_grow_linestr(pTHX_ STRLEN len)
901 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
902 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
903 linestr = PL_parser->linestr;
904 buf = SvPVX(linestr);
905 if (len <= SvLEN(linestr))
907 bufend_pos = PL_parser->bufend - buf;
908 bufptr_pos = PL_parser->bufptr - buf;
909 oldbufptr_pos = PL_parser->oldbufptr - buf;
910 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
911 linestart_pos = PL_parser->linestart - buf;
912 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
913 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
914 buf = sv_grow(linestr, len);
915 PL_parser->bufend = buf + bufend_pos;
916 PL_parser->bufptr = buf + bufptr_pos;
917 PL_parser->oldbufptr = buf + oldbufptr_pos;
918 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
919 PL_parser->linestart = buf + linestart_pos;
920 if (PL_parser->last_uni)
921 PL_parser->last_uni = buf + last_uni_pos;
922 if (PL_parser->last_lop)
923 PL_parser->last_lop = buf + last_lop_pos;
928 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
930 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
931 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
932 reallocating the buffer if necessary. This means that lexing code that
933 runs later will see the characters as if they had appeared in the input.
934 It is not recommended to do this as part of normal parsing, and most
935 uses of this facility run the risk of the inserted characters being
936 interpreted in an unintended manner.
938 The string to be inserted is represented by I<len> octets starting
939 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
940 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
941 The characters are recoded for the lexer buffer, according to how the
942 buffer is currently being interpreted (L</lex_bufutf8>). If a string
943 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
944 function is more convenient.
950 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
954 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
955 if (flags & ~(LEX_STUFF_UTF8))
956 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
958 if (flags & LEX_STUFF_UTF8) {
962 const char *p, *e = pv+len;
963 for (p = pv; p != e; p++)
964 highhalf += !!(((U8)*p) & 0x80);
967 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
968 bufptr = PL_parser->bufptr;
969 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
970 SvCUR_set(PL_parser->linestr,
971 SvCUR(PL_parser->linestr) + len+highhalf);
972 PL_parser->bufend += len+highhalf;
973 for (p = pv; p != e; p++) {
976 *bufptr++ = (char)(0xc0 | (c >> 6));
977 *bufptr++ = (char)(0x80 | (c & 0x3f));
984 if (flags & LEX_STUFF_UTF8) {
986 const char *p, *e = pv+len;
987 for (p = pv; p != e; p++) {
990 Perl_croak(aTHX_ "Lexing code attempted to stuff "
991 "non-Latin-1 character into Latin-1 input");
992 } else if (c >= 0xc2 && p+1 != e &&
993 (((U8)p[1]) & 0xc0) == 0x80) {
996 } else if (c >= 0x80) {
997 /* malformed UTF-8 */
999 SAVESPTR(PL_warnhook);
1000 PL_warnhook = PERL_WARNHOOK_FATAL;
1001 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
1007 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1008 bufptr = PL_parser->bufptr;
1009 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1010 SvCUR_set(PL_parser->linestr,
1011 SvCUR(PL_parser->linestr) + len-highhalf);
1012 PL_parser->bufend += len-highhalf;
1013 for (p = pv; p != e; p++) {
1016 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1019 *bufptr++ = (char)c;
1024 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1025 bufptr = PL_parser->bufptr;
1026 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1027 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1028 PL_parser->bufend += len;
1029 Copy(pv, bufptr, len, char);
1035 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1037 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1038 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1039 reallocating the buffer if necessary. This means that lexing code that
1040 runs later will see the characters as if they had appeared in the input.
1041 It is not recommended to do this as part of normal parsing, and most
1042 uses of this facility run the risk of the inserted characters being
1043 interpreted in an unintended manner.
1045 The string to be inserted is represented by octets starting at I<pv>
1046 and continuing to the first nul. These octets are interpreted as either
1047 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1048 in I<flags>. The characters are recoded for the lexer buffer, according
1049 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1050 If it is not convenient to nul-terminate a string to be inserted, the
1051 L</lex_stuff_pvn> function is more appropriate.
1057 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1059 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1060 lex_stuff_pvn(pv, strlen(pv), flags);
1064 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1066 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1067 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1068 reallocating the buffer if necessary. This means that lexing code that
1069 runs later will see the characters as if they had appeared in the input.
1070 It is not recommended to do this as part of normal parsing, and most
1071 uses of this facility run the risk of the inserted characters being
1072 interpreted in an unintended manner.
1074 The string to be inserted is the string value of I<sv>. The characters
1075 are recoded for the lexer buffer, according to how the buffer is currently
1076 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1077 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1078 need to construct a scalar.
1084 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1088 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1090 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1092 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1096 =for apidoc Amx|void|lex_unstuff|char *ptr
1098 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1099 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1100 This hides the discarded text from any lexing code that runs later,
1101 as if the text had never appeared.
1103 This is not the normal way to consume lexed text. For that, use
1110 Perl_lex_unstuff(pTHX_ char *ptr)
1114 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1115 buf = PL_parser->bufptr;
1117 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1120 bufend = PL_parser->bufend;
1122 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1123 unstuff_len = ptr - buf;
1124 Move(ptr, buf, bufend+1-ptr, char);
1125 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1126 PL_parser->bufend = bufend - unstuff_len;
1130 =for apidoc Amx|void|lex_read_to|char *ptr
1132 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1133 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1134 performing the correct bookkeeping whenever a newline character is passed.
1135 This is the normal way to consume lexed text.
1137 Interpretation of the buffer's octets can be abstracted out by
1138 using the slightly higher-level functions L</lex_peek_unichar> and
1139 L</lex_read_unichar>.
1145 Perl_lex_read_to(pTHX_ char *ptr)
1148 PERL_ARGS_ASSERT_LEX_READ_TO;
1149 s = PL_parser->bufptr;
1150 if (ptr < s || ptr > PL_parser->bufend)
1151 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1152 for (; s != ptr; s++)
1154 CopLINE_inc(PL_curcop);
1155 PL_parser->linestart = s+1;
1157 PL_parser->bufptr = ptr;
1161 =for apidoc Amx|void|lex_discard_to|char *ptr
1163 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1164 up to I<ptr>. The remaining content of the buffer will be moved, and
1165 all pointers into the buffer updated appropriately. I<ptr> must not
1166 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1167 it is not permitted to discard text that has yet to be lexed.
1169 Normally it is not necessarily to do this directly, because it suffices to
1170 use the implicit discarding behaviour of L</lex_next_chunk> and things
1171 based on it. However, if a token stretches across multiple lines,
1172 and the lexing code has kept multiple lines of text in the buffer for
1173 that purpose, then after completion of the token it would be wise to
1174 explicitly discard the now-unneeded earlier lines, to avoid future
1175 multi-line tokens growing the buffer without bound.
1181 Perl_lex_discard_to(pTHX_ char *ptr)
1185 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1186 buf = SvPVX(PL_parser->linestr);
1188 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1191 if (ptr > PL_parser->bufptr)
1192 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1193 discard_len = ptr - buf;
1194 if (PL_parser->oldbufptr < ptr)
1195 PL_parser->oldbufptr = ptr;
1196 if (PL_parser->oldoldbufptr < ptr)
1197 PL_parser->oldoldbufptr = ptr;
1198 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1199 PL_parser->last_uni = NULL;
1200 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1201 PL_parser->last_lop = NULL;
1202 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1203 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1204 PL_parser->bufend -= discard_len;
1205 PL_parser->bufptr -= discard_len;
1206 PL_parser->oldbufptr -= discard_len;
1207 PL_parser->oldoldbufptr -= discard_len;
1208 if (PL_parser->last_uni)
1209 PL_parser->last_uni -= discard_len;
1210 if (PL_parser->last_lop)
1211 PL_parser->last_lop -= discard_len;
1215 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1217 Reads in the next chunk of text to be lexed, appending it to
1218 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1219 looked to the end of the current chunk and wants to know more. It is
1220 usual, but not necessary, for lexing to have consumed the entirety of
1221 the current chunk at this time.
1223 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1224 chunk (i.e., the current chunk has been entirely consumed), normally the
1225 current chunk will be discarded at the same time that the new chunk is
1226 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1227 will not be discarded. If the current chunk has not been entirely
1228 consumed, then it will not be discarded regardless of the flag.
1230 Returns true if some new text was added to the buffer, or false if the
1231 buffer has reached the end of the input text.
1236 #define LEX_FAKE_EOF 0x80000000
1239 Perl_lex_next_chunk(pTHX_ U32 flags)
1243 STRLEN old_bufend_pos, new_bufend_pos;
1244 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1245 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1246 bool got_some_for_debugger = 0;
1248 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1249 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1250 linestr = PL_parser->linestr;
1251 buf = SvPVX(linestr);
1252 if (!(flags & LEX_KEEP_PREVIOUS) &&
1253 PL_parser->bufptr == PL_parser->bufend) {
1254 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1256 if (PL_parser->last_uni != PL_parser->bufend)
1257 PL_parser->last_uni = NULL;
1258 if (PL_parser->last_lop != PL_parser->bufend)
1259 PL_parser->last_lop = NULL;
1260 last_uni_pos = last_lop_pos = 0;
1264 old_bufend_pos = PL_parser->bufend - buf;
1265 bufptr_pos = PL_parser->bufptr - buf;
1266 oldbufptr_pos = PL_parser->oldbufptr - buf;
1267 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1268 linestart_pos = PL_parser->linestart - buf;
1269 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1270 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1272 if (flags & LEX_FAKE_EOF) {
1274 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
1276 } else if (filter_gets(linestr, old_bufend_pos)) {
1278 got_some_for_debugger = 1;
1280 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1281 sv_setpvs(linestr, "");
1283 /* End of real input. Close filehandle (unless it was STDIN),
1284 * then add implicit termination.
1286 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1287 PerlIO_clearerr(PL_parser->rsfp);
1288 else if (PL_parser->rsfp)
1289 (void)PerlIO_close(PL_parser->rsfp);
1290 PL_parser->rsfp = NULL;
1291 PL_parser->in_pod = PL_parser->filtered = 0;
1293 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1296 if (!PL_in_eval && PL_minus_p) {
1298 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1299 PL_minus_n = PL_minus_p = 0;
1300 } else if (!PL_in_eval && PL_minus_n) {
1301 sv_catpvs(linestr, /*{*/";}");
1304 sv_catpvs(linestr, ";");
1307 buf = SvPVX(linestr);
1308 new_bufend_pos = SvCUR(linestr);
1309 PL_parser->bufend = buf + new_bufend_pos;
1310 PL_parser->bufptr = buf + bufptr_pos;
1311 PL_parser->oldbufptr = buf + oldbufptr_pos;
1312 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1313 PL_parser->linestart = buf + linestart_pos;
1314 if (PL_parser->last_uni)
1315 PL_parser->last_uni = buf + last_uni_pos;
1316 if (PL_parser->last_lop)
1317 PL_parser->last_lop = buf + last_lop_pos;
1318 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1319 PL_curstash != PL_debstash) {
1320 /* debugger active and we're not compiling the debugger code,
1321 * so store the line into the debugger's array of lines
1323 update_debugger_info(NULL, buf+old_bufend_pos,
1324 new_bufend_pos-old_bufend_pos);
1330 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1332 Looks ahead one (Unicode) character in the text currently being lexed.
1333 Returns the codepoint (unsigned integer value) of the next character,
1334 or -1 if lexing has reached the end of the input text. To consume the
1335 peeked character, use L</lex_read_unichar>.
1337 If the next character is in (or extends into) the next chunk of input
1338 text, the next chunk will be read in. Normally the current chunk will be
1339 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1340 then the current chunk will not be discarded.
1342 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1343 is encountered, an exception is generated.
1349 Perl_lex_peek_unichar(pTHX_ U32 flags)
1353 if (flags & ~(LEX_KEEP_PREVIOUS))
1354 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1355 s = PL_parser->bufptr;
1356 bufend = PL_parser->bufend;
1362 if (!lex_next_chunk(flags))
1364 s = PL_parser->bufptr;
1365 bufend = PL_parser->bufend;
1371 len = PL_utf8skip[head];
1372 while ((STRLEN)(bufend-s) < len) {
1373 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1375 s = PL_parser->bufptr;
1376 bufend = PL_parser->bufend;
1379 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1380 if (retlen == (STRLEN)-1) {
1381 /* malformed UTF-8 */
1383 SAVESPTR(PL_warnhook);
1384 PL_warnhook = PERL_WARNHOOK_FATAL;
1385 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1391 if (!lex_next_chunk(flags))
1393 s = PL_parser->bufptr;
1400 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1402 Reads the next (Unicode) character in the text currently being lexed.
1403 Returns the codepoint (unsigned integer value) of the character read,
1404 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1405 if lexing has reached the end of the input text. To non-destructively
1406 examine the next character, use L</lex_peek_unichar> instead.
1408 If the next character is in (or extends into) the next chunk of input
1409 text, the next chunk will be read in. Normally the current chunk will be
1410 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1411 then the current chunk will not be discarded.
1413 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1414 is encountered, an exception is generated.
1420 Perl_lex_read_unichar(pTHX_ U32 flags)
1423 if (flags & ~(LEX_KEEP_PREVIOUS))
1424 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1425 c = lex_peek_unichar(flags);
1428 CopLINE_inc(PL_curcop);
1430 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1432 ++(PL_parser->bufptr);
1438 =for apidoc Amx|void|lex_read_space|U32 flags
1440 Reads optional spaces, in Perl style, in the text currently being
1441 lexed. The spaces may include ordinary whitespace characters and
1442 Perl-style comments. C<#line> directives are processed if encountered.
1443 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1444 at a non-space character (or the end of the input text).
1446 If spaces extend into the next chunk of input text, the next chunk will
1447 be read in. Normally the current chunk will be discarded at the same
1448 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1449 chunk will not be discarded.
1454 #define LEX_NO_NEXT_CHUNK 0x80000000
1457 Perl_lex_read_space(pTHX_ U32 flags)
1460 bool need_incline = 0;
1461 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1462 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1465 sv_free(PL_skipwhite);
1466 PL_skipwhite = NULL;
1469 PL_skipwhite = newSVpvs("");
1470 #endif /* PERL_MAD */
1471 s = PL_parser->bufptr;
1472 bufend = PL_parser->bufend;
1478 } while (!(c == '\n' || (c == 0 && s == bufend)));
1479 } else if (c == '\n') {
1481 PL_parser->linestart = s;
1486 } else if (isSPACE(c)) {
1488 } else if (c == 0 && s == bufend) {
1492 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1493 #endif /* PERL_MAD */
1494 if (flags & LEX_NO_NEXT_CHUNK)
1496 PL_parser->bufptr = s;
1497 CopLINE_inc(PL_curcop);
1498 got_more = lex_next_chunk(flags);
1499 CopLINE_dec(PL_curcop);
1500 s = PL_parser->bufptr;
1501 bufend = PL_parser->bufend;
1504 if (need_incline && PL_parser->rsfp) {
1514 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1515 #endif /* PERL_MAD */
1516 PL_parser->bufptr = s;
1521 * This subroutine has nothing to do with tilting, whether at windmills
1522 * or pinball tables. Its name is short for "increment line". It
1523 * increments the current line number in CopLINE(PL_curcop) and checks
1524 * to see whether the line starts with a comment of the form
1525 * # line 500 "foo.pm"
1526 * If so, it sets the current line number and file to the values in the comment.
1530 S_incline(pTHX_ const char *s)
1538 PERL_ARGS_ASSERT_INCLINE;
1540 CopLINE_inc(PL_curcop);
1543 while (SPACE_OR_TAB(*s))
1545 if (strnEQ(s, "line", 4))
1549 if (SPACE_OR_TAB(*s))
1553 while (SPACE_OR_TAB(*s))
1561 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1563 while (SPACE_OR_TAB(*s))
1565 if (*s == '"' && (t = strchr(s+1, '"'))) {
1571 while (!isSPACE(*t))
1575 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1577 if (*e != '\n' && *e != '\0')
1578 return; /* false alarm */
1580 line_num = atoi(n)-1;
1583 const STRLEN len = t - s;
1584 SV *const temp_sv = CopFILESV(PL_curcop);
1589 cf = SvPVX(temp_sv);
1590 tmplen = SvCUR(temp_sv);
1596 if (!PL_rsfp && !PL_parser->filtered) {
1597 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1598 * to *{"::_<newfilename"} */
1599 /* However, the long form of evals is only turned on by the
1600 debugger - usually they're "(eval %lu)" */
1604 STRLEN tmplen2 = len;
1605 if (tmplen + 2 <= sizeof smallbuf)
1608 Newx(tmpbuf, tmplen + 2, char);
1611 memcpy(tmpbuf + 2, cf, tmplen);
1613 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1618 if (tmplen2 + 2 <= sizeof smallbuf)
1621 Newx(tmpbuf2, tmplen2 + 2, char);
1623 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1624 /* Either they malloc'd it, or we malloc'd it,
1625 so no prefix is present in ours. */
1630 memcpy(tmpbuf2 + 2, s, tmplen2);
1633 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1635 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1636 /* adjust ${"::_<newfilename"} to store the new file name */
1637 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1638 /* The line number may differ. If that is the case,
1639 alias the saved lines that are in the array.
1640 Otherwise alias the whole array. */
1641 if (CopLINE(PL_curcop) == line_num) {
1642 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1643 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1645 else if (GvAV(*gvp)) {
1646 AV * const av = GvAV(*gvp);
1647 const I32 start = CopLINE(PL_curcop)+1;
1648 I32 items = AvFILLp(av) - start;
1650 AV * const av2 = GvAVn(gv2);
1651 SV **svp = AvARRAY(av) + start;
1652 I32 l = (I32)line_num+1;
1654 av_store(av2, l++, SvREFCNT_inc(*svp++));
1659 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1661 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1663 CopFILE_free(PL_curcop);
1664 CopFILE_setn(PL_curcop, s, len);
1666 CopLINE_set(PL_curcop, line_num);
1670 /* skip space before PL_thistoken */
1673 S_skipspace0(pTHX_ register char *s)
1675 PERL_ARGS_ASSERT_SKIPSPACE0;
1682 PL_thiswhite = newSVpvs("");
1683 sv_catsv(PL_thiswhite, PL_skipwhite);
1684 sv_free(PL_skipwhite);
1687 PL_realtokenstart = s - SvPVX(PL_linestr);
1691 /* skip space after PL_thistoken */
1694 S_skipspace1(pTHX_ register char *s)
1696 const char *start = s;
1697 I32 startoff = start - SvPVX(PL_linestr);
1699 PERL_ARGS_ASSERT_SKIPSPACE1;
1704 start = SvPVX(PL_linestr) + startoff;
1705 if (!PL_thistoken && PL_realtokenstart >= 0) {
1706 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1707 PL_thistoken = newSVpvn(tstart, start - tstart);
1709 PL_realtokenstart = -1;
1712 PL_nextwhite = newSVpvs("");
1713 sv_catsv(PL_nextwhite, PL_skipwhite);
1714 sv_free(PL_skipwhite);
1721 S_skipspace2(pTHX_ register char *s, SV **svp)
1724 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1725 const I32 startoff = s - SvPVX(PL_linestr);
1727 PERL_ARGS_ASSERT_SKIPSPACE2;
1730 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1731 if (!PL_madskills || !svp)
1733 start = SvPVX(PL_linestr) + startoff;
1734 if (!PL_thistoken && PL_realtokenstart >= 0) {
1735 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1736 PL_thistoken = newSVpvn(tstart, start - tstart);
1737 PL_realtokenstart = -1;
1741 *svp = newSVpvs("");
1742 sv_setsv(*svp, PL_skipwhite);
1743 sv_free(PL_skipwhite);
1752 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1754 AV *av = CopFILEAVx(PL_curcop);
1756 SV * const sv = newSV_type(SVt_PVMG);
1758 sv_setsv(sv, orig_sv);
1760 sv_setpvn(sv, buf, len);
1763 av_store(av, (I32)CopLINE(PL_curcop), sv);
1769 * Called to gobble the appropriate amount and type of whitespace.
1770 * Skips comments as well.
1774 S_skipspace(pTHX_ register char *s)
1778 #endif /* PERL_MAD */
1779 PERL_ARGS_ASSERT_SKIPSPACE;
1782 sv_free(PL_skipwhite);
1783 PL_skipwhite = NULL;
1785 #endif /* PERL_MAD */
1786 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1787 while (s < PL_bufend && SPACE_OR_TAB(*s))
1790 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1792 lex_read_space(LEX_KEEP_PREVIOUS |
1793 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1794 LEX_NO_NEXT_CHUNK : 0));
1796 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1797 if (PL_linestart > PL_bufptr)
1798 PL_bufptr = PL_linestart;
1803 PL_skipwhite = newSVpvn(start, s-start);
1804 #endif /* PERL_MAD */
1810 * Check the unary operators to ensure there's no ambiguity in how they're
1811 * used. An ambiguous piece of code would be:
1813 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1814 * the +5 is its argument.
1824 if (PL_oldoldbufptr != PL_last_uni)
1826 while (isSPACE(*PL_last_uni))
1829 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1831 if ((t = strchr(s, '(')) && t < PL_bufptr)
1834 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1835 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1836 (int)(s - PL_last_uni), PL_last_uni);
1840 * LOP : macro to build a list operator. Its behaviour has been replaced
1841 * with a subroutine, S_lop() for which LOP is just another name.
1844 #define LOP(f,x) return lop(f,x,s)
1848 * Build a list operator (or something that might be one). The rules:
1849 * - if we have a next token, then it's a list operator [why?]
1850 * - if the next thing is an opening paren, then it's a function
1851 * - else it's a list operator
1855 S_lop(pTHX_ I32 f, int x, char *s)
1859 PERL_ARGS_ASSERT_LOP;
1865 PL_last_lop = PL_oldbufptr;
1866 PL_last_lop_op = (OPCODE)f;
1875 return REPORT(FUNC);
1878 return REPORT(FUNC);
1881 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1882 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1883 return REPORT(LSTOP);
1890 * Sets up for an eventual force_next(). start_force(0) basically does
1891 * an unshift, while start_force(-1) does a push. yylex removes items
1896 S_start_force(pTHX_ int where)
1900 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1901 where = PL_lasttoke;
1902 assert(PL_curforce < 0 || PL_curforce == where);
1903 if (PL_curforce != where) {
1904 for (i = PL_lasttoke; i > where; --i) {
1905 PL_nexttoke[i] = PL_nexttoke[i-1];
1909 if (PL_curforce < 0) /* in case of duplicate start_force() */
1910 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1911 PL_curforce = where;
1914 curmad('^', newSVpvs(""));
1915 CURMAD('_', PL_nextwhite);
1920 S_curmad(pTHX_ char slot, SV *sv)
1926 if (PL_curforce < 0)
1927 where = &PL_thismad;
1929 where = &PL_nexttoke[PL_curforce].next_mad;
1935 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1937 else if (PL_encoding) {
1938 sv_recode_to_utf8(sv, PL_encoding);
1943 /* keep a slot open for the head of the list? */
1944 if (slot != '_' && *where && (*where)->mad_key == '^') {
1945 (*where)->mad_key = slot;
1946 sv_free(MUTABLE_SV(((*where)->mad_val)));
1947 (*where)->mad_val = (void*)sv;
1950 addmad(newMADsv(slot, sv), where, 0);
1953 # define start_force(where) NOOP
1954 # define curmad(slot, sv) NOOP
1959 * When the lexer realizes it knows the next token (for instance,
1960 * it is reordering tokens for the parser) then it can call S_force_next
1961 * to know what token to return the next time the lexer is called. Caller
1962 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1963 * and possibly PL_expect to ensure the lexer handles the token correctly.
1967 S_force_next(pTHX_ I32 type)
1972 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1973 tokereport(type, &NEXTVAL_NEXTTOKE);
1977 if (PL_curforce < 0)
1978 start_force(PL_lasttoke);
1979 PL_nexttoke[PL_curforce].next_type = type;
1980 if (PL_lex_state != LEX_KNOWNEXT)
1981 PL_lex_defer = PL_lex_state;
1982 PL_lex_state = LEX_KNOWNEXT;
1983 PL_lex_expect = PL_expect;
1986 PL_nexttype[PL_nexttoke] = type;
1988 if (PL_lex_state != LEX_KNOWNEXT) {
1989 PL_lex_defer = PL_lex_state;
1990 PL_lex_expect = PL_expect;
1991 PL_lex_state = LEX_KNOWNEXT;
1999 int yyc = PL_parser->yychar;
2000 if (yyc != YYEMPTY) {
2003 NEXTVAL_NEXTTOKE = PL_parser->yylval;
2004 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
2005 PL_lex_allbrackets--;
2007 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
2008 } else if (yyc == '('/*)*/) {
2009 PL_lex_allbrackets--;
2014 PL_parser->yychar = YYEMPTY;
2019 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
2022 SV * const sv = newSVpvn_utf8(start, len,
2025 && !is_ascii_string((const U8*)start, len)
2026 && is_utf8_string((const U8*)start, len));
2032 * When the lexer knows the next thing is a word (for instance, it has
2033 * just seen -> and it knows that the next char is a word char, then
2034 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2038 * char *start : buffer position (must be within PL_linestr)
2039 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2040 * int check_keyword : if true, Perl checks to make sure the word isn't
2041 * a keyword (do this if the word is a label, e.g. goto FOO)
2042 * int allow_pack : if true, : characters will also be allowed (require,
2043 * use, etc. do this)
2044 * int allow_initial_tick : used by the "sub" lexer only.
2048 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
2054 PERL_ARGS_ASSERT_FORCE_WORD;
2056 start = SKIPSPACE1(start);
2058 if (isIDFIRST_lazy_if(s,UTF) ||
2059 (allow_pack && *s == ':') ||
2060 (allow_initial_tick && *s == '\'') )
2062 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2063 if (check_keyword && keyword(PL_tokenbuf, len, 0))
2065 start_force(PL_curforce);
2067 curmad('X', newSVpvn(start,s-start));
2068 if (token == METHOD) {
2073 PL_expect = XOPERATOR;
2077 curmad('g', newSVpvs( "forced" ));
2078 NEXTVAL_NEXTTOKE.opval
2079 = (OP*)newSVOP(OP_CONST,0,
2080 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2081 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2089 * Called when the lexer wants $foo *foo &foo etc, but the program
2090 * text only contains the "foo" portion. The first argument is a pointer
2091 * to the "foo", and the second argument is the type symbol to prefix.
2092 * Forces the next token to be a "WORD".
2093 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2097 S_force_ident(pTHX_ register const char *s, int kind)
2101 PERL_ARGS_ASSERT_FORCE_IDENT;
2104 const STRLEN len = strlen(s);
2105 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2106 UTF ? SVf_UTF8 : 0));
2107 start_force(PL_curforce);
2108 NEXTVAL_NEXTTOKE.opval = o;
2111 o->op_private = OPpCONST_ENTERED;
2112 /* XXX see note in pp_entereval() for why we forgo typo
2113 warnings if the symbol must be introduced in an eval.
2115 gv_fetchpvn_flags(s, len,
2116 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2117 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2118 kind == '$' ? SVt_PV :
2119 kind == '@' ? SVt_PVAV :
2120 kind == '%' ? SVt_PVHV :
2128 Perl_str_to_version(pTHX_ SV *sv)
2133 const char *start = SvPV_const(sv,len);
2134 const char * const end = start + len;
2135 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2137 PERL_ARGS_ASSERT_STR_TO_VERSION;
2139 while (start < end) {
2143 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2148 retval += ((NV)n)/nshift;
2157 * Forces the next token to be a version number.
2158 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2159 * and if "guessing" is TRUE, then no new token is created (and the caller
2160 * must use an alternative parsing method).
2164 S_force_version(pTHX_ char *s, int guessing)
2170 I32 startoff = s - SvPVX(PL_linestr);
2173 PERL_ARGS_ASSERT_FORCE_VERSION;
2181 while (isDIGIT(*d) || *d == '_' || *d == '.')
2185 start_force(PL_curforce);
2186 curmad('X', newSVpvn(s,d-s));
2189 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2191 #ifdef USE_LOCALE_NUMERIC
2192 char *loc = savepv(setlocale(LC_NUMERIC, NULL));
2193 setlocale(LC_NUMERIC, "C");
2195 s = scan_num(s, &pl_yylval);
2196 #ifdef USE_LOCALE_NUMERIC
2197 setlocale(LC_NUMERIC, loc);
2200 version = pl_yylval.opval;
2201 ver = cSVOPx(version)->op_sv;
2202 if (SvPOK(ver) && !SvNIOK(ver)) {
2203 SvUPGRADE(ver, SVt_PVNV);
2204 SvNV_set(ver, str_to_version(ver));
2205 SvNOK_on(ver); /* hint that it is a version */
2208 else if (guessing) {
2211 sv_free(PL_nextwhite); /* let next token collect whitespace */
2213 s = SvPVX(PL_linestr) + startoff;
2221 if (PL_madskills && !version) {
2222 sv_free(PL_nextwhite); /* let next token collect whitespace */
2224 s = SvPVX(PL_linestr) + startoff;
2227 /* NOTE: The parser sees the package name and the VERSION swapped */
2228 start_force(PL_curforce);
2229 NEXTVAL_NEXTTOKE.opval = version;
2236 * S_force_strict_version
2237 * Forces the next token to be a version number using strict syntax rules.
2241 S_force_strict_version(pTHX_ char *s)
2246 I32 startoff = s - SvPVX(PL_linestr);
2248 const char *errstr = NULL;
2250 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2252 while (isSPACE(*s)) /* leading whitespace */
2255 if (is_STRICT_VERSION(s,&errstr)) {
2257 s = (char *)scan_version(s, ver, 0);
2258 version = newSVOP(OP_CONST, 0, ver);
2260 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2261 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2265 yyerror(errstr); /* version required */
2270 if (PL_madskills && !version) {
2271 sv_free(PL_nextwhite); /* let next token collect whitespace */
2273 s = SvPVX(PL_linestr) + startoff;
2276 /* NOTE: The parser sees the package name and the VERSION swapped */
2277 start_force(PL_curforce);
2278 NEXTVAL_NEXTTOKE.opval = version;
2286 * Tokenize a quoted string passed in as an SV. It finds the next
2287 * chunk, up to end of string or a backslash. It may make a new
2288 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2293 S_tokeq(pTHX_ SV *sv)
2297 register char *send;
2302 PERL_ARGS_ASSERT_TOKEQ;
2307 s = SvPV_force(sv, len);
2308 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2311 /* This is relying on the SV being "well formed" with a trailing '\0' */
2312 while (s < send && !(*s == '\\' && s[1] == '\\'))
2317 if ( PL_hints & HINT_NEW_STRING ) {
2318 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2322 if (s + 1 < send && (s[1] == '\\'))
2323 s++; /* all that, just for this */
2328 SvCUR_set(sv, d - SvPVX_const(sv));
2330 if ( PL_hints & HINT_NEW_STRING )
2331 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2336 * Now come three functions related to double-quote context,
2337 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2338 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2339 * interact with PL_lex_state, and create fake ( ... ) argument lists
2340 * to handle functions and concatenation.
2341 * They assume that whoever calls them will be setting up a fake
2342 * join call, because each subthing puts a ',' after it. This lets
2345 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2347 * (I'm not sure whether the spurious commas at the end of lcfirst's
2348 * arguments and join's arguments are created or not).
2353 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2355 * Pattern matching will set PL_lex_op to the pattern-matching op to
2356 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2358 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2360 * Everything else becomes a FUNC.
2362 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2363 * had an OP_CONST or OP_READLINE). This just sets us up for a
2364 * call to S_sublex_push().
2368 S_sublex_start(pTHX)
2371 register const I32 op_type = pl_yylval.ival;
2373 if (op_type == OP_NULL) {
2374 pl_yylval.opval = PL_lex_op;
2378 if (op_type == OP_CONST || op_type == OP_READLINE) {
2379 SV *sv = tokeq(PL_lex_stuff);
2381 if (SvTYPE(sv) == SVt_PVIV) {
2382 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2384 const char * const p = SvPV_const(sv, len);
2385 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2389 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2390 PL_lex_stuff = NULL;
2391 /* Allow <FH> // "foo" */
2392 if (op_type == OP_READLINE)
2393 PL_expect = XTERMORDORDOR;
2396 else if (op_type == OP_BACKTICK && PL_lex_op) {
2397 /* readpipe() vas overriden */
2398 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2399 pl_yylval.opval = PL_lex_op;
2401 PL_lex_stuff = NULL;
2405 PL_sublex_info.super_state = PL_lex_state;
2406 PL_sublex_info.sub_inwhat = (U16)op_type;
2407 PL_sublex_info.sub_op = PL_lex_op;
2408 PL_lex_state = LEX_INTERPPUSH;
2412 pl_yylval.opval = PL_lex_op;
2422 * Create a new scope to save the lexing state. The scope will be
2423 * ended in S_sublex_done. Returns a '(', starting the function arguments
2424 * to the uc, lc, etc. found before.
2425 * Sets PL_lex_state to LEX_INTERPCONCAT.
2434 PL_lex_state = PL_sublex_info.super_state;
2435 SAVEBOOL(PL_lex_dojoin);
2436 SAVEI32(PL_lex_brackets);
2437 SAVEI32(PL_lex_allbrackets);
2438 SAVEI8(PL_lex_fakeeof);
2439 SAVEI32(PL_lex_casemods);
2440 SAVEI32(PL_lex_starts);
2441 SAVEI8(PL_lex_state);
2442 SAVEVPTR(PL_lex_inpat);
2443 SAVEI16(PL_lex_inwhat);
2444 SAVECOPLINE(PL_curcop);
2445 SAVEPPTR(PL_bufptr);
2446 SAVEPPTR(PL_bufend);
2447 SAVEPPTR(PL_oldbufptr);
2448 SAVEPPTR(PL_oldoldbufptr);
2449 SAVEPPTR(PL_last_lop);
2450 SAVEPPTR(PL_last_uni);
2451 SAVEPPTR(PL_linestart);
2452 SAVESPTR(PL_linestr);
2453 SAVEGENERICPV(PL_lex_brackstack);
2454 SAVEGENERICPV(PL_lex_casestack);
2456 PL_linestr = PL_lex_stuff;
2457 PL_lex_stuff = NULL;
2459 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2460 = SvPVX(PL_linestr);
2461 PL_bufend += SvCUR(PL_linestr);
2462 PL_last_lop = PL_last_uni = NULL;
2463 SAVEFREESV(PL_linestr);
2465 PL_lex_dojoin = FALSE;
2466 PL_lex_brackets = 0;
2467 PL_lex_allbrackets = 0;
2468 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2469 Newx(PL_lex_brackstack, 120, char);
2470 Newx(PL_lex_casestack, 12, char);
2471 PL_lex_casemods = 0;
2472 *PL_lex_casestack = '\0';
2474 PL_lex_state = LEX_INTERPCONCAT;
2475 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2477 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2478 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2479 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2480 PL_lex_inpat = PL_sublex_info.sub_op;
2482 PL_lex_inpat = NULL;
2489 * Restores lexer state after a S_sublex_push.
2496 if (!PL_lex_starts++) {
2497 SV * const sv = newSVpvs("");
2498 if (SvUTF8(PL_linestr))
2500 PL_expect = XOPERATOR;
2501 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2505 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2506 PL_lex_state = LEX_INTERPCASEMOD;
2510 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2511 assert(PL_lex_inwhat != OP_TRANSR);
2512 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2513 PL_linestr = PL_lex_repl;
2515 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2516 PL_bufend += SvCUR(PL_linestr);
2517 PL_last_lop = PL_last_uni = NULL;
2518 SAVEFREESV(PL_linestr);
2519 PL_lex_dojoin = FALSE;
2520 PL_lex_brackets = 0;
2521 PL_lex_allbrackets = 0;
2522 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2523 PL_lex_casemods = 0;
2524 *PL_lex_casestack = '\0';
2526 if (SvEVALED(PL_lex_repl)) {
2527 PL_lex_state = LEX_INTERPNORMAL;
2529 /* we don't clear PL_lex_repl here, so that we can check later
2530 whether this is an evalled subst; that means we rely on the
2531 logic to ensure sublex_done() is called again only via the
2532 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2535 PL_lex_state = LEX_INTERPCONCAT;
2545 PL_endwhite = newSVpvs("");
2546 sv_catsv(PL_endwhite, PL_thiswhite);
2550 sv_setpvs(PL_thistoken,"");
2552 PL_realtokenstart = -1;
2556 PL_bufend = SvPVX(PL_linestr);
2557 PL_bufend += SvCUR(PL_linestr);
2558 PL_expect = XOPERATOR;
2559 PL_sublex_info.sub_inwhat = 0;
2567 Extracts a pattern, double-quoted string, or transliteration. This
2570 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2571 processing a pattern (PL_lex_inpat is true), a transliteration
2572 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2574 Returns a pointer to the character scanned up to. If this is
2575 advanced from the start pointer supplied (i.e. if anything was
2576 successfully parsed), will leave an OP for the substring scanned
2577 in pl_yylval. Caller must intuit reason for not parsing further
2578 by looking at the next characters herself.
2582 constants: \N{NAME} only
2583 case and quoting: \U \Q \E
2584 stops on @ and $, but not for $ as tail anchor
2586 In transliterations:
2587 characters are VERY literal, except for - not at the start or end
2588 of the string, which indicates a range. If the range is in bytes,
2589 scan_const expands the range to the full set of intermediate
2590 characters. If the range is in utf8, the hyphen is replaced with
2591 a certain range mark which will be handled by pmtrans() in op.c.
2593 In double-quoted strings:
2595 double-quoted style: \r and \n
2596 constants: \x31, etc.
2597 deprecated backrefs: \1 (in substitution replacements)
2598 case and quoting: \U \Q \E
2601 scan_const does *not* construct ops to handle interpolated strings.
2602 It stops processing as soon as it finds an embedded $ or @ variable
2603 and leaves it to the caller to work out what's going on.
2605 embedded arrays (whether in pattern or not) could be:
2606 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2608 $ in double-quoted strings must be the symbol of an embedded scalar.
2610 $ in pattern could be $foo or could be tail anchor. Assumption:
2611 it's a tail anchor if $ is the last thing in the string, or if it's
2612 followed by one of "()| \r\n\t"
2614 \1 (backreferences) are turned into $1
2616 The structure of the code is
2617 while (there's a character to process) {
2618 handle transliteration ranges
2619 skip regexp comments /(?#comment)/ and codes /(?{code})/
2620 skip #-initiated comments in //x patterns
2621 check for embedded arrays
2622 check for embedded scalars
2624 deprecate \1 in substitution replacements
2625 handle string-changing backslashes \l \U \Q \E, etc.
2626 switch (what was escaped) {
2627 handle \- in a transliteration (becomes a literal -)
2628 if a pattern and not \N{, go treat as regular character
2629 handle \132 (octal characters)
2630 handle \x15 and \x{1234} (hex characters)
2631 handle \N{name} (named characters, also \N{3,5} in a pattern)
2632 handle \cV (control characters)
2633 handle printf-style backslashes (\f, \r, \n, etc)
2636 } (end if backslash)
2637 handle regular character
2638 } (end while character to read)
2643 S_scan_const(pTHX_ char *start)
2646 register char *send = PL_bufend; /* end of the constant */
2647 SV *sv = newSV(send - start); /* sv for the constant. See
2648 note below on sizing. */
2649 register char *s = start; /* start of the constant */
2650 register char *d = SvPVX(sv); /* destination for copies */
2651 bool dorange = FALSE; /* are we in a translit range? */
2652 bool didrange = FALSE; /* did we just finish a range? */
2653 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2654 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
2655 to be UTF8? But, this can
2656 show as true when the source
2657 isn't utf8, as for example
2658 when it is entirely composed
2661 /* Note on sizing: The scanned constant is placed into sv, which is
2662 * initialized by newSV() assuming one byte of output for every byte of
2663 * input. This routine expects newSV() to allocate an extra byte for a
2664 * trailing NUL, which this routine will append if it gets to the end of
2665 * the input. There may be more bytes of input than output (eg., \N{LATIN
2666 * CAPITAL LETTER A}), or more output than input if the constant ends up
2667 * recoded to utf8, but each time a construct is found that might increase
2668 * the needed size, SvGROW() is called. Its size parameter each time is
2669 * based on the best guess estimate at the time, namely the length used so
2670 * far, plus the length the current construct will occupy, plus room for
2671 * the trailing NUL, plus one byte for every input byte still unscanned */
2675 UV literal_endpoint = 0;
2676 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2679 PERL_ARGS_ASSERT_SCAN_CONST;
2681 assert(PL_lex_inwhat != OP_TRANSR);
2682 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2683 /* If we are doing a trans and we know we want UTF8 set expectation */
2684 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2685 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2689 while (s < send || dorange) {
2691 /* get transliterations out of the way (they're most literal) */
2692 if (PL_lex_inwhat == OP_TRANS) {
2693 /* expand a range A-Z to the full set of characters. AIE! */
2695 I32 i; /* current expanded character */
2696 I32 min; /* first character in range */
2697 I32 max; /* last character in range */
2708 char * const c = (char*)utf8_hop((U8*)d, -1);
2712 *c = (char)UTF_TO_NATIVE(0xff);
2713 /* mark the range as done, and continue */
2719 i = d - SvPVX_const(sv); /* remember current offset */
2722 SvLEN(sv) + (has_utf8 ?
2723 (512 - UTF_CONTINUATION_MARK +
2726 /* How many two-byte within 0..255: 128 in UTF-8,
2727 * 96 in UTF-8-mod. */
2729 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2731 d = SvPVX(sv) + i; /* refresh d after realloc */
2735 for (j = 0; j <= 1; j++) {
2736 char * const c = (char*)utf8_hop((U8*)d, -1);
2737 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2743 max = (U8)0xff; /* only to \xff */
2744 uvmax = uv; /* \x{100} to uvmax */
2746 d = c; /* eat endpoint chars */
2751 d -= 2; /* eat the first char and the - */
2752 min = (U8)*d; /* first char in range */
2753 max = (U8)d[1]; /* last char in range */
2760 "Invalid range \"%c-%c\" in transliteration operator",
2761 (char)min, (char)max);
2765 if (literal_endpoint == 2 &&
2766 ((isLOWER(min) && isLOWER(max)) ||
2767 (isUPPER(min) && isUPPER(max)))) {
2769 for (i = min; i <= max; i++)
2771 *d++ = NATIVE_TO_NEED(has_utf8,i);
2773 for (i = min; i <= max; i++)
2775 *d++ = NATIVE_TO_NEED(has_utf8,i);
2780 for (i = min; i <= max; i++)
2783 const U8 ch = (U8)NATIVE_TO_UTF(i);
2784 if (UNI_IS_INVARIANT(ch))
2787 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2788 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2797 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2799 *d++ = (char)UTF_TO_NATIVE(0xff);
2801 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2805 /* mark the range as done, and continue */
2809 literal_endpoint = 0;
2814 /* range begins (ignore - as first or last char) */
2815 else if (*s == '-' && s+1 < send && s != start) {
2817 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2824 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2834 literal_endpoint = 0;
2835 native_range = TRUE;
2840 /* if we get here, we're not doing a transliteration */
2842 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2843 except for the last char, which will be done separately. */
2844 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2846 while (s+1 < send && *s != ')')
2847 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2849 else if (s[2] == '{' /* This should match regcomp.c */
2850 || (s[2] == '?' && s[3] == '{'))
2853 char *regparse = s + (s[2] == '{' ? 3 : 4);
2856 while (count && (c = *regparse)) {
2857 if (c == '\\' && regparse[1])
2865 if (*regparse != ')')
2866 regparse--; /* Leave one char for continuation. */
2867 while (s < regparse)
2868 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2872 /* likewise skip #-initiated comments in //x patterns */
2873 else if (*s == '#' && PL_lex_inpat &&
2874 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
2875 while (s+1 < send && *s != '\n')
2876 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2879 /* check for embedded arrays
2880 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2882 else if (*s == '@' && s[1]) {
2883 if (isALNUM_lazy_if(s+1,UTF))
2885 if (strchr(":'{$", s[1]))
2887 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2888 break; /* in regexp, neither @+ nor @- are interpolated */
2891 /* check for embedded scalars. only stop if we're sure it's a
2894 else if (*s == '$') {
2895 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2897 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2899 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2900 "Possible unintended interpolation of $\\ in regex");
2902 break; /* in regexp, $ might be tail anchor */
2906 /* End of else if chain - OP_TRANS rejoin rest */
2909 if (*s == '\\' && s+1 < send) {
2910 char* e; /* Can be used for ending '}', etc. */
2914 /* warn on \1 - \9 in substitution replacements, but note that \11
2915 * is an octal; and \19 is \1 followed by '9' */
2916 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2917 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2919 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2924 /* string-change backslash escapes */
2925 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
2929 /* In a pattern, process \N, but skip any other backslash escapes.
2930 * This is because we don't want to translate an escape sequence
2931 * into a meta symbol and have the regex compiler use the meta
2932 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2933 * in spite of this, we do have to process \N here while the proper
2934 * charnames handler is in scope. See bugs #56444 and #62056.
2935 * There is a complication because \N in a pattern may also stand
2936 * for 'match a non-nl', and not mean a charname, in which case its
2937 * processing should be deferred to the regex compiler. To be a
2938 * charname it must be followed immediately by a '{', and not look
2939 * like \N followed by a curly quantifier, i.e., not something like
2940 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2942 else if (PL_lex_inpat
2945 || regcurly(s + 1)))
2947 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2948 goto default_action;
2953 /* quoted - in transliterations */
2955 if (PL_lex_inwhat == OP_TRANS) {
2962 if ((isALPHA(*s) || isDIGIT(*s)))
2963 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2964 "Unrecognized escape \\%c passed through",
2966 /* default action is to copy the quoted character */
2967 goto default_action;
2970 /* eg. \132 indicates the octal constant 0132 */
2971 case '0': case '1': case '2': case '3':
2972 case '4': case '5': case '6': case '7':
2976 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2979 goto NUM_ESCAPE_INSERT;
2981 /* eg. \o{24} indicates the octal constant \024 */
2987 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
2993 goto NUM_ESCAPE_INSERT;
2996 /* eg. \x24 indicates the hex constant 0x24 */
3000 char* const e = strchr(s, '}');
3001 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
3002 PERL_SCAN_DISALLOW_PREFIX;
3007 yyerror("Missing right brace on \\x{}");
3011 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
3017 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
3018 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
3024 /* Insert oct or hex escaped character. There will always be
3025 * enough room in sv since such escapes will be longer than any
3026 * UTF-8 sequence they can end up as, except if they force us
3027 * to recode the rest of the string into utf8 */
3029 /* Here uv is the ordinal of the next character being added in
3030 * unicode (converted from native). */
3031 if (!UNI_IS_INVARIANT(uv)) {
3032 if (!has_utf8 && uv > 255) {
3033 /* Might need to recode whatever we have accumulated so
3034 * far if it contains any chars variant in utf8 or
3037 SvCUR_set(sv, d - SvPVX_const(sv));
3040 /* See Note on sizing above. */
3041 sv_utf8_upgrade_flags_grow(sv,
3042 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3043 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3044 d = SvPVX(sv) + SvCUR(sv);
3049 d = (char*)uvuni_to_utf8((U8*)d, uv);
3050 if (PL_lex_inwhat == OP_TRANS &&
3051 PL_sublex_info.sub_op) {
3052 PL_sublex_info.sub_op->op_private |=
3053 (PL_lex_repl ? OPpTRANS_FROM_UTF
3057 if (uv > 255 && !dorange)
3058 native_range = FALSE;
3071 /* In a non-pattern \N must be a named character, like \N{LATIN
3072 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3073 * mean to match a non-newline. For non-patterns, named
3074 * characters are converted to their string equivalents. In
3075 * patterns, named characters are not converted to their
3076 * ultimate forms for the same reasons that other escapes
3077 * aren't. Instead, they are converted to the \N{U+...} form
3078 * to get the value from the charnames that is in effect right
3079 * now, while preserving the fact that it was a named character
3080 * so that the regex compiler knows this */
3082 /* This section of code doesn't generally use the
3083 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3084 * a close examination of this macro and determined it is a
3085 * no-op except on utfebcdic variant characters. Every
3086 * character generated by this that would normally need to be
3087 * enclosed by this macro is invariant, so the macro is not
3088 * needed, and would complicate use of copy(). XXX There are
3089 * other parts of this file where the macro is used
3090 * inconsistently, but are saved by it being a no-op */
3092 /* The structure of this section of code (besides checking for
3093 * errors and upgrading to utf8) is:
3094 * Further disambiguate between the two meanings of \N, and if
3095 * not a charname, go process it elsewhere
3096 * If of form \N{U+...}, pass it through if a pattern;
3097 * otherwise convert to utf8
3098 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3099 * pattern; otherwise convert to utf8 */
3101 /* Here, s points to the 'N'; the test below is guaranteed to
3102 * succeed if we are being called on a pattern as we already
3103 * know from a test above that the next character is a '{'.
3104 * On a non-pattern \N must mean 'named sequence, which
3105 * requires braces */
3108 yyerror("Missing braces on \\N{}");
3113 /* If there is no matching '}', it is an error. */
3114 if (! (e = strchr(s, '}'))) {
3115 if (! PL_lex_inpat) {
3116 yyerror("Missing right brace on \\N{}");
3118 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3123 /* Here it looks like a named character */
3127 /* XXX This block is temporary code. \N{} implies that the
3128 * pattern is to have Unicode semantics, and therefore
3129 * currently has to be encoded in utf8. By putting it in
3130 * utf8 now, we save a whole pass in the regular expression
3131 * compiler. Once that code is changed so Unicode
3132 * semantics doesn't necessarily have to be in utf8, this
3133 * block should be removed. However, the code that parses
3134 * the output of this would have to be changed to not
3135 * necessarily expect utf8 */
3137 SvCUR_set(sv, d - SvPVX_const(sv));
3140 /* See Note on sizing above. */
3141 sv_utf8_upgrade_flags_grow(sv,
3142 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3143 /* 5 = '\N{' + cur char + NUL */
3144 (STRLEN)(send - s) + 5);
3145 d = SvPVX(sv) + SvCUR(sv);
3150 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3151 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3152 | PERL_SCAN_DISALLOW_PREFIX;
3155 /* For \N{U+...}, the '...' is a unicode value even on
3156 * EBCDIC machines */
3157 s += 2; /* Skip to next char after the 'U+' */
3159 uv = grok_hex(s, &len, &flags, NULL);
3160 if (len == 0 || len != (STRLEN)(e - s)) {
3161 yyerror("Invalid hexadecimal number in \\N{U+...}");
3168 /* On non-EBCDIC platforms, pass through to the regex
3169 * compiler unchanged. The reason we evaluated the
3170 * number above is to make sure there wasn't a syntax
3171 * error. But on EBCDIC we convert to native so
3172 * downstream code can continue to assume it's native
3174 s -= 5; /* Include the '\N{U+' */
3176 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3179 (unsigned int) UNI_TO_NATIVE(uv));
3181 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3185 else { /* Not a pattern: convert the hex to string */
3187 /* If destination is not in utf8, unconditionally
3188 * recode it to be so. This is because \N{} implies
3189 * Unicode semantics, and scalars have to be in utf8
3190 * to guarantee those semantics */
3192 SvCUR_set(sv, d - SvPVX_const(sv));
3195 /* See Note on sizing above. */
3196 sv_utf8_upgrade_flags_grow(
3198 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3199 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3200 d = SvPVX(sv) + SvCUR(sv);
3204 /* Add the string to the output */
3205 if (UNI_IS_INVARIANT(uv)) {
3208 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3211 else { /* Here is \N{NAME} but not \N{U+...}. */
3213 SV *res; /* result from charnames */
3214 const char *str; /* the string in 'res' */
3215 STRLEN len; /* its length */
3217 /* Get the value for NAME */
3218 res = newSVpvn(s, e - s);
3219 res = new_constant( NULL, 0, "charnames",
3220 /* includes all of: \N{...} */
3221 res, NULL, s - 3, e - s + 4 );
3223 /* Most likely res will be in utf8 already since the
3224 * standard charnames uses pack U, but a custom translator
3225 * can leave it otherwise, so make sure. XXX This can be
3226 * revisited to not have charnames use utf8 for characters
3227 * that don't need it when regexes don't have to be in utf8
3228 * for Unicode semantics. If doing so, remember EBCDIC */
3229 sv_utf8_upgrade(res);
3230 str = SvPV_const(res, len);
3232 /* Don't accept malformed input */
3233 if (! is_utf8_string((U8 *) str, len)) {
3234 yyerror("Malformed UTF-8 returned by \\N");
3236 else if (PL_lex_inpat) {
3238 if (! len) { /* The name resolved to an empty string */
3239 Copy("\\N{}", d, 4, char);
3243 /* In order to not lose information for the regex
3244 * compiler, pass the result in the specially made
3245 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3246 * the code points in hex of each character
3247 * returned by charnames */
3249 const char *str_end = str + len;
3250 STRLEN char_length; /* cur char's byte length */
3251 STRLEN output_length; /* and the number of bytes
3252 after this is translated
3254 const STRLEN off = d - SvPVX_const(sv);
3256 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3257 * max('U+', '.'); and 1 for NUL */
3258 char hex_string[2 * UTF8_MAXBYTES + 5];
3260 /* Get the first character of the result. */
3261 U32 uv = utf8n_to_uvuni((U8 *) str,
3266 /* The call to is_utf8_string() above hopefully
3267 * guarantees that there won't be an error. But
3268 * it's easy here to make sure. The function just
3269 * above warns and returns 0 if invalid utf8, but
3270 * it can also return 0 if the input is validly a
3271 * NUL. Disambiguate */
3272 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3273 uv = UNICODE_REPLACEMENT;
3276 /* Convert first code point to hex, including the
3277 * boiler plate before it. For all these, we
3278 * convert to native format so that downstream code
3279 * can continue to assume the input is native */
3281 my_snprintf(hex_string, sizeof(hex_string),
3283 (unsigned int) UNI_TO_NATIVE(uv));
3285 /* Make sure there is enough space to hold it */
3286 d = off + SvGROW(sv, off
3288 + (STRLEN)(send - e)
3289 + 2); /* '}' + NUL */
3291 Copy(hex_string, d, output_length, char);
3294 /* For each subsequent character, append dot and
3295 * its ordinal in hex */
3296 while ((str += char_length) < str_end) {
3297 const STRLEN off = d - SvPVX_const(sv);
3298 U32 uv = utf8n_to_uvuni((U8 *) str,
3302 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3303 uv = UNICODE_REPLACEMENT;
3307 my_snprintf(hex_string, sizeof(hex_string),
3309 (unsigned int) UNI_TO_NATIVE(uv));
3311 d = off + SvGROW(sv, off
3313 + (STRLEN)(send - e)
3314 + 2); /* '}' + NUL */
3315 Copy(hex_string, d, output_length, char);
3319 *d++ = '}'; /* Done. Add the trailing brace */
3322 else { /* Here, not in a pattern. Convert the name to a
3325 /* If destination is not in utf8, unconditionally
3326 * recode it to be so. This is because \N{} implies
3327 * Unicode semantics, and scalars have to be in utf8
3328 * to guarantee those semantics */
3330 SvCUR_set(sv, d - SvPVX_const(sv));
3333 /* See Note on sizing above. */
3334 sv_utf8_upgrade_flags_grow(sv,
3335 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3336 len + (STRLEN)(send - s) + 1);
3337 d = SvPVX(sv) + SvCUR(sv);
3339 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3341 /* See Note on sizing above. (NOTE: SvCUR() is not
3342 * set correctly here). */
3343 const STRLEN off = d - SvPVX_const(sv);
3344 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3346 Copy(str, d, len, char);
3351 /* Deprecate non-approved name syntax */
3352 if (ckWARN_d(WARN_DEPRECATED)) {
3353 bool problematic = FALSE;
3356 /* For non-ut8 input, look to see that the first
3357 * character is an alpha, then loop through the rest
3358 * checking that each is a continuation */
3360 if (! isALPHAU(*i)) problematic = TRUE;
3361 else for (i = s + 1; i < e; i++) {
3362 if (isCHARNAME_CONT(*i)) continue;
3368 /* Similarly for utf8. For invariants can check
3369 * directly. We accept anything above the latin1
3370 * range because it is immaterial to Perl if it is
3371 * correct or not, and is expensive to check. But
3372 * it is fairly easy in the latin1 range to convert
3373 * the variants into a single character and check
3375 if (UTF8_IS_INVARIANT(*i)) {
3376 if (! isALPHAU(*i)) problematic = TRUE;
3377 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3378 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*i,
3384 if (! problematic) for (i = s + UTF8SKIP(s);
3388 if (UTF8_IS_INVARIANT(*i)) {
3389 if (isCHARNAME_CONT(*i)) continue;
3390 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3392 } else if (isCHARNAME_CONT(
3394 TWO_BYTE_UTF8_TO_UNI(*i, *(i+1)))))
3403 /* The e-i passed to the final %.*s makes sure that
3404 * should the trailing NUL be missing that this
3405 * print won't run off the end of the string */
3406 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3407 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s",
3408 (int)(i - s + 1), s, (int)(e - i), i + 1);
3411 } /* End \N{NAME} */
3414 native_range = FALSE; /* \N{} is defined to be Unicode */
3416 s = e + 1; /* Point to just after the '}' */
3419 /* \c is a control character */
3423 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3426 yyerror("Missing control char name in \\c");
3430 /* printf-style backslashes, formfeeds, newlines, etc */
3432 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3435 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3438 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3441 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3444 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3447 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3450 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3456 } /* end if (backslash) */
3463 /* If we started with encoded form, or already know we want it,
3464 then encode the next character */
3465 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3469 /* One might think that it is wasted effort in the case of the
3470 * source being utf8 (this_utf8 == TRUE) to take the next character
3471 * in the source, convert it to an unsigned value, and then convert
3472 * it back again. But the source has not been validated here. The
3473 * routine that does the conversion checks for errors like
3476 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3477 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3479 SvCUR_set(sv, d - SvPVX_const(sv));
3482 /* See Note on sizing above. */
3483 sv_utf8_upgrade_flags_grow(sv,
3484 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3485 need + (STRLEN)(send - s) + 1);
3486 d = SvPVX(sv) + SvCUR(sv);
3488 } else if (need > len) {
3489 /* encoded value larger than old, may need extra space (NOTE:
3490 * SvCUR() is not set correctly here). See Note on sizing
3492 const STRLEN off = d - SvPVX_const(sv);
3493 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3497 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3499 if (uv > 255 && !dorange)
3500 native_range = FALSE;
3504 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3506 } /* while loop to process each character */
3508 /* terminate the string and set up the sv */
3510 SvCUR_set(sv, d - SvPVX_const(sv));
3511 if (SvCUR(sv) >= SvLEN(sv))
3512 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3513 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3516 if (PL_encoding && !has_utf8) {
3517 sv_recode_to_utf8(sv, PL_encoding);
3523 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3524 PL_sublex_info.sub_op->op_private |=
3525 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3529 /* shrink the sv if we allocated more than we used */
3530 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3531 SvPV_shrink_to_cur(sv);
3534 /* return the substring (via pl_yylval) only if we parsed anything */
3535 if (s > PL_bufptr) {
3536 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3537 const char *const key = PL_lex_inpat ? "qr" : "q";
3538 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3542 if (PL_lex_inwhat == OP_TRANS) {
3545 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3553 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3556 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3563 * Returns TRUE if there's more to the expression (e.g., a subscript),
3566 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3568 * ->[ and ->{ return TRUE
3569 * { and [ outside a pattern are always subscripts, so return TRUE
3570 * if we're outside a pattern and it's not { or [, then return FALSE
3571 * if we're in a pattern and the first char is a {
3572 * {4,5} (any digits around the comma) returns FALSE
3573 * if we're in a pattern and the first char is a [
3575 * [SOMETHING] has a funky algorithm to decide whether it's a
3576 * character class or not. It has to deal with things like
3577 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3578 * anything else returns TRUE
3581 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3584 S_intuit_more(pTHX_ register char *s)
3588 PERL_ARGS_ASSERT_INTUIT_MORE;
3590 if (PL_lex_brackets)
3592 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3594 if (*s != '{' && *s != '[')
3599 /* In a pattern, so maybe we have {n,m}. */
3607 /* On the other hand, maybe we have a character class */
3610 if (*s == ']' || *s == '^')
3613 /* this is terrifying, and it works */
3614 int weight = 2; /* let's weigh the evidence */
3616 unsigned char un_char = 255, last_un_char;
3617 const char * const send = strchr(s,']');
3618 char tmpbuf[sizeof PL_tokenbuf * 4];
3620 if (!send) /* has to be an expression */
3623 Zero(seen,256,char);
3626 else if (isDIGIT(*s)) {
3628 if (isDIGIT(s[1]) && s[2] == ']')
3634 for (; s < send; s++) {
3635 last_un_char = un_char;
3636 un_char = (unsigned char)*s;
3641 weight -= seen[un_char] * 10;
3642 if (isALNUM_lazy_if(s+1,UTF)) {
3644 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3645 len = (int)strlen(tmpbuf);
3646 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3647 UTF ? SVf_UTF8 : 0, SVt_PV))
3652 else if (*s == '$' && s[1] &&
3653 strchr("[#!%*<>()-=",s[1])) {
3654 if (/*{*/ strchr("])} =",s[2]))
3663 if (strchr("wds]",s[1]))
3665 else if (seen[(U8)'\''] || seen[(U8)'"'])
3667 else if (strchr("rnftbxcav",s[1]))
3669 else if (isDIGIT(s[1])) {
3671 while (s[1] && isDIGIT(s[1]))
3681 if (strchr("aA01! ",last_un_char))
3683 if (strchr("zZ79~",s[1]))
3685 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3686 weight -= 5; /* cope with negative subscript */
3689 if (!isALNUM(last_un_char)
3690 && !(last_un_char == '$' || last_un_char == '@'
3691 || last_un_char == '&')
3692 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3697 if (keyword(tmpbuf, d - tmpbuf, 0))
3700 if (un_char == last_un_char + 1)
3702 weight -= seen[un_char];
3707 if (weight >= 0) /* probably a character class */
3717 * Does all the checking to disambiguate
3719 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3720 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3722 * First argument is the stuff after the first token, e.g. "bar".
3724 * Not a method if bar is a filehandle.
3725 * Not a method if foo is a subroutine prototyped to take a filehandle.
3726 * Not a method if it's really "Foo $bar"
3727 * Method if it's "foo $bar"
3728 * Not a method if it's really "print foo $bar"
3729 * Method if it's really "foo package::" (interpreted as package->foo)
3730 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3731 * Not a method if bar is a filehandle or package, but is quoted with
3736 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3739 char *s = start + (*start == '$');
3740 char tmpbuf[sizeof PL_tokenbuf];
3747 PERL_ARGS_ASSERT_INTUIT_METHOD;
3750 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3754 const char *proto = CvPROTO(cv);
3765 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3766 /* start is the beginning of the possible filehandle/object,
3767 * and s is the end of it
3768 * tmpbuf is a copy of it
3771 if (*start == '$') {
3772 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3773 isUPPER(*PL_tokenbuf))
3776 len = start - SvPVX(PL_linestr);
3780 start = SvPVX(PL_linestr) + len;
3784 return *s == '(' ? FUNCMETH : METHOD;
3786 if (!keyword(tmpbuf, len, 0)) {
3787 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3791 soff = s - SvPVX(PL_linestr);
3795 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
3796 if (indirgv && GvCVu(indirgv))
3798 /* filehandle or package name makes it a method */
3799 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
3801 soff = s - SvPVX(PL_linestr);
3804 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3805 return 0; /* no assumptions -- "=>" quotes bareword */
3807 start_force(PL_curforce);
3808 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3809 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3810 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3812 curmad('X', newSVpvn_flags(start,SvPVX(PL_linestr) + soff - start,
3813 ( UTF ? SVf_UTF8 : 0 )));
3818 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3820 return *s == '(' ? FUNCMETH : METHOD;
3826 /* Encoded script support. filter_add() effectively inserts a
3827 * 'pre-processing' function into the current source input stream.
3828 * Note that the filter function only applies to the current source file
3829 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3831 * The datasv parameter (which may be NULL) can be used to pass
3832 * private data to this instance of the filter. The filter function
3833 * can recover the SV using the FILTER_DATA macro and use it to
3834 * store private buffers and state information.
3836 * The supplied datasv parameter is upgraded to a PVIO type
3837 * and the IoDIRP/IoANY field is used to store the function pointer,
3838 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3839 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3840 * private use must be set using malloc'd pointers.
3844 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3853 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
3854 Perl_croak(aTHX_ "Source filters apply only to byte streams");
3856 if (!PL_rsfp_filters)
3857 PL_rsfp_filters = newAV();
3860 SvUPGRADE(datasv, SVt_PVIO);
3861 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3862 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3863 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3864 FPTR2DPTR(void *, IoANY(datasv)),
3865 SvPV_nolen(datasv)));
3866 av_unshift(PL_rsfp_filters, 1);
3867 av_store(PL_rsfp_filters, 0, datasv) ;
3869 !PL_parser->filtered
3870 && PL_parser->lex_flags & LEX_EVALBYTES
3871 && PL_bufptr < PL_bufend
3873 const char *s = PL_bufptr;
3874 while (s < PL_bufend) {
3876 SV *linestr = PL_parser->linestr;
3877 char *buf = SvPVX(linestr);
3878 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
3879 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
3880 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
3881 STRLEN const linestart_pos = PL_parser->linestart - buf;
3882 STRLEN const last_uni_pos =
3883 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
3884 STRLEN const last_lop_pos =
3885 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
3886 av_push(PL_rsfp_filters, linestr);
3887 PL_parser->linestr =
3888 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
3889 buf = SvPVX(PL_parser->linestr);
3890 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
3891 PL_parser->bufptr = buf + bufptr_pos;
3892 PL_parser->oldbufptr = buf + oldbufptr_pos;
3893 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
3894 PL_parser->linestart = buf + linestart_pos;
3895 if (PL_parser->last_uni)
3896 PL_parser->last_uni = buf + last_uni_pos;
3897 if (PL_parser->last_lop)
3898 PL_parser->last_lop = buf + last_lop_pos;
3899 SvLEN(linestr) = SvCUR(linestr);
3900 SvCUR(linestr) = s-SvPVX(linestr);
3901 PL_parser->filtered = 1;
3911 /* Delete most recently added instance of this filter function. */
3913 Perl_filter_del(pTHX_ filter_t funcp)
3918 PERL_ARGS_ASSERT_FILTER_DEL;
3921 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3922 FPTR2DPTR(void*, funcp)));
3924 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3926 /* if filter is on top of stack (usual case) just pop it off */
3927 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3928 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3929 sv_free(av_pop(PL_rsfp_filters));
3933 /* we need to search for the correct entry and clear it */
3934 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3938 /* Invoke the idxth filter function for the current rsfp. */
3939 /* maxlen 0 = read one text line */
3941 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3946 /* This API is bad. It should have been using unsigned int for maxlen.
3947 Not sure if we want to change the API, but if not we should sanity
3948 check the value here. */
3949 unsigned int correct_length
3958 PERL_ARGS_ASSERT_FILTER_READ;
3960 if (!PL_parser || !PL_rsfp_filters)
3962 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3963 /* Provide a default input filter to make life easy. */
3964 /* Note that we append to the line. This is handy. */
3965 DEBUG_P(PerlIO_printf(Perl_debug_log,
3966 "filter_read %d: from rsfp\n", idx));
3967 if (correct_length) {
3970 const int old_len = SvCUR(buf_sv);
3972 /* ensure buf_sv is large enough */
3973 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3974 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3975 correct_length)) <= 0) {
3976 if (PerlIO_error(PL_rsfp))
3977 return -1; /* error */
3979 return 0 ; /* end of file */
3981 SvCUR_set(buf_sv, old_len + len) ;
3982 SvPVX(buf_sv)[old_len + len] = '\0';
3985 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3986 if (PerlIO_error(PL_rsfp))
3987 return -1; /* error */
3989 return 0 ; /* end of file */
3992 return SvCUR(buf_sv);
3994 /* Skip this filter slot if filter has been deleted */
3995 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3996 DEBUG_P(PerlIO_printf(Perl_debug_log,
3997 "filter_read %d: skipped (filter deleted)\n",
3999 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4001 if (SvTYPE(datasv) != SVt_PVIO) {
4002 if (correct_length) {
4004 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4005 if (!remainder) return 0; /* eof */
4006 if (correct_length > remainder) correct_length = remainder;
4007 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4008 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4011 const char *s = SvEND(datasv);
4012 const char *send = SvPVX(datasv) + SvLEN(datasv);
4020 if (s == send) return 0; /* eof */
4021 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4022 SvCUR_set(datasv, s-SvPVX(datasv));
4024 return SvCUR(buf_sv);
4026 /* Get function pointer hidden within datasv */
4027 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4028 DEBUG_P(PerlIO_printf(Perl_debug_log,
4029 "filter_read %d: via function %p (%s)\n",
4030 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4031 /* Call function. The function is expected to */
4032 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4033 /* Return: <0:error, =0:eof, >0:not eof */
4034 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4038 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
4042 PERL_ARGS_ASSERT_FILTER_GETS;
4044 #ifdef PERL_CR_FILTER
4045 if (!PL_rsfp_filters) {
4046 filter_add(S_cr_textfilter,NULL);
4049 if (PL_rsfp_filters) {
4051 SvCUR_set(sv, 0); /* start with empty line */
4052 if (FILTER_READ(0, sv, 0) > 0)
4053 return ( SvPVX(sv) ) ;
4058 return (sv_gets(sv, PL_rsfp, append));
4062 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4067 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4069 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4073 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4074 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4076 return GvHV(gv); /* Foo:: */
4079 /* use constant CLASS => 'MyClass' */
4080 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4081 if (gv && GvCV(gv)) {
4082 SV * const sv = cv_const_sv(GvCV(gv));
4084 pkgname = SvPV_const(sv, len);
4087 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4091 * S_readpipe_override
4092 * Check whether readpipe() is overridden, and generates the appropriate
4093 * optree, provided sublex_start() is called afterwards.
4096 S_readpipe_override(pTHX)
4099 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4100 pl_yylval.ival = OP_BACKTICK;
4102 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4104 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4105 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4106 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4108 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4109 op_append_elem(OP_LIST,
4110 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4111 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4118 * The intent of this yylex wrapper is to minimize the changes to the
4119 * tokener when we aren't interested in collecting madprops. It remains
4120 * to be seen how successful this strategy will be...
4127 char *s = PL_bufptr;
4129 /* make sure PL_thiswhite is initialized */
4133 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
4134 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4135 return S_pending_ident(aTHX);
4137 /* previous token ate up our whitespace? */
4138 if (!PL_lasttoke && PL_nextwhite) {
4139 PL_thiswhite = PL_nextwhite;
4143 /* isolate the token, and figure out where it is without whitespace */
4144 PL_realtokenstart = -1;
4148 assert(PL_curforce < 0);
4150 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4151 if (!PL_thistoken) {
4152 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4153 PL_thistoken = newSVpvs("");
4155 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4156 PL_thistoken = newSVpvn(tstart, s - tstart);
4159 if (PL_thismad) /* install head */
4160 CURMAD('X', PL_thistoken);
4163 /* last whitespace of a sublex? */
4164 if (optype == ')' && PL_endwhite) {
4165 CURMAD('X', PL_endwhite);
4170 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4171 if (!PL_thiswhite && !PL_endwhite && !optype) {
4172 sv_free(PL_thistoken);
4177 /* put off final whitespace till peg */
4178 if (optype == ';' && !PL_rsfp && !PL_parser->filtered) {
4179 PL_nextwhite = PL_thiswhite;
4182 else if (PL_thisopen) {
4183 CURMAD('q', PL_thisopen);
4185 sv_free(PL_thistoken);
4189 /* Store actual token text as madprop X */
4190 CURMAD('X', PL_thistoken);
4194 /* add preceding whitespace as madprop _ */
4195 CURMAD('_', PL_thiswhite);
4199 /* add quoted material as madprop = */
4200 CURMAD('=', PL_thisstuff);
4204 /* add terminating quote as madprop Q */
4205 CURMAD('Q', PL_thisclose);
4209 /* special processing based on optype */
4213 /* opval doesn't need a TOKEN since it can already store mp */
4223 if (pl_yylval.opval)
4224 append_madprops(PL_thismad, pl_yylval.opval, 0);
4232 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4241 /* remember any fake bracket that lexer is about to discard */
4242 if (PL_lex_brackets == 1 &&
4243 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4246 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4249 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4250 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4253 break; /* don't bother looking for trailing comment */
4262 /* attach a trailing comment to its statement instead of next token */
4266 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4268 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4270 if (*s == '\n' || *s == '#') {
4271 while (s < PL_bufend && *s != '\n')
4275 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4276 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4293 /* Create new token struct. Note: opvals return early above. */
4294 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4301 S_tokenize_use(pTHX_ int is_use, char *s) {
4304 PERL_ARGS_ASSERT_TOKENIZE_USE;
4306 if (PL_expect != XSTATE)
4307 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4308 is_use ? "use" : "no"));
4310 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4311 s = force_version(s, TRUE);
4312 if (*s == ';' || *s == '}'
4313 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4314 start_force(PL_curforce);
4315 NEXTVAL_NEXTTOKE.opval = NULL;
4318 else if (*s == 'v') {
4319 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4320 s = force_version(s, FALSE);
4324 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4325 s = force_version(s, FALSE);
4327 pl_yylval.ival = is_use;
4331 static const char* const exp_name[] =
4332 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4333 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4337 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4339 S_word_takes_any_delimeter(char *p, STRLEN len)
4341 return (len == 1 && strchr("msyq", p[0])) ||
4343 (p[0] == 't' && p[1] == 'r') ||
4344 (p[0] == 'q' && strchr("qwxr", p[1]))));
4350 Works out what to call the token just pulled out of the input
4351 stream. The yacc parser takes care of taking the ops we return and
4352 stitching them into a tree.
4358 if read an identifier
4359 if we're in a my declaration
4360 croak if they tried to say my($foo::bar)
4361 build the ops for a my() declaration
4362 if it's an access to a my() variable
4363 are we in a sort block?
4364 croak if my($a); $a <=> $b
4365 build ops for access to a my() variable
4366 if in a dq string, and they've said @foo and we can't find @foo
4368 build ops for a bareword
4369 if we already built the token before, use it.
4374 #pragma segment Perl_yylex
4380 register char *s = PL_bufptr;
4386 /* orig_keyword, gvp, and gv are initialized here because
4387 * jump to the label just_a_word_zero can bypass their
4388 * initialization later. */
4389 I32 orig_keyword = 0;
4394 SV* tmp = newSVpvs("");
4395 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4396 (IV)CopLINE(PL_curcop),
4397 lex_state_names[PL_lex_state],
4398 exp_name[PL_expect],
4399 pv_display(tmp, s, strlen(s), 0, 60));
4402 /* check if there's an identifier for us to look at */
4403 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4404 return REPORT(S_pending_ident(aTHX));
4406 /* no identifier pending identification */
4408 switch (PL_lex_state) {
4410 case LEX_NORMAL: /* Some compilers will produce faster */
4411 case LEX_INTERPNORMAL: /* code if we comment these out. */
4415 /* when we've already built the next token, just pull it out of the queue */
4419 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4421 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4422 PL_nexttoke[PL_lasttoke].next_mad = 0;
4423 if (PL_thismad && PL_thismad->mad_key == '_') {
4424 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4425 PL_thismad->mad_val = 0;
4426 mad_free(PL_thismad);
4431 PL_lex_state = PL_lex_defer;
4432 PL_expect = PL_lex_expect;
4433 PL_lex_defer = LEX_NORMAL;
4434 if (!PL_nexttoke[PL_lasttoke].next_type)
4439 pl_yylval = PL_nextval[PL_nexttoke];
4441 PL_lex_state = PL_lex_defer;
4442 PL_expect = PL_lex_expect;
4443 PL_lex_defer = LEX_NORMAL;
4449 next_type = PL_nexttoke[PL_lasttoke].next_type;
4451 next_type = PL_nexttype[PL_nexttoke];
4453 if (next_type & (7<<24)) {
4454 if (next_type & (1<<24)) {
4455 if (PL_lex_brackets > 100)
4456 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4457 PL_lex_brackstack[PL_lex_brackets++] =
4458 (char) ((next_type >> 16) & 0xff);
4460 if (next_type & (2<<24))
4461 PL_lex_allbrackets++;
4462 if (next_type & (4<<24))
4463 PL_lex_allbrackets--;
4464 next_type &= 0xffff;
4467 /* FIXME - can these be merged? */
4470 return REPORT(next_type);
4474 /* interpolated case modifiers like \L \U, including \Q and \E.
4475 when we get here, PL_bufptr is at the \
4477 case LEX_INTERPCASEMOD:
4479 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4481 "panic: INTERPCASEMOD bufptr=%p, bufend=%p, *bufptr=%u",
4482 PL_bufptr, PL_bufend, *PL_bufptr);
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'
4493 || oldmod == 'F')) {
4495 PL_lex_state = LEX_INTERPCONCAT;
4498 PL_thistoken = newSVpvs("\\E");
4501 PL_lex_allbrackets--;
4504 else if ( PL_bufptr != PL_bufend && PL_bufptr[1] == 'E' ) {
4505 /* Got an unpaired \E */
4506 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4507 "Useless use of \\E");
4510 while (PL_bufptr != PL_bufend &&
4511 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4513 PL_thiswhite = newSVpvs("");
4514 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4518 if (PL_bufptr != PL_bufend)
4521 PL_lex_state = LEX_INTERPCONCAT;
4525 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4526 "### Saw case modifier\n"); });
4528 if (s[1] == '\\' && s[2] == 'E') {
4531 PL_thiswhite = newSVpvs("");
4532 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4535 PL_lex_state = LEX_INTERPCONCAT;
4540 if (!PL_madskills) /* when just compiling don't need correct */
4541 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4542 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4543 if ((*s == 'L' || *s == 'U' || *s == 'F') &&
4544 (strchr(PL_lex_casestack, 'L')
4545 || strchr(PL_lex_casestack, 'U')
4546 || strchr(PL_lex_casestack, 'F'))) {
4547 PL_lex_casestack[--PL_lex_casemods] = '\0';
4548 PL_lex_allbrackets--;
4551 if (PL_lex_casemods > 10)
4552 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4553 PL_lex_casestack[PL_lex_casemods++] = *s;
4554 PL_lex_casestack[PL_lex_casemods] = '\0';
4555 PL_lex_state = LEX_INTERPCONCAT;
4556 start_force(PL_curforce);
4557 NEXTVAL_NEXTTOKE.ival = 0;
4558 force_next((2<<24)|'(');
4559 start_force(PL_curforce);
4561 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4563 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4565 NEXTVAL_NEXTTOKE.ival = OP_LC;
4567 NEXTVAL_NEXTTOKE.ival = OP_UC;
4569 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4571 NEXTVAL_NEXTTOKE.ival = OP_FC;
4573 Perl_croak(aTHX_ "panic: yylex, *s=%u", *s);
4575 SV* const tmpsv = newSVpvs("\\ ");
4576 /* replace the space with the character we want to escape
4578 SvPVX(tmpsv)[1] = *s;
4584 if (PL_lex_starts) {
4590 sv_free(PL_thistoken);
4591 PL_thistoken = newSVpvs("");
4594 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4595 if (PL_lex_casemods == 1 && PL_lex_inpat)
4604 case LEX_INTERPPUSH:
4605 return REPORT(sublex_push());
4607 case LEX_INTERPSTART:
4608 if (PL_bufptr == PL_bufend)
4609 return REPORT(sublex_done());
4610 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4611 "### Interpolated variable\n"); });
4613 PL_lex_dojoin = (*PL_bufptr == '@');
4614 PL_lex_state = LEX_INTERPNORMAL;
4615 if (PL_lex_dojoin) {
4616 start_force(PL_curforce);
4617 NEXTVAL_NEXTTOKE.ival = 0;
4619 start_force(PL_curforce);
4620 force_ident("\"", '$');
4621 start_force(PL_curforce);
4622 NEXTVAL_NEXTTOKE.ival = 0;
4624 start_force(PL_curforce);
4625 NEXTVAL_NEXTTOKE.ival = 0;
4626 force_next((2<<24)|'(');
4627 start_force(PL_curforce);
4628 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4631 if (PL_lex_starts++) {
4636 sv_free(PL_thistoken);
4637 PL_thistoken = newSVpvs("");
4640 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4641 if (!PL_lex_casemods && PL_lex_inpat)
4648 case LEX_INTERPENDMAYBE:
4649 if (intuit_more(PL_bufptr)) {
4650 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4656 if (PL_lex_dojoin) {
4657 PL_lex_dojoin = FALSE;
4658 PL_lex_state = LEX_INTERPCONCAT;
4662 sv_free(PL_thistoken);
4663 PL_thistoken = newSVpvs("");
4666 PL_lex_allbrackets--;
4669 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4670 && SvEVALED(PL_lex_repl))
4672 if (PL_bufptr != PL_bufend)
4673 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4677 case LEX_INTERPCONCAT:
4679 if (PL_lex_brackets)
4680 Perl_croak(aTHX_ "panic: INTERPCONCAT, lex_brackets=%ld",
4681 (long) PL_lex_brackets);
4683 if (PL_bufptr == PL_bufend)
4684 return REPORT(sublex_done());
4686 if (SvIVX(PL_linestr) == '\'') {
4687 SV *sv = newSVsv(PL_linestr);
4690 else if ( PL_hints & HINT_NEW_RE )
4691 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4692 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4696 s = scan_const(PL_bufptr);
4698 PL_lex_state = LEX_INTERPCASEMOD;
4700 PL_lex_state = LEX_INTERPSTART;
4703 if (s != PL_bufptr) {
4704 start_force(PL_curforce);
4706 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4708 NEXTVAL_NEXTTOKE = pl_yylval;
4711 if (PL_lex_starts++) {
4715 sv_free(PL_thistoken);
4716 PL_thistoken = newSVpvs("");
4719 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4720 if (!PL_lex_casemods && PL_lex_inpat)
4733 PL_lex_state = LEX_NORMAL;
4734 s = scan_formline(PL_bufptr);
4735 if (!PL_lex_formbrack)
4741 PL_oldoldbufptr = PL_oldbufptr;
4747 sv_free(PL_thistoken);
4750 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4754 if (isIDFIRST_lazy_if(s,UTF))
4757 unsigned char c = *s;
4758 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4759 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4760 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4765 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4769 goto fake_eof; /* emulate EOF on ^D or ^Z */
4775 if (!PL_rsfp && (!PL_parser->filtered || s+1 < PL_bufend)) {
4778 if (PL_lex_brackets &&
4779 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
4780 yyerror((const char *)
4782 ? "Format not terminated"
4783 : "Missing right curly or square bracket"));
4785 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4786 "### Tokener got EOF\n");
4790 if (s++ < PL_bufend)
4791 goto retry; /* ignore stray nulls */
4794 if (!PL_in_eval && !PL_preambled) {
4795 PL_preambled = TRUE;
4801 /* Generate a string of Perl code to load the debugger.
4802 * If PERL5DB is set, it will return the contents of that,
4803 * otherwise a compile-time require of perl5db.pl. */
4805 const char * const pdb = PerlEnv_getenv("PERL5DB");
4808 sv_setpv(PL_linestr, pdb);
4809 sv_catpvs(PL_linestr,";");
4811 SETERRNO(0,SS_NORMAL);
4812 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4815 sv_setpvs(PL_linestr,"");
4816 if (PL_preambleav) {
4817 SV **svp = AvARRAY(PL_preambleav);
4818 SV **const end = svp + AvFILLp(PL_preambleav);
4820 sv_catsv(PL_linestr, *svp);
4822 sv_catpvs(PL_linestr, ";");
4824 sv_free(MUTABLE_SV(PL_preambleav));
4825 PL_preambleav = NULL;
4828 sv_catpvs(PL_linestr,
4829 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4830 if (PL_minus_n || PL_minus_p) {
4831 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4833 sv_catpvs(PL_linestr,"chomp;");
4836 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4837 || *PL_splitstr == '"')
4838 && strchr(PL_splitstr + 1, *PL_splitstr))
4839 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4841 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4842 bytes can be used as quoting characters. :-) */
4843 const char *splits = PL_splitstr;
4844 sv_catpvs(PL_linestr, "our @F=split(q\0");
4847 if (*splits == '\\')
4848 sv_catpvn(PL_linestr, splits, 1);
4849 sv_catpvn(PL_linestr, splits, 1);
4850 } while (*splits++);
4851 /* This loop will embed the trailing NUL of
4852 PL_linestr as the last thing it does before
4854 sv_catpvs(PL_linestr, ");");
4858 sv_catpvs(PL_linestr,"our @F=split(' ');");
4861 sv_catpvs(PL_linestr, "\n");
4862 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4863 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4864 PL_last_lop = PL_last_uni = NULL;
4865 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4866 update_debugger_info(PL_linestr, NULL, 0);
4871 bof = PL_rsfp ? TRUE : FALSE;
4874 fake_eof = LEX_FAKE_EOF;
4876 PL_bufptr = PL_bufend;
4877 CopLINE_inc(PL_curcop);
4878 if (!lex_next_chunk(fake_eof)) {
4879 CopLINE_dec(PL_curcop);
4881 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4883 CopLINE_dec(PL_curcop);
4886 PL_realtokenstart = -1;
4889 /* If it looks like the start of a BOM or raw UTF-16,
4890 * check if it in fact is. */
4891 if (bof && PL_rsfp &&
4896 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
4897 bof = (offset == (Off_t)SvCUR(PL_linestr));
4898 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
4899 /* offset may include swallowed CR */
4901 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
4904 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4905 s = swallow_bom((U8*)s);
4908 if (PL_parser->in_pod) {
4909 /* Incest with pod. */
4912 sv_catsv(PL_thiswhite, PL_linestr);
4914 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4915 sv_setpvs(PL_linestr, "");
4916 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4917 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4918 PL_last_lop = PL_last_uni = NULL;
4919 PL_parser->in_pod = 0;
4922 if (PL_rsfp || PL_parser->filtered)
4924 } while (PL_parser->in_pod);
4925 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4926 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4927 PL_last_lop = PL_last_uni = NULL;
4928 if (CopLINE(PL_curcop) == 1) {
4929 while (s < PL_bufend && isSPACE(*s))
4931 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4935 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4939 if (*s == '#' && *(s+1) == '!')
4941 #ifdef ALTERNATE_SHEBANG
4943 static char const as[] = ALTERNATE_SHEBANG;
4944 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4945 d = s + (sizeof(as) - 1);
4947 #endif /* ALTERNATE_SHEBANG */
4956 while (*d && !isSPACE(*d))
4960 #ifdef ARG_ZERO_IS_SCRIPT
4961 if (ipathend > ipath) {
4963 * HP-UX (at least) sets argv[0] to the script name,
4964 * which makes $^X incorrect. And Digital UNIX and Linux,
4965 * at least, set argv[0] to the basename of the Perl
4966 * interpreter. So, having found "#!", we'll set it right.
4968 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4970 assert(SvPOK(x) || SvGMAGICAL(x));
4971 if (sv_eq(x, CopFILESV(PL_curcop))) {
4972 sv_setpvn(x, ipath, ipathend - ipath);
4978 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4979 const char * const lstart = SvPV_const(x,llen);
4981 bstart += blen - llen;
4982 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4983 sv_setpvn(x, ipath, ipathend - ipath);
4988 TAINT_NOT; /* $^X is always tainted, but that's OK */
4990 #endif /* ARG_ZERO_IS_SCRIPT */
4995 d = instr(s,"perl -");
4997 d = instr(s,"perl");
4999 /* avoid getting into infinite loops when shebang
5000 * line contains "Perl" rather than "perl" */
5002 for (d = ipathend-4; d >= ipath; --d) {
5003 if ((*d == 'p' || *d == 'P')
5004 && !ibcmp(d, "perl", 4))
5014 #ifdef ALTERNATE_SHEBANG
5016 * If the ALTERNATE_SHEBANG on this system starts with a
5017 * character that can be part of a Perl expression, then if
5018 * we see it but not "perl", we're probably looking at the
5019 * start of Perl code, not a request to hand off to some
5020 * other interpreter. Similarly, if "perl" is there, but
5021 * not in the first 'word' of the line, we assume the line
5022 * contains the start of the Perl program.
5024 if (d && *s != '#') {
5025 const char *c = ipath;
5026 while (*c && !strchr("; \t\r\n\f\v#", *c))
5029 d = NULL; /* "perl" not in first word; ignore */
5031 *s = '#'; /* Don't try to parse shebang line */
5033 #endif /* ALTERNATE_SHEBANG */
5038 !instr(s,"indir") &&
5039 instr(PL_origargv[0],"perl"))
5046 while (s < PL_bufend && isSPACE(*s))
5048 if (s < PL_bufend) {
5049 Newx(newargv,PL_origargc+3,char*);
5051 while (s < PL_bufend && !isSPACE(*s))
5054 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
5057 newargv = PL_origargv;
5060 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
5062 Perl_croak(aTHX_ "Can't exec %s", ipath);
5065 while (*d && !isSPACE(*d))
5067 while (SPACE_OR_TAB(*d))
5071 const bool switches_done = PL_doswitches;
5072 const U32 oldpdb = PL_perldb;
5073 const bool oldn = PL_minus_n;
5074 const bool oldp = PL_minus_p;
5078 bool baduni = FALSE;
5080 const char *d2 = d1 + 1;
5081 if (parse_unicode_opts((const char **)&d2)
5085 if (baduni || *d1 == 'M' || *d1 == 'm') {
5086 const char * const m = d1;
5087 while (*d1 && !isSPACE(*d1))
5089 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
5092 d1 = moreswitches(d1);
5094 if (PL_doswitches && !switches_done) {
5095 int argc = PL_origargc;
5096 char **argv = PL_origargv;
5099 } while (argc && argv[0][0] == '-' && argv[0][1]);
5100 init_argv_symbols(argc,argv);
5102 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
5103 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
5104 /* if we have already added "LINE: while (<>) {",
5105 we must not do it again */
5107 sv_setpvs(PL_linestr, "");
5108 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5109 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5110 PL_last_lop = PL_last_uni = NULL;
5111 PL_preambled = FALSE;
5112 if (PERLDB_LINE || PERLDB_SAVESRC)
5113 (void)gv_fetchfile(PL_origfilename);
5120 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5122 PL_lex_state = LEX_FORMLINE;
5127 #ifdef PERL_STRICT_CR
5128 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5130 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5132 case ' ': case '\t': case '\f': case 013:
5134 PL_realtokenstart = -1;
5136 PL_thiswhite = newSVpvs("");
5137 sv_catpvn(PL_thiswhite, s, 1);
5144 PL_realtokenstart = -1;
5148 if (PL_lex_state != LEX_NORMAL ||
5149 (PL_in_eval && !PL_rsfp && !PL_parser->filtered)) {
5150 if (*s == '#' && s == PL_linestart && PL_in_eval
5151 && !PL_rsfp && !PL_parser->filtered) {
5152 /* handle eval qq[#line 1 "foo"\n ...] */
5153 CopLINE_dec(PL_curcop);
5156 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
5158 if (!PL_in_eval || PL_rsfp || PL_parser->filtered)
5163 while (d < PL_bufend && *d != '\n')
5167 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5168 Perl_croak(aTHX_ "panic: input overflow, %p > %p",
5172 PL_thiswhite = newSVpvn(s, d - s);
5177 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5179 PL_lex_state = LEX_FORMLINE;
5185 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
5186 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
5189 TOKEN(PEG); /* make sure any #! line is accessible */
5194 /* if (PL_madskills && PL_lex_formbrack) { */
5196 while (d < PL_bufend && *d != '\n')
5200 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5201 Perl_croak(aTHX_ "panic: input overflow");
5202 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
5204 PL_thiswhite = newSVpvs("");
5205 if (CopLINE(PL_curcop) == 1) {
5206 sv_setpvs(PL_thiswhite, "");
5209 sv_catpvn(PL_thiswhite, s, d - s);
5223 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
5231 while (s < PL_bufend && SPACE_OR_TAB(*s))
5234 if (strnEQ(s,"=>",2)) {
5235 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
5236 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5237 OPERATOR('-'); /* unary minus */
5239 PL_last_uni = PL_oldbufptr;
5241 case 'r': ftst = OP_FTEREAD; break;
5242 case 'w': ftst = OP_FTEWRITE; break;
5243 case 'x': ftst = OP_FTEEXEC; break;
5244 case 'o': ftst = OP_FTEOWNED; break;
5245 case 'R': ftst = OP_FTRREAD; break;
5246 case 'W': ftst = OP_FTRWRITE; break;
5247 case 'X': ftst = OP_FTREXEC; break;
5248 case 'O': ftst = OP_FTROWNED; break;
5249 case 'e': ftst = OP_FTIS; break;
5250 case 'z': ftst = OP_FTZERO; break;
5251 case 's': ftst = OP_FTSIZE; break;
5252 case 'f': ftst = OP_FTFILE; break;
5253 case 'd': ftst = OP_FTDIR; break;
5254 case 'l': ftst = OP_FTLINK; break;
5255 case 'p': ftst = OP_FTPIPE; break;
5256 case 'S': ftst = OP_FTSOCK; break;
5257 case 'u': ftst = OP_FTSUID; break;
5258 case 'g': ftst = OP_FTSGID; break;
5259 case 'k': ftst = OP_FTSVTX; break;
5260 case 'b': ftst = OP_FTBLK; break;
5261 case 'c': ftst = OP_FTCHR; break;
5262 case 't': ftst = OP_FTTTY; break;
5263 case 'T': ftst = OP_FTTEXT; break;
5264 case 'B': ftst = OP_FTBINARY; break;
5265 case 'M': case 'A': case 'C':
5266 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5268 case 'M': ftst = OP_FTMTIME; break;
5269 case 'A': ftst = OP_FTATIME; break;
5270 case 'C': ftst = OP_FTCTIME; break;
5278 PL_last_lop_op = (OPCODE)ftst;
5279 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5280 "### Saw file test %c\n", (int)tmp);
5285 /* Assume it was a minus followed by a one-letter named
5286 * subroutine call (or a -bareword), then. */
5287 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5288 "### '-%c' looked like a file test but was not\n",
5295 const char tmp = *s++;
5298 if (PL_expect == XOPERATOR)
5303 else if (*s == '>') {
5306 if (isIDFIRST_lazy_if(s,UTF)) {
5307 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5315 if (PL_expect == XOPERATOR) {
5316 if (*s == '=' && !PL_lex_allbrackets &&
5317 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5324 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5326 OPERATOR('-'); /* unary minus */
5332 const char tmp = *s++;
5335 if (PL_expect == XOPERATOR)
5340 if (PL_expect == XOPERATOR) {
5341 if (*s == '=' && !PL_lex_allbrackets &&
5342 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5349 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5356 if (PL_expect != XOPERATOR) {
5357 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5358 PL_expect = XOPERATOR;
5359 force_ident(PL_tokenbuf, '*');
5367 if (*s == '=' && !PL_lex_allbrackets &&
5368 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5374 if (*s == '=' && !PL_lex_allbrackets &&
5375 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5382 if (PL_expect == XOPERATOR) {
5383 if (s[1] == '=' && !PL_lex_allbrackets &&
5384 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5389 PL_tokenbuf[0] = '%';
5390 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5391 sizeof PL_tokenbuf - 1, FALSE);
5392 if (!PL_tokenbuf[1]) {
5395 PL_pending_ident = '%';
5399 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5400 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5405 if (PL_lex_brackets > 100)
5406 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5407 PL_lex_brackstack[PL_lex_brackets++] = 0;
5408 PL_lex_allbrackets++;
5410 const char tmp = *s++;
5415 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5417 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5425 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5432 goto just_a_word_zero_gv;
5435 switch (PL_expect) {
5441 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5443 PL_bufptr = s; /* update in case we back off */
5446 "Use of := for an empty attribute list is not allowed");
5453 PL_expect = XTERMBLOCK;
5456 stuffstart = s - SvPVX(PL_linestr) - 1;
5460 while (isIDFIRST_lazy_if(s,UTF)) {
5463 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5464 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5465 if (tmp < 0) tmp = -tmp;
5480 sv = newSVpvn_flags(s, len, UTF ? SVf_UTF8 : 0);
5482 d = scan_str(d,TRUE,TRUE);
5484 /* MUST advance bufptr here to avoid bogus
5485 "at end of line" context messages from yyerror().
5487 PL_bufptr = s + len;
5488 yyerror("Unterminated attribute parameter in attribute list");
5492 return REPORT(0); /* EOF indicator */
5496 sv_catsv(sv, PL_lex_stuff);
5497 attrs = op_append_elem(OP_LIST, attrs,
5498 newSVOP(OP_CONST, 0, sv));
5499 SvREFCNT_dec(PL_lex_stuff);
5500 PL_lex_stuff = NULL;
5503 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5505 if (PL_in_my == KEY_our) {
5506 deprecate(":unique");
5509 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5512 /* NOTE: any CV attrs applied here need to be part of
5513 the CVf_BUILTIN_ATTRS define in cv.h! */
5514 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5516 CvLVALUE_on(PL_compcv);
5518 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5520 deprecate(":locked");
5522 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5524 CvMETHOD_on(PL_compcv);
5526 /* After we've set the flags, it could be argued that
5527 we don't need to do the attributes.pm-based setting
5528 process, and shouldn't bother appending recognized
5529 flags. To experiment with that, uncomment the
5530 following "else". (Note that's already been
5531 uncommented. That keeps the above-applied built-in
5532 attributes from being intercepted (and possibly
5533 rejected) by a package's attribute routines, but is
5534 justified by the performance win for the common case
5535 of applying only built-in attributes.) */
5537 attrs = op_append_elem(OP_LIST, attrs,
5538 newSVOP(OP_CONST, 0,
5542 if (*s == ':' && s[1] != ':')
5545 break; /* require real whitespace or :'s */
5546 /* XXX losing whitespace on sequential attributes here */
5550 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5551 if (*s != ';' && *s != '}' && *s != tmp
5552 && (tmp != '=' || *s != ')')) {
5553 const char q = ((*s == '\'') ? '"' : '\'');
5554 /* If here for an expression, and parsed no attrs, back
5556 if (tmp == '=' && !attrs) {
5560 /* MUST advance bufptr here to avoid bogus "at end of line"
5561 context messages from yyerror().
5564 yyerror( (const char *)
5566 ? Perl_form(aTHX_ "Invalid separator character "
5567 "%c%c%c in attribute list", q, *s, q)
5568 : "Unterminated attribute list" ) );
5576 start_force(PL_curforce);
5577 NEXTVAL_NEXTTOKE.opval = attrs;
5578 CURMAD('_', PL_nextwhite);
5583 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5584 (s - SvPVX(PL_linestr)) - stuffstart);
5589 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5593 PL_lex_allbrackets--;
5597 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5598 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5602 PL_lex_allbrackets++;
5605 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5611 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5614 PL_lex_allbrackets--;
5620 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5623 if (PL_lex_brackets <= 0)
5624 yyerror("Unmatched right square bracket");
5627 PL_lex_allbrackets--;
5628 if (PL_lex_state == LEX_INTERPNORMAL) {
5629 if (PL_lex_brackets == 0) {
5630 if (*s == '-' && s[1] == '>')
5631 PL_lex_state = LEX_INTERPENDMAYBE;
5632 else if (*s != '[' && *s != '{')
5633 PL_lex_state = LEX_INTERPEND;
5640 if (PL_lex_brackets > 100) {
5641 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5643 switch (PL_expect) {
5645 if (PL_lex_formbrack) {
5649 if (PL_oldoldbufptr == PL_last_lop)
5650 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5652 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5653 PL_lex_allbrackets++;
5654 OPERATOR(HASHBRACK);
5656 while (s < PL_bufend && SPACE_OR_TAB(*s))
5659 PL_tokenbuf[0] = '\0';
5660 if (d < PL_bufend && *d == '-') {
5661 PL_tokenbuf[0] = '-';
5663 while (d < PL_bufend && SPACE_OR_TAB(*d))
5666 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5667 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5669 while (d < PL_bufend && SPACE_OR_TAB(*d))
5672 const char minus = (PL_tokenbuf[0] == '-');
5673 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5681 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5682 PL_lex_allbrackets++;
5687 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5688 PL_lex_allbrackets++;
5693 if (PL_oldoldbufptr == PL_last_lop)
5694 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5696 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5697 PL_lex_allbrackets++;
5700 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5702 /* This hack is to get the ${} in the message. */
5704 yyerror("syntax error");
5707 OPERATOR(HASHBRACK);
5709 /* This hack serves to disambiguate a pair of curlies
5710 * as being a block or an anon hash. Normally, expectation
5711 * determines that, but in cases where we're not in a
5712 * position to expect anything in particular (like inside
5713 * eval"") we have to resolve the ambiguity. This code
5714 * covers the case where the first term in the curlies is a
5715 * quoted string. Most other cases need to be explicitly
5716 * disambiguated by prepending a "+" before the opening
5717 * curly in order to force resolution as an anon hash.
5719 * XXX should probably propagate the outer expectation
5720 * into eval"" to rely less on this hack, but that could
5721 * potentially break current behavior of eval"".
5725 if (*s == '\'' || *s == '"' || *s == '`') {
5726 /* common case: get past first string, handling escapes */
5727 for (t++; t < PL_bufend && *t != *s;)
5728 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5732 else if (*s == 'q') {
5735 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5738 /* skip q//-like construct */
5740 char open, close, term;
5743 while (t < PL_bufend && isSPACE(*t))
5745 /* check for q => */
5746 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5747 OPERATOR(HASHBRACK);
5751 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5755 for (t++; t < PL_bufend; t++) {
5756 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5758 else if (*t == open)
5762 for (t++; t < PL_bufend; t++) {
5763 if (*t == '\\' && t+1 < PL_bufend)
5765 else if (*t == close && --brackets <= 0)
5767 else if (*t == open)
5774 /* skip plain q word */
5775 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5778 else if (isALNUM_lazy_if(t,UTF)) {
5780 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5783 while (t < PL_bufend && isSPACE(*t))
5785 /* if comma follows first term, call it an anon hash */
5786 /* XXX it could be a comma expression with loop modifiers */
5787 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5788 || (*t == '=' && t[1] == '>')))
5789 OPERATOR(HASHBRACK);
5790 if (PL_expect == XREF)
5793 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5799 pl_yylval.ival = CopLINE(PL_curcop);
5800 if (isSPACE(*s) || *s == '#')
5801 PL_copline = NOLINE; /* invalidate current command line number */
5804 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5808 if (PL_lex_brackets <= 0)
5809 yyerror("Unmatched right curly bracket");
5811 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5812 PL_lex_allbrackets--;
5813 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5814 PL_lex_formbrack = 0;
5815 if (PL_lex_state == LEX_INTERPNORMAL) {
5816 if (PL_lex_brackets == 0) {
5817 if (PL_expect & XFAKEBRACK) {
5818 PL_expect &= XENUMMASK;
5819 PL_lex_state = LEX_INTERPEND;
5824 PL_thiswhite = newSVpvs("");
5825 sv_catpvs(PL_thiswhite,"}");
5828 return yylex(); /* ignore fake brackets */
5830 if (*s == '-' && s[1] == '>')
5831 PL_lex_state = LEX_INTERPENDMAYBE;
5832 else if (*s != '[' && *s != '{')
5833 PL_lex_state = LEX_INTERPEND;
5836 if (PL_expect & XFAKEBRACK) {
5837 PL_expect &= XENUMMASK;
5839 return yylex(); /* ignore fake brackets */
5841 start_force(PL_curforce);
5843 curmad('X', newSVpvn(s-1,1));
5844 CURMAD('_', PL_thiswhite);
5849 PL_thistoken = newSVpvs("");
5855 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5856 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5863 if (PL_expect == XOPERATOR) {
5864 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5865 && isIDFIRST_lazy_if(s,UTF))
5867 CopLINE_dec(PL_curcop);
5868 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5869 CopLINE_inc(PL_curcop);
5871 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5872 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5879 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5881 PL_expect = XOPERATOR;
5882 force_ident(PL_tokenbuf, '&');
5886 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5892 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5893 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5900 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5901 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5909 const char tmp = *s++;
5911 if (!PL_lex_allbrackets &&
5912 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5919 if (!PL_lex_allbrackets &&
5920 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
5928 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5929 && strchr("+-*/%.^&|<",tmp))
5930 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5931 "Reversed %c= operator",(int)tmp);
5933 if (PL_expect == XSTATE && isALPHA(tmp) &&
5934 (s == PL_linestart+1 || s[-2] == '\n') )
5936 if (PL_in_eval && !PL_rsfp && !PL_parser->filtered) {
5941 if (strnEQ(s,"=cut",4)) {
5957 PL_thiswhite = newSVpvs("");
5958 sv_catpvn(PL_thiswhite, PL_linestart,
5959 PL_bufend - PL_linestart);
5963 PL_parser->in_pod = 1;
5967 if (PL_lex_brackets < PL_lex_formbrack) {
5969 #ifdef PERL_STRICT_CR
5970 while (SPACE_OR_TAB(*t))
5972 while (SPACE_OR_TAB(*t) || *t == '\r')
5975 if (*t == '\n' || *t == '#') {
5981 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5990 const char tmp = *s++;
5992 /* was this !=~ where !~ was meant?
5993 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5995 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5996 const char *t = s+1;
5998 while (t < PL_bufend && isSPACE(*t))
6001 if (*t == '/' || *t == '?' ||
6002 ((*t == 'm' || *t == 's' || *t == 'y')
6003 && !isALNUM(t[1])) ||
6004 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
6005 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6006 "!=~ should be !~");
6008 if (!PL_lex_allbrackets &&
6009 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6021 if (PL_expect != XOPERATOR) {
6022 if (s[1] != '<' && !strchr(s,'>'))
6025 s = scan_heredoc(s);
6027 s = scan_inputsymbol(s);
6028 TERM(sublex_start());
6034 if (*s == '=' && !PL_lex_allbrackets &&
6035 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6039 SHop(OP_LEFT_SHIFT);
6044 if (!PL_lex_allbrackets &&
6045 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6052 if (!PL_lex_allbrackets &&
6053 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6061 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6069 const char tmp = *s++;
6071 if (*s == '=' && !PL_lex_allbrackets &&
6072 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6076 SHop(OP_RIGHT_SHIFT);
6078 else if (tmp == '=') {
6079 if (!PL_lex_allbrackets &&
6080 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6088 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6097 if (PL_expect == XOPERATOR) {
6098 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6099 return deprecate_commaless_var_list();
6103 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
6104 PL_tokenbuf[0] = '@';
6105 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
6106 sizeof PL_tokenbuf - 1, FALSE);
6107 if (PL_expect == XOPERATOR)
6108 no_op("Array length", s);
6109 if (!PL_tokenbuf[1])
6111 PL_expect = XOPERATOR;
6112 PL_pending_ident = '#';
6116 PL_tokenbuf[0] = '$';
6117 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
6118 sizeof PL_tokenbuf - 1, FALSE);
6119 if (PL_expect == XOPERATOR)
6121 if (!PL_tokenbuf[1]) {
6123 yyerror("Final $ should be \\$ or $name");
6129 const char tmp = *s;
6130 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6133 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6134 && intuit_more(s)) {
6136 PL_tokenbuf[0] = '@';
6137 if (ckWARN(WARN_SYNTAX)) {
6140 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
6143 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6144 while (t < PL_bufend && *t != ']')
6146 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6147 "Multidimensional syntax %.*s not supported",
6148 (int)((t - PL_bufptr) + 1), PL_bufptr);
6152 else if (*s == '{') {
6154 PL_tokenbuf[0] = '%';
6155 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6156 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6158 char tmpbuf[sizeof PL_tokenbuf];
6161 } while (isSPACE(*t));
6162 if (isIDFIRST_lazy_if(t,UTF)) {
6164 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6168 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
6169 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6170 "You need to quote \"%s\"",
6177 PL_expect = XOPERATOR;
6178 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6179 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6180 if (!islop || PL_last_lop_op == OP_GREPSTART)
6181 PL_expect = XOPERATOR;
6182 else if (strchr("$@\"'`q", *s))
6183 PL_expect = XTERM; /* e.g. print $fh "foo" */
6184 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6185 PL_expect = XTERM; /* e.g. print $fh &sub */
6186 else if (isIDFIRST_lazy_if(s,UTF)) {
6187 char tmpbuf[sizeof PL_tokenbuf];
6189 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6190 if ((t2 = keyword(tmpbuf, len, 0))) {
6191 /* binary operators exclude handle interpretations */
6203 PL_expect = XTERM; /* e.g. print $fh length() */
6208 PL_expect = XTERM; /* e.g. print $fh subr() */
6211 else if (isDIGIT(*s))
6212 PL_expect = XTERM; /* e.g. print $fh 3 */
6213 else if (*s == '.' && isDIGIT(s[1]))
6214 PL_expect = XTERM; /* e.g. print $fh .3 */
6215 else if ((*s == '?' || *s == '-' || *s == '+')
6216 && !isSPACE(s[1]) && s[1] != '=')
6217 PL_expect = XTERM; /* e.g. print $fh -1 */
6218 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6220 PL_expect = XTERM; /* e.g. print $fh /.../
6221 XXX except DORDOR operator
6223 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6225 PL_expect = XTERM; /* print $fh <<"EOF" */
6228 PL_pending_ident = '$';
6232 if (PL_expect == XOPERATOR)
6234 PL_tokenbuf[0] = '@';
6235 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6236 if (!PL_tokenbuf[1]) {
6239 if (PL_lex_state == LEX_NORMAL)
6241 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6243 PL_tokenbuf[0] = '%';
6245 /* Warn about @ where they meant $. */
6246 if (*s == '[' || *s == '{') {
6247 if (ckWARN(WARN_SYNTAX)) {
6248 const char *t = s + 1;
6249 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
6251 if (*t == '}' || *t == ']') {
6253 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6254 /* diag_listed_as: Scalar value @%s[%s] better written as $%s[%s] */
6255 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6256 "Scalar value %.*s better written as $%.*s",
6257 (int)(t-PL_bufptr), PL_bufptr,
6258 (int)(t-PL_bufptr-1), PL_bufptr+1);
6263 PL_pending_ident = '@';
6266 case '/': /* may be division, defined-or, or pattern */
6267 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
6268 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6269 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6274 case '?': /* may either be conditional or pattern */
6275 if (PL_expect == XOPERATOR) {
6278 if (!PL_lex_allbrackets &&
6279 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6283 PL_lex_allbrackets++;
6289 /* A // operator. */
6290 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6291 (*s == '=' ? LEX_FAKEEOF_ASSIGN :
6292 LEX_FAKEEOF_LOGIC)) {
6300 if (*s == '=' && !PL_lex_allbrackets &&
6301 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6310 /* Disable warning on "study /blah/" */
6311 if (PL_oldoldbufptr == PL_last_uni
6312 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6313 || memNE(PL_last_uni, "study", 5)
6314 || isALNUM_lazy_if(PL_last_uni+5,UTF)
6318 deprecate("?PATTERN? without explicit operator");
6319 s = scan_pat(s,OP_MATCH);
6320 TERM(sublex_start());
6324 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6325 #ifdef PERL_STRICT_CR
6328 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6330 && (s == PL_linestart || s[-1] == '\n') )
6332 PL_lex_formbrack = 0;
6336 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6340 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6343 if (!PL_lex_allbrackets &&
6344 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6351 pl_yylval.ival = OPf_SPECIAL;
6357 if (*s == '=' && !PL_lex_allbrackets &&
6358 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6365 case '0': case '1': case '2': case '3': case '4':
6366 case '5': case '6': case '7': case '8': case '9':
6367 s = scan_num(s, &pl_yylval);
6368 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6369 if (PL_expect == XOPERATOR)
6374 s = scan_str(s,!!PL_madskills,FALSE);
6375 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6376 if (PL_expect == XOPERATOR) {
6377 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6378 return deprecate_commaless_var_list();
6385 pl_yylval.ival = OP_CONST;
6386 TERM(sublex_start());
6389 s = scan_str(s,!!PL_madskills,FALSE);
6390 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6391 if (PL_expect == XOPERATOR) {
6392 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6393 return deprecate_commaless_var_list();
6400 pl_yylval.ival = OP_CONST;
6401 /* FIXME. I think that this can be const if char *d is replaced by
6402 more localised variables. */
6403 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6404 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6405 pl_yylval.ival = OP_STRINGIFY;
6409 TERM(sublex_start());
6412 s = scan_str(s,!!PL_madskills,FALSE);
6413 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6414 if (PL_expect == XOPERATOR)
6415 no_op("Backticks",s);
6418 readpipe_override();
6419 TERM(sublex_start());
6423 if (PL_lex_inwhat && isDIGIT(*s))
6424 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6426 if (PL_expect == XOPERATOR)
6427 no_op("Backslash",s);
6431 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6432 char *start = s + 2;
6433 while (isDIGIT(*start) || *start == '_')
6435 if (*start == '.' && isDIGIT(start[1])) {
6436 s = scan_num(s, &pl_yylval);
6439 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6440 else if (!isALPHA(*start) && (PL_expect == XTERM
6441 || PL_expect == XREF || PL_expect == XSTATE
6442 || PL_expect == XTERMORDORDOR)) {
6443 GV *const gv = gv_fetchpvn_flags(s, start - s,
6444 UTF ? SVf_UTF8 : 0, SVt_PVCV);
6446 s = scan_num(s, &pl_yylval);
6453 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6496 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6498 /* Some keywords can be followed by any delimiter, including ':' */
6499 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6501 /* x::* is just a word, unless x is "CORE" */
6502 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6506 while (d < PL_bufend && isSPACE(*d))
6507 d++; /* no comments skipped here, or s### is misparsed */
6509 /* Is this a word before a => operator? */
6510 if (*d == '=' && d[1] == '>') {
6513 = (OP*)newSVOP(OP_CONST, 0,
6514 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6515 pl_yylval.opval->op_private = OPpCONST_BARE;
6519 /* Check for plugged-in keyword */
6523 char *saved_bufptr = PL_bufptr;
6525 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6527 if (result == KEYWORD_PLUGIN_DECLINE) {
6528 /* not a plugged-in keyword */
6529 PL_bufptr = saved_bufptr;
6530 } else if (result == KEYWORD_PLUGIN_STMT) {
6531 pl_yylval.opval = o;
6534 return REPORT(PLUGSTMT);
6535 } else if (result == KEYWORD_PLUGIN_EXPR) {
6536 pl_yylval.opval = o;
6538 PL_expect = XOPERATOR;
6539 return REPORT(PLUGEXPR);
6541 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6546 /* Check for built-in keyword */
6547 tmp = keyword(PL_tokenbuf, len, 0);
6549 /* Is this a label? */
6550 if (!anydelim && PL_expect == XSTATE
6551 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6553 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
6558 if (tmp < 0) { /* second-class keyword? */
6559 GV *ogv = NULL; /* override (winner) */
6560 GV *hgv = NULL; /* hidden (loser) */
6561 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6563 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6564 UTF ? SVf_UTF8 : 0, SVt_PVCV)) &&
6567 if (GvIMPORTED_CV(gv))
6569 else if (! CvMETHOD(cv))
6573 (gvp = (GV**)hv_fetch(PL_globalstash, PL_tokenbuf,
6574 UTF ? -(I32)len : (I32)len, FALSE)) &&
6575 (gv = *gvp) && isGV_with_GP(gv) &&
6576 GvCVu(gv) && GvIMPORTED_CV(gv))
6583 tmp = 0; /* overridden by import or by GLOBAL */
6586 && -tmp==KEY_lock /* XXX generalizable kludge */
6589 tmp = 0; /* any sub overrides "weak" keyword */
6591 else { /* no override */
6593 if (tmp == KEY_dump) {
6594 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6595 "dump() better written as CORE::dump()");
6599 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6600 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6601 "Ambiguous call resolved as CORE::%s(), "
6602 "qualify as such or use &",
6610 default: /* not a keyword */
6611 /* Trade off - by using this evil construction we can pull the
6612 variable gv into the block labelled keylookup. If not, then
6613 we have to give it function scope so that the goto from the
6614 earlier ':' case doesn't bypass the initialisation. */
6616 just_a_word_zero_gv:
6624 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6628 SV *nextPL_nextwhite = 0;
6632 /* Get the rest if it looks like a package qualifier */
6634 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6636 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6639 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
6640 *s == '\'' ? "'" : "::");
6645 if (PL_expect == XOPERATOR) {
6646 if (PL_bufptr == PL_linestart) {
6647 CopLINE_dec(PL_curcop);
6648 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6649 CopLINE_inc(PL_curcop);
6652 no_op("Bareword",s);
6655 /* Look for a subroutine with this name in current package,
6656 unless name is "Foo::", in which case Foo is a bareword
6657 (and a package name). */
6659 if (len > 2 && !PL_madskills &&
6660 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6662 if (ckWARN(WARN_BAREWORD)
6663 && ! gv_fetchpvn_flags(PL_tokenbuf, len, UTF ? SVf_UTF8 : 0, SVt_PVHV))
6664 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6665 "Bareword \"%s\" refers to nonexistent package",
6668 PL_tokenbuf[len] = '\0';
6674 /* Mustn't actually add anything to a symbol table.
6675 But also don't want to "initialise" any placeholder
6676 constants that might already be there into full
6677 blown PVGVs with attached PVCV. */
6678 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6679 GV_NOADD_NOINIT | ( UTF ? SVf_UTF8 : 0 ),
6685 /* if we saw a global override before, get the right name */
6687 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
6688 len ? len : strlen(PL_tokenbuf));
6690 SV * const tmp_sv = sv;
6691 sv = newSVpvs("CORE::GLOBAL::");
6692 sv_catsv(sv, tmp_sv);
6693 SvREFCNT_dec(tmp_sv);
6697 if (PL_madskills && !PL_thistoken) {
6698 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6699 PL_thistoken = newSVpvn(start,s - start);
6700 PL_realtokenstart = s - SvPVX(PL_linestr);
6704 /* Presume this is going to be a bareword of some sort. */
6706 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6707 pl_yylval.opval->op_private = OPpCONST_BARE;
6709 /* And if "Foo::", then that's what it certainly is. */
6714 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
6715 const_op->op_private = OPpCONST_BARE;
6716 rv2cv_op = newCVREF(0, const_op);
6718 cv = rv2cv_op_cv(rv2cv_op, 0);
6720 /* See if it's the indirect object for a list operator. */
6722 if (PL_oldoldbufptr &&
6723 PL_oldoldbufptr < PL_bufptr &&
6724 (PL_oldoldbufptr == PL_last_lop
6725 || PL_oldoldbufptr == PL_last_uni) &&
6726 /* NO SKIPSPACE BEFORE HERE! */
6727 (PL_expect == XREF ||
6728 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6730 bool immediate_paren = *s == '(';
6732 /* (Now we can afford to cross potential line boundary.) */
6733 s = SKIPSPACE2(s,nextPL_nextwhite);
6735 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6738 /* Two barewords in a row may indicate method call. */
6740 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6741 (tmp = intuit_method(s, gv, cv))) {
6743 if (tmp == METHOD && !PL_lex_allbrackets &&
6744 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6745 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6749 /* If not a declared subroutine, it's an indirect object. */
6750 /* (But it's an indir obj regardless for sort.) */
6751 /* Also, if "_" follows a filetest operator, it's a bareword */
6754 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6756 (PL_last_lop_op != OP_MAPSTART &&
6757 PL_last_lop_op != OP_GREPSTART))))
6758 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6759 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6762 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6767 PL_expect = XOPERATOR;
6770 s = SKIPSPACE2(s,nextPL_nextwhite);
6771 PL_nextwhite = nextPL_nextwhite;
6776 /* Is this a word before a => operator? */
6777 if (*s == '=' && s[1] == '>' && !pkgname) {
6780 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6781 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6782 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6786 /* If followed by a paren, it's certainly a subroutine. */
6791 while (SPACE_OR_TAB(*d))
6793 if (*d == ')' && (sv = cv_const_sv(cv))) {
6800 PL_nextwhite = PL_thiswhite;
6803 start_force(PL_curforce);
6805 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6806 PL_expect = XOPERATOR;
6809 PL_nextwhite = nextPL_nextwhite;
6810 curmad('X', PL_thistoken);
6811 PL_thistoken = newSVpvs("");
6820 /* If followed by var or block, call it a method (unless sub) */
6822 if ((*s == '$' || *s == '{') && !cv) {
6824 PL_last_lop = PL_oldbufptr;
6825 PL_last_lop_op = OP_METHOD;
6826 if (!PL_lex_allbrackets &&
6827 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6828 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6832 /* If followed by a bareword, see if it looks like indir obj. */
6835 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6836 && (tmp = intuit_method(s, gv, cv))) {
6838 if (tmp == METHOD && !PL_lex_allbrackets &&
6839 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6840 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6844 /* Not a method, so call it a subroutine (if defined) */
6847 if (lastchar == '-')
6848 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6849 "Ambiguous use of -%s resolved as -&%s()",
6850 PL_tokenbuf, PL_tokenbuf);
6851 /* Check for a constant sub */
6852 if ((sv = cv_const_sv(cv))) {
6855 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6856 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6857 pl_yylval.opval->op_private = 0;
6858 pl_yylval.opval->op_flags |= OPf_SPECIAL;
6862 op_free(pl_yylval.opval);
6863 pl_yylval.opval = rv2cv_op;
6864 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6865 PL_last_lop = PL_oldbufptr;
6866 PL_last_lop_op = OP_ENTERSUB;
6867 /* Is there a prototype? */
6874 STRLEN protolen = CvPROTOLEN(cv);
6875 const char *proto = CvPROTO(cv);
6879 if ((optional = *proto == ';'))
6882 while (*proto == ';');
6886 *proto == '$' || *proto == '_'
6887 || *proto == '*' || *proto == '+'
6892 *proto == '\\' && proto[1] && proto[2] == '\0'
6895 UNIPROTO(UNIOPSUB,optional);
6896 if (*proto == '\\' && proto[1] == '[') {
6897 const char *p = proto + 2;
6898 while(*p && *p != ']')
6900 if(*p == ']' && !p[1])
6901 UNIPROTO(UNIOPSUB,optional);
6903 if (*proto == '&' && *s == '{') {
6905 sv_setpvs(PL_subname, "__ANON__");
6907 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6908 if (!PL_lex_allbrackets &&
6909 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6910 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6917 PL_nextwhite = PL_thiswhite;
6920 start_force(PL_curforce);
6921 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6924 PL_nextwhite = nextPL_nextwhite;
6925 curmad('X', PL_thistoken);
6926 PL_thistoken = newSVpvs("");
6929 if (!PL_lex_allbrackets &&
6930 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6931 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6936 /* Guess harder when madskills require "best effort". */
6937 if (PL_madskills && (!gv || !GvCVu(gv))) {
6938 int probable_sub = 0;
6939 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6941 else if (isALPHA(*s)) {
6945 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6946 if (!keyword(tmpbuf, tmplen, 0))
6949 while (d < PL_bufend && isSPACE(*d))
6951 if (*d == '=' && d[1] == '>')
6956 gv = gv_fetchpv(PL_tokenbuf, GV_ADD | ( UTF ? SVf_UTF8 : 0 ),
6958 op_free(pl_yylval.opval);
6959 pl_yylval.opval = rv2cv_op;
6960 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6961 PL_last_lop = PL_oldbufptr;
6962 PL_last_lop_op = OP_ENTERSUB;
6963 PL_nextwhite = PL_thiswhite;
6965 start_force(PL_curforce);
6966 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6968 PL_nextwhite = nextPL_nextwhite;
6969 curmad('X', PL_thistoken);
6970 PL_thistoken = newSVpvs("");
6972 if (!PL_lex_allbrackets &&
6973 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6974 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6978 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6981 if (!PL_lex_allbrackets &&
6982 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6983 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6988 /* Call it a bare word */
6990 if (PL_hints & HINT_STRICT_SUBS)
6991 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6994 /* after "print" and similar functions (corresponding to
6995 * "F? L" in opcode.pl), whatever wasn't already parsed as
6996 * a filehandle should be subject to "strict subs".
6997 * Likewise for the optional indirect-object argument to system
6998 * or exec, which can't be a bareword */
6999 if ((PL_last_lop_op == OP_PRINT
7000 || PL_last_lop_op == OP_PRTF
7001 || PL_last_lop_op == OP_SAY
7002 || PL_last_lop_op == OP_SYSTEM
7003 || PL_last_lop_op == OP_EXEC)
7004 && (PL_hints & HINT_STRICT_SUBS))
7005 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7006 if (lastchar != '-') {
7007 if (ckWARN(WARN_RESERVED)) {
7011 if (!*d && !gv_stashpv(PL_tokenbuf, UTF ? SVf_UTF8 : 0))
7012 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
7020 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
7021 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7022 "Operator or semicolon missing before %c%s",
7023 lastchar, PL_tokenbuf);
7024 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7025 "Ambiguous use of %c resolved as operator %c",
7026 lastchar, lastchar);
7033 (OP*)newSVOP(OP_CONST, 0, newSVpv(CopFILE(PL_curcop),0))
7038 (OP*)newSVOP(OP_CONST, 0,
7039 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
7042 case KEY___PACKAGE__:
7044 (OP*)newSVOP(OP_CONST, 0,
7046 ? newSVhek(HvNAME_HEK(PL_curstash))
7053 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
7054 const char *pname = "main";
7057 if (PL_tokenbuf[2] == 'D')
7060 PL_curstash ? PL_curstash : PL_defstash;
7061 pname = HvNAME_get(stash);
7062 plen = HvNAMELEN (stash);
7063 if(HvNAMEUTF8(stash)) putf8 = SVf_UTF8;
7065 gv = gv_fetchpvn_flags(
7066 Perl_form(aTHX_ "%*s::DATA", (int)plen, pname),
7067 plen+6, GV_ADD|putf8, SVt_PVIO
7071 GvIOp(gv) = newIO();
7072 IoIFP(GvIOp(gv)) = PL_rsfp;
7073 #if defined(HAS_FCNTL) && defined(F_SETFD)
7075 const int fd = PerlIO_fileno(PL_rsfp);
7076 fcntl(fd,F_SETFD,fd >= 3);
7079 /* Mark this internal pseudo-handle as clean */
7080 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
7081 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
7082 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
7084 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
7085 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
7086 /* if the script was opened in binmode, we need to revert
7087 * it to text mode for compatibility; but only iff it has CRs
7088 * XXX this is a questionable hack at best. */
7089 if (PL_bufend-PL_bufptr > 2
7090 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
7093 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
7094 loc = PerlIO_tell(PL_rsfp);
7095 (void)PerlIO_seek(PL_rsfp, 0L, 0);
7098 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
7100 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
7101 #endif /* NETWARE */
7103 PerlIO_seek(PL_rsfp, loc, 0);
7107 #ifdef PERLIO_LAYERS
7110 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
7111 else if (PL_encoding) {
7118 XPUSHs(PL_encoding);
7120 call_method("name", G_SCALAR);
7124 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7125 Perl_form(aTHX_ ":encoding(%"SVf")",
7134 if (PL_realtokenstart >= 0) {
7135 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7137 PL_endwhite = newSVpvs("");
7138 sv_catsv(PL_endwhite, PL_thiswhite);
7140 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
7141 PL_realtokenstart = -1;
7143 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
7153 FUN0OP(newPVOP(OP_RUNCV,0,NULL));
7162 if (PL_expect == XSTATE) {
7169 if (*s == ':' && s[1] == ':') {
7172 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7173 if (!(tmp = keyword(PL_tokenbuf, len, 1)))
7174 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
7177 else if (tmp == KEY_require || tmp == KEY_do
7179 /* that's a way to remember we saw "CORE::" */
7192 LOP(OP_ACCEPT,XTERM);
7195 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7200 LOP(OP_ATAN2,XTERM);
7206 LOP(OP_BINMODE,XTERM);
7209 LOP(OP_BLESS,XTERM);
7218 /* We have to disambiguate the two senses of
7219 "continue". If the next token is a '{' then
7220 treat it as the start of a continue block;
7221 otherwise treat it as a control operator.
7231 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7241 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7250 if (!PL_cryptseen) {
7251 PL_cryptseen = TRUE;
7255 LOP(OP_CRYPT,XTERM);
7258 LOP(OP_CHMOD,XTERM);
7261 LOP(OP_CHOWN,XTERM);
7264 LOP(OP_CONNECT,XTERM);
7283 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7284 if (orig_keyword == KEY_do) {
7293 PL_hints |= HINT_BLOCK_SCOPE;
7303 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7304 STR_WITH_LEN("NDBM_File::"),
7305 STR_WITH_LEN("DB_File::"),
7306 STR_WITH_LEN("GDBM_File::"),
7307 STR_WITH_LEN("SDBM_File::"),
7308 STR_WITH_LEN("ODBM_File::"),
7310 LOP(OP_DBMOPEN,XTERM);
7316 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7323 pl_yylval.ival = CopLINE(PL_curcop);
7327 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7341 if (*s == '{') { /* block eval */
7342 PL_expect = XTERMBLOCK;
7343 UNIBRACK(OP_ENTERTRY);
7345 else { /* string eval */
7347 UNIBRACK(OP_ENTEREVAL);
7352 UNIBRACK(-OP_ENTEREVAL);
7366 case KEY_endhostent:
7372 case KEY_endservent:
7375 case KEY_endprotoent:
7386 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7388 pl_yylval.ival = CopLINE(PL_curcop);
7390 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7393 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
7396 if ((PL_bufend - p) >= 3 &&
7397 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7399 else if ((PL_bufend - p) >= 4 &&
7400 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7403 if (isIDFIRST_lazy_if(p,UTF)) {
7404 p = scan_ident(p, PL_bufend,
7405 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
7409 Perl_croak(aTHX_ "Missing $ on loop variable");
7411 s = SvPVX(PL_linestr) + soff;
7417 LOP(OP_FORMLINE,XTERM);
7426 LOP(OP_FCNTL,XTERM);
7432 LOP(OP_FLOCK,XTERM);
7435 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7440 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7445 LOP(OP_GREPSTART, XREF);
7448 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7463 case KEY_getpriority:
7464 LOP(OP_GETPRIORITY,XTERM);
7466 case KEY_getprotobyname:
7469 case KEY_getprotobynumber:
7470 LOP(OP_GPBYNUMBER,XTERM);
7472 case KEY_getprotoent:
7484 case KEY_getpeername:
7485 UNI(OP_GETPEERNAME);
7487 case KEY_gethostbyname:
7490 case KEY_gethostbyaddr:
7491 LOP(OP_GHBYADDR,XTERM);
7493 case KEY_gethostent:
7496 case KEY_getnetbyname:
7499 case KEY_getnetbyaddr:
7500 LOP(OP_GNBYADDR,XTERM);
7505 case KEY_getservbyname:
7506 LOP(OP_GSBYNAME,XTERM);
7508 case KEY_getservbyport:
7509 LOP(OP_GSBYPORT,XTERM);
7511 case KEY_getservent:
7514 case KEY_getsockname:
7515 UNI(OP_GETSOCKNAME);
7517 case KEY_getsockopt:
7518 LOP(OP_GSOCKOPT,XTERM);
7533 pl_yylval.ival = CopLINE(PL_curcop);
7538 orig_keyword==KEY_glob ? (orig_keyword=0, -OP_GLOB) : OP_GLOB,
7546 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7548 pl_yylval.ival = CopLINE(PL_curcop);
7552 LOP(OP_INDEX,XTERM);
7558 LOP(OP_IOCTL,XTERM);
7570 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7587 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7592 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7606 LOP(OP_LISTEN,XTERM);
7615 s = scan_pat(s,OP_MATCH);
7616 TERM(sublex_start());
7619 LOP(OP_MAPSTART, XREF);
7622 LOP(OP_MKDIR,XTERM);
7625 LOP(OP_MSGCTL,XTERM);
7628 LOP(OP_MSGGET,XTERM);
7631 LOP(OP_MSGRCV,XTERM);
7634 LOP(OP_MSGSND,XTERM);
7639 PL_in_my = (U16)tmp;
7641 if (isIDFIRST_lazy_if(s,UTF)) {
7645 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7646 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7648 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7649 if (!PL_in_my_stash) {
7652 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7656 if (PL_madskills) { /* just add type to declarator token */
7657 sv_catsv(PL_thistoken, PL_nextwhite);
7659 sv_catpvn(PL_thistoken, start, s - start);
7667 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7671 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7676 s = tokenize_use(0, s);
7680 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7683 if (!PL_lex_allbrackets &&
7684 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7685 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7691 if (isIDFIRST_lazy_if(s,UTF)) {
7693 for (d = s; isALNUM_lazy_if(d,UTF);)
7695 for (t=d; isSPACE(*t);)
7697 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7699 && !(t[0] == '=' && t[1] == '>')
7700 && !(t[0] == ':' && t[1] == ':')
7701 && !keyword(s, d-s, 0)
7703 int parms_len = (int)(d-s);
7704 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7705 "Precedence problem: open %.*s should be open(%.*s)",
7706 parms_len, s, parms_len, s);
7712 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7714 pl_yylval.ival = OP_OR;
7724 LOP(OP_OPEN_DIR,XTERM);
7727 checkcomma(s,PL_tokenbuf,"filehandle");
7731 checkcomma(s,PL_tokenbuf,"filehandle");
7750 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7752 s = force_strict_version(s);
7753 PL_lex_expect = XBLOCK;
7757 LOP(OP_PIPE_OP,XTERM);
7760 s = scan_str(s,!!PL_madskills,FALSE);
7763 pl_yylval.ival = OP_CONST;
7764 TERM(sublex_start());
7771 s = scan_str(s,!!PL_madskills,FALSE);
7774 PL_expect = XOPERATOR;
7775 if (SvCUR(PL_lex_stuff)) {
7776 int warned_comma = !ckWARN(WARN_QW);
7777 int warned_comment = warned_comma;
7778 d = SvPV_force(PL_lex_stuff, len);
7780 for (; isSPACE(*d) && len; --len, ++d)
7785 if (!warned_comma || !warned_comment) {
7786 for (; !isSPACE(*d) && len; --len, ++d) {
7787 if (!warned_comma && *d == ',') {
7788 Perl_warner(aTHX_ packWARN(WARN_QW),
7789 "Possible attempt to separate words with commas");
7792 else if (!warned_comment && *d == '#') {
7793 Perl_warner(aTHX_ packWARN(WARN_QW),
7794 "Possible attempt to put comments in qw() list");
7800 for (; !isSPACE(*d) && len; --len, ++d)
7803 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7804 words = op_append_elem(OP_LIST, words,
7805 newSVOP(OP_CONST, 0, tokeq(sv)));
7810 words = newNULLLIST();
7812 SvREFCNT_dec(PL_lex_stuff);
7813 PL_lex_stuff = NULL;
7815 PL_expect = XOPERATOR;
7816 pl_yylval.opval = sawparens(words);
7821 s = scan_str(s,!!PL_madskills,FALSE);
7824 pl_yylval.ival = OP_STRINGIFY;
7825 if (SvIVX(PL_lex_stuff) == '\'')
7826 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
7827 TERM(sublex_start());
7830 s = scan_pat(s,OP_QR);
7831 TERM(sublex_start());
7834 s = scan_str(s,!!PL_madskills,FALSE);
7837 readpipe_override();
7838 TERM(sublex_start());
7846 s = force_version(s, FALSE);
7848 else if (*s != 'v' || !isDIGIT(s[1])
7849 || (s = force_version(s, TRUE), *s == 'v'))
7851 *PL_tokenbuf = '\0';
7852 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7853 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7854 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf),
7855 GV_ADD | (UTF ? SVf_UTF8 : 0));
7857 yyerror("<> should be quotes");
7859 if (orig_keyword == KEY_require) {
7867 PL_last_uni = PL_oldbufptr;
7868 PL_last_lop_op = OP_REQUIRE;
7870 return REPORT( (int)REQUIRE );
7876 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7880 LOP(OP_RENAME,XTERM);
7889 LOP(OP_RINDEX,XTERM);
7898 UNIDOR(OP_READLINE);
7901 UNIDOR(OP_BACKTICK);
7910 LOP(OP_REVERSE,XTERM);
7913 UNIDOR(OP_READLINK);
7920 if (pl_yylval.opval)
7921 TERM(sublex_start());
7923 TOKEN(1); /* force error */
7926 checkcomma(s,PL_tokenbuf,"filehandle");
7936 LOP(OP_SELECT,XTERM);
7942 LOP(OP_SEMCTL,XTERM);
7945 LOP(OP_SEMGET,XTERM);
7948 LOP(OP_SEMOP,XTERM);
7954 LOP(OP_SETPGRP,XTERM);
7956 case KEY_setpriority:
7957 LOP(OP_SETPRIORITY,XTERM);
7959 case KEY_sethostent:
7965 case KEY_setservent:
7968 case KEY_setprotoent:
7978 LOP(OP_SEEKDIR,XTERM);
7980 case KEY_setsockopt:
7981 LOP(OP_SSOCKOPT,XTERM);
7987 LOP(OP_SHMCTL,XTERM);
7990 LOP(OP_SHMGET,XTERM);
7993 LOP(OP_SHMREAD,XTERM);
7996 LOP(OP_SHMWRITE,XTERM);
7999 LOP(OP_SHUTDOWN,XTERM);
8008 LOP(OP_SOCKET,XTERM);
8010 case KEY_socketpair:
8011 LOP(OP_SOCKPAIR,XTERM);
8014 checkcomma(s,PL_tokenbuf,"subroutine name");
8017 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8021 LOP(OP_SPLIT,XTERM);
8024 LOP(OP_SPRINTF,XTERM);
8027 LOP(OP_SPLICE,XTERM);
8042 LOP(OP_SUBSTR,XTERM);
8048 char tmpbuf[sizeof PL_tokenbuf];
8049 SSize_t tboffset = 0;
8050 expectation attrful;
8051 bool have_name, have_proto;
8052 const int key = tmp;
8057 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
8058 SV *subtoken = newSVpvn_flags(tstart, s - tstart, SvUTF8(PL_linestr));
8062 s = SKIPSPACE2(s,tmpwhite);
8067 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
8068 (*s == ':' && s[1] == ':'))
8071 SV *nametoke = NULL;
8075 attrful = XATTRBLOCK;
8076 /* remember buffer pos'n for later force_word */
8077 tboffset = s - PL_oldbufptr;
8078 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
8081 nametoke = newSVpvn_flags(s, d - s, SvUTF8(PL_linestr));
8083 if (memchr(tmpbuf, ':', len))
8084 sv_setpvn(PL_subname, tmpbuf, len);
8086 sv_setsv(PL_subname,PL_curstname);
8087 sv_catpvs(PL_subname,"::");
8088 sv_catpvn(PL_subname,tmpbuf,len);
8090 if (SvUTF8(PL_linestr))
8091 SvUTF8_on(PL_subname);
8097 CURMAD('X', nametoke);
8098 CURMAD('_', tmpwhite);
8099 (void) force_word(PL_oldbufptr + tboffset, WORD,
8102 s = SKIPSPACE2(d,tmpwhite);
8109 Perl_croak(aTHX_ "Missing name in \"my sub\"");
8110 PL_expect = XTERMBLOCK;
8111 attrful = XATTRTERM;
8112 sv_setpvs(PL_subname,"?");
8116 if (key == KEY_format) {
8118 PL_lex_formbrack = PL_lex_brackets + 1;
8120 PL_thistoken = subtoken;
8124 (void) force_word(PL_oldbufptr + tboffset, WORD,
8130 /* Look for a prototype */
8133 bool bad_proto = FALSE;
8134 bool in_brackets = FALSE;
8135 char greedy_proto = ' ';
8136 bool proto_after_greedy_proto = FALSE;
8137 bool must_be_last = FALSE;
8138 bool underscore = FALSE;
8139 bool seen_underscore = FALSE;
8140 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
8143 s = scan_str(s,!!PL_madskills,FALSE);
8145 Perl_croak(aTHX_ "Prototype not terminated");
8146 /* strip spaces and check for bad characters */
8147 d = SvPV(PL_lex_stuff, tmplen);
8149 for (p = d; tmplen; tmplen--, ++p) {
8153 if (warnillegalproto) {
8155 proto_after_greedy_proto = TRUE;
8156 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
8161 if ( !strchr(";@%", *p) )
8168 else if ( *p == ']' ) {
8169 in_brackets = FALSE;
8171 else if ( (*p == '@' || *p == '%') &&
8172 ( tmp < 2 || d[tmp-2] != '\\' ) &&
8174 must_be_last = TRUE;
8177 else if ( *p == '_' ) {
8178 underscore = seen_underscore = TRUE;
8185 if (proto_after_greedy_proto)
8186 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8187 "Prototype after '%c' for %"SVf" : %s",
8188 greedy_proto, SVfARG(PL_subname), d);
8190 SV *dsv = newSVpvs_flags("", SVs_TEMP);
8191 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8192 "Illegal character %sin prototype for %"SVf" : %s",
8193 seen_underscore ? "after '_' " : "",
8196 newSVpvn_flags(d, tmp, SVs_TEMP | SvUTF8(PL_lex_stuff)),
8197 tmp, UNI_DISPLAY_ISPRINT));
8199 SvCUR_set(PL_lex_stuff, tmp);
8204 CURMAD('q', PL_thisopen);
8205 CURMAD('_', tmpwhite);
8206 CURMAD('=', PL_thisstuff);
8207 CURMAD('Q', PL_thisclose);
8208 NEXTVAL_NEXTTOKE.opval =
8209 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8210 PL_lex_stuff = NULL;
8213 s = SKIPSPACE2(s,tmpwhite);
8221 if (*s == ':' && s[1] != ':')
8222 PL_expect = attrful;
8223 else if (*s != '{' && key == KEY_sub) {
8225 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8226 else if (*s != ';' && *s != '}')
8227 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8234 curmad('^', newSVpvs(""));
8235 CURMAD('_', tmpwhite);
8239 PL_thistoken = subtoken;
8242 NEXTVAL_NEXTTOKE.opval =
8243 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8244 PL_lex_stuff = NULL;
8250 sv_setpvs(PL_subname, "__ANON__");
8252 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8256 (void) force_word(PL_oldbufptr + tboffset, WORD,
8265 LOP(OP_SYSTEM,XREF);
8268 LOP(OP_SYMLINK,XTERM);
8271 LOP(OP_SYSCALL,XTERM);
8274 LOP(OP_SYSOPEN,XTERM);
8277 LOP(OP_SYSSEEK,XTERM);
8280 LOP(OP_SYSREAD,XTERM);
8283 LOP(OP_SYSWRITE,XTERM);
8287 TERM(sublex_start());
8308 LOP(OP_TRUNCATE,XTERM);
8320 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8322 pl_yylval.ival = CopLINE(PL_curcop);
8326 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8328 pl_yylval.ival = CopLINE(PL_curcop);
8332 LOP(OP_UNLINK,XTERM);
8338 LOP(OP_UNPACK,XTERM);
8341 LOP(OP_UTIME,XTERM);
8347 LOP(OP_UNSHIFT,XTERM);
8350 s = tokenize_use(1, s);
8360 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8362 pl_yylval.ival = CopLINE(PL_curcop);
8366 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8368 pl_yylval.ival = CopLINE(PL_curcop);
8372 PL_hints |= HINT_BLOCK_SCOPE;
8379 LOP(OP_WAITPID,XTERM);
8388 ctl_l[0] = toCTRL('L');
8390 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
8393 /* Make sure $^L is defined */
8394 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
8399 if (PL_expect == XOPERATOR) {
8400 if (*s == '=' && !PL_lex_allbrackets &&
8401 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8409 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8411 pl_yylval.ival = OP_XOR;
8416 TERM(sublex_start());
8421 #pragma segment Main
8425 S_pending_ident(pTHX)
8430 /* pit holds the identifier we read and pending_ident is reset */
8431 char pit = PL_pending_ident;
8432 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8433 /* All routes through this function want to know if there is a colon. */
8434 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8435 PL_pending_ident = 0;
8437 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
8438 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8439 "### Pending identifier '%s'\n", PL_tokenbuf); });
8441 /* if we're in a my(), we can't allow dynamics here.
8442 $foo'bar has already been turned into $foo::bar, so
8443 just check for colons.
8445 if it's a legal name, the OP is a PADANY.
8448 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8450 yyerror(Perl_form(aTHX_ "No package name allowed for "
8451 "variable %s in \"our\"",
8453 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8457 yyerror(Perl_form(aTHX_ PL_no_myglob,
8458 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
8460 pl_yylval.opval = newOP(OP_PADANY, 0);
8461 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8462 UTF ? SVf_UTF8 : 0);
8468 build the ops for accesses to a my() variable.
8470 Deny my($a) or my($b) in a sort block, *if* $a or $b is
8471 then used in a comparison. This catches most, but not
8472 all cases. For instance, it catches
8473 sort { my($a); $a <=> $b }
8475 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
8476 (although why you'd do that is anyone's guess).
8481 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8482 UTF ? SVf_UTF8 : 0);
8483 if (tmp != NOT_IN_PAD) {
8484 /* might be an "our" variable" */
8485 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8486 /* build ops for a bareword */
8487 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8488 HEK * const stashname = HvNAME_HEK(stash);
8489 SV * const sym = newSVhek(stashname);
8490 sv_catpvs(sym, "::");
8491 sv_catpvn_flags(sym, PL_tokenbuf+1, tokenbuf_len - 1, (UTF ? SV_CATUTF8 : SV_CATBYTES ));
8492 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8493 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8496 ? (GV_ADDMULTI | GV_ADDINEVAL)
8499 ((PL_tokenbuf[0] == '$') ? SVt_PV
8500 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8505 /* if it's a sort block and they're naming $a or $b */
8506 if (PL_last_lop_op == OP_SORT &&
8507 PL_tokenbuf[0] == '$' &&
8508 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
8511 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
8512 d < PL_bufend && *d != '\n';
8515 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
8516 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
8522 pl_yylval.opval = newOP(OP_PADANY, 0);
8523 pl_yylval.opval->op_targ = tmp;
8529 Whine if they've said @foo in a doublequoted string,
8530 and @foo isn't a variable we can find in the symbol
8533 if (ckWARN(WARN_AMBIGUOUS) &&
8534 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8535 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1,
8536 ( UTF ? SVf_UTF8 : 0 ), SVt_PVAV);
8537 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8538 /* DO NOT warn for @- and @+ */
8539 && !( PL_tokenbuf[2] == '\0' &&
8540 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8543 /* Downgraded from fatal to warning 20000522 mjd */
8544 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8545 "Possible unintended interpolation of %s in string",
8550 /* build ops for a bareword */
8551 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(PL_tokenbuf + 1,
8553 UTF ? SVf_UTF8 : 0 ));
8554 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8555 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8556 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD)
8557 | ( UTF ? SVf_UTF8 : 0 ),
8558 ((PL_tokenbuf[0] == '$') ? SVt_PV
8559 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8565 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8569 PERL_ARGS_ASSERT_CHECKCOMMA;
8571 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8572 if (ckWARN(WARN_SYNTAX)) {
8575 for (w = s+2; *w && level; w++) {
8583 /* the list of chars below is for end of statements or
8584 * block / parens, boolean operators (&&, ||, //) and branch
8585 * constructs (or, and, if, until, unless, while, err, for).
8586 * Not a very solid hack... */
8587 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8588 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8589 "%s (...) interpreted as function",name);
8592 while (s < PL_bufend && isSPACE(*s))
8596 while (s < PL_bufend && isSPACE(*s))
8598 if (isIDFIRST_lazy_if(s,UTF)) {
8599 const char * const w = s++;
8600 while (isALNUM_lazy_if(s,UTF))
8602 while (s < PL_bufend && isSPACE(*s))
8606 if (keyword(w, s - w, 0))
8609 gv = gv_fetchpvn_flags(w, s - w, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
8610 if (gv && GvCVu(gv))
8612 Perl_croak(aTHX_ "No comma allowed after %s", what);
8617 /* Either returns sv, or mortalizes sv and returns a new SV*.
8618 Best used as sv=new_constant(..., sv, ...).
8619 If s, pv are NULL, calls subroutine with one argument,
8620 and type is used with error messages only. */
8623 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
8624 SV *sv, SV *pv, const char *type, STRLEN typelen)
8627 HV * table = GvHV(PL_hintgv); /* ^H */
8631 const char *why1 = "", *why2 = "", *why3 = "";
8633 PERL_ARGS_ASSERT_NEW_CONSTANT;
8635 /* charnames doesn't work well if there have been errors found */
8636 if (PL_error_count > 0 && strEQ(key,"charnames"))
8637 return &PL_sv_undef;
8640 || ! (PL_hints & HINT_LOCALIZE_HH)
8641 || ! (cvp = hv_fetch(table, key, keylen, FALSE))
8646 /* Here haven't found what we're looking for. If it is charnames,
8647 * perhaps it needs to be loaded. Try doing that before giving up */
8648 if (strEQ(key,"charnames")) {
8649 Perl_load_module(aTHX_
8651 newSVpvs("_charnames"),
8652 /* version parameter; no need to specify it, as if
8653 * we get too early a version, will fail anyway,
8654 * not being able to find '_charnames' */
8660 table = GvHV(PL_hintgv);
8662 && (PL_hints & HINT_LOCALIZE_HH)
8663 && (cvp = hv_fetch(table, key, keylen, FALSE))
8669 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8670 msg = Perl_newSVpvf(aTHX_
8671 "Constant(%s) unknown", (type ? type: "undef"));
8676 why3 = "} is not defined";
8678 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
8679 (type ? type: "undef"), why1, why2, why3);
8681 yyerror(SvPVX_const(msg));
8686 sv_2mortal(sv); /* Parent created it permanently */
8689 pv = newSVpvn_flags(s, len, SVs_TEMP);
8691 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
8693 typesv = &PL_sv_undef;
8695 PUSHSTACKi(PERLSI_OVERLOAD);
8707 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
8711 /* Check the eval first */
8712 if (!PL_in_eval && SvTRUE(ERRSV)) {
8713 sv_catpvs(ERRSV, "Propagated");
8714 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
8716 res = SvREFCNT_inc_simple(sv);
8720 SvREFCNT_inc_simple_void(res);
8729 why1 = "Call to &{$^H{";
8731 why3 = "}} did not return a defined value";
8739 /* Returns a NUL terminated string, with the length of the string written to
8743 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
8746 register char *d = dest;
8747 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
8749 PERL_ARGS_ASSERT_SCAN_WORD;
8753 Perl_croak(aTHX_ ident_too_long);
8754 if (isALNUM(*s)) /* UTF handled below */
8756 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
8761 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
8765 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
8766 char *t = s + UTF8SKIP(s);
8768 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
8772 Perl_croak(aTHX_ ident_too_long);
8773 Copy(s, d, len, char);
8786 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
8789 char *bracket = NULL;
8791 register char *d = dest;
8792 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
8794 PERL_ARGS_ASSERT_SCAN_IDENT;
8799 while (isDIGIT(*s)) {
8801 Perl_croak(aTHX_ ident_too_long);
8808 Perl_croak(aTHX_ ident_too_long);
8809 if (isALNUM(*s)) /* UTF handled below */
8811 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
8816 else if (*s == ':' && s[1] == ':') {
8820 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
8821 char *t = s + UTF8SKIP(s);
8822 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
8824 if (d + (t - s) > e)
8825 Perl_croak(aTHX_ ident_too_long);
8826 Copy(s, d, t - s, char);
8837 if (PL_lex_state != LEX_NORMAL)
8838 PL_lex_state = LEX_INTERPENDMAYBE;
8841 if (*s == '$' && s[1] &&
8842 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
8854 const STRLEN skip = UTF8SKIP(s);
8857 for ( i = 0; i < skip; i++ )
8865 if (*d == '^' && *s && isCONTROLVAR(*s)) {
8870 if (isSPACE(s[-1])) {
8872 const char ch = *s++;
8873 if (!SPACE_OR_TAB(ch)) {
8879 if (isIDFIRST_lazy_if(d,UTF)) {
8883 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
8884 end += UTF8SKIP(end);
8885 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
8886 end += UTF8SKIP(end);
8888 Copy(s, d, end - s, char);
8893 while ((isALNUM(*s) || *s == ':') && d < e)
8896 Perl_croak(aTHX_ ident_too_long);
8899 while (s < send && SPACE_OR_TAB(*s))
8901 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
8902 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
8903 const char * const brack =
8905 ((*s == '[') ? "[...]" : "{...}");
8906 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
8907 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8908 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
8909 funny, dest, brack, funny, dest, brack);
8912 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
8913 PL_lex_allbrackets++;
8917 /* Handle extended ${^Foo} variables
8918 * 1999-02-27 mjd-perl-patch@plover.com */
8919 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
8923 while (isALNUM(*s) && d < e) {
8927 Perl_croak(aTHX_ ident_too_long);
8932 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
8933 PL_lex_state = LEX_INTERPEND;
8936 if (PL_lex_state == LEX_NORMAL) {
8937 if (ckWARN(WARN_AMBIGUOUS) &&
8938 (keyword(dest, d - dest, 0)
8939 || get_cvn_flags(dest, d - dest, 0)))
8943 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8944 "Ambiguous use of %c{%s} resolved to %c%s",
8945 funny, dest, funny, dest);
8950 s = bracket; /* let the parser handle it */
8954 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
8955 PL_lex_state = LEX_INTERPEND;
8960 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
8962 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
8963 * the parse starting at 's', based on the subset that are valid in this
8964 * context input to this routine in 'valid_flags'. Advances s. Returns
8965 * TRUE if the input was a valid flag, so the next char may be as well;
8966 * otherwise FALSE. 'charset' should point to a NUL upon first call on the
8967 * current regex. This routine will set it to any charset modifier found.
8968 * The caller shouldn't change it. This way, another charset modifier
8969 * encountered in the parse can be detected as an error, as we have decided
8974 if (! strchr(valid_flags, c)) {
8983 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
8984 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
8985 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
8986 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
8987 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
8988 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
8989 case LOCALE_PAT_MOD:
8991 /* In 5.14, qr//lt is legal but deprecated; the 't' means they
8992 * can't be regex modifiers.
8993 * In 5.14, s///le is legal and ambiguous. Try to disambiguate as
8994 * much as easily done. s///lei, for example, has to mean regex
8995 * modifiers if it's not an error (as does any word character
8996 * following the 'e'). Otherwise, we resolve to the backwards-
8997 * compatible, but less likely 's/// le ...', i.e. as meaning
8998 * less-than-or-equal. The reason it's not likely is that s//
8999 * returns a number for code in the field (/r returns a string, but
9000 * that wasn't added until the 5.13 series), and so '<=' should be
9001 * used for comparing, not 'le'. */
9002 if (*((*s) + 1) == 't') {
9005 else if (*((*s) + 1) == 'e' && ! isALNUM(*((*s) + 2))) {
9007 /* 'e' is valid only for substitutes, s///e. If it is not
9008 * valid in the current context, then 'm//le' must mean the
9009 * comparison operator, so use the regular deprecation message.
9011 if (! strchr(valid_flags, 'e')) {
9014 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
9015 "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");
9019 goto multiple_charsets;
9021 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
9024 case UNICODE_PAT_MOD:
9025 /* In 5.14, qr//unless and qr//until are legal but deprecated; the
9026 * 'n' means they can't be regex modifiers */
9027 if (*((*s) + 1) == 'n') {
9031 goto multiple_charsets;
9033 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
9036 case ASCII_RESTRICT_PAT_MOD:
9037 /* In 5.14, qr//and is legal but deprecated; the 'n' means they
9038 * can't be regex modifiers */
9039 if (*((*s) + 1) == 'n') {
9044 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
9048 /* Error if previous modifier wasn't an 'a', but if it was, see
9049 * if, and accept, a second occurrence (only) */
9051 || get_regex_charset(*pmfl)
9052 != REGEX_ASCII_RESTRICTED_CHARSET)
9054 goto multiple_charsets;
9056 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
9060 case DEPENDS_PAT_MOD:
9062 goto multiple_charsets;
9064 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
9073 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX),
9074 "Having no space between pattern and following word is deprecated");
9078 if (*charset != c) {
9079 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
9081 else if (c == 'a') {
9082 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
9085 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
9088 /* Pretend that it worked, so will continue processing before dieing */
9094 S_scan_pat(pTHX_ char *start, I32 type)
9098 char *s = scan_str(start,!!PL_madskills,FALSE);
9099 const char * const valid_flags =
9100 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
9101 char charset = '\0'; /* character set modifier */
9106 PERL_ARGS_ASSERT_SCAN_PAT;
9109 const char * const delimiter = skipspace(start);
9113 ? "Search pattern not terminated or ternary operator parsed as search pattern"
9114 : "Search pattern not terminated" ));
9117 pm = (PMOP*)newPMOP(type, 0);
9118 if (PL_multi_open == '?') {
9119 /* This is the only point in the code that sets PMf_ONCE: */
9120 pm->op_pmflags |= PMf_ONCE;
9122 /* Hence it's safe to do this bit of PMOP book-keeping here, which
9123 allows us to restrict the list needed by reset to just the ??
9125 assert(type != OP_TRANS);
9127 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
9130 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
9133 elements = mg->mg_len / sizeof(PMOP**);
9134 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
9135 ((PMOP**)mg->mg_ptr) [elements++] = pm;
9136 mg->mg_len = elements * sizeof(PMOP**);
9137 PmopSTASH_set(pm,PL_curstash);
9143 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
9145 if (PL_madskills && modstart != s) {
9146 SV* tmptoken = newSVpvn(modstart, s - modstart);
9147 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
9150 /* issue a warning if /c is specified,but /g is not */
9151 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
9153 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
9154 "Use of /c modifier is meaningless without /g" );
9157 PL_lex_op = (OP*)pm;
9158 pl_yylval.ival = OP_MATCH;
9163 S_scan_subst(pTHX_ char *start)
9170 char charset = '\0'; /* character set modifier */
9175 PERL_ARGS_ASSERT_SCAN_SUBST;
9177 pl_yylval.ival = OP_NULL;
9179 s = scan_str(start,!!PL_madskills,FALSE);
9182 Perl_croak(aTHX_ "Substitution pattern not terminated");
9184 if (s[-1] == PL_multi_open)
9188 CURMAD('q', PL_thisopen);
9189 CURMAD('_', PL_thiswhite);
9190 CURMAD('E', PL_thisstuff);
9191 CURMAD('Q', PL_thisclose);
9192 PL_realtokenstart = s - SvPVX(PL_linestr);
9196 first_start = PL_multi_start;
9197 s = scan_str(s,!!PL_madskills,FALSE);
9200 SvREFCNT_dec(PL_lex_stuff);
9201 PL_lex_stuff = NULL;
9203 Perl_croak(aTHX_ "Substitution replacement not terminated");
9205 PL_multi_start = first_start; /* so whole substitution is taken together */
9207 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9211 CURMAD('z', PL_thisopen);
9212 CURMAD('R', PL_thisstuff);
9213 CURMAD('Z', PL_thisclose);
9219 if (*s == EXEC_PAT_MOD) {
9223 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
9232 curmad('m', newSVpvn(modstart, s - modstart));
9233 append_madprops(PL_thismad, (OP*)pm, 0);
9237 if ((pm->op_pmflags & PMf_CONTINUE)) {
9238 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9242 SV * const repl = newSVpvs("");
9244 PL_sublex_info.super_bufptr = s;
9245 PL_sublex_info.super_bufend = PL_bufend;
9247 pm->op_pmflags |= PMf_EVAL;
9250 sv_catpvs(repl, "eval ");
9252 sv_catpvs(repl, "do ");
9254 sv_catpvs(repl, "{");
9255 sv_catsv(repl, PL_lex_repl);
9256 if (strchr(SvPVX(PL_lex_repl), '#'))
9257 sv_catpvs(repl, "\n");
9258 sv_catpvs(repl, "}");
9260 SvREFCNT_dec(PL_lex_repl);
9264 PL_lex_op = (OP*)pm;
9265 pl_yylval.ival = OP_SUBST;
9270 S_scan_trans(pTHX_ char *start)
9279 bool nondestruct = 0;
9284 PERL_ARGS_ASSERT_SCAN_TRANS;
9286 pl_yylval.ival = OP_NULL;
9288 s = scan_str(start,!!PL_madskills,FALSE);
9290 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9292 if (s[-1] == PL_multi_open)
9296 CURMAD('q', PL_thisopen);
9297 CURMAD('_', PL_thiswhite);
9298 CURMAD('E', PL_thisstuff);
9299 CURMAD('Q', PL_thisclose);
9300 PL_realtokenstart = s - SvPVX(PL_linestr);
9304 s = scan_str(s,!!PL_madskills,FALSE);
9307 SvREFCNT_dec(PL_lex_stuff);
9308 PL_lex_stuff = NULL;
9310 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9313 CURMAD('z', PL_thisopen);
9314 CURMAD('R', PL_thisstuff);
9315 CURMAD('Z', PL_thisclose);
9318 complement = del = squash = 0;
9325 complement = OPpTRANS_COMPLEMENT;
9328 del = OPpTRANS_DELETE;
9331 squash = OPpTRANS_SQUASH;
9343 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
9344 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)tbl);
9345 o->op_private &= ~OPpTRANS_ALL;
9346 o->op_private |= del|squash|complement|
9347 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9348 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
9351 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9356 curmad('m', newSVpvn(modstart, s - modstart));
9357 append_madprops(PL_thismad, o, 0);
9366 S_scan_heredoc(pTHX_ register char *s)
9370 I32 op_type = OP_SCALAR;
9374 const char *found_newline;
9378 const int outer = (PL_rsfp || PL_parser->filtered)
9379 && !(PL_lex_inwhat == OP_SCALAR);
9381 I32 stuffstart = s - SvPVX(PL_linestr);
9384 PL_realtokenstart = -1;
9387 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9391 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9395 while (SPACE_OR_TAB(*peek))
9397 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9400 s = delimcpy(d, e, s, PL_bufend, term, &len);
9410 if (!isALNUM_lazy_if(s,UTF))
9411 deprecate("bare << to mean <<\"\"");
9412 for (; isALNUM_lazy_if(s,UTF); s++) {
9417 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9418 Perl_croak(aTHX_ "Delimiter for here document is too long");
9421 len = d - PL_tokenbuf;
9425 tstart = PL_tokenbuf + !outer;
9426 PL_thisclose = newSVpvn(tstart, len - !outer);
9427 tstart = SvPVX(PL_linestr) + stuffstart;
9428 PL_thisopen = newSVpvn(tstart, s - tstart);
9429 stuffstart = s - SvPVX(PL_linestr);
9432 #ifndef PERL_STRICT_CR
9433 d = strchr(s, '\r');
9435 char * const olds = s;
9437 while (s < PL_bufend) {
9443 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9452 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9459 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
9460 herewas = newSVpvn(s,PL_bufend-s);
9464 herewas = newSVpvn(s-1,found_newline-s+1);
9467 herewas = newSVpvn(s,found_newline-s);
9472 tstart = SvPVX(PL_linestr) + stuffstart;
9474 sv_catpvn(PL_thisstuff, tstart, s - tstart);
9476 PL_thisstuff = newSVpvn(tstart, s - tstart);
9479 s += SvCUR(herewas);
9482 stuffstart = s - SvPVX(PL_linestr);
9488 tmpstr = newSV_type(SVt_PVIV);
9492 SvIV_set(tmpstr, -1);
9494 else if (term == '`') {
9495 op_type = OP_BACKTICK;
9496 SvIV_set(tmpstr, '\\');
9500 PL_multi_start = CopLINE(PL_curcop);
9501 PL_multi_open = PL_multi_close = '<';
9502 term = *PL_tokenbuf;
9503 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp
9504 && !PL_parser->filtered) {
9505 char * const bufptr = PL_sublex_info.super_bufptr;
9506 char * const bufend = PL_sublex_info.super_bufend;
9507 char * const olds = s - SvCUR(herewas);
9508 s = strchr(bufptr, '\n');
9512 while (s < bufend &&
9513 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9515 CopLINE_inc(PL_curcop);
9518 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9519 missingterm(PL_tokenbuf);
9521 sv_setpvn(herewas,bufptr,d-bufptr+1);
9522 sv_setpvn(tmpstr,d+1,s-d);
9524 sv_catpvn(herewas,s,bufend-s);
9525 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
9532 while (s < PL_bufend &&
9533 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9535 CopLINE_inc(PL_curcop);
9537 if (s >= PL_bufend) {
9538 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9539 missingterm(PL_tokenbuf);
9541 sv_setpvn(tmpstr,d+1,s-d);
9545 sv_catpvn(PL_thisstuff, d + 1, s - d);
9547 PL_thisstuff = newSVpvn(d + 1, s - d);
9548 stuffstart = s - SvPVX(PL_linestr);
9552 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
9554 sv_catpvn(herewas,s,PL_bufend-s);
9555 sv_setsv(PL_linestr,herewas);
9556 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
9557 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9558 PL_last_lop = PL_last_uni = NULL;
9561 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9562 while (s >= PL_bufend) { /* multiple line string? */
9565 tstart = SvPVX(PL_linestr) + stuffstart;
9567 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
9569 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
9573 CopLINE_inc(PL_curcop);
9574 if (!outer || !lex_next_chunk(0)) {
9575 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9576 missingterm(PL_tokenbuf);
9578 CopLINE_dec(PL_curcop);
9581 stuffstart = s - SvPVX(PL_linestr);
9583 CopLINE_inc(PL_curcop);
9584 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9585 PL_last_lop = PL_last_uni = NULL;
9586 #ifndef PERL_STRICT_CR
9587 if (PL_bufend - PL_linestart >= 2) {
9588 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
9589 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9591 PL_bufend[-2] = '\n';
9593 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9595 else if (PL_bufend[-1] == '\r')
9596 PL_bufend[-1] = '\n';
9598 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9599 PL_bufend[-1] = '\n';
9601 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
9602 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
9603 *(SvPVX(PL_linestr) + off ) = ' ';
9604 lex_grow_linestr(SvCUR(PL_linestr) + SvCUR(herewas) + 1);
9605 sv_catsv(PL_linestr,herewas);
9606 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9607 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
9611 sv_catsv(tmpstr,PL_linestr);
9616 PL_multi_end = CopLINE(PL_curcop);
9617 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9618 SvPV_shrink_to_cur(tmpstr);
9620 SvREFCNT_dec(herewas);
9622 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9624 else if (PL_encoding)
9625 sv_recode_to_utf8(tmpstr, PL_encoding);
9627 PL_lex_stuff = tmpstr;
9628 pl_yylval.ival = op_type;
9633 takes: current position in input buffer
9634 returns: new position in input buffer
9635 side-effects: pl_yylval and lex_op are set.
9640 <FH> read from filehandle
9641 <pkg::FH> read from package qualified filehandle
9642 <pkg'FH> read from package qualified filehandle
9643 <$fh> read from filehandle in $fh
9649 S_scan_inputsymbol(pTHX_ char *start)
9652 register char *s = start; /* current position in buffer */
9655 char *d = PL_tokenbuf; /* start of temp holding space */
9656 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9658 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
9660 end = strchr(s, '\n');
9663 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9665 /* die if we didn't have space for the contents of the <>,
9666 or if it didn't end, or if we see a newline
9669 if (len >= (I32)sizeof PL_tokenbuf)
9670 Perl_croak(aTHX_ "Excessively long <> operator");
9672 Perl_croak(aTHX_ "Unterminated <> operator");
9677 Remember, only scalar variables are interpreted as filehandles by
9678 this code. Anything more complex (e.g., <$fh{$num}>) will be
9679 treated as a glob() call.
9680 This code makes use of the fact that except for the $ at the front,
9681 a scalar variable and a filehandle look the same.
9683 if (*d == '$' && d[1]) d++;
9685 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9686 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9687 d += UTF ? UTF8SKIP(d) : 1;
9689 /* If we've tried to read what we allow filehandles to look like, and
9690 there's still text left, then it must be a glob() and not a getline.
9691 Use scan_str to pull out the stuff between the <> and treat it
9692 as nothing more than a string.
9695 if (d - PL_tokenbuf != len) {
9696 pl_yylval.ival = OP_GLOB;
9697 s = scan_str(start,!!PL_madskills,FALSE);
9699 Perl_croak(aTHX_ "Glob not terminated");
9703 bool readline_overriden = FALSE;
9706 /* we're in a filehandle read situation */
9709 /* turn <> into <ARGV> */
9711 Copy("ARGV",d,5,char);
9713 /* Check whether readline() is overriden */
9714 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
9716 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
9718 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
9719 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
9720 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
9721 readline_overriden = TRUE;
9723 /* if <$fh>, create the ops to turn the variable into a
9727 /* try to find it in the pad for this block, otherwise find
9728 add symbol table ops
9730 const PADOFFSET tmp = pad_findmy_pvn(d, len, UTF ? SVf_UTF8 : 0);
9731 if (tmp != NOT_IN_PAD) {
9732 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
9733 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
9734 HEK * const stashname = HvNAME_HEK(stash);
9735 SV * const sym = sv_2mortal(newSVhek(stashname));
9736 sv_catpvs(sym, "::");
9742 OP * const o = newOP(OP_PADSV, 0);
9744 PL_lex_op = readline_overriden
9745 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9746 op_append_elem(OP_LIST, o,
9747 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
9748 : (OP*)newUNOP(OP_READLINE, 0, o);
9757 ? (GV_ADDMULTI | GV_ADDINEVAL)
9758 : GV_ADDMULTI) | ( UTF ? SVf_UTF8 : 0 ),
9760 PL_lex_op = readline_overriden
9761 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9762 op_append_elem(OP_LIST,
9763 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
9764 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9765 : (OP*)newUNOP(OP_READLINE, 0,
9766 newUNOP(OP_RV2SV, 0,
9767 newGVOP(OP_GV, 0, gv)));
9769 if (!readline_overriden)
9770 PL_lex_op->op_flags |= OPf_SPECIAL;
9771 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
9772 pl_yylval.ival = OP_NULL;
9775 /* If it's none of the above, it must be a literal filehandle
9776 (<Foo::BAR> or <FOO>) so build a simple readline OP */
9778 GV * const gv = gv_fetchpv(d, GV_ADD | ( UTF ? SVf_UTF8 : 0 ), SVt_PVIO);
9779 PL_lex_op = readline_overriden
9780 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9781 op_append_elem(OP_LIST,
9782 newGVOP(OP_GV, 0, gv),
9783 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9784 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
9785 pl_yylval.ival = OP_NULL;
9794 takes: start position in buffer
9795 keep_quoted preserve \ on the embedded delimiter(s)
9796 keep_delims preserve the delimiters around the string
9797 returns: position to continue reading from buffer
9798 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
9799 updates the read buffer.
9801 This subroutine pulls a string out of the input. It is called for:
9802 q single quotes q(literal text)
9803 ' single quotes 'literal text'
9804 qq double quotes qq(interpolate $here please)
9805 " double quotes "interpolate $here please"
9806 qx backticks qx(/bin/ls -l)
9807 ` backticks `/bin/ls -l`
9808 qw quote words @EXPORT_OK = qw( func() $spam )
9809 m// regexp match m/this/
9810 s/// regexp substitute s/this/that/
9811 tr/// string transliterate tr/this/that/
9812 y/// string transliterate y/this/that/
9813 ($*@) sub prototypes sub foo ($)
9814 (stuff) sub attr parameters sub foo : attr(stuff)
9815 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
9817 In most of these cases (all but <>, patterns and transliterate)
9818 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
9819 calls scan_str(). s/// makes yylex() call scan_subst() which calls
9820 scan_str(). tr/// and y/// make yylex() call scan_trans() which
9823 It skips whitespace before the string starts, and treats the first
9824 character as the delimiter. If the delimiter is one of ([{< then
9825 the corresponding "close" character )]}> is used as the closing
9826 delimiter. It allows quoting of delimiters, and if the string has
9827 balanced delimiters ([{<>}]) it allows nesting.
9829 On success, the SV with the resulting string is put into lex_stuff or,
9830 if that is already non-NULL, into lex_repl. The second case occurs only
9831 when parsing the RHS of the special constructs s/// and tr/// (y///).
9832 For convenience, the terminating delimiter character is stuffed into
9837 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
9840 SV *sv; /* scalar value: string */
9841 const char *tmps; /* temp string, used for delimiter matching */
9842 register char *s = start; /* current position in the buffer */
9843 register char term; /* terminating character */
9844 register char *to; /* current position in the sv's data */
9845 I32 brackets = 1; /* bracket nesting level */
9846 bool has_utf8 = FALSE; /* is there any utf8 content? */
9847 I32 termcode; /* terminating char. code */
9848 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
9849 STRLEN termlen; /* length of terminating string */
9850 int last_off = 0; /* last position for nesting bracket */
9856 PERL_ARGS_ASSERT_SCAN_STR;
9858 /* skip space before the delimiter */
9864 if (PL_realtokenstart >= 0) {
9865 stuffstart = PL_realtokenstart;
9866 PL_realtokenstart = -1;
9869 stuffstart = start - SvPVX(PL_linestr);
9871 /* mark where we are, in case we need to report errors */
9874 /* after skipping whitespace, the next character is the terminator */
9877 termcode = termstr[0] = term;
9881 termcode = utf8_to_uvchr((U8*)s, &termlen);
9882 Copy(s, termstr, termlen, U8);
9883 if (!UTF8_IS_INVARIANT(term))
9887 /* mark where we are */
9888 PL_multi_start = CopLINE(PL_curcop);
9889 PL_multi_open = term;
9891 /* find corresponding closing delimiter */
9892 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
9893 termcode = termstr[0] = term = tmps[5];
9895 PL_multi_close = term;
9897 /* create a new SV to hold the contents. 79 is the SV's initial length.
9898 What a random number. */
9899 sv = newSV_type(SVt_PVIV);
9901 SvIV_set(sv, termcode);
9902 (void)SvPOK_only(sv); /* validate pointer */
9904 /* move past delimiter and try to read a complete string */
9906 sv_catpvn(sv, s, termlen);
9909 tstart = SvPVX(PL_linestr) + stuffstart;
9910 if (!PL_thisopen && !keep_delims) {
9911 PL_thisopen = newSVpvn(tstart, s - tstart);
9912 stuffstart = s - SvPVX(PL_linestr);
9916 if (PL_encoding && !UTF) {
9920 int offset = s - SvPVX_const(PL_linestr);
9921 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
9922 &offset, (char*)termstr, termlen);
9923 const char * const ns = SvPVX_const(PL_linestr) + offset;
9924 char * const svlast = SvEND(sv) - 1;
9926 for (; s < ns; s++) {
9927 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9928 CopLINE_inc(PL_curcop);
9931 goto read_more_line;
9933 /* handle quoted delimiters */
9934 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
9936 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
9938 if ((svlast-1 - t) % 2) {
9942 SvCUR_set(sv, SvCUR(sv) - 1);
9947 if (PL_multi_open == PL_multi_close) {
9953 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
9954 /* At here, all closes are "was quoted" one,
9955 so we don't check PL_multi_close. */
9957 if (!keep_quoted && *(t+1) == PL_multi_open)
9962 else if (*t == PL_multi_open)
9970 SvCUR_set(sv, w - SvPVX_const(sv));
9972 last_off = w - SvPVX(sv);
9973 if (--brackets <= 0)
9979 SvCUR_set(sv, SvCUR(sv) - 1);
9985 /* extend sv if need be */
9986 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
9987 /* set 'to' to the next character in the sv's string */
9988 to = SvPVX(sv)+SvCUR(sv);
9990 /* if open delimiter is the close delimiter read unbridle */
9991 if (PL_multi_open == PL_multi_close) {
9992 for (; s < PL_bufend; s++,to++) {
9993 /* embedded newlines increment the current line number */
9994 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
9995 CopLINE_inc(PL_curcop);
9996 /* handle quoted delimiters */
9997 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
9998 if (!keep_quoted && s[1] == term)
10000 /* any other quotes are simply copied straight through */
10004 /* terminate when run out of buffer (the for() condition), or
10005 have found the terminator */
10006 else if (*s == term) {
10009 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
10012 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10018 /* if the terminator isn't the same as the start character (e.g.,
10019 matched brackets), we have to allow more in the quoting, and
10020 be prepared for nested brackets.
10023 /* read until we run out of string, or we find the terminator */
10024 for (; s < PL_bufend; s++,to++) {
10025 /* embedded newlines increment the line count */
10026 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10027 CopLINE_inc(PL_curcop);
10028 /* backslashes can escape the open or closing characters */
10029 if (*s == '\\' && s+1 < PL_bufend) {
10030 if (!keep_quoted &&
10031 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
10036 /* allow nested opens and closes */
10037 else if (*s == PL_multi_close && --brackets <= 0)
10039 else if (*s == PL_multi_open)
10041 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10046 /* terminate the copied string and update the sv's end-of-string */
10048 SvCUR_set(sv, to - SvPVX_const(sv));
10051 * this next chunk reads more into the buffer if we're not done yet
10055 break; /* handle case where we are done yet :-) */
10057 #ifndef PERL_STRICT_CR
10058 if (to - SvPVX_const(sv) >= 2) {
10059 if ((to[-2] == '\r' && to[-1] == '\n') ||
10060 (to[-2] == '\n' && to[-1] == '\r'))
10064 SvCUR_set(sv, to - SvPVX_const(sv));
10066 else if (to[-1] == '\r')
10069 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
10074 /* if we're out of file, or a read fails, bail and reset the current
10075 line marker so we can report where the unterminated string began
10078 if (PL_madskills) {
10079 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10081 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10083 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10086 CopLINE_inc(PL_curcop);
10087 PL_bufptr = PL_bufend;
10088 if (!lex_next_chunk(0)) {
10090 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10099 /* at this point, we have successfully read the delimited string */
10101 if (!PL_encoding || UTF) {
10103 if (PL_madskills) {
10104 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10105 const int len = s - tstart;
10107 sv_catpvn(PL_thisstuff, tstart, len);
10109 PL_thisstuff = newSVpvn(tstart, len);
10110 if (!PL_thisclose && !keep_delims)
10111 PL_thisclose = newSVpvn(s,termlen);
10116 sv_catpvn(sv, s, termlen);
10121 if (PL_madskills) {
10122 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10123 const int len = s - tstart - termlen;
10125 sv_catpvn(PL_thisstuff, tstart, len);
10127 PL_thisstuff = newSVpvn(tstart, len);
10128 if (!PL_thisclose && !keep_delims)
10129 PL_thisclose = newSVpvn(s - termlen,termlen);
10133 if (has_utf8 || PL_encoding)
10136 PL_multi_end = CopLINE(PL_curcop);
10138 /* if we allocated too much space, give some back */
10139 if (SvCUR(sv) + 5 < SvLEN(sv)) {
10140 SvLEN_set(sv, SvCUR(sv) + 1);
10141 SvPV_renew(sv, SvLEN(sv));
10144 /* decide whether this is the first or second quoted string we've read
10157 takes: pointer to position in buffer
10158 returns: pointer to new position in buffer
10159 side-effects: builds ops for the constant in pl_yylval.op
10161 Read a number in any of the formats that Perl accepts:
10163 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
10164 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
10167 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
10169 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
10172 If it reads a number without a decimal point or an exponent, it will
10173 try converting the number to an integer and see if it can do so
10174 without loss of precision.
10178 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10181 register const char *s = start; /* current position in buffer */
10182 register char *d; /* destination in temp buffer */
10183 register char *e; /* end of temp buffer */
10184 NV nv; /* number read, as a double */
10185 SV *sv = NULL; /* place to put the converted number */
10186 bool floatit; /* boolean: int or float? */
10187 const char *lastub = NULL; /* position of last underbar */
10188 static char const number_too_long[] = "Number too long";
10190 PERL_ARGS_ASSERT_SCAN_NUM;
10192 /* We use the first character to decide what type of number this is */
10196 Perl_croak(aTHX_ "panic: scan_num, *s=%d", *s);
10198 /* if it starts with a 0, it could be an octal number, a decimal in
10199 0.13 disguise, or a hexadecimal number, or a binary number. */
10203 u holds the "number so far"
10204 shift the power of 2 of the base
10205 (hex == 4, octal == 3, binary == 1)
10206 overflowed was the number more than we can hold?
10208 Shift is used when we add a digit. It also serves as an "are
10209 we in octal/hex/binary?" indicator to disallow hex characters
10210 when in octal mode.
10215 bool overflowed = FALSE;
10216 bool just_zero = TRUE; /* just plain 0 or binary number? */
10217 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10218 static const char* const bases[5] =
10219 { "", "binary", "", "octal", "hexadecimal" };
10220 static const char* const Bases[5] =
10221 { "", "Binary", "", "Octal", "Hexadecimal" };
10222 static const char* const maxima[5] =
10224 "0b11111111111111111111111111111111",
10228 const char *base, *Base, *max;
10230 /* check for hex */
10231 if (s[1] == 'x' || s[1] == 'X') {
10235 } else if (s[1] == 'b' || s[1] == 'B') {
10240 /* check for a decimal in disguise */
10241 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
10243 /* so it must be octal */
10250 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10251 "Misplaced _ in number");
10255 base = bases[shift];
10256 Base = Bases[shift];
10257 max = maxima[shift];
10259 /* read the rest of the number */
10261 /* x is used in the overflow test,
10262 b is the digit we're adding on. */
10267 /* if we don't mention it, we're done */
10271 /* _ are ignored -- but warned about if consecutive */
10273 if (lastub && s == lastub + 1)
10274 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10275 "Misplaced _ in number");
10279 /* 8 and 9 are not octal */
10280 case '8': case '9':
10282 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10286 case '2': case '3': case '4':
10287 case '5': case '6': case '7':
10289 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10292 case '0': case '1':
10293 b = *s++ & 15; /* ASCII digit -> value of digit */
10297 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10298 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10299 /* make sure they said 0x */
10302 b = (*s++ & 7) + 9;
10304 /* Prepare to put the digit we have onto the end
10305 of the number so far. We check for overflows.
10311 x = u << shift; /* make room for the digit */
10313 if ((x >> shift) != u
10314 && !(PL_hints & HINT_NEW_BINARY)) {
10317 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10318 "Integer overflow in %s number",
10321 u = x | b; /* add the digit to the end */
10324 n *= nvshift[shift];
10325 /* If an NV has not enough bits in its
10326 * mantissa to represent an UV this summing of
10327 * small low-order numbers is a waste of time
10328 * (because the NV cannot preserve the
10329 * low-order bits anyway): we could just
10330 * remember when did we overflow and in the
10331 * end just multiply n by the right
10339 /* if we get here, we had success: make a scalar value from
10344 /* final misplaced underbar check */
10345 if (s[-1] == '_') {
10346 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10350 if (n > 4294967295.0)
10351 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10352 "%s number > %s non-portable",
10358 if (u > 0xffffffff)
10359 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10360 "%s number > %s non-portable",
10365 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10366 sv = new_constant(start, s - start, "integer",
10367 sv, NULL, NULL, 0);
10368 else if (PL_hints & HINT_NEW_BINARY)
10369 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10374 handle decimal numbers.
10375 we're also sent here when we read a 0 as the first digit
10377 case '1': case '2': case '3': case '4': case '5':
10378 case '6': case '7': case '8': case '9': case '.':
10381 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10384 /* read next group of digits and _ and copy into d */
10385 while (isDIGIT(*s) || *s == '_') {
10386 /* skip underscores, checking for misplaced ones
10390 if (lastub && s == lastub + 1)
10391 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10392 "Misplaced _ in number");
10396 /* check for end of fixed-length buffer */
10398 Perl_croak(aTHX_ number_too_long);
10399 /* if we're ok, copy the character */
10404 /* final misplaced underbar check */
10405 if (lastub && s == lastub + 1) {
10406 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10409 /* read a decimal portion if there is one. avoid
10410 3..5 being interpreted as the number 3. followed
10413 if (*s == '.' && s[1] != '.') {
10418 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10419 "Misplaced _ in number");
10423 /* copy, ignoring underbars, until we run out of digits.
10425 for (; isDIGIT(*s) || *s == '_'; s++) {
10426 /* fixed length buffer check */
10428 Perl_croak(aTHX_ number_too_long);
10430 if (lastub && s == lastub + 1)
10431 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10432 "Misplaced _ in number");
10438 /* fractional part ending in underbar? */
10439 if (s[-1] == '_') {
10440 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10441 "Misplaced _ in number");
10443 if (*s == '.' && isDIGIT(s[1])) {
10444 /* oops, it's really a v-string, but without the "v" */
10450 /* read exponent part, if present */
10451 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10455 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10456 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10458 /* stray preinitial _ */
10460 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10461 "Misplaced _ in number");
10465 /* allow positive or negative exponent */
10466 if (*s == '+' || *s == '-')
10469 /* stray initial _ */
10471 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10472 "Misplaced _ in number");
10476 /* read digits of exponent */
10477 while (isDIGIT(*s) || *s == '_') {
10480 Perl_croak(aTHX_ number_too_long);
10484 if (((lastub && s == lastub + 1) ||
10485 (!isDIGIT(s[1]) && s[1] != '_')))
10486 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10487 "Misplaced _ in number");
10495 We try to do an integer conversion first if no characters
10496 indicating "float" have been found.
10501 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10503 if (flags == IS_NUMBER_IN_UV) {
10505 sv = newSViv(uv); /* Prefer IVs over UVs. */
10508 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10509 if (uv <= (UV) IV_MIN)
10510 sv = newSViv(-(IV)uv);
10517 /* terminate the string */
10519 nv = Atof(PL_tokenbuf);
10524 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10525 const char *const key = floatit ? "float" : "integer";
10526 const STRLEN keylen = floatit ? 5 : 7;
10527 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10528 key, keylen, sv, NULL, NULL, 0);
10532 /* if it starts with a v, it could be a v-string */
10535 sv = newSV(5); /* preallocate storage space */
10536 s = scan_vstring(s, PL_bufend, sv);
10540 /* make the op for the constant and return */
10543 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10545 lvalp->opval = NULL;
10551 S_scan_formline(pTHX_ register char *s)
10554 register char *eol;
10556 SV * const stuff = newSVpvs("");
10557 bool needargs = FALSE;
10558 bool eofmt = FALSE;
10560 char *tokenstart = s;
10561 SV* savewhite = NULL;
10563 if (PL_madskills) {
10564 savewhite = PL_thiswhite;
10569 PERL_ARGS_ASSERT_SCAN_FORMLINE;
10571 while (!needargs) {
10574 #ifdef PERL_STRICT_CR
10575 while (SPACE_OR_TAB(*t))
10578 while (SPACE_OR_TAB(*t) || *t == '\r')
10581 if (*t == '\n' || t == PL_bufend) {
10586 if (PL_in_eval && !PL_rsfp && !PL_parser->filtered) {
10587 eol = (char *) memchr(s,'\n',PL_bufend-s);
10592 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10594 for (t = s; t < eol; t++) {
10595 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10597 goto enough; /* ~~ must be first line in formline */
10599 if (*t == '@' || *t == '^')
10603 sv_catpvn(stuff, s, eol-s);
10604 #ifndef PERL_STRICT_CR
10605 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10606 char *end = SvPVX(stuff) + SvCUR(stuff);
10609 SvCUR_set(stuff, SvCUR(stuff) - 1);
10617 if (PL_rsfp || PL_parser->filtered) {
10620 if (PL_madskills) {
10622 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
10624 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
10627 PL_bufptr = PL_bufend;
10628 CopLINE_inc(PL_curcop);
10629 got_some = lex_next_chunk(0);
10630 CopLINE_dec(PL_curcop);
10633 tokenstart = PL_bufptr;
10641 if (SvCUR(stuff)) {
10644 PL_lex_state = LEX_NORMAL;
10645 start_force(PL_curforce);
10646 NEXTVAL_NEXTTOKE.ival = 0;
10650 PL_lex_state = LEX_FORMLINE;
10652 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10654 else if (PL_encoding)
10655 sv_recode_to_utf8(stuff, PL_encoding);
10657 start_force(PL_curforce);
10658 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10660 start_force(PL_curforce);
10661 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
10665 SvREFCNT_dec(stuff);
10667 PL_lex_formbrack = 0;
10671 if (PL_madskills) {
10673 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
10675 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
10676 PL_thiswhite = savewhite;
10683 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10686 const I32 oldsavestack_ix = PL_savestack_ix;
10687 CV* const outsidecv = PL_compcv;
10690 assert(SvTYPE(PL_compcv) == SVt_PVCV);
10692 SAVEI32(PL_subline);
10693 save_item(PL_subname);
10694 SAVESPTR(PL_compcv);
10696 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
10697 CvFLAGS(PL_compcv) |= flags;
10699 PL_subline = CopLINE(PL_curcop);
10700 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10701 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
10702 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10704 return oldsavestack_ix;
10708 #pragma segment Perl_yylex
10711 S_yywarn(pTHX_ const char *const s)
10715 PERL_ARGS_ASSERT_YYWARN;
10717 PL_in_eval |= EVAL_WARNONLY;
10719 PL_in_eval &= ~EVAL_WARNONLY;
10724 Perl_yyerror(pTHX_ const char *const s)
10727 const char *where = NULL;
10728 const char *context = NULL;
10731 int yychar = PL_parser->yychar;
10733 PERL_ARGS_ASSERT_YYERROR;
10735 if (!yychar || (yychar == ';' && !PL_rsfp))
10737 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
10738 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
10739 PL_oldbufptr != PL_bufptr) {
10742 The code below is removed for NetWare because it abends/crashes on NetWare
10743 when the script has error such as not having the closing quotes like:
10744 if ($var eq "value)
10745 Checking of white spaces is anyway done in NetWare code.
10748 while (isSPACE(*PL_oldoldbufptr))
10751 context = PL_oldoldbufptr;
10752 contlen = PL_bufptr - PL_oldoldbufptr;
10754 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
10755 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
10758 The code below is removed for NetWare because it abends/crashes on NetWare
10759 when the script has error such as not having the closing quotes like:
10760 if ($var eq "value)
10761 Checking of white spaces is anyway done in NetWare code.
10764 while (isSPACE(*PL_oldbufptr))
10767 context = PL_oldbufptr;
10768 contlen = PL_bufptr - PL_oldbufptr;
10770 else if (yychar > 255)
10771 where = "next token ???";
10772 else if (yychar == -2) { /* YYEMPTY */
10773 if (PL_lex_state == LEX_NORMAL ||
10774 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
10775 where = "at end of line";
10776 else if (PL_lex_inpat)
10777 where = "within pattern";
10779 where = "within string";
10782 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
10784 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
10785 else if (isPRINT_LC(yychar)) {
10786 const char string = yychar;
10787 sv_catpvn(where_sv, &string, 1);
10790 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
10791 where = SvPVX_const(where_sv);
10793 msg = sv_2mortal(newSVpv(s, 0));
10794 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
10795 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
10797 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
10799 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
10800 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
10801 Perl_sv_catpvf(aTHX_ msg,
10802 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
10803 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
10806 if (PL_in_eval & EVAL_WARNONLY) {
10807 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
10811 if (PL_error_count >= 10) {
10812 if (PL_in_eval && SvCUR(ERRSV))
10813 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
10814 SVfARG(ERRSV), OutCopFILE(PL_curcop));
10816 Perl_croak(aTHX_ "%s has too many errors.\n",
10817 OutCopFILE(PL_curcop));
10820 PL_in_my_stash = NULL;
10824 #pragma segment Main
10828 S_swallow_bom(pTHX_ U8 *s)
10831 const STRLEN slen = SvCUR(PL_linestr);
10833 PERL_ARGS_ASSERT_SWALLOW_BOM;
10837 if (s[1] == 0xFE) {
10838 /* UTF-16 little-endian? (or UTF-32LE?) */
10839 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
10840 /* diag_listed_as: Unsupported script encoding %s */
10841 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
10842 #ifndef PERL_NO_UTF16_FILTER
10843 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
10845 if (PL_bufend > (char*)s) {
10846 s = add_utf16_textfilter(s, TRUE);
10849 /* diag_listed_as: Unsupported script encoding %s */
10850 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10855 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
10856 #ifndef PERL_NO_UTF16_FILTER
10857 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
10859 if (PL_bufend > (char *)s) {
10860 s = add_utf16_textfilter(s, FALSE);
10863 /* diag_listed_as: Unsupported script encoding %s */
10864 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10869 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
10870 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10871 s += 3; /* UTF-8 */
10877 if (s[2] == 0xFE && s[3] == 0xFF) {
10878 /* UTF-32 big-endian */
10879 /* diag_listed_as: Unsupported script encoding %s */
10880 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
10883 else if (s[2] == 0 && s[3] != 0) {
10886 * are a good indicator of UTF-16BE. */
10887 #ifndef PERL_NO_UTF16_FILTER
10888 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
10889 s = add_utf16_textfilter(s, FALSE);
10891 /* diag_listed_as: Unsupported script encoding %s */
10892 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10898 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
10899 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10900 s += 4; /* UTF-8 */
10906 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
10909 * are a good indicator of UTF-16LE. */
10910 #ifndef PERL_NO_UTF16_FILTER
10911 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
10912 s = add_utf16_textfilter(s, TRUE);
10914 /* diag_listed_as: Unsupported script encoding %s */
10915 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10923 #ifndef PERL_NO_UTF16_FILTER
10925 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
10928 SV *const filter = FILTER_DATA(idx);
10929 /* We re-use this each time round, throwing the contents away before we
10931 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
10932 SV *const utf8_buffer = filter;
10933 IV status = IoPAGE(filter);
10934 const bool reverse = cBOOL(IoLINES(filter));
10937 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
10939 /* As we're automatically added, at the lowest level, and hence only called
10940 from this file, we can be sure that we're not called in block mode. Hence
10941 don't bother writing code to deal with block mode. */
10943 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
10946 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
10948 DEBUG_P(PerlIO_printf(Perl_debug_log,
10949 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10950 FPTR2DPTR(void *, S_utf16_textfilter),
10951 reverse ? 'l' : 'b', idx, maxlen, status,
10952 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10959 /* First, look in our buffer of existing UTF-8 data: */
10960 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
10964 } else if (status == 0) {
10966 IoPAGE(filter) = 0;
10967 nl = SvEND(utf8_buffer);
10970 STRLEN got = nl - SvPVX(utf8_buffer);
10971 /* Did we have anything to append? */
10973 sv_catpvn(sv, SvPVX(utf8_buffer), got);
10974 /* Everything else in this code works just fine if SVp_POK isn't
10975 set. This, however, needs it, and we need it to work, else
10976 we loop infinitely because the buffer is never consumed. */
10977 sv_chop(utf8_buffer, nl);
10981 /* OK, not a complete line there, so need to read some more UTF-16.
10982 Read an extra octect if the buffer currently has an odd number. */
10986 if (SvCUR(utf16_buffer) >= 2) {
10987 /* Location of the high octet of the last complete code point.
10988 Gosh, UTF-16 is a pain. All the benefits of variable length,
10989 *coupled* with all the benefits of partial reads and
10991 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
10992 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
10994 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
10998 /* We have the first half of a surrogate. Read more. */
10999 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
11002 status = FILTER_READ(idx + 1, utf16_buffer,
11003 160 + (SvCUR(utf16_buffer) & 1));
11004 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
11005 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
11008 IoPAGE(filter) = status;
11013 chars = SvCUR(utf16_buffer) >> 1;
11014 have = SvCUR(utf8_buffer);
11015 SvGROW(utf8_buffer, have + chars * 3 + 1);
11018 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
11019 (U8*)SvPVX_const(utf8_buffer) + have,
11020 chars * 2, &newlen);
11022 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
11023 (U8*)SvPVX_const(utf8_buffer) + have,
11024 chars * 2, &newlen);
11026 SvCUR_set(utf8_buffer, have + newlen);
11029 /* No need to keep this SV "well-formed" with a '\0' after the end, as
11030 it's private to us, and utf16_to_utf8{,reversed} take a
11031 (pointer,length) pair, rather than a NUL-terminated string. */
11032 if(SvCUR(utf16_buffer) & 1) {
11033 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
11034 SvCUR_set(utf16_buffer, 1);
11036 SvCUR_set(utf16_buffer, 0);
11039 DEBUG_P(PerlIO_printf(Perl_debug_log,
11040 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11042 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11043 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
11048 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
11050 SV *filter = filter_add(S_utf16_textfilter, NULL);
11052 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
11054 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
11055 sv_setpvs(filter, "");
11056 IoLINES(filter) = reversed;
11057 IoPAGE(filter) = 1; /* Not EOF */
11059 /* Sadly, we have to return a valid pointer, come what may, so we have to
11060 ignore any error return from this. */
11061 SvCUR_set(PL_linestr, 0);
11062 if (FILTER_READ(0, PL_linestr, 0)) {
11063 SvUTF8_on(PL_linestr);
11065 SvUTF8_on(PL_linestr);
11067 PL_bufend = SvEND(PL_linestr);
11068 return (U8*)SvPVX(PL_linestr);
11073 Returns a pointer to the next character after the parsed
11074 vstring, as well as updating the passed in sv.
11076 Function must be called like
11079 s = scan_vstring(s,e,sv);
11081 where s and e are the start and end of the string.
11082 The sv should already be large enough to store the vstring
11083 passed in, for performance reasons.
11088 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
11091 const char *pos = s;
11092 const char *start = s;
11094 PERL_ARGS_ASSERT_SCAN_VSTRING;
11096 if (*pos == 'v') pos++; /* get past 'v' */
11097 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11099 if ( *pos != '.') {
11100 /* this may not be a v-string if followed by => */
11101 const char *next = pos;
11102 while (next < e && isSPACE(*next))
11104 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
11105 /* return string not v-string */
11106 sv_setpvn(sv,(char *)s,pos-s);
11107 return (char *)pos;
11111 if (!isALPHA(*pos)) {
11112 U8 tmpbuf[UTF8_MAXBYTES+1];
11115 s++; /* get past 'v' */
11120 /* this is atoi() that tolerates underscores */
11123 const char *end = pos;
11125 while (--end >= s) {
11127 const UV orev = rev;
11128 rev += (*end - '0') * mult;
11131 /* diag_listed_as: Integer overflow in %s number */
11132 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
11133 "Integer overflow in decimal number");
11137 if (rev > 0x7FFFFFFF)
11138 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
11140 /* Append native character for the rev point */
11141 tmpend = uvchr_to_utf8(tmpbuf, rev);
11142 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
11143 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
11145 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
11151 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11155 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
11162 Perl_keyword_plugin_standard(pTHX_
11163 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
11165 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
11166 PERL_UNUSED_CONTEXT;
11167 PERL_UNUSED_ARG(keyword_ptr);
11168 PERL_UNUSED_ARG(keyword_len);
11169 PERL_UNUSED_ARG(op_ptr);
11170 return KEYWORD_PLUGIN_DECLINE;
11173 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
11175 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
11177 SAVEI32(PL_lex_brackets);
11178 if (PL_lex_brackets > 100)
11179 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
11180 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
11181 SAVEI32(PL_lex_allbrackets);
11182 PL_lex_allbrackets = 0;
11183 SAVEI8(PL_lex_fakeeof);
11184 PL_lex_fakeeof = (U8)fakeeof;
11185 if(yyparse(gramtype) && !PL_parser->error_count)
11186 qerror(Perl_mess(aTHX_ "Parse error"));
11189 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11191 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11195 SAVEVPTR(PL_eval_root);
11196 PL_eval_root = NULL;
11197 parse_recdescent(gramtype, fakeeof);
11203 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11205 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11208 if (flags & ~PARSE_OPTIONAL)
11209 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11210 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11211 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11212 if (!PL_parser->error_count)
11213 qerror(Perl_mess(aTHX_ "Parse error"));
11214 exprop = newOP(OP_NULL, 0);
11220 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11222 Parse a Perl arithmetic expression. This may contain operators of precedence
11223 down to the bit shift operators. The expression must be followed (and thus
11224 terminated) either by a comparison or lower-precedence operator or by
11225 something that would normally terminate an expression such as semicolon.
11226 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11227 otherwise it is mandatory. It is up to the caller to ensure that the
11228 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11229 the source of the code to be parsed and the lexical context for the
11232 The op tree representing the expression is returned. If an optional
11233 expression is absent, a null pointer is returned, otherwise the pointer
11236 If an error occurs in parsing or compilation, in most cases a valid op
11237 tree is returned anyway. The error is reflected in the parser state,
11238 normally resulting in a single exception at the top level of parsing
11239 which covers all the compilation errors that occurred. Some compilation
11240 errors, however, will throw an exception immediately.
11246 Perl_parse_arithexpr(pTHX_ U32 flags)
11248 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11252 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11254 Parse a Perl term expression. This may contain operators of precedence
11255 down to the assignment operators. The expression must be followed (and thus
11256 terminated) either by a comma or lower-precedence operator or by
11257 something that would normally terminate an expression such as semicolon.
11258 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11259 otherwise it is mandatory. It is up to the caller to ensure that the
11260 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11261 the source of the code to be parsed and the lexical context for the
11264 The op tree representing the expression is returned. If an optional
11265 expression is absent, a null pointer is returned, otherwise the pointer
11268 If an error occurs in parsing or compilation, in most cases a valid op
11269 tree is returned anyway. The error is reflected in the parser state,
11270 normally resulting in a single exception at the top level of parsing
11271 which covers all the compilation errors that occurred. Some compilation
11272 errors, however, will throw an exception immediately.
11278 Perl_parse_termexpr(pTHX_ U32 flags)
11280 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11284 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11286 Parse a Perl list expression. This may contain operators of precedence
11287 down to the comma operator. The expression must be followed (and thus
11288 terminated) either by a low-precedence logic operator such as C<or> or by
11289 something that would normally terminate an expression such as semicolon.
11290 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11291 otherwise it is mandatory. It is up to the caller to ensure that the
11292 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11293 the source of the code to be parsed and the lexical context for the
11296 The op tree representing the expression is returned. If an optional
11297 expression is absent, a null pointer is returned, otherwise the pointer
11300 If an error occurs in parsing or compilation, in most cases a valid op
11301 tree is returned anyway. The error is reflected in the parser state,
11302 normally resulting in a single exception at the top level of parsing
11303 which covers all the compilation errors that occurred. Some compilation
11304 errors, however, will throw an exception immediately.
11310 Perl_parse_listexpr(pTHX_ U32 flags)
11312 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11316 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11318 Parse a single complete Perl expression. This allows the full
11319 expression grammar, including the lowest-precedence operators such
11320 as C<or>. The expression must be followed (and thus terminated) by a
11321 token that an expression would normally be terminated by: end-of-file,
11322 closing bracketing punctuation, semicolon, or one of the keywords that
11323 signals a postfix expression-statement modifier. If I<flags> includes
11324 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11325 mandatory. It is up to the caller to ensure that the dynamic parser
11326 state (L</PL_parser> et al) is correctly set to reflect the source of
11327 the code to be parsed and the lexical context for the expression.
11329 The op tree representing the expression is returned. If an optional
11330 expression is absent, a null pointer is returned, otherwise the pointer
11333 If an error occurs in parsing or compilation, in most cases a valid op
11334 tree is returned anyway. The error is reflected in the parser state,
11335 normally resulting in a single exception at the top level of parsing
11336 which covers all the compilation errors that occurred. Some compilation
11337 errors, however, will throw an exception immediately.
11343 Perl_parse_fullexpr(pTHX_ U32 flags)
11345 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11349 =for apidoc Amx|OP *|parse_block|U32 flags
11351 Parse a single complete Perl code block. This consists of an opening
11352 brace, a sequence of statements, and a closing brace. The block
11353 constitutes a lexical scope, so C<my> variables and various compile-time
11354 effects can be contained within it. It is up to the caller to ensure
11355 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11356 reflect the source of the code to be parsed and the lexical context for
11359 The op tree representing the code block is returned. This is always a
11360 real op, never a null pointer. It will normally be a C<lineseq> list,
11361 including C<nextstate> or equivalent ops. No ops to construct any kind
11362 of runtime scope are included by virtue of it being a block.
11364 If an error occurs in parsing or compilation, in most cases a valid op
11365 tree (most likely null) is returned anyway. The error is reflected in
11366 the parser state, normally resulting in a single exception at the top
11367 level of parsing which covers all the compilation errors that occurred.
11368 Some compilation errors, however, will throw an exception immediately.
11370 The I<flags> parameter is reserved for future use, and must always
11377 Perl_parse_block(pTHX_ U32 flags)
11380 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11381 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11385 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11387 Parse a single unadorned Perl statement. This may be a normal imperative
11388 statement or a declaration that has compile-time effect. It does not
11389 include any label or other affixture. It is up to the caller to ensure
11390 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11391 reflect the source of the code to be parsed and the lexical context for
11394 The op tree representing the statement is returned. This may be a
11395 null pointer if the statement is null, for example if it was actually
11396 a subroutine definition (which has compile-time side effects). If not
11397 null, it will be ops directly implementing the statement, suitable to
11398 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11399 equivalent op (except for those embedded in a scope contained entirely
11400 within the statement).
11402 If an error occurs in parsing or compilation, in most cases a valid op
11403 tree (most likely null) is returned anyway. The error is reflected in
11404 the parser state, normally resulting in a single exception at the top
11405 level of parsing which covers all the compilation errors that occurred.
11406 Some compilation errors, however, will throw an exception immediately.
11408 The I<flags> parameter is reserved for future use, and must always
11415 Perl_parse_barestmt(pTHX_ U32 flags)
11418 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11419 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11423 =for apidoc Amx|SV *|parse_label|U32 flags
11425 Parse a single label, possibly optional, of the type that may prefix a
11426 Perl statement. It is up to the caller to ensure that the dynamic parser
11427 state (L</PL_parser> et al) is correctly set to reflect the source of
11428 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11429 label is optional, otherwise it is mandatory.
11431 The name of the label is returned in the form of a fresh scalar. If an
11432 optional label is absent, a null pointer is returned.
11434 If an error occurs in parsing, which can only occur if the label is
11435 mandatory, a valid label is returned anyway. The error is reflected in
11436 the parser state, normally resulting in a single exception at the top
11437 level of parsing which covers all the compilation errors that occurred.
11443 Perl_parse_label(pTHX_ U32 flags)
11445 if (flags & ~PARSE_OPTIONAL)
11446 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11447 if (PL_lex_state == LEX_KNOWNEXT) {
11448 PL_parser->yychar = yylex();
11449 if (PL_parser->yychar == LABEL) {
11450 char *lpv = pl_yylval.pval;
11451 STRLEN llen = strlen(lpv);
11453 PL_parser->yychar = YYEMPTY;
11454 lsv = newSV_type(SVt_PV);
11455 SvPV_set(lsv, lpv);
11456 SvCUR_set(lsv, llen);
11457 SvLEN_set(lsv, llen+1);
11467 STRLEN wlen, bufptr_pos;
11471 if (!isIDFIRST_A(c))
11475 } while(isWORDCHAR_A(c));
11477 if (word_takes_any_delimeter(s, wlen))
11479 bufptr_pos = s - SvPVX(PL_linestr);
11481 lex_read_space(LEX_KEEP_PREVIOUS);
11483 s = SvPVX(PL_linestr) + bufptr_pos;
11484 if (t[0] == ':' && t[1] != ':') {
11485 PL_oldoldbufptr = PL_oldbufptr;
11488 return newSVpvn(s, wlen);
11492 if (flags & PARSE_OPTIONAL) {
11495 qerror(Perl_mess(aTHX_ "Parse error"));
11496 return newSVpvs("x");
11503 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11505 Parse a single complete Perl statement. This may be a normal imperative
11506 statement or a declaration that has compile-time effect, and may include
11507 optional labels. It is up to the caller to ensure that the dynamic
11508 parser state (L</PL_parser> et al) is correctly set to reflect the source
11509 of the code to be parsed and the lexical context for the statement.
11511 The op tree representing the statement is returned. This may be a
11512 null pointer if the statement is null, for example if it was actually
11513 a subroutine definition (which has compile-time side effects). If not
11514 null, it will be the result of a L</newSTATEOP> call, normally including
11515 a C<nextstate> or equivalent op.
11517 If an error occurs in parsing or compilation, in most cases a valid op
11518 tree (most likely null) is returned anyway. The error is reflected in
11519 the parser state, normally resulting in a single exception at the top
11520 level of parsing which covers all the compilation errors that occurred.
11521 Some compilation errors, however, will throw an exception immediately.
11523 The I<flags> parameter is reserved for future use, and must always
11530 Perl_parse_fullstmt(pTHX_ U32 flags)
11533 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11534 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11538 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11540 Parse a sequence of zero or more Perl statements. These may be normal
11541 imperative statements, including optional labels, or declarations
11542 that have compile-time effect, or any mixture thereof. The statement
11543 sequence ends when a closing brace or end-of-file is encountered in a
11544 place where a new statement could have validly started. It is up to
11545 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11546 is correctly set to reflect the source of the code to be parsed and the
11547 lexical context for the statements.
11549 The op tree representing the statement sequence is returned. This may
11550 be a null pointer if the statements were all null, for example if there
11551 were no statements or if there were only subroutine definitions (which
11552 have compile-time side effects). If not null, it will be a C<lineseq>
11553 list, normally including C<nextstate> or equivalent ops.
11555 If an error occurs in parsing or compilation, in most cases a valid op
11556 tree is returned anyway. The error is reflected in the parser state,
11557 normally resulting in a single exception at the top level of parsing
11558 which covers all the compilation errors that occurred. Some compilation
11559 errors, however, will throw an exception immediately.
11561 The I<flags> parameter is reserved for future use, and must always
11568 Perl_parse_stmtseq(pTHX_ U32 flags)
11573 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
11574 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
11575 c = lex_peek_unichar(0);
11576 if (c != -1 && c != /*{*/'}')
11577 qerror(Perl_mess(aTHX_ "Parse error"));
11582 Perl_munge_qwlist_to_paren_list(pTHX_ OP *qwlist)
11584 PERL_ARGS_ASSERT_MUNGE_QWLIST_TO_PAREN_LIST;
11585 deprecate("qw(...) as parentheses");
11586 force_next((4<<24)|')');
11587 if (qwlist->op_type == OP_STUB) {
11591 start_force(PL_curforce);
11592 NEXTVAL_NEXTTOKE.opval = qwlist;
11595 force_next((2<<24)|'(');
11600 * c-indentation-style: bsd
11601 * c-basic-offset: 4
11602 * indent-tabs-mode: t
11605 * ex: set ts=8 sts=4 sw=4 noet: