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_preambled (PL_parser->preambled)
70 #define PL_sublex_info (PL_parser->sublex_info)
71 #define PL_linestr (PL_parser->linestr)
72 #define PL_expect (PL_parser->expect)
73 #define PL_copline (PL_parser->copline)
74 #define PL_bufptr (PL_parser->bufptr)
75 #define PL_oldbufptr (PL_parser->oldbufptr)
76 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
77 #define PL_linestart (PL_parser->linestart)
78 #define PL_bufend (PL_parser->bufend)
79 #define PL_last_uni (PL_parser->last_uni)
80 #define PL_last_lop (PL_parser->last_lop)
81 #define PL_last_lop_op (PL_parser->last_lop_op)
82 #define PL_lex_state (PL_parser->lex_state)
83 #define PL_rsfp (PL_parser->rsfp)
84 #define PL_rsfp_filters (PL_parser->rsfp_filters)
85 #define PL_in_my (PL_parser->in_my)
86 #define PL_in_my_stash (PL_parser->in_my_stash)
87 #define PL_tokenbuf (PL_parser->tokenbuf)
88 #define PL_multi_end (PL_parser->multi_end)
89 #define PL_error_count (PL_parser->error_count)
92 # define PL_endwhite (PL_parser->endwhite)
93 # define PL_faketokens (PL_parser->faketokens)
94 # define PL_lasttoke (PL_parser->lasttoke)
95 # define PL_nextwhite (PL_parser->nextwhite)
96 # define PL_realtokenstart (PL_parser->realtokenstart)
97 # define PL_skipwhite (PL_parser->skipwhite)
98 # define PL_thisclose (PL_parser->thisclose)
99 # define PL_thismad (PL_parser->thismad)
100 # define PL_thisopen (PL_parser->thisopen)
101 # define PL_thisstuff (PL_parser->thisstuff)
102 # define PL_thistoken (PL_parser->thistoken)
103 # define PL_thiswhite (PL_parser->thiswhite)
104 # define PL_thiswhite (PL_parser->thiswhite)
105 # define PL_nexttoke (PL_parser->nexttoke)
106 # define PL_curforce (PL_parser->curforce)
108 # define PL_nexttoke (PL_parser->nexttoke)
109 # define PL_nexttype (PL_parser->nexttype)
110 # define PL_nextval (PL_parser->nextval)
113 static const char ident_too_long[] = "Identifier too long";
116 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
117 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
119 # define CURMAD(slot,sv)
120 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
123 #define XENUMMASK 0x3f
124 #define XFAKEEOF 0x40
125 #define XFAKEBRACK 0x80
127 #ifdef USE_UTF8_SCRIPTS
128 # define UTF (!IN_BYTES)
130 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || ( !(PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS) && (PL_hints & HINT_UTF8)))
133 /* The maximum number of characters preceding the unrecognized one to display */
134 #define UNRECOGNIZED_PRECEDE_COUNT 10
136 /* In variables named $^X, these are the legal values for X.
137 * 1999-02-27 mjd-perl-patch@plover.com */
138 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
140 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
142 /* LEX_* are values for PL_lex_state, the state of the lexer.
143 * They are arranged oddly so that the guard on the switch statement
144 * can get by with a single comparison (if the compiler is smart enough).
146 * These values refer to the various states within a sublex parse,
147 * i.e. within a double quotish string
150 /* #define LEX_NOTPARSING 11 is done in perl.h. */
152 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
153 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
154 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
155 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
156 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
158 /* at end of code, eg "$x" followed by: */
159 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
160 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
162 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
163 string or after \E, $foo, etc */
164 #define LEX_INTERPCONST 2 /* NOT USED */
165 #define LEX_FORMLINE 1 /* expecting a format line */
166 #define LEX_KNOWNEXT 0 /* next token known; just return it */
170 static const char* const lex_state_names[] = {
189 #include "keywords.h"
191 /* CLINE is a macro that ensures PL_copline has a sane value */
196 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
199 # define SKIPSPACE0(s) skipspace0(s)
200 # define SKIPSPACE1(s) skipspace1(s)
201 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
202 # define PEEKSPACE(s) skipspace2(s,0)
204 # define SKIPSPACE0(s) skipspace(s)
205 # define SKIPSPACE1(s) skipspace(s)
206 # define SKIPSPACE2(s,tsv) skipspace(s)
207 # define PEEKSPACE(s) skipspace(s)
211 * Convenience functions to return different tokens and prime the
212 * lexer for the next token. They all take an argument.
214 * TOKEN : generic token (used for '(', DOLSHARP, etc)
215 * OPERATOR : generic operator
216 * AOPERATOR : assignment operator
217 * PREBLOCK : beginning the block after an if, while, foreach, ...
218 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
219 * PREREF : *EXPR where EXPR is not a simple identifier
220 * TERM : expression term
221 * LOOPX : loop exiting command (goto, last, dump, etc)
222 * FTST : file test operator
223 * FUN0 : zero-argument function
224 * FUN0OP : zero-argument function, with its op created in this file
225 * FUN1 : not used, except for not, which isn't a UNIOP
226 * BOop : bitwise or or xor
228 * SHop : shift operator
229 * PWop : power operator
230 * PMop : pattern-matching operator
231 * Aop : addition-level operator
232 * Mop : multiplication-level operator
233 * Eop : equality-testing operator
234 * Rop : relational operator <= != gt
236 * Also see LOP and lop() below.
239 #ifdef DEBUGGING /* Serve -DT. */
240 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
242 # define REPORT(retval) (retval)
245 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
246 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
247 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
248 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
249 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
250 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
251 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
252 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
253 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
254 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
255 #define FUN0OP(f) return (pl_yylval.opval=f, CLINE, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0OP))
256 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
257 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
258 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
259 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
260 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
261 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
262 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
263 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
264 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
265 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
267 /* This bit of chicanery makes a unary function followed by
268 * a parenthesis into a function with one argument, highest precedence.
269 * The UNIDOR macro is for unary functions that can be followed by the //
270 * operator (such as C<shift // 0>).
272 #define UNI3(f,x,have_x) { \
273 pl_yylval.ival = f; \
274 if (have_x) PL_expect = x; \
276 PL_last_uni = PL_oldbufptr; \
277 PL_last_lop_op = f; \
279 return REPORT( (int)FUNC1 ); \
281 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
283 #define UNI(f) UNI3(f,XTERM,1)
284 #define UNIDOR(f) UNI3(f,XTERMORDORDOR,1)
285 #define UNIPROTO(f,optional) { \
286 if (optional) PL_last_uni = PL_oldbufptr; \
290 #define UNIBRACK(f) UNI3(f,0,0)
292 /* grandfather return to old style */
295 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
296 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
297 pl_yylval.ival = (f); \
303 #define COPLINE_INC_WITH_HERELINES \
305 CopLINE_inc(PL_curcop); \
306 if (PL_parser->lex_shared->herelines) \
307 CopLINE(PL_curcop) += PL_parser->lex_shared->herelines, \
308 PL_parser->lex_shared->herelines = 0; \
314 /* how to interpret the pl_yylval associated with the token */
318 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
323 static struct debug_tokens {
325 enum token_type type;
327 } const debug_tokens[] =
329 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
330 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
331 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
332 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
333 { ARROW, TOKENTYPE_NONE, "ARROW" },
334 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
335 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
336 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
337 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
338 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
339 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
340 { DO, TOKENTYPE_NONE, "DO" },
341 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
342 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
343 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
344 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
345 { ELSE, TOKENTYPE_NONE, "ELSE" },
346 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
347 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
348 { FOR, TOKENTYPE_IVAL, "FOR" },
349 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
350 { FORMLBRACK, TOKENTYPE_NONE, "FORMLBRACK" },
351 { FORMRBRACK, TOKENTYPE_NONE, "FORMRBRACK" },
352 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
353 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
354 { FUNC0OP, TOKENTYPE_OPVAL, "FUNC0OP" },
355 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
356 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
357 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
358 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
359 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
360 { IF, TOKENTYPE_IVAL, "IF" },
361 { LABEL, TOKENTYPE_OPVAL, "LABEL" },
362 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
363 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
364 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
365 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
366 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
367 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
368 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
369 { MY, TOKENTYPE_IVAL, "MY" },
370 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
371 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
372 { OROP, TOKENTYPE_IVAL, "OROP" },
373 { OROR, TOKENTYPE_NONE, "OROR" },
374 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
375 { PEG, TOKENTYPE_NONE, "PEG" },
376 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
377 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
378 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
379 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
380 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
381 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
382 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
383 { PREINC, TOKENTYPE_NONE, "PREINC" },
384 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
385 { QWLIST, TOKENTYPE_OPVAL, "QWLIST" },
386 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
387 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
388 { REQUIRE, TOKENTYPE_NONE, "REQUIRE" },
389 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
390 { SUB, TOKENTYPE_NONE, "SUB" },
391 { THING, TOKENTYPE_OPVAL, "THING" },
392 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
393 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
394 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
395 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
396 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
397 { USE, TOKENTYPE_IVAL, "USE" },
398 { WHEN, TOKENTYPE_IVAL, "WHEN" },
399 { WHILE, TOKENTYPE_IVAL, "WHILE" },
400 { WORD, TOKENTYPE_OPVAL, "WORD" },
401 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
402 { 0, TOKENTYPE_NONE, NULL }
405 /* dump the returned token in rv, plus any optional arg in pl_yylval */
408 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
412 PERL_ARGS_ASSERT_TOKEREPORT;
415 const char *name = NULL;
416 enum token_type type = TOKENTYPE_NONE;
417 const struct debug_tokens *p;
418 SV* const report = newSVpvs("<== ");
420 for (p = debug_tokens; p->token; p++) {
421 if (p->token == (int)rv) {
428 Perl_sv_catpv(aTHX_ report, name);
429 else if ((char)rv > ' ' && (char)rv <= '~')
430 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
432 sv_catpvs(report, "EOF");
434 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
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), UTF ? SVf_UTF8 : 0);
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 == ':');
548 t += UTF ? UTF8SKIP(t) : 1)
550 if (t < PL_bufptr && isSPACE(*t))
551 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
552 "\t(Do you need to predeclare %"SVf"?)\n",
553 SVfARG(newSVpvn_flags(PL_oldoldbufptr, (STRLEN)(t - PL_oldoldbufptr),
554 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
558 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
559 "\t(Missing operator before %"SVf"?)\n",
560 SVfARG(newSVpvn_flags(oldbp, (STRLEN)(s - oldbp),
561 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
569 * Complain about missing quote/regexp/heredoc terminator.
570 * If it's called with NULL then it cauterizes the line buffer.
571 * If we're in a delimited string and the delimiter is a control
572 * character, it's reformatted into a two-char sequence like ^C.
577 S_missingterm(pTHX_ char *s)
583 char * const nl = strrchr(s,'\n');
587 else if (isCNTRL(PL_multi_close)) {
589 tmpbuf[1] = (char)toCTRL(PL_multi_close);
594 *tmpbuf = (char)PL_multi_close;
598 q = strchr(s,'"') ? '\'' : '"';
599 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
605 * Check whether the named feature is enabled.
608 Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
611 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
613 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
615 assert(CURRENT_FEATURE_BUNDLE == FEATURE_BUNDLE_CUSTOM);
617 if (namelen > MAX_FEATURE_LEN)
619 memcpy(&he_name[8], name, namelen);
621 return cBOOL(cop_hints_fetch_pvn(PL_curcop, he_name, 8 + namelen, 0,
622 REFCOUNTED_HE_EXISTS));
626 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
627 * utf16-to-utf8-reversed.
630 #ifdef PERL_CR_FILTER
634 const char *s = SvPVX_const(sv);
635 const char * const e = s + SvCUR(sv);
637 PERL_ARGS_ASSERT_STRIP_RETURN;
639 /* outer loop optimized to do nothing if there are no CR-LFs */
641 if (*s++ == '\r' && *s == '\n') {
642 /* hit a CR-LF, need to copy the rest */
646 if (*s == '\r' && s[1] == '\n')
657 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
659 const I32 count = FILTER_READ(idx+1, sv, maxlen);
660 if (count > 0 && !maxlen)
667 =for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
669 Creates and initialises a new lexer/parser state object, supplying
670 a context in which to lex and parse from a new source of Perl code.
671 A pointer to the new state object is placed in L</PL_parser>. An entry
672 is made on the save stack so that upon unwinding the new state object
673 will be destroyed and the former value of L</PL_parser> will be restored.
674 Nothing else need be done to clean up the parsing context.
676 The code to be parsed comes from I<line> and I<rsfp>. I<line>, if
677 non-null, provides a string (in SV form) containing code to be parsed.
678 A copy of the string is made, so subsequent modification of I<line>
679 does not affect parsing. I<rsfp>, if non-null, provides an input stream
680 from which code will be read to be parsed. If both are non-null, the
681 code in I<line> comes first and must consist of complete lines of input,
682 and I<rsfp> supplies the remainder of the source.
684 The I<flags> parameter is reserved for future use. Currently it is only
685 used by perl internally, so extensions should always pass zero.
690 /* LEX_START_SAME_FILTER indicates that this is not a new file, so it
691 can share filters with the current parser.
692 LEX_START_DONT_CLOSE indicates that the file handle wasn't opened by the
693 caller, hence isn't owned by the parser, so shouldn't be closed on parser
694 destruction. This is used to handle the case of defaulting to reading the
695 script from the standard input because no filename was given on the command
696 line (without getting confused by situation where STDIN has been closed, so
697 the script handle is opened on fd 0) */
700 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
703 const char *s = NULL;
704 yy_parser *parser, *oparser;
705 if (flags && flags & ~LEX_START_FLAGS)
706 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
708 /* create and initialise a parser */
710 Newxz(parser, 1, yy_parser);
711 parser->old_parser = oparser = PL_parser;
714 parser->stack = NULL;
716 parser->stack_size = 0;
718 /* on scope exit, free this parser and restore any outer one */
720 parser->saved_curcop = PL_curcop;
722 /* initialise lexer state */
725 parser->curforce = -1;
727 parser->nexttoke = 0;
729 parser->error_count = oparser ? oparser->error_count : 0;
730 parser->copline = NOLINE;
731 parser->lex_state = LEX_NORMAL;
732 parser->expect = XSTATE;
734 parser->rsfp_filters =
735 !(flags & LEX_START_SAME_FILTER) || !oparser
737 : MUTABLE_AV(SvREFCNT_inc(
738 oparser->rsfp_filters
739 ? oparser->rsfp_filters
740 : (oparser->rsfp_filters = newAV())
743 Newx(parser->lex_brackstack, 120, char);
744 Newx(parser->lex_casestack, 12, char);
745 *parser->lex_casestack = '\0';
746 Newxz(parser->lex_shared, 1, LEXSHARED);
750 s = SvPV_const(line, len);
751 parser->linestr = flags & LEX_START_COPIED
752 ? SvREFCNT_inc_simple_NN(line)
753 : newSVpvn_flags(s, len, SvUTF8(line));
754 sv_catpvs(parser->linestr, "\n;");
756 parser->linestr = newSVpvs("\n;");
758 parser->oldoldbufptr =
761 parser->linestart = SvPVX(parser->linestr);
762 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
763 parser->last_lop = parser->last_uni = NULL;
764 parser->lex_flags = flags & (LEX_IGNORE_UTF8_HINTS|LEX_EVALBYTES
765 |LEX_DONT_CLOSE_RSFP);
767 parser->in_pod = parser->filtered = 0;
771 /* delete a parser object */
774 Perl_parser_free(pTHX_ const yy_parser *parser)
776 PERL_ARGS_ASSERT_PARSER_FREE;
778 PL_curcop = parser->saved_curcop;
779 SvREFCNT_dec(parser->linestr);
781 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
782 PerlIO_clearerr(parser->rsfp);
783 else if (parser->rsfp && (!parser->old_parser ||
784 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
785 PerlIO_close(parser->rsfp);
786 SvREFCNT_dec(parser->rsfp_filters);
788 Safefree(parser->lex_brackstack);
789 Safefree(parser->lex_casestack);
790 Safefree(parser->lex_shared);
791 PL_parser = parser->old_parser;
797 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
799 Buffer scalar containing the chunk currently under consideration of the
800 text currently being lexed. This is always a plain string scalar (for
801 which C<SvPOK> is true). It is not intended to be used as a scalar by
802 normal scalar means; instead refer to the buffer directly by the pointer
803 variables described below.
805 The lexer maintains various C<char*> pointers to things in the
806 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
807 reallocated, all of these pointers must be updated. Don't attempt to
808 do this manually, but rather use L</lex_grow_linestr> if you need to
809 reallocate the buffer.
811 The content of the text chunk in the buffer is commonly exactly one
812 complete line of input, up to and including a newline terminator,
813 but there are situations where it is otherwise. The octets of the
814 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
815 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
816 flag on this scalar, which may disagree with it.
818 For direct examination of the buffer, the variable
819 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
820 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
821 of these pointers is usually preferable to examination of the scalar
822 through normal scalar means.
824 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
826 Direct pointer to the end of the chunk of text currently being lexed, the
827 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
828 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
829 always located at the end of the buffer, and does not count as part of
830 the buffer's contents.
832 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
834 Points to the current position of lexing inside the lexer buffer.
835 Characters around this point may be freely examined, within
836 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
837 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
838 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
840 Lexing code (whether in the Perl core or not) moves this pointer past
841 the characters that it consumes. It is also expected to perform some
842 bookkeeping whenever a newline character is consumed. This movement
843 can be more conveniently performed by the function L</lex_read_to>,
844 which handles newlines appropriately.
846 Interpretation of the buffer's octets can be abstracted out by
847 using the slightly higher-level functions L</lex_peek_unichar> and
848 L</lex_read_unichar>.
850 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
852 Points to the start of the current line inside the lexer buffer.
853 This is useful for indicating at which column an error occurred, and
854 not much else. This must be updated by any lexing code that consumes
855 a newline; the function L</lex_read_to> handles this detail.
861 =for apidoc Amx|bool|lex_bufutf8
863 Indicates whether the octets in the lexer buffer
864 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
865 of Unicode characters. If not, they should be interpreted as Latin-1
866 characters. This is analogous to the C<SvUTF8> flag for scalars.
868 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
869 contains valid UTF-8. Lexing code must be robust in the face of invalid
872 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
873 is significant, but not the whole story regarding the input character
874 encoding. Normally, when a file is being read, the scalar contains octets
875 and its C<SvUTF8> flag is off, but the octets should be interpreted as
876 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
877 however, the scalar may have the C<SvUTF8> flag on, and in this case its
878 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
879 is in effect. This logic may change in the future; use this function
880 instead of implementing the logic yourself.
886 Perl_lex_bufutf8(pTHX)
892 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
894 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
895 at least I<len> octets (including terminating NUL). Returns a
896 pointer to the reallocated buffer. This is necessary before making
897 any direct modification of the buffer that would increase its length.
898 L</lex_stuff_pvn> provides a more convenient way to insert text into
901 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
902 this function updates all of the lexer's variables that point directly
909 Perl_lex_grow_linestr(pTHX_ STRLEN len)
913 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
914 STRLEN linestart_pos, last_uni_pos, last_lop_pos, re_eval_start_pos;
915 linestr = PL_parser->linestr;
916 buf = SvPVX(linestr);
917 if (len <= SvLEN(linestr))
919 bufend_pos = PL_parser->bufend - buf;
920 bufptr_pos = PL_parser->bufptr - buf;
921 oldbufptr_pos = PL_parser->oldbufptr - buf;
922 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
923 linestart_pos = PL_parser->linestart - buf;
924 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
925 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
926 re_eval_start_pos = PL_parser->lex_shared->re_eval_start ?
927 PL_parser->lex_shared->re_eval_start - buf : 0;
929 buf = sv_grow(linestr, len);
931 PL_parser->bufend = buf + bufend_pos;
932 PL_parser->bufptr = buf + bufptr_pos;
933 PL_parser->oldbufptr = buf + oldbufptr_pos;
934 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
935 PL_parser->linestart = buf + linestart_pos;
936 if (PL_parser->last_uni)
937 PL_parser->last_uni = buf + last_uni_pos;
938 if (PL_parser->last_lop)
939 PL_parser->last_lop = buf + last_lop_pos;
940 if (PL_parser->lex_shared->re_eval_start)
941 PL_parser->lex_shared->re_eval_start = buf + re_eval_start_pos;
946 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
948 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
949 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
950 reallocating the buffer if necessary. This means that lexing code that
951 runs later will see the characters as if they had appeared in the input.
952 It is not recommended to do this as part of normal parsing, and most
953 uses of this facility run the risk of the inserted characters being
954 interpreted in an unintended manner.
956 The string to be inserted is represented by I<len> octets starting
957 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
958 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
959 The characters are recoded for the lexer buffer, according to how the
960 buffer is currently being interpreted (L</lex_bufutf8>). If a string
961 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
962 function is more convenient.
968 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
972 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
973 if (flags & ~(LEX_STUFF_UTF8))
974 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
976 if (flags & LEX_STUFF_UTF8) {
980 const char *p, *e = pv+len;
981 for (p = pv; p != e; p++)
982 highhalf += !!(((U8)*p) & 0x80);
985 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
986 bufptr = PL_parser->bufptr;
987 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
988 SvCUR_set(PL_parser->linestr,
989 SvCUR(PL_parser->linestr) + len+highhalf);
990 PL_parser->bufend += len+highhalf;
991 for (p = pv; p != e; p++) {
994 *bufptr++ = (char)(0xc0 | (c >> 6));
995 *bufptr++ = (char)(0x80 | (c & 0x3f));
1002 if (flags & LEX_STUFF_UTF8) {
1003 STRLEN highhalf = 0;
1004 const char *p, *e = pv+len;
1005 for (p = pv; p != e; p++) {
1008 Perl_croak(aTHX_ "Lexing code attempted to stuff "
1009 "non-Latin-1 character into Latin-1 input");
1010 } else if (c >= 0xc2 && p+1 != e &&
1011 (((U8)p[1]) & 0xc0) == 0x80) {
1014 } else if (c >= 0x80) {
1015 /* malformed UTF-8 */
1017 SAVESPTR(PL_warnhook);
1018 PL_warnhook = PERL_WARNHOOK_FATAL;
1019 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
1025 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
1026 bufptr = PL_parser->bufptr;
1027 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
1028 SvCUR_set(PL_parser->linestr,
1029 SvCUR(PL_parser->linestr) + len-highhalf);
1030 PL_parser->bufend += len-highhalf;
1031 for (p = pv; p != e; p++) {
1034 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1037 *bufptr++ = (char)c;
1042 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1043 bufptr = PL_parser->bufptr;
1044 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1045 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1046 PL_parser->bufend += len;
1047 Copy(pv, bufptr, len, char);
1053 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1055 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1056 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1057 reallocating the buffer if necessary. This means that lexing code that
1058 runs later will see the characters as if they had appeared in the input.
1059 It is not recommended to do this as part of normal parsing, and most
1060 uses of this facility run the risk of the inserted characters being
1061 interpreted in an unintended manner.
1063 The string to be inserted is represented by octets starting at I<pv>
1064 and continuing to the first nul. These octets are interpreted as either
1065 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1066 in I<flags>. The characters are recoded for the lexer buffer, according
1067 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1068 If it is not convenient to nul-terminate a string to be inserted, the
1069 L</lex_stuff_pvn> function is more appropriate.
1075 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1077 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1078 lex_stuff_pvn(pv, strlen(pv), flags);
1082 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1084 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1085 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1086 reallocating the buffer if necessary. This means that lexing code that
1087 runs later will see the characters as if they had appeared in the input.
1088 It is not recommended to do this as part of normal parsing, and most
1089 uses of this facility run the risk of the inserted characters being
1090 interpreted in an unintended manner.
1092 The string to be inserted is the string value of I<sv>. The characters
1093 are recoded for the lexer buffer, according to how the buffer is currently
1094 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1095 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1096 need to construct a scalar.
1102 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1106 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1108 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1110 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1114 =for apidoc Amx|void|lex_unstuff|char *ptr
1116 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1117 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1118 This hides the discarded text from any lexing code that runs later,
1119 as if the text had never appeared.
1121 This is not the normal way to consume lexed text. For that, use
1128 Perl_lex_unstuff(pTHX_ char *ptr)
1132 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1133 buf = PL_parser->bufptr;
1135 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1138 bufend = PL_parser->bufend;
1140 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1141 unstuff_len = ptr - buf;
1142 Move(ptr, buf, bufend+1-ptr, char);
1143 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1144 PL_parser->bufend = bufend - unstuff_len;
1148 =for apidoc Amx|void|lex_read_to|char *ptr
1150 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1151 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1152 performing the correct bookkeeping whenever a newline character is passed.
1153 This is the normal way to consume lexed text.
1155 Interpretation of the buffer's octets can be abstracted out by
1156 using the slightly higher-level functions L</lex_peek_unichar> and
1157 L</lex_read_unichar>.
1163 Perl_lex_read_to(pTHX_ char *ptr)
1166 PERL_ARGS_ASSERT_LEX_READ_TO;
1167 s = PL_parser->bufptr;
1168 if (ptr < s || ptr > PL_parser->bufend)
1169 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1170 for (; s != ptr; s++)
1172 COPLINE_INC_WITH_HERELINES;
1173 PL_parser->linestart = s+1;
1175 PL_parser->bufptr = ptr;
1179 =for apidoc Amx|void|lex_discard_to|char *ptr
1181 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1182 up to I<ptr>. The remaining content of the buffer will be moved, and
1183 all pointers into the buffer updated appropriately. I<ptr> must not
1184 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1185 it is not permitted to discard text that has yet to be lexed.
1187 Normally it is not necessarily to do this directly, because it suffices to
1188 use the implicit discarding behaviour of L</lex_next_chunk> and things
1189 based on it. However, if a token stretches across multiple lines,
1190 and the lexing code has kept multiple lines of text in the buffer for
1191 that purpose, then after completion of the token it would be wise to
1192 explicitly discard the now-unneeded earlier lines, to avoid future
1193 multi-line tokens growing the buffer without bound.
1199 Perl_lex_discard_to(pTHX_ char *ptr)
1203 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1204 buf = SvPVX(PL_parser->linestr);
1206 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1209 if (ptr > PL_parser->bufptr)
1210 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1211 discard_len = ptr - buf;
1212 if (PL_parser->oldbufptr < ptr)
1213 PL_parser->oldbufptr = ptr;
1214 if (PL_parser->oldoldbufptr < ptr)
1215 PL_parser->oldoldbufptr = ptr;
1216 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1217 PL_parser->last_uni = NULL;
1218 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1219 PL_parser->last_lop = NULL;
1220 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1221 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1222 PL_parser->bufend -= discard_len;
1223 PL_parser->bufptr -= discard_len;
1224 PL_parser->oldbufptr -= discard_len;
1225 PL_parser->oldoldbufptr -= discard_len;
1226 if (PL_parser->last_uni)
1227 PL_parser->last_uni -= discard_len;
1228 if (PL_parser->last_lop)
1229 PL_parser->last_lop -= discard_len;
1233 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1235 Reads in the next chunk of text to be lexed, appending it to
1236 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1237 looked to the end of the current chunk and wants to know more. It is
1238 usual, but not necessary, for lexing to have consumed the entirety of
1239 the current chunk at this time.
1241 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1242 chunk (i.e., the current chunk has been entirely consumed), normally the
1243 current chunk will be discarded at the same time that the new chunk is
1244 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1245 will not be discarded. If the current chunk has not been entirely
1246 consumed, then it will not be discarded regardless of the flag.
1248 Returns true if some new text was added to the buffer, or false if the
1249 buffer has reached the end of the input text.
1254 #define LEX_FAKE_EOF 0x80000000
1255 #define LEX_NO_TERM 0x40000000
1258 Perl_lex_next_chunk(pTHX_ U32 flags)
1262 STRLEN old_bufend_pos, new_bufend_pos;
1263 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1264 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1265 bool got_some_for_debugger = 0;
1267 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF|LEX_NO_TERM))
1268 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1269 linestr = PL_parser->linestr;
1270 buf = SvPVX(linestr);
1271 if (!(flags & LEX_KEEP_PREVIOUS) &&
1272 PL_parser->bufptr == PL_parser->bufend) {
1273 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1275 if (PL_parser->last_uni != PL_parser->bufend)
1276 PL_parser->last_uni = NULL;
1277 if (PL_parser->last_lop != PL_parser->bufend)
1278 PL_parser->last_lop = NULL;
1279 last_uni_pos = last_lop_pos = 0;
1283 old_bufend_pos = PL_parser->bufend - buf;
1284 bufptr_pos = PL_parser->bufptr - buf;
1285 oldbufptr_pos = PL_parser->oldbufptr - buf;
1286 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1287 linestart_pos = PL_parser->linestart - buf;
1288 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1289 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1291 if (flags & LEX_FAKE_EOF) {
1293 } else if (!PL_parser->rsfp && !PL_parser->filtered) {
1295 } else if (filter_gets(linestr, old_bufend_pos)) {
1297 got_some_for_debugger = 1;
1298 } else if (flags & LEX_NO_TERM) {
1301 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1302 sv_setpvs(linestr, "");
1304 /* End of real input. Close filehandle (unless it was STDIN),
1305 * then add implicit termination.
1307 if (PL_parser->lex_flags & LEX_DONT_CLOSE_RSFP)
1308 PerlIO_clearerr(PL_parser->rsfp);
1309 else if (PL_parser->rsfp)
1310 (void)PerlIO_close(PL_parser->rsfp);
1311 PL_parser->rsfp = NULL;
1312 PL_parser->in_pod = PL_parser->filtered = 0;
1314 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1317 if (!PL_in_eval && PL_minus_p) {
1319 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1320 PL_minus_n = PL_minus_p = 0;
1321 } else if (!PL_in_eval && PL_minus_n) {
1322 sv_catpvs(linestr, /*{*/";}");
1325 sv_catpvs(linestr, ";");
1328 buf = SvPVX(linestr);
1329 new_bufend_pos = SvCUR(linestr);
1330 PL_parser->bufend = buf + new_bufend_pos;
1331 PL_parser->bufptr = buf + bufptr_pos;
1332 PL_parser->oldbufptr = buf + oldbufptr_pos;
1333 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1334 PL_parser->linestart = buf + linestart_pos;
1335 if (PL_parser->last_uni)
1336 PL_parser->last_uni = buf + last_uni_pos;
1337 if (PL_parser->last_lop)
1338 PL_parser->last_lop = buf + last_lop_pos;
1339 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1340 PL_curstash != PL_debstash) {
1341 /* debugger active and we're not compiling the debugger code,
1342 * so store the line into the debugger's array of lines
1344 update_debugger_info(NULL, buf+old_bufend_pos,
1345 new_bufend_pos-old_bufend_pos);
1351 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1353 Looks ahead one (Unicode) character in the text currently being lexed.
1354 Returns the codepoint (unsigned integer value) of the next character,
1355 or -1 if lexing has reached the end of the input text. To consume the
1356 peeked character, use L</lex_read_unichar>.
1358 If the next character is in (or extends into) the next chunk of input
1359 text, the next chunk will be read in. Normally the current chunk will be
1360 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1361 then the current chunk will not be discarded.
1363 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1364 is encountered, an exception is generated.
1370 Perl_lex_peek_unichar(pTHX_ U32 flags)
1374 if (flags & ~(LEX_KEEP_PREVIOUS))
1375 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1376 s = PL_parser->bufptr;
1377 bufend = PL_parser->bufend;
1383 if (!lex_next_chunk(flags))
1385 s = PL_parser->bufptr;
1386 bufend = PL_parser->bufend;
1392 len = PL_utf8skip[head];
1393 while ((STRLEN)(bufend-s) < len) {
1394 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1396 s = PL_parser->bufptr;
1397 bufend = PL_parser->bufend;
1400 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1401 if (retlen == (STRLEN)-1) {
1402 /* malformed UTF-8 */
1404 SAVESPTR(PL_warnhook);
1405 PL_warnhook = PERL_WARNHOOK_FATAL;
1406 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1412 if (!lex_next_chunk(flags))
1414 s = PL_parser->bufptr;
1421 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1423 Reads the next (Unicode) character in the text currently being lexed.
1424 Returns the codepoint (unsigned integer value) of the character read,
1425 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1426 if lexing has reached the end of the input text. To non-destructively
1427 examine the next character, use L</lex_peek_unichar> instead.
1429 If the next character is in (or extends into) the next chunk of input
1430 text, the next chunk will be read in. Normally the current chunk will be
1431 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1432 then the current chunk will not be discarded.
1434 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1435 is encountered, an exception is generated.
1441 Perl_lex_read_unichar(pTHX_ U32 flags)
1444 if (flags & ~(LEX_KEEP_PREVIOUS))
1445 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1446 c = lex_peek_unichar(flags);
1449 COPLINE_INC_WITH_HERELINES;
1451 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1453 ++(PL_parser->bufptr);
1459 =for apidoc Amx|void|lex_read_space|U32 flags
1461 Reads optional spaces, in Perl style, in the text currently being
1462 lexed. The spaces may include ordinary whitespace characters and
1463 Perl-style comments. C<#line> directives are processed if encountered.
1464 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1465 at a non-space character (or the end of the input text).
1467 If spaces extend into the next chunk of input text, the next chunk will
1468 be read in. Normally the current chunk will be discarded at the same
1469 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1470 chunk will not be discarded.
1475 #define LEX_NO_NEXT_CHUNK 0x80000000
1478 Perl_lex_read_space(pTHX_ U32 flags)
1481 bool need_incline = 0;
1482 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1483 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1486 sv_free(PL_skipwhite);
1487 PL_skipwhite = NULL;
1490 PL_skipwhite = newSVpvs("");
1491 #endif /* PERL_MAD */
1492 s = PL_parser->bufptr;
1493 bufend = PL_parser->bufend;
1499 } while (!(c == '\n' || (c == 0 && s == bufend)));
1500 } else if (c == '\n') {
1502 PL_parser->linestart = s;
1507 } else if (isSPACE(c)) {
1509 } else if (c == 0 && s == bufend) {
1513 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1514 #endif /* PERL_MAD */
1515 if (flags & LEX_NO_NEXT_CHUNK)
1517 PL_parser->bufptr = s;
1518 COPLINE_INC_WITH_HERELINES;
1519 got_more = lex_next_chunk(flags);
1520 CopLINE_dec(PL_curcop);
1521 s = PL_parser->bufptr;
1522 bufend = PL_parser->bufend;
1525 if (need_incline && PL_parser->rsfp) {
1535 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1536 #endif /* PERL_MAD */
1537 PL_parser->bufptr = s;
1542 * This subroutine has nothing to do with tilting, whether at windmills
1543 * or pinball tables. Its name is short for "increment line". It
1544 * increments the current line number in CopLINE(PL_curcop) and checks
1545 * to see whether the line starts with a comment of the form
1546 * # line 500 "foo.pm"
1547 * If so, it sets the current line number and file to the values in the comment.
1551 S_incline(pTHX_ const char *s)
1559 PERL_ARGS_ASSERT_INCLINE;
1561 COPLINE_INC_WITH_HERELINES;
1564 while (SPACE_OR_TAB(*s))
1566 if (strnEQ(s, "line", 4))
1570 if (SPACE_OR_TAB(*s))
1574 while (SPACE_OR_TAB(*s))
1582 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1584 while (SPACE_OR_TAB(*s))
1586 if (*s == '"' && (t = strchr(s+1, '"'))) {
1592 while (!isSPACE(*t))
1596 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1598 if (*e != '\n' && *e != '\0')
1599 return; /* false alarm */
1601 line_num = atoi(n)-1;
1604 const STRLEN len = t - s;
1605 SV *const temp_sv = CopFILESV(PL_curcop);
1610 cf = SvPVX(temp_sv);
1611 tmplen = SvCUR(temp_sv);
1617 if (!PL_rsfp && !PL_parser->filtered) {
1618 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1619 * to *{"::_<newfilename"} */
1620 /* However, the long form of evals is only turned on by the
1621 debugger - usually they're "(eval %lu)" */
1625 STRLEN tmplen2 = len;
1626 if (tmplen + 2 <= sizeof smallbuf)
1629 Newx(tmpbuf, tmplen + 2, char);
1632 memcpy(tmpbuf + 2, cf, tmplen);
1634 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1639 if (tmplen2 + 2 <= sizeof smallbuf)
1642 Newx(tmpbuf2, tmplen2 + 2, char);
1644 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1645 /* Either they malloc'd it, or we malloc'd it,
1646 so no prefix is present in ours. */
1651 memcpy(tmpbuf2 + 2, s, tmplen2);
1654 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1656 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1657 /* adjust ${"::_<newfilename"} to store the new file name */
1658 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1659 /* The line number may differ. If that is the case,
1660 alias the saved lines that are in the array.
1661 Otherwise alias the whole array. */
1662 if (CopLINE(PL_curcop) == line_num) {
1663 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1664 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1666 else if (GvAV(*gvp)) {
1667 AV * const av = GvAV(*gvp);
1668 const I32 start = CopLINE(PL_curcop)+1;
1669 I32 items = AvFILLp(av) - start;
1671 AV * const av2 = GvAVn(gv2);
1672 SV **svp = AvARRAY(av) + start;
1673 I32 l = (I32)line_num+1;
1675 av_store(av2, l++, SvREFCNT_inc(*svp++));
1680 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1682 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1684 CopFILE_free(PL_curcop);
1685 CopFILE_setn(PL_curcop, s, len);
1687 CopLINE_set(PL_curcop, line_num);
1691 /* skip space before PL_thistoken */
1694 S_skipspace0(pTHX_ register char *s)
1696 PERL_ARGS_ASSERT_SKIPSPACE0;
1703 PL_thiswhite = newSVpvs("");
1704 sv_catsv(PL_thiswhite, PL_skipwhite);
1705 sv_free(PL_skipwhite);
1708 PL_realtokenstart = s - SvPVX(PL_linestr);
1712 /* skip space after PL_thistoken */
1715 S_skipspace1(pTHX_ register char *s)
1717 const char *start = s;
1718 I32 startoff = start - SvPVX(PL_linestr);
1720 PERL_ARGS_ASSERT_SKIPSPACE1;
1725 start = SvPVX(PL_linestr) + startoff;
1726 if (!PL_thistoken && PL_realtokenstart >= 0) {
1727 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1728 PL_thistoken = newSVpvn(tstart, start - tstart);
1730 PL_realtokenstart = -1;
1733 PL_nextwhite = newSVpvs("");
1734 sv_catsv(PL_nextwhite, PL_skipwhite);
1735 sv_free(PL_skipwhite);
1742 S_skipspace2(pTHX_ register char *s, SV **svp)
1745 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1746 const I32 startoff = s - SvPVX(PL_linestr);
1748 PERL_ARGS_ASSERT_SKIPSPACE2;
1751 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1752 if (!PL_madskills || !svp)
1754 start = SvPVX(PL_linestr) + startoff;
1755 if (!PL_thistoken && PL_realtokenstart >= 0) {
1756 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1757 PL_thistoken = newSVpvn(tstart, start - tstart);
1758 PL_realtokenstart = -1;
1762 *svp = newSVpvs("");
1763 sv_setsv(*svp, PL_skipwhite);
1764 sv_free(PL_skipwhite);
1773 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1775 AV *av = CopFILEAVx(PL_curcop);
1777 SV * const sv = newSV_type(SVt_PVMG);
1779 sv_setsv(sv, orig_sv);
1781 sv_setpvn(sv, buf, len);
1784 av_store(av, (I32)CopLINE(PL_curcop), sv);
1790 * Called to gobble the appropriate amount and type of whitespace.
1791 * Skips comments as well.
1795 S_skipspace(pTHX_ register char *s)
1799 #endif /* PERL_MAD */
1800 PERL_ARGS_ASSERT_SKIPSPACE;
1803 sv_free(PL_skipwhite);
1804 PL_skipwhite = NULL;
1806 #endif /* PERL_MAD */
1807 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1808 while (s < PL_bufend && SPACE_OR_TAB(*s))
1811 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1813 lex_read_space(LEX_KEEP_PREVIOUS |
1814 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1815 LEX_NO_NEXT_CHUNK : 0));
1817 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1818 if (PL_linestart > PL_bufptr)
1819 PL_bufptr = PL_linestart;
1824 PL_skipwhite = newSVpvn(start, s-start);
1825 #endif /* PERL_MAD */
1831 * Check the unary operators to ensure there's no ambiguity in how they're
1832 * used. An ambiguous piece of code would be:
1834 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1835 * the +5 is its argument.
1845 if (PL_oldoldbufptr != PL_last_uni)
1847 while (isSPACE(*PL_last_uni))
1850 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1852 if ((t = strchr(s, '(')) && t < PL_bufptr)
1855 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1856 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1857 (int)(s - PL_last_uni), PL_last_uni);
1861 * LOP : macro to build a list operator. Its behaviour has been replaced
1862 * with a subroutine, S_lop() for which LOP is just another name.
1865 #define LOP(f,x) return lop(f,x,s)
1869 * Build a list operator (or something that might be one). The rules:
1870 * - if we have a next token, then it's a list operator [why?]
1871 * - if the next thing is an opening paren, then it's a function
1872 * - else it's a list operator
1876 S_lop(pTHX_ I32 f, int x, char *s)
1880 PERL_ARGS_ASSERT_LOP;
1886 PL_last_lop = PL_oldbufptr;
1887 PL_last_lop_op = (OPCODE)f;
1896 return REPORT(FUNC);
1899 return REPORT(FUNC);
1902 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1903 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1904 return REPORT(LSTOP);
1911 * Sets up for an eventual force_next(). start_force(0) basically does
1912 * an unshift, while start_force(-1) does a push. yylex removes items
1917 S_start_force(pTHX_ int where)
1921 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1922 where = PL_lasttoke;
1923 assert(PL_curforce < 0 || PL_curforce == where);
1924 if (PL_curforce != where) {
1925 for (i = PL_lasttoke; i > where; --i) {
1926 PL_nexttoke[i] = PL_nexttoke[i-1];
1930 if (PL_curforce < 0) /* in case of duplicate start_force() */
1931 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1932 PL_curforce = where;
1935 curmad('^', newSVpvs(""));
1936 CURMAD('_', PL_nextwhite);
1941 S_curmad(pTHX_ char slot, SV *sv)
1947 if (PL_curforce < 0)
1948 where = &PL_thismad;
1950 where = &PL_nexttoke[PL_curforce].next_mad;
1956 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1958 else if (PL_encoding) {
1959 sv_recode_to_utf8(sv, PL_encoding);
1964 /* keep a slot open for the head of the list? */
1965 if (slot != '_' && *where && (*where)->mad_key == '^') {
1966 (*where)->mad_key = slot;
1967 sv_free(MUTABLE_SV(((*where)->mad_val)));
1968 (*where)->mad_val = (void*)sv;
1971 addmad(newMADsv(slot, sv), where, 0);
1974 # define start_force(where) NOOP
1975 # define curmad(slot, sv) NOOP
1980 * When the lexer realizes it knows the next token (for instance,
1981 * it is reordering tokens for the parser) then it can call S_force_next
1982 * to know what token to return the next time the lexer is called. Caller
1983 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1984 * and possibly PL_expect to ensure the lexer handles the token correctly.
1988 S_force_next(pTHX_ I32 type)
1993 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1994 tokereport(type, &NEXTVAL_NEXTTOKE);
1997 /* Don’t let opslab_force_free snatch it */
1998 if (S_is_opval_token(type & 0xffff) && NEXTVAL_NEXTTOKE.opval) {
1999 assert(!NEXTVAL_NEXTTOKE.opval->op_savefree);
2000 NEXTVAL_NEXTTOKE.opval->op_savefree = 1;
2003 if (PL_curforce < 0)
2004 start_force(PL_lasttoke);
2005 PL_nexttoke[PL_curforce].next_type = type;
2006 if (PL_lex_state != LEX_KNOWNEXT)
2007 PL_lex_defer = PL_lex_state;
2008 PL_lex_state = LEX_KNOWNEXT;
2009 PL_lex_expect = PL_expect;
2012 PL_nexttype[PL_nexttoke] = type;
2014 if (PL_lex_state != LEX_KNOWNEXT) {
2015 PL_lex_defer = PL_lex_state;
2016 PL_lex_expect = PL_expect;
2017 PL_lex_state = LEX_KNOWNEXT;
2025 int yyc = PL_parser->yychar;
2026 if (yyc != YYEMPTY) {
2029 NEXTVAL_NEXTTOKE = PL_parser->yylval;
2030 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
2031 PL_lex_allbrackets--;
2033 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
2034 } else if (yyc == '('/*)*/) {
2035 PL_lex_allbrackets--;
2040 PL_parser->yychar = YYEMPTY;
2045 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
2048 SV * const sv = newSVpvn_utf8(start, len,
2051 && !is_ascii_string((const U8*)start, len)
2052 && is_utf8_string((const U8*)start, len));
2058 * When the lexer knows the next thing is a word (for instance, it has
2059 * just seen -> and it knows that the next char is a word char, then
2060 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2064 * char *start : buffer position (must be within PL_linestr)
2065 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2066 * int check_keyword : if true, Perl checks to make sure the word isn't
2067 * a keyword (do this if the word is a label, e.g. goto FOO)
2068 * int allow_pack : if true, : characters will also be allowed (require,
2069 * use, etc. do this)
2070 * int allow_initial_tick : used by the "sub" lexer only.
2074 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
2080 PERL_ARGS_ASSERT_FORCE_WORD;
2082 start = SKIPSPACE1(start);
2084 if (isIDFIRST_lazy_if(s,UTF) ||
2085 (allow_pack && *s == ':') ||
2086 (allow_initial_tick && *s == '\'') )
2088 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2089 if (check_keyword && keyword(PL_tokenbuf, len, 0))
2091 start_force(PL_curforce);
2093 curmad('X', newSVpvn(start,s-start));
2094 if (token == METHOD) {
2099 PL_expect = XOPERATOR;
2103 curmad('g', newSVpvs( "forced" ));
2104 NEXTVAL_NEXTTOKE.opval
2105 = (OP*)newSVOP(OP_CONST,0,
2106 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2107 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2115 * Called when the lexer wants $foo *foo &foo etc, but the program
2116 * text only contains the "foo" portion. The first argument is a pointer
2117 * to the "foo", and the second argument is the type symbol to prefix.
2118 * Forces the next token to be a "WORD".
2119 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2123 S_force_ident(pTHX_ register const char *s, int kind)
2127 PERL_ARGS_ASSERT_FORCE_IDENT;
2130 const STRLEN len = strlen(s);
2131 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn_flags(s, len,
2132 UTF ? SVf_UTF8 : 0));
2133 start_force(PL_curforce);
2134 NEXTVAL_NEXTTOKE.opval = o;
2137 o->op_private = OPpCONST_ENTERED;
2138 /* XXX see note in pp_entereval() for why we forgo typo
2139 warnings if the symbol must be introduced in an eval.
2141 gv_fetchpvn_flags(s, len,
2142 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2143 : GV_ADD) | ( UTF ? SVf_UTF8 : 0 ),
2144 kind == '$' ? SVt_PV :
2145 kind == '@' ? SVt_PVAV :
2146 kind == '%' ? SVt_PVHV :
2154 S_force_ident_maybe_lex(pTHX_ char pit)
2156 start_force(PL_curforce);
2157 NEXTVAL_NEXTTOKE.ival = pit;
2162 Perl_str_to_version(pTHX_ SV *sv)
2167 const char *start = SvPV_const(sv,len);
2168 const char * const end = start + len;
2169 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2171 PERL_ARGS_ASSERT_STR_TO_VERSION;
2173 while (start < end) {
2177 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2182 retval += ((NV)n)/nshift;
2191 * Forces the next token to be a version number.
2192 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2193 * and if "guessing" is TRUE, then no new token is created (and the caller
2194 * must use an alternative parsing method).
2198 S_force_version(pTHX_ char *s, int guessing)
2204 I32 startoff = s - SvPVX(PL_linestr);
2207 PERL_ARGS_ASSERT_FORCE_VERSION;
2215 while (isDIGIT(*d) || *d == '_' || *d == '.')
2219 start_force(PL_curforce);
2220 curmad('X', newSVpvn(s,d-s));
2223 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2225 #ifdef USE_LOCALE_NUMERIC
2226 char *loc = savepv(setlocale(LC_NUMERIC, NULL));
2227 setlocale(LC_NUMERIC, "C");
2229 s = scan_num(s, &pl_yylval);
2230 #ifdef USE_LOCALE_NUMERIC
2231 setlocale(LC_NUMERIC, loc);
2234 version = pl_yylval.opval;
2235 ver = cSVOPx(version)->op_sv;
2236 if (SvPOK(ver) && !SvNIOK(ver)) {
2237 SvUPGRADE(ver, SVt_PVNV);
2238 SvNV_set(ver, str_to_version(ver));
2239 SvNOK_on(ver); /* hint that it is a version */
2242 else if (guessing) {
2245 sv_free(PL_nextwhite); /* let next token collect whitespace */
2247 s = SvPVX(PL_linestr) + startoff;
2255 if (PL_madskills && !version) {
2256 sv_free(PL_nextwhite); /* let next token collect whitespace */
2258 s = SvPVX(PL_linestr) + startoff;
2261 /* NOTE: The parser sees the package name and the VERSION swapped */
2262 start_force(PL_curforce);
2263 NEXTVAL_NEXTTOKE.opval = version;
2270 * S_force_strict_version
2271 * Forces the next token to be a version number using strict syntax rules.
2275 S_force_strict_version(pTHX_ char *s)
2280 I32 startoff = s - SvPVX(PL_linestr);
2282 const char *errstr = NULL;
2284 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2286 while (isSPACE(*s)) /* leading whitespace */
2289 if (is_STRICT_VERSION(s,&errstr)) {
2291 s = (char *)scan_version(s, ver, 0);
2292 version = newSVOP(OP_CONST, 0, ver);
2294 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2295 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2299 yyerror(errstr); /* version required */
2304 if (PL_madskills && !version) {
2305 sv_free(PL_nextwhite); /* let next token collect whitespace */
2307 s = SvPVX(PL_linestr) + startoff;
2310 /* NOTE: The parser sees the package name and the VERSION swapped */
2311 start_force(PL_curforce);
2312 NEXTVAL_NEXTTOKE.opval = version;
2320 * Tokenize a quoted string passed in as an SV. It finds the next
2321 * chunk, up to end of string or a backslash. It may make a new
2322 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2327 S_tokeq(pTHX_ SV *sv)
2336 PERL_ARGS_ASSERT_TOKEQ;
2341 s = SvPV_force(sv, len);
2342 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2345 /* This is relying on the SV being "well formed" with a trailing '\0' */
2346 while (s < send && !(*s == '\\' && s[1] == '\\'))
2351 if ( PL_hints & HINT_NEW_STRING ) {
2352 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2356 if (s + 1 < send && (s[1] == '\\'))
2357 s++; /* all that, just for this */
2362 SvCUR_set(sv, d - SvPVX_const(sv));
2364 if ( PL_hints & HINT_NEW_STRING )
2365 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2370 * Now come three functions related to double-quote context,
2371 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2372 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2373 * interact with PL_lex_state, and create fake ( ... ) argument lists
2374 * to handle functions and concatenation.
2378 * stringify ( const[foo] concat lcfirst ( const[bar] ) )
2383 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2385 * Pattern matching will set PL_lex_op to the pattern-matching op to
2386 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2388 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2390 * Everything else becomes a FUNC.
2392 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2393 * had an OP_CONST or OP_READLINE). This just sets us up for a
2394 * call to S_sublex_push().
2398 S_sublex_start(pTHX)
2401 const I32 op_type = pl_yylval.ival;
2403 if (op_type == OP_NULL) {
2404 pl_yylval.opval = PL_lex_op;
2408 if (op_type == OP_CONST || op_type == OP_READLINE) {
2409 SV *sv = tokeq(PL_lex_stuff);
2411 if (SvTYPE(sv) == SVt_PVIV) {
2412 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2414 const char * const p = SvPV_const(sv, len);
2415 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2419 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2420 PL_lex_stuff = NULL;
2421 /* Allow <FH> // "foo" */
2422 if (op_type == OP_READLINE)
2423 PL_expect = XTERMORDORDOR;
2426 else if (op_type == OP_BACKTICK && PL_lex_op) {
2427 /* readpipe() vas overriden */
2428 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2429 pl_yylval.opval = PL_lex_op;
2431 PL_lex_stuff = NULL;
2435 PL_sublex_info.super_state = PL_lex_state;
2436 PL_sublex_info.sub_inwhat = (U16)op_type;
2437 PL_sublex_info.sub_op = PL_lex_op;
2438 PL_lex_state = LEX_INTERPPUSH;
2442 pl_yylval.opval = PL_lex_op;
2452 * Create a new scope to save the lexing state. The scope will be
2453 * ended in S_sublex_done. Returns a '(', starting the function arguments
2454 * to the uc, lc, etc. found before.
2455 * Sets PL_lex_state to LEX_INTERPCONCAT.
2465 PL_lex_state = PL_sublex_info.super_state;
2466 SAVEBOOL(PL_lex_dojoin);
2467 SAVEI32(PL_lex_brackets);
2468 SAVEI32(PL_lex_allbrackets);
2469 SAVEI32(PL_lex_formbrack);
2470 SAVEI8(PL_lex_fakeeof);
2471 SAVEI32(PL_lex_casemods);
2472 SAVEI32(PL_lex_starts);
2473 SAVEI8(PL_lex_state);
2474 SAVESPTR(PL_lex_repl);
2475 SAVEVPTR(PL_lex_inpat);
2476 SAVEI16(PL_lex_inwhat);
2477 SAVECOPLINE(PL_curcop);
2478 SAVEPPTR(PL_bufptr);
2479 SAVEPPTR(PL_bufend);
2480 SAVEPPTR(PL_oldbufptr);
2481 SAVEPPTR(PL_oldoldbufptr);
2482 SAVEPPTR(PL_last_lop);
2483 SAVEPPTR(PL_last_uni);
2484 SAVEPPTR(PL_linestart);
2485 SAVESPTR(PL_linestr);
2486 SAVEGENERICPV(PL_lex_brackstack);
2487 SAVEGENERICPV(PL_lex_casestack);
2488 SAVEGENERICPV(PL_parser->lex_shared);
2490 /* The here-doc parser needs to be able to peek into outer lexing
2491 scopes to find the body of the here-doc. So we put PL_linestr and
2492 PL_bufptr into lex_shared, to ‘share’ those values.
2494 PL_parser->lex_shared->ls_linestr = PL_linestr;
2495 PL_parser->lex_shared->ls_bufptr = PL_bufptr;
2497 PL_linestr = PL_lex_stuff;
2498 PL_lex_repl = PL_sublex_info.repl;
2499 PL_lex_stuff = NULL;
2500 PL_sublex_info.repl = NULL;
2502 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2503 = SvPVX(PL_linestr);
2504 PL_bufend += SvCUR(PL_linestr);
2505 PL_last_lop = PL_last_uni = NULL;
2506 SAVEFREESV(PL_linestr);
2507 if (PL_lex_repl) SAVEFREESV(PL_lex_repl);
2509 PL_lex_dojoin = FALSE;
2510 PL_lex_brackets = PL_lex_formbrack = 0;
2511 PL_lex_allbrackets = 0;
2512 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2513 Newx(PL_lex_brackstack, 120, char);
2514 Newx(PL_lex_casestack, 12, char);
2515 PL_lex_casemods = 0;
2516 *PL_lex_casestack = '\0';
2518 PL_lex_state = LEX_INTERPCONCAT;
2519 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2521 Newxz(shared, 1, LEXSHARED);
2522 shared->ls_prev = PL_parser->lex_shared;
2523 PL_parser->lex_shared = shared;
2525 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2526 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2527 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2528 PL_lex_inpat = PL_sublex_info.sub_op;
2530 PL_lex_inpat = NULL;
2537 * Restores lexer state after a S_sublex_push.
2544 if (!PL_lex_starts++) {
2545 SV * const sv = newSVpvs("");
2546 if (SvUTF8(PL_linestr))
2548 PL_expect = XOPERATOR;
2549 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2553 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2554 PL_lex_state = LEX_INTERPCASEMOD;
2558 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2559 assert(PL_lex_inwhat != OP_TRANSR);
2560 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2561 PL_linestr = PL_lex_repl;
2563 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2564 PL_bufend += SvCUR(PL_linestr);
2565 PL_last_lop = PL_last_uni = NULL;
2566 PL_lex_dojoin = FALSE;
2567 PL_lex_brackets = 0;
2568 PL_lex_allbrackets = 0;
2569 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2570 PL_lex_casemods = 0;
2571 *PL_lex_casestack = '\0';
2573 if (SvEVALED(PL_lex_repl)) {
2574 PL_lex_state = LEX_INTERPNORMAL;
2576 /* we don't clear PL_lex_repl here, so that we can check later
2577 whether this is an evalled subst; that means we rely on the
2578 logic to ensure sublex_done() is called again only via the
2579 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2582 PL_lex_state = LEX_INTERPCONCAT;
2592 PL_endwhite = newSVpvs("");
2593 sv_catsv(PL_endwhite, PL_thiswhite);
2597 sv_setpvs(PL_thistoken,"");
2599 PL_realtokenstart = -1;
2603 PL_bufend = SvPVX(PL_linestr);
2604 PL_bufend += SvCUR(PL_linestr);
2605 PL_expect = XOPERATOR;
2606 PL_sublex_info.sub_inwhat = 0;
2614 Extracts the next constant part of a pattern, double-quoted string,
2615 or transliteration. This is terrifying code.
2617 For example, in parsing the double-quoted string "ab\x63$d", it would
2618 stop at the '$' and return an OP_CONST containing 'abc'.
2620 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2621 processing a pattern (PL_lex_inpat is true), a transliteration
2622 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2624 Returns a pointer to the character scanned up to. If this is
2625 advanced from the start pointer supplied (i.e. if anything was
2626 successfully parsed), will leave an OP_CONST for the substring scanned
2627 in pl_yylval. Caller must intuit reason for not parsing further
2628 by looking at the next characters herself.
2632 \N{ABC} => \N{U+41.42.43}
2635 all other \-char, including \N and \N{ apart from \N{ABC}
2638 @ and $ where it appears to be a var, but not for $ as tail anchor
2643 In transliterations:
2644 characters are VERY literal, except for - not at the start or end
2645 of the string, which indicates a range. If the range is in bytes,
2646 scan_const expands the range to the full set of intermediate
2647 characters. If the range is in utf8, the hyphen is replaced with
2648 a certain range mark which will be handled by pmtrans() in op.c.
2650 In double-quoted strings:
2652 double-quoted style: \r and \n
2653 constants: \x31, etc.
2654 deprecated backrefs: \1 (in substitution replacements)
2655 case and quoting: \U \Q \E
2658 scan_const does *not* construct ops to handle interpolated strings.
2659 It stops processing as soon as it finds an embedded $ or @ variable
2660 and leaves it to the caller to work out what's going on.
2662 embedded arrays (whether in pattern or not) could be:
2663 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2665 $ in double-quoted strings must be the symbol of an embedded scalar.
2667 $ in pattern could be $foo or could be tail anchor. Assumption:
2668 it's a tail anchor if $ is the last thing in the string, or if it's
2669 followed by one of "()| \r\n\t"
2671 \1 (backreferences) are turned into $1 in substitutions
2673 The structure of the code is
2674 while (there's a character to process) {
2675 handle transliteration ranges
2676 skip regexp comments /(?#comment)/ and codes /(?{code})/
2677 skip #-initiated comments in //x patterns
2678 check for embedded arrays
2679 check for embedded scalars
2681 deprecate \1 in substitution replacements
2682 handle string-changing backslashes \l \U \Q \E, etc.
2683 switch (what was escaped) {
2684 handle \- in a transliteration (becomes a literal -)
2685 if a pattern and not \N{, go treat as regular character
2686 handle \132 (octal characters)
2687 handle \x15 and \x{1234} (hex characters)
2688 handle \N{name} (named characters, also \N{3,5} in a pattern)
2689 handle \cV (control characters)
2690 handle printf-style backslashes (\f, \r, \n, etc)
2693 } (end if backslash)
2694 handle regular character
2695 } (end while character to read)
2700 S_scan_const(pTHX_ char *start)
2703 char *send = PL_bufend; /* end of the constant */
2704 SV *sv = newSV(send - start); /* sv for the constant. See
2705 note below on sizing. */
2706 char *s = start; /* start of the constant */
2707 char *d = SvPVX(sv); /* destination for copies */
2708 bool dorange = FALSE; /* are we in a translit range? */
2709 bool didrange = FALSE; /* did we just finish a range? */
2710 bool in_charclass = FALSE; /* within /[...]/ */
2711 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2712 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
2713 to be UTF8? But, this can
2714 show as true when the source
2715 isn't utf8, as for example
2716 when it is entirely composed
2719 /* Note on sizing: The scanned constant is placed into sv, which is
2720 * initialized by newSV() assuming one byte of output for every byte of
2721 * input. This routine expects newSV() to allocate an extra byte for a
2722 * trailing NUL, which this routine will append if it gets to the end of
2723 * the input. There may be more bytes of input than output (eg., \N{LATIN
2724 * CAPITAL LETTER A}), or more output than input if the constant ends up
2725 * recoded to utf8, but each time a construct is found that might increase
2726 * the needed size, SvGROW() is called. Its size parameter each time is
2727 * based on the best guess estimate at the time, namely the length used so
2728 * far, plus the length the current construct will occupy, plus room for
2729 * the trailing NUL, plus one byte for every input byte still unscanned */
2733 UV literal_endpoint = 0;
2734 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2737 PERL_ARGS_ASSERT_SCAN_CONST;
2739 assert(PL_lex_inwhat != OP_TRANSR);
2740 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2741 /* If we are doing a trans and we know we want UTF8 set expectation */
2742 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2743 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2747 while (s < send || dorange) {
2749 /* get transliterations out of the way (they're most literal) */
2750 if (PL_lex_inwhat == OP_TRANS) {
2751 /* expand a range A-Z to the full set of characters. AIE! */
2753 I32 i; /* current expanded character */
2754 I32 min; /* first character in range */
2755 I32 max; /* last character in range */
2766 char * const c = (char*)utf8_hop((U8*)d, -1);
2770 *c = (char)UTF_TO_NATIVE(0xff);
2771 /* mark the range as done, and continue */
2777 i = d - SvPVX_const(sv); /* remember current offset */
2780 SvLEN(sv) + (has_utf8 ?
2781 (512 - UTF_CONTINUATION_MARK +
2784 /* How many two-byte within 0..255: 128 in UTF-8,
2785 * 96 in UTF-8-mod. */
2787 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2789 d = SvPVX(sv) + i; /* refresh d after realloc */
2793 for (j = 0; j <= 1; j++) {
2794 char * const c = (char*)utf8_hop((U8*)d, -1);
2795 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2801 max = (U8)0xff; /* only to \xff */
2802 uvmax = uv; /* \x{100} to uvmax */
2804 d = c; /* eat endpoint chars */
2809 d -= 2; /* eat the first char and the - */
2810 min = (U8)*d; /* first char in range */
2811 max = (U8)d[1]; /* last char in range */
2819 "Invalid range \"%c-%c\" in transliteration operator",
2820 (char)min, (char)max);
2824 if (literal_endpoint == 2 &&
2825 ((isLOWER(min) && isLOWER(max)) ||
2826 (isUPPER(min) && isUPPER(max)))) {
2828 for (i = min; i <= max; i++)
2830 *d++ = NATIVE_TO_NEED(has_utf8,i);
2832 for (i = min; i <= max; i++)
2834 *d++ = NATIVE_TO_NEED(has_utf8,i);
2839 for (i = min; i <= max; i++)
2842 const U8 ch = (U8)NATIVE_TO_UTF(i);
2843 if (UNI_IS_INVARIANT(ch))
2846 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2847 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2856 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2858 *d++ = (char)UTF_TO_NATIVE(0xff);
2860 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2864 /* mark the range as done, and continue */
2868 literal_endpoint = 0;
2873 /* range begins (ignore - as first or last char) */
2874 else if (*s == '-' && s+1 < send && s != start) {
2877 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2884 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2894 literal_endpoint = 0;
2895 native_range = TRUE;
2900 /* if we get here, we're not doing a transliteration */
2902 else if (*s == '[' && PL_lex_inpat && !in_charclass) {
2905 while (s1 >= start && *s1-- == '\\')
2908 in_charclass = TRUE;
2911 else if (*s == ']' && PL_lex_inpat && in_charclass) {
2914 while (s1 >= start && *s1-- == '\\')
2917 in_charclass = FALSE;
2920 /* skip for regexp comments /(?#comment)/, except for the last
2921 * char, which will be done separately.
2922 * Stop on (?{..}) and friends */
2924 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2926 while (s+1 < send && *s != ')')
2927 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2929 else if (!PL_lex_casemods && !in_charclass &&
2930 ( s[2] == '{' /* This should match regcomp.c */
2931 || (s[2] == '?' && s[3] == '{')))
2937 /* likewise skip #-initiated comments in //x patterns */
2938 else if (*s == '#' && PL_lex_inpat &&
2939 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
2940 while (s+1 < send && *s != '\n')
2941 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2944 /* no further processing of single-quoted regex */
2945 else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'')
2946 goto default_action;
2948 /* check for embedded arrays
2949 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2951 else if (*s == '@' && s[1]) {
2952 if (isALNUM_lazy_if(s+1,UTF))
2954 if (strchr(":'{$", s[1]))
2956 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2957 break; /* in regexp, neither @+ nor @- are interpolated */
2960 /* check for embedded scalars. only stop if we're sure it's a
2963 else if (*s == '$') {
2964 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2966 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2968 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2969 "Possible unintended interpolation of $\\ in regex");
2971 break; /* in regexp, $ might be tail anchor */
2975 /* End of else if chain - OP_TRANS rejoin rest */
2978 if (*s == '\\' && s+1 < send) {
2979 char* e; /* Can be used for ending '}', etc. */
2983 /* warn on \1 - \9 in substitution replacements, but note that \11
2984 * is an octal; and \19 is \1 followed by '9' */
2985 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2986 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2988 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2993 /* string-change backslash escapes */
2994 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQF", *s)) {
2998 /* In a pattern, process \N, but skip any other backslash escapes.
2999 * This is because we don't want to translate an escape sequence
3000 * into a meta symbol and have the regex compiler use the meta
3001 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
3002 * in spite of this, we do have to process \N here while the proper
3003 * charnames handler is in scope. See bugs #56444 and #62056.
3004 * There is a complication because \N in a pattern may also stand
3005 * for 'match a non-nl', and not mean a charname, in which case its
3006 * processing should be deferred to the regex compiler. To be a
3007 * charname it must be followed immediately by a '{', and not look
3008 * like \N followed by a curly quantifier, i.e., not something like
3009 * \N{3,}. regcurly returns a boolean indicating if it is a legal
3011 else if (PL_lex_inpat
3014 || regcurly(s + 1)))
3016 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
3017 goto default_action;
3022 /* quoted - in transliterations */
3024 if (PL_lex_inwhat == OP_TRANS) {
3032 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
3033 "Unrecognized escape \\%c passed through",
3035 /* default action is to copy the quoted character */
3036 goto default_action;
3039 /* eg. \132 indicates the octal constant 0132 */
3040 case '0': case '1': case '2': case '3':
3041 case '4': case '5': case '6': case '7':
3045 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
3048 goto NUM_ESCAPE_INSERT;
3050 /* eg. \o{24} indicates the octal constant \024 */
3056 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
3062 goto NUM_ESCAPE_INSERT;
3065 /* eg. \x24 indicates the hex constant 0x24 */
3071 bool valid = grok_bslash_x(s, &uv, &len, &error, 1);
3080 /* Insert oct or hex escaped character. There will always be
3081 * enough room in sv since such escapes will be longer than any
3082 * UTF-8 sequence they can end up as, except if they force us
3083 * to recode the rest of the string into utf8 */
3085 /* Here uv is the ordinal of the next character being added in
3086 * unicode (converted from native). */
3087 if (!UNI_IS_INVARIANT(uv)) {
3088 if (!has_utf8 && uv > 255) {
3089 /* Might need to recode whatever we have accumulated so
3090 * far if it contains any chars variant in utf8 or
3093 SvCUR_set(sv, d - SvPVX_const(sv));
3096 /* See Note on sizing above. */
3097 sv_utf8_upgrade_flags_grow(sv,
3098 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3099 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3100 d = SvPVX(sv) + SvCUR(sv);
3105 d = (char*)uvuni_to_utf8((U8*)d, uv);
3106 if (PL_lex_inwhat == OP_TRANS &&
3107 PL_sublex_info.sub_op) {
3108 PL_sublex_info.sub_op->op_private |=
3109 (PL_lex_repl ? OPpTRANS_FROM_UTF
3113 if (uv > 255 && !dorange)
3114 native_range = FALSE;
3127 /* In a non-pattern \N must be a named character, like \N{LATIN
3128 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3129 * mean to match a non-newline. For non-patterns, named
3130 * characters are converted to their string equivalents. In
3131 * patterns, named characters are not converted to their
3132 * ultimate forms for the same reasons that other escapes
3133 * aren't. Instead, they are converted to the \N{U+...} form
3134 * to get the value from the charnames that is in effect right
3135 * now, while preserving the fact that it was a named character
3136 * so that the regex compiler knows this */
3138 /* This section of code doesn't generally use the
3139 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3140 * a close examination of this macro and determined it is a
3141 * no-op except on utfebcdic variant characters. Every
3142 * character generated by this that would normally need to be
3143 * enclosed by this macro is invariant, so the macro is not
3144 * needed, and would complicate use of copy(). XXX There are
3145 * other parts of this file where the macro is used
3146 * inconsistently, but are saved by it being a no-op */
3148 /* The structure of this section of code (besides checking for
3149 * errors and upgrading to utf8) is:
3150 * Further disambiguate between the two meanings of \N, and if
3151 * not a charname, go process it elsewhere
3152 * If of form \N{U+...}, pass it through if a pattern;
3153 * otherwise convert to utf8
3154 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3155 * pattern; otherwise convert to utf8 */
3157 /* Here, s points to the 'N'; the test below is guaranteed to
3158 * succeed if we are being called on a pattern as we already
3159 * know from a test above that the next character is a '{'.
3160 * On a non-pattern \N must mean 'named sequence, which
3161 * requires braces */
3164 yyerror("Missing braces on \\N{}");
3169 /* If there is no matching '}', it is an error. */
3170 if (! (e = strchr(s, '}'))) {
3171 if (! PL_lex_inpat) {
3172 yyerror("Missing right brace on \\N{}");
3174 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3179 /* Here it looks like a named character */
3183 /* XXX This block is temporary code. \N{} implies that the
3184 * pattern is to have Unicode semantics, and therefore
3185 * currently has to be encoded in utf8. By putting it in
3186 * utf8 now, we save a whole pass in the regular expression
3187 * compiler. Once that code is changed so Unicode
3188 * semantics doesn't necessarily have to be in utf8, this
3189 * block should be removed. However, the code that parses
3190 * the output of this would have to be changed to not
3191 * necessarily expect utf8 */
3193 SvCUR_set(sv, d - SvPVX_const(sv));
3196 /* See Note on sizing above. */
3197 sv_utf8_upgrade_flags_grow(sv,
3198 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3199 /* 5 = '\N{' + cur char + NUL */
3200 (STRLEN)(send - s) + 5);
3201 d = SvPVX(sv) + SvCUR(sv);
3206 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3207 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3208 | PERL_SCAN_DISALLOW_PREFIX;
3211 /* For \N{U+...}, the '...' is a unicode value even on
3212 * EBCDIC machines */
3213 s += 2; /* Skip to next char after the 'U+' */
3215 uv = grok_hex(s, &len, &flags, NULL);
3216 if (len == 0 || len != (STRLEN)(e - s)) {
3217 yyerror("Invalid hexadecimal number in \\N{U+...}");
3224 /* On non-EBCDIC platforms, pass through to the regex
3225 * compiler unchanged. The reason we evaluated the
3226 * number above is to make sure there wasn't a syntax
3227 * error. But on EBCDIC we convert to native so
3228 * downstream code can continue to assume it's native
3230 s -= 5; /* Include the '\N{U+' */
3232 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3235 (unsigned int) UNI_TO_NATIVE(uv));
3237 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3241 else { /* Not a pattern: convert the hex to string */
3243 /* If destination is not in utf8, unconditionally
3244 * recode it to be so. This is because \N{} implies
3245 * Unicode semantics, and scalars have to be in utf8
3246 * to guarantee those semantics */
3248 SvCUR_set(sv, d - SvPVX_const(sv));
3251 /* See Note on sizing above. */
3252 sv_utf8_upgrade_flags_grow(
3254 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3255 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3256 d = SvPVX(sv) + SvCUR(sv);
3260 /* Add the string to the output */
3261 if (UNI_IS_INVARIANT(uv)) {
3264 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3267 else { /* Here is \N{NAME} but not \N{U+...}. */
3269 SV *res; /* result from charnames */
3270 const char *str; /* the string in 'res' */
3271 STRLEN len; /* its length */
3273 /* Get the value for NAME */
3274 res = newSVpvn(s, e - s);
3275 res = new_constant( NULL, 0, "charnames",
3276 /* includes all of: \N{...} */
3277 res, NULL, s - 3, e - s + 4 );
3279 /* Most likely res will be in utf8 already since the
3280 * standard charnames uses pack U, but a custom translator
3281 * can leave it otherwise, so make sure. XXX This can be
3282 * revisited to not have charnames use utf8 for characters
3283 * that don't need it when regexes don't have to be in utf8
3284 * for Unicode semantics. If doing so, remember EBCDIC */
3285 sv_utf8_upgrade(res);
3286 str = SvPV_const(res, len);
3288 /* Don't accept malformed input */
3289 if (! is_utf8_string((U8 *) str, len)) {
3290 yyerror("Malformed UTF-8 returned by \\N");
3292 else if (PL_lex_inpat) {
3294 if (! len) { /* The name resolved to an empty string */
3295 Copy("\\N{}", d, 4, char);
3299 /* In order to not lose information for the regex
3300 * compiler, pass the result in the specially made
3301 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3302 * the code points in hex of each character
3303 * returned by charnames */
3305 const char *str_end = str + len;
3306 STRLEN char_length; /* cur char's byte length */
3307 STRLEN output_length; /* and the number of bytes
3308 after this is translated
3310 const STRLEN off = d - SvPVX_const(sv);
3312 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3313 * max('U+', '.'); and 1 for NUL */
3314 char hex_string[2 * UTF8_MAXBYTES + 5];
3316 /* Get the first character of the result. */
3317 U32 uv = utf8n_to_uvuni((U8 *) str,
3322 /* The call to is_utf8_string() above hopefully
3323 * guarantees that there won't be an error. But
3324 * it's easy here to make sure. The function just
3325 * above warns and returns 0 if invalid utf8, but
3326 * it can also return 0 if the input is validly a
3327 * NUL. Disambiguate */
3328 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3329 uv = UNICODE_REPLACEMENT;
3332 /* Convert first code point to hex, including the
3333 * boiler plate before it. For all these, we
3334 * convert to native format so that downstream code
3335 * can continue to assume the input is native */
3337 my_snprintf(hex_string, sizeof(hex_string),
3339 (unsigned int) UNI_TO_NATIVE(uv));
3341 /* Make sure there is enough space to hold it */
3342 d = off + SvGROW(sv, off
3344 + (STRLEN)(send - e)
3345 + 2); /* '}' + NUL */
3347 Copy(hex_string, d, output_length, char);
3350 /* For each subsequent character, append dot and
3351 * its ordinal in hex */
3352 while ((str += char_length) < str_end) {
3353 const STRLEN off = d - SvPVX_const(sv);
3354 U32 uv = utf8n_to_uvuni((U8 *) str,
3358 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3359 uv = UNICODE_REPLACEMENT;
3363 my_snprintf(hex_string, sizeof(hex_string),
3365 (unsigned int) UNI_TO_NATIVE(uv));
3367 d = off + SvGROW(sv, off
3369 + (STRLEN)(send - e)
3370 + 2); /* '}' + NUL */
3371 Copy(hex_string, d, output_length, char);
3375 *d++ = '}'; /* Done. Add the trailing brace */
3378 else { /* Here, not in a pattern. Convert the name to a
3381 /* If destination is not in utf8, unconditionally
3382 * recode it to be so. This is because \N{} implies
3383 * Unicode semantics, and scalars have to be in utf8
3384 * to guarantee those semantics */
3386 SvCUR_set(sv, d - SvPVX_const(sv));
3389 /* See Note on sizing above. */
3390 sv_utf8_upgrade_flags_grow(sv,
3391 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3392 len + (STRLEN)(send - s) + 1);
3393 d = SvPVX(sv) + SvCUR(sv);
3395 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3397 /* See Note on sizing above. (NOTE: SvCUR() is not
3398 * set correctly here). */
3399 const STRLEN off = d - SvPVX_const(sv);
3400 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3402 Copy(str, d, len, char);
3407 /* Deprecate non-approved name syntax */
3408 if (ckWARN_d(WARN_DEPRECATED)) {
3409 bool problematic = FALSE;
3412 /* For non-ut8 input, look to see that the first
3413 * character is an alpha, then loop through the rest
3414 * checking that each is a continuation */
3416 if (! isALPHAU(*i)) problematic = TRUE;
3417 else for (i = s + 1; i < e; i++) {
3418 if (isCHARNAME_CONT(*i)) continue;
3424 /* Similarly for utf8. For invariants can check
3425 * directly. We accept anything above the latin1
3426 * range because it is immaterial to Perl if it is
3427 * correct or not, and is expensive to check. But
3428 * it is fairly easy in the latin1 range to convert
3429 * the variants into a single character and check
3431 if (UTF8_IS_INVARIANT(*i)) {
3432 if (! isALPHAU(*i)) problematic = TRUE;
3433 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3434 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*i,
3440 if (! problematic) for (i = s + UTF8SKIP(s);
3444 if (UTF8_IS_INVARIANT(*i)) {
3445 if (isCHARNAME_CONT(*i)) continue;
3446 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3448 } else if (isCHARNAME_CONT(
3450 TWO_BYTE_UTF8_TO_UNI(*i, *(i+1)))))
3459 /* The e-i passed to the final %.*s makes sure that
3460 * should the trailing NUL be missing that this
3461 * print won't run off the end of the string */
3462 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3463 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s",
3464 (int)(i - s + 1), s, (int)(e - i), i + 1);
3467 } /* End \N{NAME} */
3470 native_range = FALSE; /* \N{} is defined to be Unicode */
3472 s = e + 1; /* Point to just after the '}' */
3475 /* \c is a control character */
3479 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3482 yyerror("Missing control char name in \\c");
3486 /* printf-style backslashes, formfeeds, newlines, etc */
3488 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3491 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3494 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3497 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3500 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3503 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3506 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3512 } /* end if (backslash) */
3519 /* If we started with encoded form, or already know we want it,
3520 then encode the next character */
3521 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3525 /* One might think that it is wasted effort in the case of the
3526 * source being utf8 (this_utf8 == TRUE) to take the next character
3527 * in the source, convert it to an unsigned value, and then convert
3528 * it back again. But the source has not been validated here. The
3529 * routine that does the conversion checks for errors like
3532 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3533 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3535 SvCUR_set(sv, d - SvPVX_const(sv));
3538 /* See Note on sizing above. */
3539 sv_utf8_upgrade_flags_grow(sv,
3540 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3541 need + (STRLEN)(send - s) + 1);
3542 d = SvPVX(sv) + SvCUR(sv);
3544 } else if (need > len) {
3545 /* encoded value larger than old, may need extra space (NOTE:
3546 * SvCUR() is not set correctly here). See Note on sizing
3548 const STRLEN off = d - SvPVX_const(sv);
3549 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3553 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3555 if (uv > 255 && !dorange)
3556 native_range = FALSE;
3560 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3562 } /* while loop to process each character */
3564 /* terminate the string and set up the sv */
3566 SvCUR_set(sv, d - SvPVX_const(sv));
3567 if (SvCUR(sv) >= SvLEN(sv))
3568 Perl_croak(aTHX_ "panic: constant overflowed allocated space, %"UVuf
3569 " >= %"UVuf, (UV)SvCUR(sv), (UV)SvLEN(sv));
3572 if (PL_encoding && !has_utf8) {
3573 sv_recode_to_utf8(sv, PL_encoding);
3579 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3580 PL_sublex_info.sub_op->op_private |=
3581 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3585 /* shrink the sv if we allocated more than we used */
3586 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3587 SvPV_shrink_to_cur(sv);
3590 /* return the substring (via pl_yylval) only if we parsed anything */
3591 if (s > PL_bufptr) {
3592 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3593 const char *const key = PL_lex_inpat ? "qr" : "q";
3594 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3598 if (PL_lex_inwhat == OP_TRANS) {
3601 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3604 } else if (PL_lex_inpat && SvIVX(PL_linestr) == '\'') {
3612 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3615 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3622 * Returns TRUE if there's more to the expression (e.g., a subscript),
3625 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3627 * ->[ and ->{ return TRUE
3628 * { and [ outside a pattern are always subscripts, so return TRUE
3629 * if we're outside a pattern and it's not { or [, then return FALSE
3630 * if we're in a pattern and the first char is a {
3631 * {4,5} (any digits around the comma) returns FALSE
3632 * if we're in a pattern and the first char is a [
3634 * [SOMETHING] has a funky algorithm to decide whether it's a
3635 * character class or not. It has to deal with things like
3636 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3637 * anything else returns TRUE
3640 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3643 S_intuit_more(pTHX_ register char *s)
3647 PERL_ARGS_ASSERT_INTUIT_MORE;
3649 if (PL_lex_brackets)
3651 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3653 if (*s != '{' && *s != '[')
3658 /* In a pattern, so maybe we have {n,m}. */
3666 /* On the other hand, maybe we have a character class */
3669 if (*s == ']' || *s == '^')
3672 /* this is terrifying, and it works */
3673 int weight = 2; /* let's weigh the evidence */
3675 unsigned char un_char = 255, last_un_char;
3676 const char * const send = strchr(s,']');
3677 char tmpbuf[sizeof PL_tokenbuf * 4];
3679 if (!send) /* has to be an expression */
3682 Zero(seen,256,char);
3685 else if (isDIGIT(*s)) {
3687 if (isDIGIT(s[1]) && s[2] == ']')
3693 for (; s < send; s++) {
3694 last_un_char = un_char;
3695 un_char = (unsigned char)*s;
3700 weight -= seen[un_char] * 10;
3701 if (isALNUM_lazy_if(s+1,UTF)) {
3703 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3704 len = (int)strlen(tmpbuf);
3705 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len,
3706 UTF ? SVf_UTF8 : 0, SVt_PV))
3711 else if (*s == '$' && s[1] &&
3712 strchr("[#!%*<>()-=",s[1])) {
3713 if (/*{*/ strchr("])} =",s[2]))
3722 if (strchr("wds]",s[1]))
3724 else if (seen[(U8)'\''] || seen[(U8)'"'])
3726 else if (strchr("rnftbxcav",s[1]))
3728 else if (isDIGIT(s[1])) {
3730 while (s[1] && isDIGIT(s[1]))
3740 if (strchr("aA01! ",last_un_char))
3742 if (strchr("zZ79~",s[1]))
3744 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3745 weight -= 5; /* cope with negative subscript */
3748 if (!isALNUM(last_un_char)
3749 && !(last_un_char == '$' || last_un_char == '@'
3750 || last_un_char == '&')
3751 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3756 if (keyword(tmpbuf, d - tmpbuf, 0))
3759 if (un_char == last_un_char + 1)
3761 weight -= seen[un_char];
3766 if (weight >= 0) /* probably a character class */
3776 * Does all the checking to disambiguate
3778 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3779 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3781 * First argument is the stuff after the first token, e.g. "bar".
3783 * Not a method if foo is a filehandle.
3784 * Not a method if foo is a subroutine prototyped to take a filehandle.
3785 * Not a method if it's really "Foo $bar"
3786 * Method if it's "foo $bar"
3787 * Not a method if it's really "print foo $bar"
3788 * Method if it's really "foo package::" (interpreted as package->foo)
3789 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3790 * Not a method if bar is a filehandle or package, but is quoted with
3795 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3798 char *s = start + (*start == '$');
3799 char tmpbuf[sizeof PL_tokenbuf];
3806 PERL_ARGS_ASSERT_INTUIT_METHOD;
3808 if (gv && SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3810 if (cv && SvPOK(cv)) {
3811 const char *proto = CvPROTO(cv);
3819 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3820 /* start is the beginning of the possible filehandle/object,
3821 * and s is the end of it
3822 * tmpbuf is a copy of it
3825 if (*start == '$') {
3826 if (cv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3827 isUPPER(*PL_tokenbuf))
3830 len = start - SvPVX(PL_linestr);
3834 start = SvPVX(PL_linestr) + len;
3838 return *s == '(' ? FUNCMETH : METHOD;
3840 if (!keyword(tmpbuf, len, 0)) {
3841 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3845 soff = s - SvPVX(PL_linestr);
3849 indirgv = gv_fetchpvn_flags(tmpbuf, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
3850 if (indirgv && GvCVu(indirgv))
3852 /* filehandle or package name makes it a method */
3853 if (!cv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, UTF ? SVf_UTF8 : 0)) {
3855 soff = s - SvPVX(PL_linestr);
3858 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3859 return 0; /* no assumptions -- "=>" quotes bareword */
3861 start_force(PL_curforce);
3862 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3863 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3864 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3866 curmad('X', newSVpvn_flags(start,SvPVX(PL_linestr) + soff - start,
3867 ( UTF ? SVf_UTF8 : 0 )));
3872 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3874 return *s == '(' ? FUNCMETH : METHOD;
3880 /* Encoded script support. filter_add() effectively inserts a
3881 * 'pre-processing' function into the current source input stream.
3882 * Note that the filter function only applies to the current source file
3883 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3885 * The datasv parameter (which may be NULL) can be used to pass
3886 * private data to this instance of the filter. The filter function
3887 * can recover the SV using the FILTER_DATA macro and use it to
3888 * store private buffers and state information.
3890 * The supplied datasv parameter is upgraded to a PVIO type
3891 * and the IoDIRP/IoANY field is used to store the function pointer,
3892 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3893 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3894 * private use must be set using malloc'd pointers.
3898 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3907 if (PL_parser->lex_flags & LEX_IGNORE_UTF8_HINTS)
3908 Perl_croak(aTHX_ "Source filters apply only to byte streams");
3910 if (!PL_rsfp_filters)
3911 PL_rsfp_filters = newAV();
3914 SvUPGRADE(datasv, SVt_PVIO);
3915 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3916 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3917 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3918 FPTR2DPTR(void *, IoANY(datasv)),
3919 SvPV_nolen(datasv)));
3920 av_unshift(PL_rsfp_filters, 1);
3921 av_store(PL_rsfp_filters, 0, datasv) ;
3923 !PL_parser->filtered
3924 && PL_parser->lex_flags & LEX_EVALBYTES
3925 && PL_bufptr < PL_bufend
3927 const char *s = PL_bufptr;
3928 while (s < PL_bufend) {
3930 SV *linestr = PL_parser->linestr;
3931 char *buf = SvPVX(linestr);
3932 STRLEN const bufptr_pos = PL_parser->bufptr - buf;
3933 STRLEN const oldbufptr_pos = PL_parser->oldbufptr - buf;
3934 STRLEN const oldoldbufptr_pos=PL_parser->oldoldbufptr-buf;
3935 STRLEN const linestart_pos = PL_parser->linestart - buf;
3936 STRLEN const last_uni_pos =
3937 PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
3938 STRLEN const last_lop_pos =
3939 PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
3940 av_push(PL_rsfp_filters, linestr);
3941 PL_parser->linestr =
3942 newSVpvn(SvPVX(linestr), ++s-SvPVX(linestr));
3943 buf = SvPVX(PL_parser->linestr);
3944 PL_parser->bufend = buf + SvCUR(PL_parser->linestr);
3945 PL_parser->bufptr = buf + bufptr_pos;
3946 PL_parser->oldbufptr = buf + oldbufptr_pos;
3947 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
3948 PL_parser->linestart = buf + linestart_pos;
3949 if (PL_parser->last_uni)
3950 PL_parser->last_uni = buf + last_uni_pos;
3951 if (PL_parser->last_lop)
3952 PL_parser->last_lop = buf + last_lop_pos;
3953 SvLEN(linestr) = SvCUR(linestr);
3954 SvCUR(linestr) = s-SvPVX(linestr);
3955 PL_parser->filtered = 1;
3965 /* Delete most recently added instance of this filter function. */
3967 Perl_filter_del(pTHX_ filter_t funcp)
3972 PERL_ARGS_ASSERT_FILTER_DEL;
3975 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3976 FPTR2DPTR(void*, funcp)));
3978 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3980 /* if filter is on top of stack (usual case) just pop it off */
3981 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3982 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3983 sv_free(av_pop(PL_rsfp_filters));
3987 /* we need to search for the correct entry and clear it */
3988 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3992 /* Invoke the idxth filter function for the current rsfp. */
3993 /* maxlen 0 = read one text line */
3995 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
4000 /* This API is bad. It should have been using unsigned int for maxlen.
4001 Not sure if we want to change the API, but if not we should sanity
4002 check the value here. */
4003 unsigned int correct_length
4012 PERL_ARGS_ASSERT_FILTER_READ;
4014 if (!PL_parser || !PL_rsfp_filters)
4016 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
4017 /* Provide a default input filter to make life easy. */
4018 /* Note that we append to the line. This is handy. */
4019 DEBUG_P(PerlIO_printf(Perl_debug_log,
4020 "filter_read %d: from rsfp\n", idx));
4021 if (correct_length) {
4024 const int old_len = SvCUR(buf_sv);
4026 /* ensure buf_sv is large enough */
4027 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
4028 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
4029 correct_length)) <= 0) {
4030 if (PerlIO_error(PL_rsfp))
4031 return -1; /* error */
4033 return 0 ; /* end of file */
4035 SvCUR_set(buf_sv, old_len + len) ;
4036 SvPVX(buf_sv)[old_len + len] = '\0';
4039 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
4040 if (PerlIO_error(PL_rsfp))
4041 return -1; /* error */
4043 return 0 ; /* end of file */
4046 return SvCUR(buf_sv);
4048 /* Skip this filter slot if filter has been deleted */
4049 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
4050 DEBUG_P(PerlIO_printf(Perl_debug_log,
4051 "filter_read %d: skipped (filter deleted)\n",
4053 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
4055 if (SvTYPE(datasv) != SVt_PVIO) {
4056 if (correct_length) {
4058 const STRLEN remainder = SvLEN(datasv) - SvCUR(datasv);
4059 if (!remainder) return 0; /* eof */
4060 if (correct_length > remainder) correct_length = remainder;
4061 sv_catpvn(buf_sv, SvEND(datasv), correct_length);
4062 SvCUR_set(datasv, SvCUR(datasv) + correct_length);
4065 const char *s = SvEND(datasv);
4066 const char *send = SvPVX(datasv) + SvLEN(datasv);
4074 if (s == send) return 0; /* eof */
4075 sv_catpvn(buf_sv, SvEND(datasv), s-SvEND(datasv));
4076 SvCUR_set(datasv, s-SvPVX(datasv));
4078 return SvCUR(buf_sv);
4080 /* Get function pointer hidden within datasv */
4081 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
4082 DEBUG_P(PerlIO_printf(Perl_debug_log,
4083 "filter_read %d: via function %p (%s)\n",
4084 idx, (void*)datasv, SvPV_nolen_const(datasv)));
4085 /* Call function. The function is expected to */
4086 /* call "FILTER_READ(idx+1, buf_sv)" first. */
4087 /* Return: <0:error, =0:eof, >0:not eof */
4088 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
4092 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
4096 PERL_ARGS_ASSERT_FILTER_GETS;
4098 #ifdef PERL_CR_FILTER
4099 if (!PL_rsfp_filters) {
4100 filter_add(S_cr_textfilter,NULL);
4103 if (PL_rsfp_filters) {
4105 SvCUR_set(sv, 0); /* start with empty line */
4106 if (FILTER_READ(0, sv, 0) > 0)
4107 return ( SvPVX(sv) ) ;
4112 return (sv_gets(sv, PL_rsfp, append));
4116 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
4121 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
4123 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
4127 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
4128 (gv = gv_fetchpvn_flags(pkgname, len, ( UTF ? SVf_UTF8 : 0 ), SVt_PVHV)))
4130 return GvHV(gv); /* Foo:: */
4133 /* use constant CLASS => 'MyClass' */
4134 gv = gv_fetchpvn_flags(pkgname, len, UTF ? SVf_UTF8 : 0, SVt_PVCV);
4135 if (gv && GvCV(gv)) {
4136 SV * const sv = cv_const_sv(GvCV(gv));
4138 pkgname = SvPV_const(sv, len);
4141 return gv_stashpvn(pkgname, len, UTF ? SVf_UTF8 : 0);
4145 * S_readpipe_override
4146 * Check whether readpipe() is overridden, and generates the appropriate
4147 * optree, provided sublex_start() is called afterwards.
4150 S_readpipe_override(pTHX)
4153 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4154 pl_yylval.ival = OP_BACKTICK;
4156 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4158 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4159 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4160 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4162 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4163 op_append_elem(OP_LIST,
4164 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4165 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4172 * The intent of this yylex wrapper is to minimize the changes to the
4173 * tokener when we aren't interested in collecting madprops. It remains
4174 * to be seen how successful this strategy will be...
4181 char *s = PL_bufptr;
4183 /* make sure PL_thiswhite is initialized */
4187 /* previous token ate up our whitespace? */
4188 if (!PL_lasttoke && PL_nextwhite) {
4189 PL_thiswhite = PL_nextwhite;
4193 /* isolate the token, and figure out where it is without whitespace */
4194 PL_realtokenstart = -1;
4198 assert(PL_curforce < 0);
4200 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4201 if (!PL_thistoken) {
4202 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4203 PL_thistoken = newSVpvs("");
4205 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4206 PL_thistoken = newSVpvn(tstart, s - tstart);
4209 if (PL_thismad) /* install head */
4210 CURMAD('X', PL_thistoken);
4213 /* last whitespace of a sublex? */
4214 if (optype == ')' && PL_endwhite) {
4215 CURMAD('X', PL_endwhite);
4220 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4221 if (!PL_thiswhite && !PL_endwhite && !optype) {
4222 sv_free(PL_thistoken);
4227 /* put off final whitespace till peg */
4228 if (optype == ';' && !PL_rsfp && !PL_parser->filtered) {
4229 PL_nextwhite = PL_thiswhite;
4232 else if (PL_thisopen) {
4233 CURMAD('q', PL_thisopen);
4235 sv_free(PL_thistoken);
4239 /* Store actual token text as madprop X */
4240 CURMAD('X', PL_thistoken);
4244 /* add preceding whitespace as madprop _ */
4245 CURMAD('_', PL_thiswhite);
4249 /* add quoted material as madprop = */
4250 CURMAD('=', PL_thisstuff);
4254 /* add terminating quote as madprop Q */
4255 CURMAD('Q', PL_thisclose);
4259 /* special processing based on optype */
4263 /* opval doesn't need a TOKEN since it can already store mp */
4274 if (pl_yylval.opval)
4275 append_madprops(PL_thismad, pl_yylval.opval, 0);
4283 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4292 /* remember any fake bracket that lexer is about to discard */
4293 if (PL_lex_brackets == 1 &&
4294 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4297 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4300 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4301 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4304 break; /* don't bother looking for trailing comment */
4313 /* attach a trailing comment to its statement instead of next token */
4317 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4319 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4321 if (*s == '\n' || *s == '#') {
4322 while (s < PL_bufend && *s != '\n')
4326 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4327 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4340 /* Create new token struct. Note: opvals return early above. */
4341 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4348 S_tokenize_use(pTHX_ int is_use, char *s) {
4351 PERL_ARGS_ASSERT_TOKENIZE_USE;
4353 if (PL_expect != XSTATE)
4354 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4355 is_use ? "use" : "no"));
4358 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4359 s = force_version(s, TRUE);
4360 if (*s == ';' || *s == '}'
4361 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4362 start_force(PL_curforce);
4363 NEXTVAL_NEXTTOKE.opval = NULL;
4366 else if (*s == 'v') {
4367 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4368 s = force_version(s, FALSE);
4372 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4373 s = force_version(s, FALSE);
4375 pl_yylval.ival = is_use;
4379 static const char* const exp_name[] =
4380 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4381 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4385 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4387 S_word_takes_any_delimeter(char *p, STRLEN len)
4389 return (len == 1 && strchr("msyq", p[0])) ||
4391 (p[0] == 't' && p[1] == 'r') ||
4392 (p[0] == 'q' && strchr("qwxr", p[1]))));
4398 Works out what to call the token just pulled out of the input
4399 stream. The yacc parser takes care of taking the ops we return and
4400 stitching them into a tree.
4406 if read an identifier
4407 if we're in a my declaration
4408 croak if they tried to say my($foo::bar)
4409 build the ops for a my() declaration
4410 if it's an access to a my() variable
4411 are we in a sort block?
4412 croak if my($a); $a <=> $b
4413 build ops for access to a my() variable
4414 if in a dq string, and they've said @foo and we can't find @foo
4416 build ops for a bareword
4417 if we already built the token before, use it.
4422 #pragma segment Perl_yylex
4428 char *s = PL_bufptr;
4435 /* orig_keyword, gvp, and gv are initialized here because
4436 * jump to the label just_a_word_zero can bypass their
4437 * initialization later. */
4438 I32 orig_keyword = 0;
4443 SV* tmp = newSVpvs("");
4444 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4445 (IV)CopLINE(PL_curcop),
4446 lex_state_names[PL_lex_state],
4447 exp_name[PL_expect],
4448 pv_display(tmp, s, strlen(s), 0, 60));
4452 switch (PL_lex_state) {
4454 case LEX_NORMAL: /* Some compilers will produce faster */
4455 case LEX_INTERPNORMAL: /* code if we comment these out. */
4459 /* when we've already built the next token, just pull it out of the queue */
4463 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4465 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4466 PL_nexttoke[PL_lasttoke].next_mad = 0;
4467 if (PL_thismad && PL_thismad->mad_key == '_') {
4468 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4469 PL_thismad->mad_val = 0;
4470 mad_free(PL_thismad);
4475 PL_lex_state = PL_lex_defer;
4476 PL_expect = PL_lex_expect;
4477 PL_lex_defer = LEX_NORMAL;
4478 if (!PL_nexttoke[PL_lasttoke].next_type)
4483 pl_yylval = PL_nextval[PL_nexttoke];
4485 PL_lex_state = PL_lex_defer;
4486 PL_expect = PL_lex_expect;
4487 PL_lex_defer = LEX_NORMAL;
4493 next_type = PL_nexttoke[PL_lasttoke].next_type;
4495 next_type = PL_nexttype[PL_nexttoke];
4497 if (next_type & (7<<24)) {
4498 if (next_type & (1<<24)) {
4499 if (PL_lex_brackets > 100)
4500 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4501 PL_lex_brackstack[PL_lex_brackets++] =
4502 (char) ((next_type >> 16) & 0xff);
4504 if (next_type & (2<<24))
4505 PL_lex_allbrackets++;
4506 if (next_type & (4<<24))
4507 PL_lex_allbrackets--;
4508 next_type &= 0xffff;
4510 if (S_is_opval_token(next_type) && pl_yylval.opval)
4511 pl_yylval.opval->op_savefree = 0; /* release */
4512 return REPORT(next_type == 'p' ? pending_ident() : next_type);
4515 /* interpolated case modifiers like \L \U, including \Q and \E.
4516 when we get here, PL_bufptr is at the \
4518 case LEX_INTERPCASEMOD:
4520 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4522 "panic: INTERPCASEMOD bufptr=%p, bufend=%p, *bufptr=%u",
4523 PL_bufptr, PL_bufend, *PL_bufptr);
4525 /* handle \E or end of string */
4526 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4528 if (PL_lex_casemods) {
4529 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4530 PL_lex_casestack[PL_lex_casemods] = '\0';
4532 if (PL_bufptr != PL_bufend
4533 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q'
4534 || oldmod == 'F')) {
4536 PL_lex_state = LEX_INTERPCONCAT;
4539 PL_thistoken = newSVpvs("\\E");
4542 PL_lex_allbrackets--;
4545 else if ( PL_bufptr != PL_bufend && PL_bufptr[1] == 'E' ) {
4546 /* Got an unpaired \E */
4547 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4548 "Useless use of \\E");
4551 while (PL_bufptr != PL_bufend &&
4552 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4554 PL_thiswhite = newSVpvs("");
4555 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4559 if (PL_bufptr != PL_bufend)
4562 PL_lex_state = LEX_INTERPCONCAT;
4566 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4567 "### Saw case modifier\n"); });
4569 if (s[1] == '\\' && s[2] == 'E') {
4572 PL_thiswhite = newSVpvs("");
4573 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4576 PL_lex_state = LEX_INTERPCONCAT;
4581 if (!PL_madskills) /* when just compiling don't need correct */
4582 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4583 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4584 if ((*s == 'L' || *s == 'U' || *s == 'F') &&
4585 (strchr(PL_lex_casestack, 'L')
4586 || strchr(PL_lex_casestack, 'U')
4587 || strchr(PL_lex_casestack, 'F'))) {
4588 PL_lex_casestack[--PL_lex_casemods] = '\0';
4589 PL_lex_allbrackets--;
4592 if (PL_lex_casemods > 10)
4593 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4594 PL_lex_casestack[PL_lex_casemods++] = *s;
4595 PL_lex_casestack[PL_lex_casemods] = '\0';
4596 PL_lex_state = LEX_INTERPCONCAT;
4597 start_force(PL_curforce);
4598 NEXTVAL_NEXTTOKE.ival = 0;
4599 force_next((2<<24)|'(');
4600 start_force(PL_curforce);
4602 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4604 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4606 NEXTVAL_NEXTTOKE.ival = OP_LC;
4608 NEXTVAL_NEXTTOKE.ival = OP_UC;
4610 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4612 NEXTVAL_NEXTTOKE.ival = OP_FC;
4614 Perl_croak(aTHX_ "panic: yylex, *s=%u", *s);
4616 SV* const tmpsv = newSVpvs("\\ ");
4617 /* replace the space with the character we want to escape
4619 SvPVX(tmpsv)[1] = *s;
4625 if (PL_lex_starts) {
4631 sv_free(PL_thistoken);
4632 PL_thistoken = newSVpvs("");
4635 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4636 if (PL_lex_casemods == 1 && PL_lex_inpat)
4645 case LEX_INTERPPUSH:
4646 return REPORT(sublex_push());
4648 case LEX_INTERPSTART:
4649 if (PL_bufptr == PL_bufend)
4650 return REPORT(sublex_done());
4651 DEBUG_T({ if(*PL_bufptr != '(') PerlIO_printf(Perl_debug_log,
4652 "### Interpolated variable\n"); });
4654 PL_lex_dojoin = (*PL_bufptr == '@');
4655 PL_lex_state = LEX_INTERPNORMAL;
4656 if (PL_lex_dojoin) {
4657 start_force(PL_curforce);
4658 NEXTVAL_NEXTTOKE.ival = 0;
4660 start_force(PL_curforce);
4661 force_ident("\"", '$');
4662 start_force(PL_curforce);
4663 NEXTVAL_NEXTTOKE.ival = 0;
4665 start_force(PL_curforce);
4666 NEXTVAL_NEXTTOKE.ival = 0;
4667 force_next((2<<24)|'(');
4668 start_force(PL_curforce);
4669 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4672 /* Convert (?{...}) and friends to 'do {...}' */
4673 if (PL_lex_inpat && *PL_bufptr == '(') {
4674 PL_parser->lex_shared->re_eval_start = PL_bufptr;
4676 if (*PL_bufptr != '{')
4678 start_force(PL_curforce);
4679 /* XXX probably need a CURMAD(something) here */
4680 PL_expect = XTERMBLOCK;
4684 if (PL_lex_starts++) {
4689 sv_free(PL_thistoken);
4690 PL_thistoken = newSVpvs("");
4693 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4694 if (!PL_lex_casemods && PL_lex_inpat)
4701 case LEX_INTERPENDMAYBE:
4702 if (intuit_more(PL_bufptr)) {
4703 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4709 if (PL_lex_dojoin) {
4710 PL_lex_dojoin = FALSE;
4711 PL_lex_state = LEX_INTERPCONCAT;
4715 sv_free(PL_thistoken);
4716 PL_thistoken = newSVpvs("");
4719 PL_lex_allbrackets--;
4722 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4723 && SvEVALED(PL_lex_repl))
4725 if (PL_bufptr != PL_bufend)
4726 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4729 /* Paranoia. re_eval_start is adjusted when S_scan_heredoc sets
4730 re_eval_str. If the here-doc body’s length equals the previous
4731 value of re_eval_start, re_eval_start will now be null. So
4732 check re_eval_str as well. */
4733 if (PL_parser->lex_shared->re_eval_start
4734 || PL_parser->lex_shared->re_eval_str) {
4736 if (*PL_bufptr != ')')
4737 Perl_croak(aTHX_ "Sequence (?{...}) not terminated with ')'");
4739 /* having compiled a (?{..}) expression, return the original
4740 * text too, as a const */
4741 if (PL_parser->lex_shared->re_eval_str) {
4742 sv = PL_parser->lex_shared->re_eval_str;
4743 PL_parser->lex_shared->re_eval_str = NULL;
4745 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4746 SvPV_shrink_to_cur(sv);
4748 else sv = newSVpvn(PL_parser->lex_shared->re_eval_start,
4749 PL_bufptr - PL_parser->lex_shared->re_eval_start);
4750 start_force(PL_curforce);
4751 /* XXX probably need a CURMAD(something) here */
4752 NEXTVAL_NEXTTOKE.opval =
4753 (OP*)newSVOP(OP_CONST, 0,
4756 PL_parser->lex_shared->re_eval_start = NULL;
4762 case LEX_INTERPCONCAT:
4764 if (PL_lex_brackets)
4765 Perl_croak(aTHX_ "panic: INTERPCONCAT, lex_brackets=%ld",
4766 (long) PL_lex_brackets);
4768 if (PL_bufptr == PL_bufend)
4769 return REPORT(sublex_done());
4771 /* m'foo' still needs to be parsed for possible (?{...}) */
4772 if (SvIVX(PL_linestr) == '\'' && !PL_lex_inpat) {
4773 SV *sv = newSVsv(PL_linestr);
4775 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4779 s = scan_const(PL_bufptr);
4781 PL_lex_state = LEX_INTERPCASEMOD;
4783 PL_lex_state = LEX_INTERPSTART;
4786 if (s != PL_bufptr) {
4787 start_force(PL_curforce);
4789 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4791 NEXTVAL_NEXTTOKE = pl_yylval;
4794 if (PL_lex_starts++) {
4798 sv_free(PL_thistoken);
4799 PL_thistoken = newSVpvs("");
4802 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4803 if (!PL_lex_casemods && PL_lex_inpat)
4816 s = scan_formline(PL_bufptr);
4817 if (!PL_lex_formbrack)
4827 PL_oldoldbufptr = PL_oldbufptr;
4833 sv_free(PL_thistoken);
4836 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4840 if (isIDFIRST_lazy_if(s,UTF))
4843 SV *dsv = newSVpvs_flags("", SVs_TEMP);
4844 const char *c = UTF ? savepv(sv_uni_display(dsv, newSVpvn_flags(s,
4846 SVs_TEMP | SVf_UTF8),
4847 10, UNI_DISPLAY_ISPRINT))
4848 : Perl_form(aTHX_ "\\x%02X", (unsigned char)*s);
4849 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4850 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4851 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4859 Perl_croak(aTHX_ "Unrecognized character %s; marked by <-- HERE after %"SVf"<-- HERE near column %d", c, SVfARG(dsv), (int) len + 1);
4863 goto fake_eof; /* emulate EOF on ^D or ^Z */
4869 if (!PL_rsfp && (!PL_parser->filtered || s+1 < PL_bufend)) {
4872 if (PL_lex_brackets &&
4873 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
4874 yyerror((const char *)
4876 ? "Format not terminated"
4877 : "Missing right curly or square bracket"));
4879 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4880 "### Tokener got EOF\n");
4884 if (s++ < PL_bufend)
4885 goto retry; /* ignore stray nulls */
4888 if (!PL_in_eval && !PL_preambled) {
4889 PL_preambled = TRUE;
4895 /* Generate a string of Perl code to load the debugger.
4896 * If PERL5DB is set, it will return the contents of that,
4897 * otherwise a compile-time require of perl5db.pl. */
4899 const char * const pdb = PerlEnv_getenv("PERL5DB");
4902 sv_setpv(PL_linestr, pdb);
4903 sv_catpvs(PL_linestr,";");
4905 SETERRNO(0,SS_NORMAL);
4906 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4909 sv_setpvs(PL_linestr,"");
4910 if (PL_preambleav) {
4911 SV **svp = AvARRAY(PL_preambleav);
4912 SV **const end = svp + AvFILLp(PL_preambleav);
4914 sv_catsv(PL_linestr, *svp);
4916 sv_catpvs(PL_linestr, ";");
4918 sv_free(MUTABLE_SV(PL_preambleav));
4919 PL_preambleav = NULL;
4922 sv_catpvs(PL_linestr,
4923 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4924 if (PL_minus_n || PL_minus_p) {
4925 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4927 sv_catpvs(PL_linestr,"chomp;");
4930 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4931 || *PL_splitstr == '"')
4932 && strchr(PL_splitstr + 1, *PL_splitstr))
4933 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4935 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4936 bytes can be used as quoting characters. :-) */
4937 const char *splits = PL_splitstr;
4938 sv_catpvs(PL_linestr, "our @F=split(q\0");
4941 if (*splits == '\\')
4942 sv_catpvn(PL_linestr, splits, 1);
4943 sv_catpvn(PL_linestr, splits, 1);
4944 } while (*splits++);
4945 /* This loop will embed the trailing NUL of
4946 PL_linestr as the last thing it does before
4948 sv_catpvs(PL_linestr, ");");
4952 sv_catpvs(PL_linestr,"our @F=split(' ');");
4955 sv_catpvs(PL_linestr, "\n");
4956 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4957 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4958 PL_last_lop = PL_last_uni = NULL;
4959 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4960 update_debugger_info(PL_linestr, NULL, 0);
4965 bof = PL_rsfp ? TRUE : FALSE;
4968 fake_eof = LEX_FAKE_EOF;
4970 PL_bufptr = PL_bufend;
4971 COPLINE_INC_WITH_HERELINES;
4972 if (!lex_next_chunk(fake_eof)) {
4973 CopLINE_dec(PL_curcop);
4975 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4977 CopLINE_dec(PL_curcop);
4980 PL_realtokenstart = -1;
4983 /* If it looks like the start of a BOM or raw UTF-16,
4984 * check if it in fact is. */
4985 if (bof && PL_rsfp &&
4990 Off_t offset = (IV)PerlIO_tell(PL_rsfp);
4991 bof = (offset == (Off_t)SvCUR(PL_linestr));
4992 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
4993 /* offset may include swallowed CR */
4995 bof = (offset == (Off_t)SvCUR(PL_linestr)+1);
4998 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4999 s = swallow_bom((U8*)s);
5002 if (PL_parser->in_pod) {
5003 /* Incest with pod. */
5006 sv_catsv(PL_thiswhite, PL_linestr);
5008 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
5009 sv_setpvs(PL_linestr, "");
5010 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5011 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5012 PL_last_lop = PL_last_uni = NULL;
5013 PL_parser->in_pod = 0;
5016 if (PL_rsfp || PL_parser->filtered)
5018 } while (PL_parser->in_pod);
5019 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
5020 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5021 PL_last_lop = PL_last_uni = NULL;
5022 if (CopLINE(PL_curcop) == 1) {
5023 while (s < PL_bufend && isSPACE(*s))
5025 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
5029 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
5033 if (*s == '#' && *(s+1) == '!')
5035 #ifdef ALTERNATE_SHEBANG
5037 static char const as[] = ALTERNATE_SHEBANG;
5038 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
5039 d = s + (sizeof(as) - 1);
5041 #endif /* ALTERNATE_SHEBANG */
5050 while (*d && !isSPACE(*d))
5054 #ifdef ARG_ZERO_IS_SCRIPT
5055 if (ipathend > ipath) {
5057 * HP-UX (at least) sets argv[0] to the script name,
5058 * which makes $^X incorrect. And Digital UNIX and Linux,
5059 * at least, set argv[0] to the basename of the Perl
5060 * interpreter. So, having found "#!", we'll set it right.
5062 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
5064 assert(SvPOK(x) || SvGMAGICAL(x));
5065 if (sv_eq(x, CopFILESV(PL_curcop))) {
5066 sv_setpvn(x, ipath, ipathend - ipath);
5072 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
5073 const char * const lstart = SvPV_const(x,llen);
5075 bstart += blen - llen;
5076 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
5077 sv_setpvn(x, ipath, ipathend - ipath);
5082 TAINT_NOT; /* $^X is always tainted, but that's OK */
5084 #endif /* ARG_ZERO_IS_SCRIPT */
5089 d = instr(s,"perl -");
5091 d = instr(s,"perl");
5093 /* avoid getting into infinite loops when shebang
5094 * line contains "Perl" rather than "perl" */
5096 for (d = ipathend-4; d >= ipath; --d) {
5097 if ((*d == 'p' || *d == 'P')
5098 && !ibcmp(d, "perl", 4))
5108 #ifdef ALTERNATE_SHEBANG
5110 * If the ALTERNATE_SHEBANG on this system starts with a
5111 * character that can be part of a Perl expression, then if
5112 * we see it but not "perl", we're probably looking at the
5113 * start of Perl code, not a request to hand off to some
5114 * other interpreter. Similarly, if "perl" is there, but
5115 * not in the first 'word' of the line, we assume the line
5116 * contains the start of the Perl program.
5118 if (d && *s != '#') {
5119 const char *c = ipath;
5120 while (*c && !strchr("; \t\r\n\f\v#", *c))
5123 d = NULL; /* "perl" not in first word; ignore */
5125 *s = '#'; /* Don't try to parse shebang line */
5127 #endif /* ALTERNATE_SHEBANG */
5132 !instr(s,"indir") &&
5133 instr(PL_origargv[0],"perl"))
5140 while (s < PL_bufend && isSPACE(*s))
5142 if (s < PL_bufend) {
5143 Newx(newargv,PL_origargc+3,char*);
5145 while (s < PL_bufend && !isSPACE(*s))
5148 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
5151 newargv = PL_origargv;
5154 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
5156 Perl_croak(aTHX_ "Can't exec %s", ipath);
5159 while (*d && !isSPACE(*d))
5161 while (SPACE_OR_TAB(*d))
5165 const bool switches_done = PL_doswitches;
5166 const U32 oldpdb = PL_perldb;
5167 const bool oldn = PL_minus_n;
5168 const bool oldp = PL_minus_p;
5172 bool baduni = FALSE;
5174 const char *d2 = d1 + 1;
5175 if (parse_unicode_opts((const char **)&d2)
5179 if (baduni || *d1 == 'M' || *d1 == 'm') {
5180 const char * const m = d1;
5181 while (*d1 && !isSPACE(*d1))
5183 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
5186 d1 = moreswitches(d1);
5188 if (PL_doswitches && !switches_done) {
5189 int argc = PL_origargc;
5190 char **argv = PL_origargv;
5193 } while (argc && argv[0][0] == '-' && argv[0][1]);
5194 init_argv_symbols(argc,argv);
5196 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
5197 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
5198 /* if we have already added "LINE: while (<>) {",
5199 we must not do it again */
5201 sv_setpvs(PL_linestr, "");
5202 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5203 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5204 PL_last_lop = PL_last_uni = NULL;
5205 PL_preambled = FALSE;
5206 if (PERLDB_LINE || PERLDB_SAVESRC)
5207 (void)gv_fetchfile(PL_origfilename);
5214 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5215 PL_lex_state = LEX_FORMLINE;
5216 start_force(PL_curforce);
5217 NEXTVAL_NEXTTOKE.ival = 0;
5218 force_next(FORMRBRACK);
5223 #ifdef PERL_STRICT_CR
5224 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5226 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5228 case ' ': case '\t': case '\f': case 013:
5230 PL_realtokenstart = -1;
5232 PL_thiswhite = newSVpvs("");
5233 sv_catpvn(PL_thiswhite, s, 1);
5240 PL_realtokenstart = -1;
5244 if (PL_lex_state != LEX_NORMAL ||
5245 (PL_in_eval && !PL_rsfp && !PL_parser->filtered)) {
5246 if (*s == '#' && s == PL_linestart && PL_in_eval
5247 && !PL_rsfp && !PL_parser->filtered) {
5248 /* handle eval qq[#line 1 "foo"\n ...] */
5249 CopLINE_dec(PL_curcop);
5252 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
5254 if (!PL_in_eval || PL_rsfp || PL_parser->filtered)
5258 const bool in_comment = *s == '#';
5260 while (d < PL_bufend && *d != '\n')
5264 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5265 Perl_croak(aTHX_ "panic: input overflow, %p > %p",
5269 PL_thiswhite = newSVpvn(s, d - s);
5272 if (in_comment && d == PL_bufend
5273 && PL_lex_state == LEX_INTERPNORMAL
5274 && PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5275 && SvEVALED(PL_lex_repl) && d[-1] == '}') s--;
5278 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5279 PL_lex_state = LEX_FORMLINE;
5280 start_force(PL_curforce);
5281 NEXTVAL_NEXTTOKE.ival = 0;
5282 force_next(FORMRBRACK);
5288 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
5289 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
5292 TOKEN(PEG); /* make sure any #! line is accessible */
5297 /* if (PL_madskills && PL_lex_formbrack) { */
5299 while (d < PL_bufend && *d != '\n')
5303 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5304 Perl_croak(aTHX_ "panic: input overflow");
5305 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
5307 PL_thiswhite = newSVpvs("");
5308 if (CopLINE(PL_curcop) == 1) {
5309 sv_setpvs(PL_thiswhite, "");
5312 sv_catpvn(PL_thiswhite, s, d - s);
5326 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
5334 while (s < PL_bufend && SPACE_OR_TAB(*s))
5337 if (strnEQ(s,"=>",2)) {
5338 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
5339 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5340 OPERATOR('-'); /* unary minus */
5342 PL_last_uni = PL_oldbufptr;
5344 case 'r': ftst = OP_FTEREAD; break;
5345 case 'w': ftst = OP_FTEWRITE; break;
5346 case 'x': ftst = OP_FTEEXEC; break;
5347 case 'o': ftst = OP_FTEOWNED; break;
5348 case 'R': ftst = OP_FTRREAD; break;
5349 case 'W': ftst = OP_FTRWRITE; break;
5350 case 'X': ftst = OP_FTREXEC; break;
5351 case 'O': ftst = OP_FTROWNED; break;
5352 case 'e': ftst = OP_FTIS; break;
5353 case 'z': ftst = OP_FTZERO; break;
5354 case 's': ftst = OP_FTSIZE; break;
5355 case 'f': ftst = OP_FTFILE; break;
5356 case 'd': ftst = OP_FTDIR; break;
5357 case 'l': ftst = OP_FTLINK; break;
5358 case 'p': ftst = OP_FTPIPE; break;
5359 case 'S': ftst = OP_FTSOCK; break;
5360 case 'u': ftst = OP_FTSUID; break;
5361 case 'g': ftst = OP_FTSGID; break;
5362 case 'k': ftst = OP_FTSVTX; break;
5363 case 'b': ftst = OP_FTBLK; break;
5364 case 'c': ftst = OP_FTCHR; break;
5365 case 't': ftst = OP_FTTTY; break;
5366 case 'T': ftst = OP_FTTEXT; break;
5367 case 'B': ftst = OP_FTBINARY; break;
5368 case 'M': case 'A': case 'C':
5369 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5371 case 'M': ftst = OP_FTMTIME; break;
5372 case 'A': ftst = OP_FTATIME; break;
5373 case 'C': ftst = OP_FTCTIME; break;
5381 PL_last_lop_op = (OPCODE)ftst;
5382 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5383 "### Saw file test %c\n", (int)tmp);
5388 /* Assume it was a minus followed by a one-letter named
5389 * subroutine call (or a -bareword), then. */
5390 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5391 "### '-%c' looked like a file test but was not\n",
5398 const char tmp = *s++;
5401 if (PL_expect == XOPERATOR)
5406 else if (*s == '>') {
5409 if (isIDFIRST_lazy_if(s,UTF)) {
5410 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5418 if (PL_expect == XOPERATOR) {
5419 if (*s == '=' && !PL_lex_allbrackets &&
5420 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5427 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5429 OPERATOR('-'); /* unary minus */
5435 const char tmp = *s++;
5438 if (PL_expect == XOPERATOR)
5443 if (PL_expect == XOPERATOR) {
5444 if (*s == '=' && !PL_lex_allbrackets &&
5445 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5452 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5459 if (PL_expect != XOPERATOR) {
5460 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5461 PL_expect = XOPERATOR;
5462 force_ident(PL_tokenbuf, '*');
5470 if (*s == '=' && !PL_lex_allbrackets &&
5471 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5477 if (*s == '=' && !PL_lex_allbrackets &&
5478 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5485 if (PL_expect == XOPERATOR) {
5486 if (s[1] == '=' && !PL_lex_allbrackets &&
5487 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5492 PL_tokenbuf[0] = '%';
5493 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5494 sizeof PL_tokenbuf - 1, FALSE);
5495 if (!PL_tokenbuf[1]) {
5498 PL_expect = XOPERATOR;
5499 force_ident_maybe_lex('%');
5503 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5504 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5509 if (PL_lex_brackets > 100)
5510 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5511 PL_lex_brackstack[PL_lex_brackets++] = 0;
5512 PL_lex_allbrackets++;
5514 const char tmp = *s++;
5519 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5521 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5529 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5536 goto just_a_word_zero_gv;
5539 switch (PL_expect) {
5545 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5547 PL_bufptr = s; /* update in case we back off */
5550 "Use of := for an empty attribute list is not allowed");
5557 PL_expect = XTERMBLOCK;
5560 stuffstart = s - SvPVX(PL_linestr) - 1;
5564 while (isIDFIRST_lazy_if(s,UTF)) {
5567 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5568 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5569 if (tmp < 0) tmp = -tmp;
5584 sv = newSVpvn_flags(s, len, UTF ? SVf_UTF8 : 0);
5586 d = scan_str(d,TRUE,TRUE,FALSE);
5588 /* MUST advance bufptr here to avoid bogus
5589 "at end of line" context messages from yyerror().
5591 PL_bufptr = s + len;
5592 yyerror("Unterminated attribute parameter in attribute list");
5596 return REPORT(0); /* EOF indicator */
5600 sv_catsv(sv, PL_lex_stuff);
5601 attrs = op_append_elem(OP_LIST, attrs,
5602 newSVOP(OP_CONST, 0, sv));
5603 SvREFCNT_dec(PL_lex_stuff);
5604 PL_lex_stuff = NULL;
5607 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5609 if (PL_in_my == KEY_our) {
5610 deprecate(":unique");
5613 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5616 /* NOTE: any CV attrs applied here need to be part of
5617 the CVf_BUILTIN_ATTRS define in cv.h! */
5618 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5620 CvLVALUE_on(PL_compcv);
5622 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5624 deprecate(":locked");
5626 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5628 CvMETHOD_on(PL_compcv);
5630 /* After we've set the flags, it could be argued that
5631 we don't need to do the attributes.pm-based setting
5632 process, and shouldn't bother appending recognized
5633 flags. To experiment with that, uncomment the
5634 following "else". (Note that's already been
5635 uncommented. That keeps the above-applied built-in
5636 attributes from being intercepted (and possibly
5637 rejected) by a package's attribute routines, but is
5638 justified by the performance win for the common case
5639 of applying only built-in attributes.) */
5641 attrs = op_append_elem(OP_LIST, attrs,
5642 newSVOP(OP_CONST, 0,
5646 if (*s == ':' && s[1] != ':')
5649 break; /* require real whitespace or :'s */
5650 /* XXX losing whitespace on sequential attributes here */
5654 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5655 if (*s != ';' && *s != '}' && *s != tmp
5656 && (tmp != '=' || *s != ')')) {
5657 const char q = ((*s == '\'') ? '"' : '\'');
5658 /* If here for an expression, and parsed no attrs, back
5660 if (tmp == '=' && !attrs) {
5664 /* MUST advance bufptr here to avoid bogus "at end of line"
5665 context messages from yyerror().
5668 yyerror( (const char *)
5670 ? Perl_form(aTHX_ "Invalid separator character "
5671 "%c%c%c in attribute list", q, *s, q)
5672 : "Unterminated attribute list" ) );
5680 start_force(PL_curforce);
5681 NEXTVAL_NEXTTOKE.opval = attrs;
5682 CURMAD('_', PL_nextwhite);
5687 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5688 (s - SvPVX(PL_linestr)) - stuffstart);
5693 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5697 PL_lex_allbrackets--;
5701 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5702 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5706 PL_lex_allbrackets++;
5709 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5715 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5718 PL_lex_allbrackets--;
5724 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5727 if (PL_lex_brackets <= 0)
5728 yyerror("Unmatched right square bracket");
5731 PL_lex_allbrackets--;
5732 if (PL_lex_state == LEX_INTERPNORMAL) {
5733 if (PL_lex_brackets == 0) {
5734 if (*s == '-' && s[1] == '>')
5735 PL_lex_state = LEX_INTERPENDMAYBE;
5736 else if (*s != '[' && *s != '{')
5737 PL_lex_state = LEX_INTERPEND;
5744 if (PL_lex_brackets > 100) {
5745 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5747 switch (PL_expect) {
5749 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5750 PL_lex_allbrackets++;
5751 OPERATOR(HASHBRACK);
5753 while (s < PL_bufend && SPACE_OR_TAB(*s))
5756 PL_tokenbuf[0] = '\0';
5757 if (d < PL_bufend && *d == '-') {
5758 PL_tokenbuf[0] = '-';
5760 while (d < PL_bufend && SPACE_OR_TAB(*d))
5763 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5764 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5766 while (d < PL_bufend && SPACE_OR_TAB(*d))
5769 const char minus = (PL_tokenbuf[0] == '-');
5770 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5778 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5779 PL_lex_allbrackets++;
5784 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5785 PL_lex_allbrackets++;
5790 if (PL_oldoldbufptr == PL_last_lop)
5791 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5793 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5794 PL_lex_allbrackets++;
5797 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5799 /* This hack is to get the ${} in the message. */
5801 yyerror("syntax error");
5804 OPERATOR(HASHBRACK);
5806 /* This hack serves to disambiguate a pair of curlies
5807 * as being a block or an anon hash. Normally, expectation
5808 * determines that, but in cases where we're not in a
5809 * position to expect anything in particular (like inside
5810 * eval"") we have to resolve the ambiguity. This code
5811 * covers the case where the first term in the curlies is a
5812 * quoted string. Most other cases need to be explicitly
5813 * disambiguated by prepending a "+" before the opening
5814 * curly in order to force resolution as an anon hash.
5816 * XXX should probably propagate the outer expectation
5817 * into eval"" to rely less on this hack, but that could
5818 * potentially break current behavior of eval"".
5822 if (*s == '\'' || *s == '"' || *s == '`') {
5823 /* common case: get past first string, handling escapes */
5824 for (t++; t < PL_bufend && *t != *s;)
5825 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5829 else if (*s == 'q') {
5832 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5835 /* skip q//-like construct */
5837 char open, close, term;
5840 while (t < PL_bufend && isSPACE(*t))
5842 /* check for q => */
5843 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5844 OPERATOR(HASHBRACK);
5848 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5852 for (t++; t < PL_bufend; t++) {
5853 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5855 else if (*t == open)
5859 for (t++; t < PL_bufend; t++) {
5860 if (*t == '\\' && t+1 < PL_bufend)
5862 else if (*t == close && --brackets <= 0)
5864 else if (*t == open)
5871 /* skip plain q word */
5872 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5875 else if (isALNUM_lazy_if(t,UTF)) {
5877 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5880 while (t < PL_bufend && isSPACE(*t))
5882 /* if comma follows first term, call it an anon hash */
5883 /* XXX it could be a comma expression with loop modifiers */
5884 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5885 || (*t == '=' && t[1] == '>')))
5886 OPERATOR(HASHBRACK);
5887 if (PL_expect == XREF)
5890 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5896 pl_yylval.ival = CopLINE(PL_curcop);
5897 if (isSPACE(*s) || *s == '#')
5898 PL_copline = NOLINE; /* invalidate current command line number */
5899 TOKEN(formbrack ? '=' : '{');
5901 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5905 if (PL_lex_brackets <= 0)
5906 yyerror("Unmatched right curly bracket");
5908 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5909 PL_lex_allbrackets--;
5910 if (PL_lex_state == LEX_INTERPNORMAL) {
5911 if (PL_lex_brackets == 0) {
5912 if (PL_expect & XFAKEBRACK) {
5913 PL_expect &= XENUMMASK;
5914 PL_lex_state = LEX_INTERPEND;
5919 PL_thiswhite = newSVpvs("");
5920 sv_catpvs(PL_thiswhite,"}");
5923 return yylex(); /* ignore fake brackets */
5925 if (PL_lex_inwhat == OP_SUBST && PL_lex_repl == PL_linestr
5926 && SvEVALED(PL_lex_repl))
5927 PL_lex_state = LEX_INTERPEND;
5928 else if (*s == '-' && s[1] == '>')
5929 PL_lex_state = LEX_INTERPENDMAYBE;
5930 else if (*s != '[' && *s != '{')
5931 PL_lex_state = LEX_INTERPEND;
5934 if (PL_expect & XFAKEBRACK) {
5935 PL_expect &= XENUMMASK;
5937 return yylex(); /* ignore fake brackets */
5939 start_force(PL_curforce);
5941 curmad('X', newSVpvn(s-1,1));
5942 CURMAD('_', PL_thiswhite);
5944 force_next(formbrack ? '.' : '}');
5945 if (formbrack) LEAVE;
5948 PL_thistoken = newSVpvs("");
5950 if (formbrack == 2) { /* means . where arguments were expected */
5951 start_force(PL_curforce);
5959 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5960 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5967 if (PL_expect == XOPERATOR) {
5968 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5969 && isIDFIRST_lazy_if(s,UTF))
5971 CopLINE_dec(PL_curcop);
5972 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5973 CopLINE_inc(PL_curcop);
5975 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5976 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5983 PL_tokenbuf[0] = '&';
5984 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf + 1,
5985 sizeof PL_tokenbuf - 1, TRUE);
5986 if (PL_tokenbuf[1]) {
5987 PL_expect = XOPERATOR;
5988 force_ident_maybe_lex('&');
5992 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5998 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5999 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
6006 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6007 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
6015 const char tmp = *s++;
6017 if (!PL_lex_allbrackets &&
6018 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6025 if (!PL_lex_allbrackets &&
6026 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
6034 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
6035 && strchr("+-*/%.^&|<",tmp))
6036 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6037 "Reversed %c= operator",(int)tmp);
6039 if (PL_expect == XSTATE && isALPHA(tmp) &&
6040 (s == PL_linestart+1 || s[-2] == '\n') )
6042 if ((PL_in_eval && !PL_rsfp && !PL_parser->filtered)
6043 || PL_lex_state != LEX_NORMAL) {
6048 if (strnEQ(s,"=cut",4)) {
6064 PL_thiswhite = newSVpvs("");
6065 sv_catpvn(PL_thiswhite, PL_linestart,
6066 PL_bufend - PL_linestart);
6070 PL_parser->in_pod = 1;
6074 if (PL_expect == XBLOCK) {
6076 #ifdef PERL_STRICT_CR
6077 while (SPACE_OR_TAB(*t))
6079 while (SPACE_OR_TAB(*t) || *t == '\r')
6082 if (*t == '\n' || *t == '#') {
6085 SAVEI8(PL_parser->form_lex_state);
6086 SAVEI32(PL_lex_formbrack);
6087 PL_parser->form_lex_state = PL_lex_state;
6088 PL_lex_formbrack = PL_lex_brackets + 1;
6092 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6101 const char tmp = *s++;
6103 /* was this !=~ where !~ was meant?
6104 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
6106 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
6107 const char *t = s+1;
6109 while (t < PL_bufend && isSPACE(*t))
6112 if (*t == '/' || *t == '?' ||
6113 ((*t == 'm' || *t == 's' || *t == 'y')
6114 && !isALNUM(t[1])) ||
6115 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
6116 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6117 "!=~ should be !~");
6119 if (!PL_lex_allbrackets &&
6120 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6132 if (PL_expect != XOPERATOR) {
6133 if (s[1] != '<' && !strchr(s,'>'))
6136 s = scan_heredoc(s);
6138 s = scan_inputsymbol(s);
6139 PL_expect = XOPERATOR;
6140 TOKEN(sublex_start());
6146 if (*s == '=' && !PL_lex_allbrackets &&
6147 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6151 SHop(OP_LEFT_SHIFT);
6156 if (!PL_lex_allbrackets &&
6157 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6164 if (!PL_lex_allbrackets &&
6165 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6173 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6181 const char tmp = *s++;
6183 if (*s == '=' && !PL_lex_allbrackets &&
6184 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6188 SHop(OP_RIGHT_SHIFT);
6190 else if (tmp == '=') {
6191 if (!PL_lex_allbrackets &&
6192 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6200 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
6209 if (PL_expect == XOPERATOR) {
6210 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6211 return deprecate_commaless_var_list();
6215 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
6216 PL_tokenbuf[0] = '@';
6217 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
6218 sizeof PL_tokenbuf - 1, FALSE);
6219 if (PL_expect == XOPERATOR)
6220 no_op("Array length", s);
6221 if (!PL_tokenbuf[1])
6223 PL_expect = XOPERATOR;
6224 force_ident_maybe_lex('#');
6228 PL_tokenbuf[0] = '$';
6229 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
6230 sizeof PL_tokenbuf - 1, FALSE);
6231 if (PL_expect == XOPERATOR)
6233 if (!PL_tokenbuf[1]) {
6235 yyerror("Final $ should be \\$ or $name");
6241 const char tmp = *s;
6242 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6245 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6246 && intuit_more(s)) {
6248 PL_tokenbuf[0] = '@';
6249 if (ckWARN(WARN_SYNTAX)) {
6252 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
6255 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6256 while (t < PL_bufend && *t != ']')
6258 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6259 "Multidimensional syntax %.*s not supported",
6260 (int)((t - PL_bufptr) + 1), PL_bufptr);
6264 else if (*s == '{') {
6266 PL_tokenbuf[0] = '%';
6267 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6268 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6270 char tmpbuf[sizeof PL_tokenbuf];
6273 } while (isSPACE(*t));
6274 if (isIDFIRST_lazy_if(t,UTF)) {
6276 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6281 && get_cvn_flags(tmpbuf, len, UTF ? SVf_UTF8 : 0))
6282 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6283 "You need to quote \"%"SVf"\"",
6284 SVfARG(newSVpvn_flags(tmpbuf, len,
6285 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
6291 PL_expect = XOPERATOR;
6292 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6293 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6294 if (!islop || PL_last_lop_op == OP_GREPSTART)
6295 PL_expect = XOPERATOR;
6296 else if (strchr("$@\"'`q", *s))
6297 PL_expect = XTERM; /* e.g. print $fh "foo" */
6298 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6299 PL_expect = XTERM; /* e.g. print $fh &sub */
6300 else if (isIDFIRST_lazy_if(s,UTF)) {
6301 char tmpbuf[sizeof PL_tokenbuf];
6303 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6304 if ((t2 = keyword(tmpbuf, len, 0))) {
6305 /* binary operators exclude handle interpretations */
6317 PL_expect = XTERM; /* e.g. print $fh length() */
6322 PL_expect = XTERM; /* e.g. print $fh subr() */
6325 else if (isDIGIT(*s))
6326 PL_expect = XTERM; /* e.g. print $fh 3 */
6327 else if (*s == '.' && isDIGIT(s[1]))
6328 PL_expect = XTERM; /* e.g. print $fh .3 */
6329 else if ((*s == '?' || *s == '-' || *s == '+')
6330 && !isSPACE(s[1]) && s[1] != '=')
6331 PL_expect = XTERM; /* e.g. print $fh -1 */
6332 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6334 PL_expect = XTERM; /* e.g. print $fh /.../
6335 XXX except DORDOR operator
6337 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6339 PL_expect = XTERM; /* print $fh <<"EOF" */
6342 force_ident_maybe_lex('$');
6346 if (PL_expect == XOPERATOR)
6348 PL_tokenbuf[0] = '@';
6349 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6350 if (!PL_tokenbuf[1]) {
6353 if (PL_lex_state == LEX_NORMAL)
6355 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6357 PL_tokenbuf[0] = '%';
6359 /* Warn about @ where they meant $. */
6360 if (*s == '[' || *s == '{') {
6361 if (ckWARN(WARN_SYNTAX)) {
6362 const char *t = s + 1;
6363 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
6364 t += UTF ? UTF8SKIP(t) : 1;
6365 if (*t == '}' || *t == ']') {
6367 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6368 /* diag_listed_as: Scalar value @%s[%s] better written as $%s[%s] */
6369 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6370 "Scalar value %"SVf" better written as $%"SVf,
6371 SVfARG(newSVpvn_flags(PL_bufptr, (STRLEN)(t-PL_bufptr),
6372 SVs_TEMP | (UTF ? SVf_UTF8 : 0 ))),
6373 SVfARG(newSVpvn_flags(PL_bufptr+1, (STRLEN)(t-PL_bufptr-1),
6374 SVs_TEMP | (UTF ? SVf_UTF8 : 0 ))));
6379 PL_expect = XOPERATOR;
6380 force_ident_maybe_lex('@');
6383 case '/': /* may be division, defined-or, or pattern */
6384 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
6385 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6386 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6391 case '?': /* may either be conditional or pattern */
6392 if (PL_expect == XOPERATOR) {
6395 if (!PL_lex_allbrackets &&
6396 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6400 PL_lex_allbrackets++;
6406 /* A // operator. */
6407 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6408 (*s == '=' ? LEX_FAKEEOF_ASSIGN :
6409 LEX_FAKEEOF_LOGIC)) {
6417 if (*s == '=' && !PL_lex_allbrackets &&
6418 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6427 /* Disable warning on "study /blah/" */
6428 if (PL_oldoldbufptr == PL_last_uni
6429 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6430 || memNE(PL_last_uni, "study", 5)
6431 || isALNUM_lazy_if(PL_last_uni+5,UTF)
6435 deprecate("?PATTERN? without explicit operator");
6436 s = scan_pat(s,OP_MATCH);
6437 TERM(sublex_start());
6441 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6442 #ifdef PERL_STRICT_CR
6445 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6447 && (s == PL_linestart || s[-1] == '\n') )
6450 formbrack = 2; /* dot seen where arguments expected */
6453 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6457 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6460 if (!PL_lex_allbrackets &&
6461 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6468 pl_yylval.ival = OPf_SPECIAL;
6474 if (*s == '=' && !PL_lex_allbrackets &&
6475 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6482 case '0': case '1': case '2': case '3': case '4':
6483 case '5': case '6': case '7': case '8': case '9':
6484 s = scan_num(s, &pl_yylval);
6485 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6486 if (PL_expect == XOPERATOR)
6491 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6492 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6493 if (PL_expect == XOPERATOR) {
6494 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6495 return deprecate_commaless_var_list();
6502 pl_yylval.ival = OP_CONST;
6503 TERM(sublex_start());
6506 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6507 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6508 if (PL_expect == XOPERATOR) {
6509 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6510 return deprecate_commaless_var_list();
6517 pl_yylval.ival = OP_CONST;
6518 /* FIXME. I think that this can be const if char *d is replaced by
6519 more localised variables. */
6520 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6521 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6522 pl_yylval.ival = OP_STRINGIFY;
6526 TERM(sublex_start());
6529 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
6530 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6531 if (PL_expect == XOPERATOR)
6532 no_op("Backticks",s);
6535 readpipe_override();
6536 TERM(sublex_start());
6540 if (PL_lex_inwhat && isDIGIT(*s))
6541 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6543 if (PL_expect == XOPERATOR)
6544 no_op("Backslash",s);
6548 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6549 char *start = s + 2;
6550 while (isDIGIT(*start) || *start == '_')
6552 if (*start == '.' && isDIGIT(start[1])) {
6553 s = scan_num(s, &pl_yylval);
6556 else if ((*start == ':' && start[1] == ':')
6557 || (PL_expect == XSTATE && *start == ':'))
6559 else if (PL_expect == XSTATE) {
6561 while (d < PL_bufend && isSPACE(*d)) d++;
6562 if (*d == ':') goto keylookup;
6564 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6565 if (!isALPHA(*start) && (PL_expect == XTERM
6566 || PL_expect == XREF || PL_expect == XSTATE
6567 || PL_expect == XTERMORDORDOR)) {
6568 GV *const gv = gv_fetchpvn_flags(s, start - s,
6569 UTF ? SVf_UTF8 : 0, SVt_PVCV);
6571 s = scan_num(s, &pl_yylval);
6578 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6631 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6633 /* Some keywords can be followed by any delimiter, including ':' */
6634 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6636 /* x::* is just a word, unless x is "CORE" */
6637 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6641 while (d < PL_bufend && isSPACE(*d))
6642 d++; /* no comments skipped here, or s### is misparsed */
6644 /* Is this a word before a => operator? */
6645 if (*d == '=' && d[1] == '>') {
6648 = (OP*)newSVOP(OP_CONST, 0,
6649 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6650 pl_yylval.opval->op_private = OPpCONST_BARE;
6654 /* Check for plugged-in keyword */
6658 char *saved_bufptr = PL_bufptr;
6660 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6662 if (result == KEYWORD_PLUGIN_DECLINE) {
6663 /* not a plugged-in keyword */
6664 PL_bufptr = saved_bufptr;
6665 } else if (result == KEYWORD_PLUGIN_STMT) {
6666 pl_yylval.opval = o;
6669 return REPORT(PLUGSTMT);
6670 } else if (result == KEYWORD_PLUGIN_EXPR) {
6671 pl_yylval.opval = o;
6673 PL_expect = XOPERATOR;
6674 return REPORT(PLUGEXPR);
6676 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6681 /* Check for built-in keyword */
6682 tmp = keyword(PL_tokenbuf, len, 0);
6684 /* Is this a label? */
6685 if (!anydelim && PL_expect == XSTATE
6686 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6688 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6689 newSVpvn_flags(PL_tokenbuf,
6690 len, UTF ? SVf_UTF8 : 0));
6695 /* Check for lexical sub */
6696 if (PL_expect != XOPERATOR) {
6697 char tmpbuf[sizeof PL_tokenbuf + 1];
6699 Copy(PL_tokenbuf, tmpbuf+1, len, char);
6700 off = pad_findmy_pvn(tmpbuf, len+1, UTF ? SVf_UTF8 : 0);
6701 if (off != NOT_IN_PAD) {
6702 assert(off); /* we assume this is boolean-true below */
6703 if (PAD_COMPNAME_FLAGS_isOUR(off)) {
6704 HV * const stash = PAD_COMPNAME_OURSTASH(off);
6705 HEK * const stashname = HvNAME_HEK(stash);
6706 sv = newSVhek(stashname);
6707 sv_catpvs(sv, "::");
6708 sv_catpvn_flags(sv, PL_tokenbuf, len,
6709 (UTF ? SV_CATUTF8 : SV_CATBYTES));
6710 gv = gv_fetchsv(sv, GV_NOADD_NOINIT | SvUTF8(sv),
6715 rv2cv_op = newOP(OP_PADANY, 0);
6716 rv2cv_op->op_targ = off;
6717 rv2cv_op = (OP*)newCVREF(0, rv2cv_op);
6718 cv = (CV *)PAD_SV(off);
6726 if (tmp < 0) { /* second-class keyword? */
6727 GV *ogv = NULL; /* override (winner) */
6728 GV *hgv = NULL; /* hidden (loser) */
6729 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6731 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6732 UTF ? SVf_UTF8 : 0, SVt_PVCV)) &&
6735 if (GvIMPORTED_CV(gv))
6737 else if (! CvMETHOD(cv))
6741 (gvp = (GV**)hv_fetch(PL_globalstash, PL_tokenbuf,
6742 UTF ? -(I32)len : (I32)len, FALSE)) &&
6743 (gv = *gvp) && isGV_with_GP(gv) &&
6744 GvCVu(gv) && GvIMPORTED_CV(gv))
6751 tmp = 0; /* overridden by import or by GLOBAL */
6754 && -tmp==KEY_lock /* XXX generalizable kludge */
6757 tmp = 0; /* any sub overrides "weak" keyword */
6759 else { /* no override */
6761 if (tmp == KEY_dump) {
6762 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6763 "dump() better written as CORE::dump()");
6767 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6768 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6769 "Ambiguous call resolved as CORE::%s(), "
6770 "qualify as such or use &",
6778 default: /* not a keyword */
6779 /* Trade off - by using this evil construction we can pull the
6780 variable gv into the block labelled keylookup. If not, then
6781 we have to give it function scope so that the goto from the
6782 earlier ':' case doesn't bypass the initialisation. */
6784 just_a_word_zero_gv:
6796 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6797 const char penultchar =
6798 lastchar && PL_bufptr - 2 >= PL_linestart
6802 SV *nextPL_nextwhite = 0;
6806 /* Get the rest if it looks like a package qualifier */
6808 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6810 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6813 Perl_croak(aTHX_ "Bad name after %"SVf"%s",
6814 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
6815 (UTF ? SVf_UTF8 : 0) | SVs_TEMP )),
6816 *s == '\'' ? "'" : "::");
6821 if (PL_expect == XOPERATOR) {
6822 if (PL_bufptr == PL_linestart) {
6823 CopLINE_dec(PL_curcop);
6824 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6825 CopLINE_inc(PL_curcop);
6828 no_op("Bareword",s);
6831 /* Look for a subroutine with this name in current package,
6832 unless this is a lexical sub, or name is "Foo::",
6833 in which case Foo is a bareword
6834 (and a package name). */
6836 if (len > 2 && !PL_madskills &&
6837 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6839 if (ckWARN(WARN_BAREWORD)
6840 && ! gv_fetchpvn_flags(PL_tokenbuf, len, UTF ? SVf_UTF8 : 0, SVt_PVHV))
6841 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6842 "Bareword \"%"SVf"\" refers to nonexistent package",
6843 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
6844 (UTF ? SVf_UTF8 : 0) | SVs_TEMP)));
6846 PL_tokenbuf[len] = '\0';
6852 /* Mustn't actually add anything to a symbol table.
6853 But also don't want to "initialise" any placeholder
6854 constants that might already be there into full
6855 blown PVGVs with attached PVCV. */
6856 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6857 GV_NOADD_NOINIT | ( UTF ? SVf_UTF8 : 0 ),
6863 /* if we saw a global override before, get the right name */
6866 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
6867 len ? len : strlen(PL_tokenbuf));
6869 SV * const tmp_sv = sv;
6870 sv = newSVpvs("CORE::GLOBAL::");
6871 sv_catsv(sv, tmp_sv);
6872 SvREFCNT_dec(tmp_sv);
6876 if (PL_madskills && !PL_thistoken) {
6877 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6878 PL_thistoken = newSVpvn(start,s - start);
6879 PL_realtokenstart = s - SvPVX(PL_linestr);
6883 /* Presume this is going to be a bareword of some sort. */
6885 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6886 pl_yylval.opval->op_private = OPpCONST_BARE;
6888 /* And if "Foo::", then that's what it certainly is. */
6894 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
6895 const_op->op_private = OPpCONST_BARE;
6896 rv2cv_op = newCVREF(0, const_op);
6897 cv = lex ? GvCV(gv) : rv2cv_op_cv(rv2cv_op, 0);
6900 /* See if it's the indirect object for a list operator. */
6902 if (PL_oldoldbufptr &&
6903 PL_oldoldbufptr < PL_bufptr &&
6904 (PL_oldoldbufptr == PL_last_lop
6905 || PL_oldoldbufptr == PL_last_uni) &&
6906 /* NO SKIPSPACE BEFORE HERE! */
6907 (PL_expect == XREF ||
6908 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6910 bool immediate_paren = *s == '(';
6912 /* (Now we can afford to cross potential line boundary.) */
6913 s = SKIPSPACE2(s,nextPL_nextwhite);
6915 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6918 /* Two barewords in a row may indicate method call. */
6920 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6921 (tmp = intuit_method(s, gv, cv))) {
6923 if (tmp == METHOD && !PL_lex_allbrackets &&
6924 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6925 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6929 /* If not a declared subroutine, it's an indirect object. */
6930 /* (But it's an indir obj regardless for sort.) */
6931 /* Also, if "_" follows a filetest operator, it's a bareword */
6934 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6936 (PL_last_lop_op != OP_MAPSTART &&
6937 PL_last_lop_op != OP_GREPSTART))))
6938 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6939 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6942 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6947 PL_expect = XOPERATOR;
6950 s = SKIPSPACE2(s,nextPL_nextwhite);
6951 PL_nextwhite = nextPL_nextwhite;
6956 /* Is this a word before a => operator? */
6957 if (*s == '=' && s[1] == '>' && !pkgname) {
6960 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6961 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6962 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6966 /* If followed by a paren, it's certainly a subroutine. */
6971 while (SPACE_OR_TAB(*d))
6973 if (*d == ')' && (sv = cv_const_sv(cv))) {
6980 PL_nextwhite = PL_thiswhite;
6983 start_force(PL_curforce);
6985 NEXTVAL_NEXTTOKE.opval =
6986 off ? rv2cv_op : pl_yylval.opval;
6987 PL_expect = XOPERATOR;
6990 PL_nextwhite = nextPL_nextwhite;
6991 curmad('X', PL_thistoken);
6992 PL_thistoken = newSVpvs("");
6996 op_free(pl_yylval.opval), force_next(PRIVATEREF);
6997 else op_free(rv2cv_op), force_next(WORD);
7002 /* If followed by var or block, call it a method (unless sub) */
7004 if ((*s == '$' || *s == '{') && !cv) {
7006 PL_last_lop = PL_oldbufptr;
7007 PL_last_lop_op = OP_METHOD;
7008 if (!PL_lex_allbrackets &&
7009 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7010 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7014 /* If followed by a bareword, see if it looks like indir obj. */
7017 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
7018 && (tmp = intuit_method(s, gv, cv))) {
7020 if (tmp == METHOD && !PL_lex_allbrackets &&
7021 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7022 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7026 /* Not a method, so call it a subroutine (if defined) */
7029 if (lastchar == '-' && penultchar != '-') {
7030 const SV *tmpsv = newSVpvn_flags( PL_tokenbuf, len ? len : strlen(PL_tokenbuf), (UTF ? SVf_UTF8 : 0) | SVs_TEMP );
7031 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7032 "Ambiguous use of -%"SVf" resolved as -&%"SVf"()",
7033 SVfARG(tmpsv), SVfARG(tmpsv));
7035 /* Check for a constant sub */
7036 if ((sv = cv_const_sv(cv))) {
7039 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
7040 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
7041 pl_yylval.opval->op_private = OPpCONST_FOLDED;
7042 pl_yylval.opval->op_flags |= OPf_SPECIAL;
7046 op_free(pl_yylval.opval);
7047 pl_yylval.opval = rv2cv_op;
7048 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
7049 PL_last_lop = PL_oldbufptr;
7050 PL_last_lop_op = OP_ENTERSUB;
7051 /* Is there a prototype? */
7058 STRLEN protolen = CvPROTOLEN(cv);
7059 const char *proto = CvPROTO(cv);
7063 if ((optional = *proto == ';'))
7066 while (*proto == ';');
7070 *proto == '$' || *proto == '_'
7071 || *proto == '*' || *proto == '+'
7076 *proto == '\\' && proto[1] && proto[2] == '\0'
7079 UNIPROTO(UNIOPSUB,optional);
7080 if (*proto == '\\' && proto[1] == '[') {
7081 const char *p = proto + 2;
7082 while(*p && *p != ']')
7084 if(*p == ']' && !p[1])
7085 UNIPROTO(UNIOPSUB,optional);
7087 if (*proto == '&' && *s == '{') {
7089 sv_setpvs(PL_subname, "__ANON__");
7091 sv_setpvs(PL_subname, "__ANON__::__ANON__");
7092 if (!PL_lex_allbrackets &&
7093 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7094 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7101 PL_nextwhite = PL_thiswhite;
7104 start_force(PL_curforce);
7105 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7108 PL_nextwhite = nextPL_nextwhite;
7109 curmad('X', PL_thistoken);
7110 PL_thistoken = newSVpvs("");
7112 force_next(off ? PRIVATEREF : WORD);
7113 if (!PL_lex_allbrackets &&
7114 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7115 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7120 /* Guess harder when madskills require "best effort". */
7121 if (PL_madskills && (!gv || !GvCVu(gv))) {
7122 int probable_sub = 0;
7123 if (strchr("\"'`$@%0123456789!*+{[<", *s))
7125 else if (isALPHA(*s)) {
7129 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
7130 if (!keyword(tmpbuf, tmplen, 0))
7133 while (d < PL_bufend && isSPACE(*d))
7135 if (*d == '=' && d[1] == '>')
7140 gv = gv_fetchpv(PL_tokenbuf, GV_ADD | ( UTF ? SVf_UTF8 : 0 ),
7142 op_free(pl_yylval.opval);
7143 pl_yylval.opval = rv2cv_op;
7144 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
7145 PL_last_lop = PL_oldbufptr;
7146 PL_last_lop_op = OP_ENTERSUB;
7147 PL_nextwhite = PL_thiswhite;
7149 start_force(PL_curforce);
7150 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7152 PL_nextwhite = nextPL_nextwhite;
7153 curmad('X', PL_thistoken);
7154 PL_thistoken = newSVpvs("");
7155 force_next(off ? PRIVATEREF : WORD);
7156 if (!PL_lex_allbrackets &&
7157 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7158 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7162 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
7164 force_next(off ? PRIVATEREF : WORD);
7165 if (!PL_lex_allbrackets &&
7166 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7167 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7172 /* Call it a bare word */
7174 if (PL_hints & HINT_STRICT_SUBS)
7175 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7178 /* after "print" and similar functions (corresponding to
7179 * "F? L" in opcode.pl), whatever wasn't already parsed as
7180 * a filehandle should be subject to "strict subs".
7181 * Likewise for the optional indirect-object argument to system
7182 * or exec, which can't be a bareword */
7183 if ((PL_last_lop_op == OP_PRINT
7184 || PL_last_lop_op == OP_PRTF
7185 || PL_last_lop_op == OP_SAY
7186 || PL_last_lop_op == OP_SYSTEM
7187 || PL_last_lop_op == OP_EXEC)
7188 && (PL_hints & HINT_STRICT_SUBS))
7189 pl_yylval.opval->op_private |= OPpCONST_STRICT;
7190 if (lastchar != '-') {
7191 if (ckWARN(WARN_RESERVED)) {
7195 if (!*d && !gv_stashpv(PL_tokenbuf, UTF ? SVf_UTF8 : 0))
7196 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
7204 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
7205 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7206 "Operator or semicolon missing before %c%"SVf,
7207 lastchar, SVfARG(newSVpvn_flags(PL_tokenbuf,
7208 strlen(PL_tokenbuf),
7209 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
7210 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
7211 "Ambiguous use of %c resolved as operator %c",
7212 lastchar, lastchar);
7219 (OP*)newSVOP(OP_CONST, 0, newSVpv(CopFILE(PL_curcop),0))
7224 (OP*)newSVOP(OP_CONST, 0,
7225 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)))
7228 case KEY___PACKAGE__:
7230 (OP*)newSVOP(OP_CONST, 0,
7232 ? newSVhek(HvNAME_HEK(PL_curstash))
7239 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
7240 const char *pname = "main";
7243 if (PL_tokenbuf[2] == 'D')
7246 PL_curstash ? PL_curstash : PL_defstash;
7247 pname = HvNAME_get(stash);
7248 plen = HvNAMELEN (stash);
7249 if(HvNAMEUTF8(stash)) putf8 = SVf_UTF8;
7251 gv = gv_fetchpvn_flags(
7252 Perl_form(aTHX_ "%*s::DATA", (int)plen, pname),
7253 plen+6, GV_ADD|putf8, SVt_PVIO
7257 GvIOp(gv) = newIO();
7258 IoIFP(GvIOp(gv)) = PL_rsfp;
7259 #if defined(HAS_FCNTL) && defined(F_SETFD)
7261 const int fd = PerlIO_fileno(PL_rsfp);
7262 fcntl(fd,F_SETFD,fd >= 3);
7265 /* Mark this internal pseudo-handle as clean */
7266 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
7267 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
7268 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
7270 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
7271 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
7272 /* if the script was opened in binmode, we need to revert
7273 * it to text mode for compatibility; but only iff it has CRs
7274 * XXX this is a questionable hack at best. */
7275 if (PL_bufend-PL_bufptr > 2
7276 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
7279 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
7280 loc = PerlIO_tell(PL_rsfp);
7281 (void)PerlIO_seek(PL_rsfp, 0L, 0);
7284 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
7286 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
7287 #endif /* NETWARE */
7289 PerlIO_seek(PL_rsfp, loc, 0);
7293 #ifdef PERLIO_LAYERS
7296 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
7297 else if (PL_encoding) {
7304 XPUSHs(PL_encoding);
7306 call_method("name", G_SCALAR);
7310 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7311 Perl_form(aTHX_ ":encoding(%"SVf")",
7320 if (PL_realtokenstart >= 0) {
7321 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7323 PL_endwhite = newSVpvs("");
7324 sv_catsv(PL_endwhite, PL_thiswhite);
7326 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
7327 PL_realtokenstart = -1;
7329 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
7339 FUN0OP(newPVOP(OP_RUNCV,0,NULL));
7348 if (PL_expect == XSTATE) {
7355 if (*s == ':' && s[1] == ':') {
7359 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7360 if ((*s == ':' && s[1] == ':')
7361 || (!(tmp = keyword(PL_tokenbuf, len, 1)) && *s == '\''))
7365 Copy(PL_bufptr, PL_tokenbuf, olen, char);
7369 Perl_croak(aTHX_ "CORE::%"SVf" is not a keyword",
7370 SVfARG(newSVpvn_flags(PL_tokenbuf, len,
7371 (UTF ? SVf_UTF8 : 0) | SVs_TEMP)));
7374 else if (tmp == KEY_require || tmp == KEY_do
7376 /* that's a way to remember we saw "CORE::" */
7389 LOP(OP_ACCEPT,XTERM);
7392 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7397 LOP(OP_ATAN2,XTERM);
7403 LOP(OP_BINMODE,XTERM);
7406 LOP(OP_BLESS,XTERM);
7415 /* We have to disambiguate the two senses of
7416 "continue". If the next token is a '{' then
7417 treat it as the start of a continue block;
7418 otherwise treat it as a control operator.
7428 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7438 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7447 if (!PL_cryptseen) {
7448 PL_cryptseen = TRUE;
7452 LOP(OP_CRYPT,XTERM);
7455 LOP(OP_CHMOD,XTERM);
7458 LOP(OP_CHOWN,XTERM);
7461 LOP(OP_CONNECT,XTERM);
7481 d = scan_word(s, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
7483 if (len && !keyword(PL_tokenbuf + 1, len, 0)) {
7486 force_ident_maybe_lex('&');
7491 if (orig_keyword == KEY_do) {
7500 PL_hints |= HINT_BLOCK_SCOPE;
7510 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7511 STR_WITH_LEN("NDBM_File::"),
7512 STR_WITH_LEN("DB_File::"),
7513 STR_WITH_LEN("GDBM_File::"),
7514 STR_WITH_LEN("SDBM_File::"),
7515 STR_WITH_LEN("ODBM_File::"),
7517 LOP(OP_DBMOPEN,XTERM);
7523 PL_expect = XOPERATOR;
7524 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7531 pl_yylval.ival = CopLINE(PL_curcop);
7535 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7549 if (*s == '{') { /* block eval */
7550 PL_expect = XTERMBLOCK;
7551 UNIBRACK(OP_ENTERTRY);
7553 else { /* string eval */
7555 UNIBRACK(OP_ENTEREVAL);
7560 UNIBRACK(-OP_ENTEREVAL);
7574 case KEY_endhostent:
7580 case KEY_endservent:
7583 case KEY_endprotoent:
7594 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7596 pl_yylval.ival = CopLINE(PL_curcop);
7598 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7601 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
7604 if ((PL_bufend - p) >= 3 &&
7605 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7607 else if ((PL_bufend - p) >= 4 &&
7608 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7611 if (isIDFIRST_lazy_if(p,UTF)) {
7612 p = scan_ident(p, PL_bufend,
7613 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
7617 Perl_croak(aTHX_ "Missing $ on loop variable");
7619 s = SvPVX(PL_linestr) + soff;
7625 LOP(OP_FORMLINE,XTERM);
7634 LOP(OP_FCNTL,XTERM);
7640 LOP(OP_FLOCK,XTERM);
7643 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7648 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7653 LOP(OP_GREPSTART, XREF);
7656 PL_expect = XOPERATOR;
7657 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7672 case KEY_getpriority:
7673 LOP(OP_GETPRIORITY,XTERM);
7675 case KEY_getprotobyname:
7678 case KEY_getprotobynumber:
7679 LOP(OP_GPBYNUMBER,XTERM);
7681 case KEY_getprotoent:
7693 case KEY_getpeername:
7694 UNI(OP_GETPEERNAME);
7696 case KEY_gethostbyname:
7699 case KEY_gethostbyaddr:
7700 LOP(OP_GHBYADDR,XTERM);
7702 case KEY_gethostent:
7705 case KEY_getnetbyname:
7708 case KEY_getnetbyaddr:
7709 LOP(OP_GNBYADDR,XTERM);
7714 case KEY_getservbyname:
7715 LOP(OP_GSBYNAME,XTERM);
7717 case KEY_getservbyport:
7718 LOP(OP_GSBYPORT,XTERM);
7720 case KEY_getservent:
7723 case KEY_getsockname:
7724 UNI(OP_GETSOCKNAME);
7726 case KEY_getsockopt:
7727 LOP(OP_GSOCKOPT,XTERM);
7742 pl_yylval.ival = CopLINE(PL_curcop);
7747 orig_keyword==KEY_glob ? (orig_keyword=0, -OP_GLOB) : OP_GLOB,
7755 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7757 pl_yylval.ival = CopLINE(PL_curcop);
7761 LOP(OP_INDEX,XTERM);
7767 LOP(OP_IOCTL,XTERM);
7779 PL_expect = XOPERATOR;
7780 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7797 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7802 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7816 LOP(OP_LISTEN,XTERM);
7825 s = scan_pat(s,OP_MATCH);
7826 TERM(sublex_start());
7829 LOP(OP_MAPSTART, XREF);
7832 LOP(OP_MKDIR,XTERM);
7835 LOP(OP_MSGCTL,XTERM);
7838 LOP(OP_MSGGET,XTERM);
7841 LOP(OP_MSGRCV,XTERM);
7844 LOP(OP_MSGSND,XTERM);
7849 PL_in_my = (U16)tmp;
7851 if (isIDFIRST_lazy_if(s,UTF)) {
7855 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7856 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7858 if (!FEATURE_LEXSUBS_IS_ENABLED)
7860 "Experimental \"%s\" subs not enabled",
7861 tmp == KEY_my ? "my" :
7862 tmp == KEY_state ? "state" : "our");
7865 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7866 if (!PL_in_my_stash) {
7869 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7870 yyerror_pv(tmpbuf, UTF ? SVf_UTF8 : 0);
7873 if (PL_madskills) { /* just add type to declarator token */
7874 sv_catsv(PL_thistoken, PL_nextwhite);
7876 sv_catpvn(PL_thistoken, start, s - start);
7884 PL_expect = XOPERATOR;
7885 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7889 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7894 s = tokenize_use(0, s);
7898 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7901 if (!PL_lex_allbrackets &&
7902 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7903 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7909 if (isIDFIRST_lazy_if(s,UTF)) {
7911 for (d = s; isALNUM_lazy_if(d,UTF);) {
7912 d += UTF ? UTF8SKIP(d) : 1;
7914 while (UTF8_IS_CONTINUED(*d) && is_utf8_mark((U8*)d)) {
7915 d += UTF ? UTF8SKIP(d) : 1;
7919 for (t=d; isSPACE(*t);)
7921 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7923 && !(t[0] == '=' && t[1] == '>')
7924 && !(t[0] == ':' && t[1] == ':')
7925 && !keyword(s, d-s, 0)
7927 SV *tmpsv = newSVpvn_flags(s, (STRLEN)(d-s),
7928 SVs_TEMP | (UTF ? SVf_UTF8 : 0));
7929 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7930 "Precedence problem: open %"SVf" should be open(%"SVf")",
7931 SVfARG(tmpsv), SVfARG(tmpsv));
7937 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7939 pl_yylval.ival = OP_OR;
7949 LOP(OP_OPEN_DIR,XTERM);
7952 checkcomma(s,PL_tokenbuf,"filehandle");
7956 checkcomma(s,PL_tokenbuf,"filehandle");
7975 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7977 s = force_strict_version(s);
7978 PL_lex_expect = XBLOCK;
7982 LOP(OP_PIPE_OP,XTERM);
7985 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
7988 pl_yylval.ival = OP_CONST;
7989 TERM(sublex_start());
7996 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
7999 PL_expect = XOPERATOR;
8000 if (SvCUR(PL_lex_stuff)) {
8001 int warned_comma = !ckWARN(WARN_QW);
8002 int warned_comment = warned_comma;
8003 d = SvPV_force(PL_lex_stuff, len);
8005 for (; isSPACE(*d) && len; --len, ++d)
8010 if (!warned_comma || !warned_comment) {
8011 for (; !isSPACE(*d) && len; --len, ++d) {
8012 if (!warned_comma && *d == ',') {
8013 Perl_warner(aTHX_ packWARN(WARN_QW),
8014 "Possible attempt to separate words with commas");
8017 else if (!warned_comment && *d == '#') {
8018 Perl_warner(aTHX_ packWARN(WARN_QW),
8019 "Possible attempt to put comments in qw() list");
8025 for (; !isSPACE(*d) && len; --len, ++d)
8028 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
8029 words = op_append_elem(OP_LIST, words,
8030 newSVOP(OP_CONST, 0, tokeq(sv)));
8035 words = newNULLLIST();
8037 SvREFCNT_dec(PL_lex_stuff);
8038 PL_lex_stuff = NULL;
8040 PL_expect = XOPERATOR;
8041 pl_yylval.opval = sawparens(words);
8046 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8049 pl_yylval.ival = OP_STRINGIFY;
8050 if (SvIVX(PL_lex_stuff) == '\'')
8051 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
8052 TERM(sublex_start());
8055 s = scan_pat(s,OP_QR);
8056 TERM(sublex_start());
8059 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8062 readpipe_override();
8063 TERM(sublex_start());
8070 PL_expect = XOPERATOR;
8072 s = force_version(s, FALSE);
8074 else if (*s != 'v' || !isDIGIT(s[1])
8075 || (s = force_version(s, TRUE), *s == 'v'))
8077 *PL_tokenbuf = '\0';
8078 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8079 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
8080 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf),
8081 GV_ADD | (UTF ? SVf_UTF8 : 0));
8083 yyerror("<> should be quotes");
8085 if (orig_keyword == KEY_require) {
8093 PL_last_uni = PL_oldbufptr;
8094 PL_last_lop_op = OP_REQUIRE;
8096 return REPORT( (int)REQUIRE );
8102 PL_expect = XOPERATOR;
8103 s = force_word(s,WORD,TRUE,FALSE,FALSE);
8107 LOP(OP_RENAME,XTERM);
8116 LOP(OP_RINDEX,XTERM);
8125 UNIDOR(OP_READLINE);
8128 UNIDOR(OP_BACKTICK);
8137 LOP(OP_REVERSE,XTERM);
8140 UNIDOR(OP_READLINK);
8147 if (pl_yylval.opval)
8148 TERM(sublex_start());
8150 TOKEN(1); /* force error */
8153 checkcomma(s,PL_tokenbuf,"filehandle");
8163 LOP(OP_SELECT,XTERM);
8169 LOP(OP_SEMCTL,XTERM);
8172 LOP(OP_SEMGET,XTERM);
8175 LOP(OP_SEMOP,XTERM);
8181 LOP(OP_SETPGRP,XTERM);
8183 case KEY_setpriority:
8184 LOP(OP_SETPRIORITY,XTERM);
8186 case KEY_sethostent:
8192 case KEY_setservent:
8195 case KEY_setprotoent:
8205 LOP(OP_SEEKDIR,XTERM);
8207 case KEY_setsockopt:
8208 LOP(OP_SSOCKOPT,XTERM);
8214 LOP(OP_SHMCTL,XTERM);
8217 LOP(OP_SHMGET,XTERM);
8220 LOP(OP_SHMREAD,XTERM);
8223 LOP(OP_SHMWRITE,XTERM);
8226 LOP(OP_SHUTDOWN,XTERM);
8235 LOP(OP_SOCKET,XTERM);
8237 case KEY_socketpair:
8238 LOP(OP_SOCKPAIR,XTERM);
8241 checkcomma(s,PL_tokenbuf,"subroutine name");
8244 s = force_word(s,WORD,TRUE,TRUE,FALSE);
8248 LOP(OP_SPLIT,XTERM);
8251 LOP(OP_SPRINTF,XTERM);
8254 LOP(OP_SPLICE,XTERM);
8269 LOP(OP_SUBSTR,XTERM);
8275 char * const tmpbuf = PL_tokenbuf + 1;
8276 SSize_t tboffset = 0;
8277 expectation attrful;
8278 bool have_name, have_proto;
8279 const int key = tmp;
8284 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
8285 SV *subtoken = newSVpvn_flags(tstart, s - tstart, SvUTF8(PL_linestr));
8289 s = SKIPSPACE2(s,tmpwhite);
8295 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
8296 (*s == ':' && s[1] == ':'))
8299 SV *nametoke = NULL;
8303 attrful = XATTRBLOCK;
8304 /* remember buffer pos'n for later force_word */
8305 tboffset = s - PL_oldbufptr;
8306 d = scan_word(s, tmpbuf, sizeof PL_tokenbuf - 1, TRUE,
8310 nametoke = newSVpvn_flags(s, d - s, SvUTF8(PL_linestr));
8313 if (memchr(tmpbuf, ':', len) || key != KEY_sub
8315 PL_tokenbuf, len + 1, UTF ? SVf_UTF8 : 0
8317 sv_setpvn(PL_subname, tmpbuf, len);
8319 sv_setsv(PL_subname,PL_curstname);
8320 sv_catpvs(PL_subname,"::");
8321 sv_catpvn(PL_subname,tmpbuf,len);
8323 if (SvUTF8(PL_linestr))
8324 SvUTF8_on(PL_subname);
8330 CURMAD('X', nametoke);
8331 CURMAD('_', tmpwhite);
8332 force_ident_maybe_lex('&');
8334 s = SKIPSPACE2(d,tmpwhite);
8340 if (key == KEY_my || key == KEY_our || key==KEY_state)
8343 /* diag_listed_as: Missing name in "%s sub" */
8345 "Missing name in \"%s\"", PL_bufptr);
8347 PL_expect = XTERMBLOCK;
8348 attrful = XATTRTERM;
8349 sv_setpvs(PL_subname,"?");
8353 if (key == KEY_format) {
8355 PL_thistoken = subtoken;
8359 (void) force_word(PL_oldbufptr + tboffset, WORD,
8365 /* Look for a prototype */
8368 bool bad_proto = FALSE;
8369 bool in_brackets = FALSE;
8370 char greedy_proto = ' ';
8371 bool proto_after_greedy_proto = FALSE;
8372 bool must_be_last = FALSE;
8373 bool underscore = FALSE;
8374 bool seen_underscore = FALSE;
8375 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
8378 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
8380 Perl_croak(aTHX_ "Prototype not terminated");
8381 /* strip spaces and check for bad characters */
8382 d = SvPV(PL_lex_stuff, tmplen);
8384 for (p = d; tmplen; tmplen--, ++p) {
8388 if (warnillegalproto) {
8390 proto_after_greedy_proto = TRUE;
8391 if (!strchr("$@%*;[]&\\_+", *p) || *p == '\0') {
8396 if ( !strchr(";@%", *p) )
8403 else if ( *p == ']' ) {
8404 in_brackets = FALSE;
8406 else if ( (*p == '@' || *p == '%') &&
8407 ( tmp < 2 || d[tmp-2] != '\\' ) &&
8409 must_be_last = TRUE;
8412 else if ( *p == '_' ) {
8413 underscore = seen_underscore = TRUE;
8420 if (proto_after_greedy_proto)
8421 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8422 "Prototype after '%c' for %"SVf" : %s",
8423 greedy_proto, SVfARG(PL_subname), d);
8425 SV *dsv = newSVpvs_flags("", SVs_TEMP);
8426 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8427 "Illegal character %sin prototype for %"SVf" : %s",
8428 seen_underscore ? "after '_' " : "",
8430 SvUTF8(PL_lex_stuff)
8431 ? sv_uni_display(dsv,
8432 newSVpvn_flags(d, tmp, SVs_TEMP | SVf_UTF8),
8434 UNI_DISPLAY_ISPRINT)
8435 : pv_pretty(dsv, d, tmp, 60, NULL, NULL,
8436 PERL_PV_ESCAPE_NONASCII));
8438 SvCUR_set(PL_lex_stuff, tmp);
8443 CURMAD('q', PL_thisopen);
8444 CURMAD('_', tmpwhite);
8445 CURMAD('=', PL_thisstuff);
8446 CURMAD('Q', PL_thisclose);
8447 NEXTVAL_NEXTTOKE.opval =
8448 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8449 PL_lex_stuff = NULL;
8452 s = SKIPSPACE2(s,tmpwhite);
8460 if (*s == ':' && s[1] != ':')
8461 PL_expect = attrful;
8462 else if (*s != '{' && key == KEY_sub) {
8464 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8465 else if (*s != ';' && *s != '}')
8466 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8473 curmad('^', newSVpvs(""));
8474 CURMAD('_', tmpwhite);
8478 PL_thistoken = subtoken;
8481 NEXTVAL_NEXTTOKE.opval =
8482 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8483 PL_lex_stuff = NULL;
8489 sv_setpvs(PL_subname, "__ANON__");
8491 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8495 force_ident_maybe_lex('&');
8501 LOP(OP_SYSTEM,XREF);
8504 LOP(OP_SYMLINK,XTERM);
8507 LOP(OP_SYSCALL,XTERM);
8510 LOP(OP_SYSOPEN,XTERM);
8513 LOP(OP_SYSSEEK,XTERM);
8516 LOP(OP_SYSREAD,XTERM);
8519 LOP(OP_SYSWRITE,XTERM);
8524 TERM(sublex_start());
8545 LOP(OP_TRUNCATE,XTERM);
8557 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8559 pl_yylval.ival = CopLINE(PL_curcop);
8563 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8565 pl_yylval.ival = CopLINE(PL_curcop);
8569 LOP(OP_UNLINK,XTERM);
8575 LOP(OP_UNPACK,XTERM);
8578 LOP(OP_UTIME,XTERM);
8584 LOP(OP_UNSHIFT,XTERM);
8587 s = tokenize_use(1, s);
8597 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8599 pl_yylval.ival = CopLINE(PL_curcop);
8603 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8605 pl_yylval.ival = CopLINE(PL_curcop);
8609 PL_hints |= HINT_BLOCK_SCOPE;
8616 LOP(OP_WAITPID,XTERM);
8625 ctl_l[0] = toCTRL('L');
8627 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
8630 /* Make sure $^L is defined */
8631 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
8636 if (PL_expect == XOPERATOR) {
8637 if (*s == '=' && !PL_lex_allbrackets &&
8638 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8646 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8648 pl_yylval.ival = OP_XOR;
8654 #pragma segment Main
8658 S_pending_ident(pTHX)
8662 const char pit = (char)pl_yylval.ival;
8663 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8664 /* All routes through this function want to know if there is a colon. */
8665 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8667 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8668 "### Pending identifier '%s'\n", PL_tokenbuf); });
8670 /* if we're in a my(), we can't allow dynamics here.
8671 $foo'bar has already been turned into $foo::bar, so
8672 just check for colons.
8674 if it's a legal name, the OP is a PADANY.
8677 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8679 yyerror_pv(Perl_form(aTHX_ "No package name allowed for "
8680 "variable %s in \"our\"",
8681 PL_tokenbuf), UTF ? SVf_UTF8 : 0);
8682 tmp = allocmy(PL_tokenbuf, tokenbuf_len, UTF ? SVf_UTF8 : 0);
8686 yyerror_pv(Perl_form(aTHX_ PL_no_myglob,
8687 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf),
8688 UTF ? SVf_UTF8 : 0);
8690 pl_yylval.opval = newOP(OP_PADANY, 0);
8691 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len,
8692 UTF ? SVf_UTF8 : 0);
8698 build the ops for accesses to a my() variable.
8703 tmp = pad_findmy_pvn(PL_tokenbuf, tokenbuf_len,
8704 UTF ? SVf_UTF8 : 0);
8705 if (tmp != NOT_IN_PAD) {
8706 /* might be an "our" variable" */
8707 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8708 /* build ops for a bareword */
8709 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8710 HEK * const stashname = HvNAME_HEK(stash);
8711 SV * const sym = newSVhek(stashname);
8712 sv_catpvs(sym, "::");
8713 sv_catpvn_flags(sym, PL_tokenbuf+1, tokenbuf_len - 1, (UTF ? SV_CATUTF8 : SV_CATBYTES ));
8714 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8715 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8719 ? (GV_ADDMULTI | GV_ADDINEVAL)
8722 ((PL_tokenbuf[0] == '$') ? SVt_PV
8723 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8728 pl_yylval.opval = newOP(OP_PADANY, 0);
8729 pl_yylval.opval->op_targ = tmp;
8735 Whine if they've said @foo in a doublequoted string,
8736 and @foo isn't a variable we can find in the symbol
8739 if (ckWARN(WARN_AMBIGUOUS) &&
8740 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8741 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1,
8742 ( UTF ? SVf_UTF8 : 0 ), SVt_PVAV);
8743 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8744 /* DO NOT warn for @- and @+ */
8745 && !( PL_tokenbuf[2] == '\0' &&
8746 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8749 /* Downgraded from fatal to warning 20000522 mjd */
8750 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8751 "Possible unintended interpolation of %"SVf" in string",
8752 SVfARG(newSVpvn_flags(PL_tokenbuf, tokenbuf_len,
8753 SVs_TEMP | ( UTF ? SVf_UTF8 : 0 ))));
8757 /* build ops for a bareword */
8758 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
8759 newSVpvn_flags(PL_tokenbuf + 1,
8761 UTF ? SVf_UTF8 : 0 ));
8762 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8764 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8765 (PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD)
8766 | ( UTF ? SVf_UTF8 : 0 ),
8767 ((PL_tokenbuf[0] == '$') ? SVt_PV
8768 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8774 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8778 PERL_ARGS_ASSERT_CHECKCOMMA;
8780 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8781 if (ckWARN(WARN_SYNTAX)) {
8784 for (w = s+2; *w && level; w++) {
8792 /* the list of chars below is for end of statements or
8793 * block / parens, boolean operators (&&, ||, //) and branch
8794 * constructs (or, and, if, until, unless, while, err, for).
8795 * Not a very solid hack... */
8796 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8797 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8798 "%s (...) interpreted as function",name);
8801 while (s < PL_bufend && isSPACE(*s))
8805 while (s < PL_bufend && isSPACE(*s))
8807 if (isIDFIRST_lazy_if(s,UTF)) {
8808 const char * const w = s;
8809 s += UTF ? UTF8SKIP(s) : 1;
8810 while (isALNUM_lazy_if(s,UTF))
8811 s += UTF ? UTF8SKIP(s) : 1;
8812 while (s < PL_bufend && isSPACE(*s))
8816 if (keyword(w, s - w, 0))
8819 gv = gv_fetchpvn_flags(w, s - w, ( UTF ? SVf_UTF8 : 0 ), SVt_PVCV);
8820 if (gv && GvCVu(gv))
8822 Perl_croak(aTHX_ "No comma allowed after %s", what);
8827 /* Either returns sv, or mortalizes sv and returns a new SV*.
8828 Best used as sv=new_constant(..., sv, ...).
8829 If s, pv are NULL, calls subroutine with one argument,
8830 and type is used with error messages only. */
8833 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
8834 SV *sv, SV *pv, const char *type, STRLEN typelen)
8837 HV * table = GvHV(PL_hintgv); /* ^H */
8841 const char *why1 = "", *why2 = "", *why3 = "";
8843 PERL_ARGS_ASSERT_NEW_CONSTANT;
8845 /* charnames doesn't work well if there have been errors found */
8846 if (PL_error_count > 0 && strEQ(key,"charnames"))
8847 return &PL_sv_undef;
8850 || ! (PL_hints & HINT_LOCALIZE_HH)
8851 || ! (cvp = hv_fetch(table, key, keylen, FALSE))
8856 /* Here haven't found what we're looking for. If it is charnames,
8857 * perhaps it needs to be loaded. Try doing that before giving up */
8858 if (strEQ(key,"charnames")) {
8859 Perl_load_module(aTHX_
8861 newSVpvs("_charnames"),
8862 /* version parameter; no need to specify it, as if
8863 * we get too early a version, will fail anyway,
8864 * not being able to find '_charnames' */
8870 table = GvHV(PL_hintgv);
8872 && (PL_hints & HINT_LOCALIZE_HH)
8873 && (cvp = hv_fetch(table, key, keylen, FALSE))
8879 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8880 msg = Perl_newSVpvf(aTHX_
8881 "Constant(%s) unknown", (type ? type: "undef"));
8886 why3 = "} is not defined";
8888 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
8889 (type ? type: "undef"), why1, why2, why3);
8891 yyerror(SvPVX_const(msg));
8896 sv_2mortal(sv); /* Parent created it permanently */
8899 pv = newSVpvn_flags(s, len, SVs_TEMP);
8901 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
8903 typesv = &PL_sv_undef;
8905 PUSHSTACKi(PERLSI_OVERLOAD);
8917 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
8921 /* Check the eval first */
8922 if (!PL_in_eval && SvTRUE(ERRSV)) {
8923 sv_catpvs(ERRSV, "Propagated");
8924 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
8926 res = SvREFCNT_inc_simple(sv);
8930 SvREFCNT_inc_simple_void(res);
8939 why1 = "Call to &{$^H{";
8941 why3 = "}} did not return a defined value";
8949 /* Returns a NUL terminated string, with the length of the string written to
8953 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
8957 char * const e = d + destlen - 3; /* two-character token, ending NUL */
8959 PERL_ARGS_ASSERT_SCAN_WORD;
8963 Perl_croak(aTHX_ ident_too_long);
8964 if (isALNUM(*s) || (!UTF && isALNUMC_L1(*s))) /* UTF handled below */
8966 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
8971 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
8975 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
8976 char *t = s + UTF8SKIP(s);
8978 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
8982 Perl_croak(aTHX_ ident_too_long);
8983 Copy(s, d, len, char);
8996 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
8999 char *bracket = NULL;
9002 char * const e = d + destlen - 3; /* two-character token, ending NUL */
9004 PERL_ARGS_ASSERT_SCAN_IDENT;
9009 while (isDIGIT(*s)) {
9011 Perl_croak(aTHX_ ident_too_long);
9018 Perl_croak(aTHX_ ident_too_long);
9019 if (isALNUM(*s)) /* UTF handled below */
9021 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
9026 else if (*s == ':' && s[1] == ':') {
9030 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
9031 char *t = s + UTF8SKIP(s);
9032 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
9034 if (d + (t - s) > e)
9035 Perl_croak(aTHX_ ident_too_long);
9036 Copy(s, d, t - s, char);
9047 if (PL_lex_state != LEX_NORMAL)
9048 PL_lex_state = LEX_INTERPENDMAYBE;
9051 if (*s == '$' && s[1] &&
9052 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
9062 const STRLEN skip = UTF8SKIP(s);
9065 for ( i = 0; i < skip; i++ )
9073 if (*d == '^' && *s && isCONTROLVAR(*s)) {
9077 else if (ck_uni && !bracket)
9080 if (isSPACE(s[-1])) {
9082 const char ch = *s++;
9083 if (!SPACE_OR_TAB(ch)) {
9089 if (isIDFIRST_lazy_if(d,UTF)) {
9093 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
9094 end += UTF8SKIP(end);
9095 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
9096 end += UTF8SKIP(end);
9098 Copy(s, d, end - s, char);
9103 while ((isALNUM(*s) || *s == ':') && d < e)
9106 Perl_croak(aTHX_ ident_too_long);
9109 while (s < send && SPACE_OR_TAB(*s))
9111 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
9112 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
9113 const char * const brack =
9115 ((*s == '[') ? "[...]" : "{...}");
9116 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
9117 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9118 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
9119 funny, dest, brack, funny, dest, brack);
9122 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
9123 PL_lex_allbrackets++;
9127 /* Handle extended ${^Foo} variables
9128 * 1999-02-27 mjd-perl-patch@plover.com */
9129 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
9133 while (isALNUM(*s) && d < e) {
9137 Perl_croak(aTHX_ ident_too_long);
9142 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
9143 PL_lex_state = LEX_INTERPEND;
9146 if (PL_lex_state == LEX_NORMAL) {
9147 if (ckWARN(WARN_AMBIGUOUS) &&
9148 (keyword(dest, d - dest, 0)
9149 || get_cvn_flags(dest, d - dest, UTF ? SVf_UTF8 : 0)))
9151 SV *tmp = newSVpvn_flags( dest, d - dest,
9152 SVs_TEMP | (UTF ? SVf_UTF8 : 0) );
9155 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
9156 "Ambiguous use of %c{%"SVf"} resolved to %c%"SVf,
9157 funny, tmp, funny, tmp);
9162 s = bracket; /* let the parser handle it */
9166 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
9167 PL_lex_state = LEX_INTERPEND;
9172 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
9174 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
9175 * the parse starting at 's', based on the subset that are valid in this
9176 * context input to this routine in 'valid_flags'. Advances s. Returns
9177 * TRUE if the input should be treated as a valid flag, so the next char
9178 * may be as well; otherwise FALSE. 'charset' should point to a NUL upon
9179 * first call on the current regex. This routine will set it to any
9180 * charset modifier found. The caller shouldn't change it. This way,
9181 * another charset modifier encountered in the parse can be detected as an
9182 * error, as we have decided to allow only one */
9185 STRLEN charlen = UTF ? UTF8SKIP(*s) : 1;
9187 if ( charlen != 1 || ! strchr(valid_flags, c) ) {
9188 if (isALNUM_lazy_if(*s, UTF)) {
9189 yyerror_pv(Perl_form(aTHX_ "Unknown regexp modifier \"/%.*s\"", (int)charlen, *s),
9190 UTF ? SVf_UTF8 : 0);
9192 /* Pretend that it worked, so will continue processing before
9201 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
9202 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
9203 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
9204 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
9205 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
9206 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
9207 case LOCALE_PAT_MOD:
9209 goto multiple_charsets;
9211 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
9214 case UNICODE_PAT_MOD:
9216 goto multiple_charsets;
9218 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
9221 case ASCII_RESTRICT_PAT_MOD:
9223 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
9227 /* Error if previous modifier wasn't an 'a', but if it was, see
9228 * if, and accept, a second occurrence (only) */
9230 || get_regex_charset(*pmfl)
9231 != REGEX_ASCII_RESTRICTED_CHARSET)
9233 goto multiple_charsets;
9235 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
9239 case DEPENDS_PAT_MOD:
9241 goto multiple_charsets;
9243 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
9252 if (*charset != c) {
9253 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
9255 else if (c == 'a') {
9256 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
9259 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
9262 /* Pretend that it worked, so will continue processing before dieing */
9268 S_scan_pat(pTHX_ char *start, I32 type)
9272 char *s = scan_str(start,!!PL_madskills,FALSE, PL_reg_state.re_reparsing);
9273 const char * const valid_flags =
9274 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
9275 char charset = '\0'; /* character set modifier */
9280 PERL_ARGS_ASSERT_SCAN_PAT;
9282 /* this was only needed for the initial scan_str; set it to false
9283 * so that any (?{}) code blocks etc are parsed normally */
9284 PL_reg_state.re_reparsing = FALSE;
9286 const char * const delimiter = skipspace(start);
9290 ? "Search pattern not terminated or ternary operator parsed as search pattern"
9291 : "Search pattern not terminated" ));
9294 pm = (PMOP*)newPMOP(type, 0);
9295 if (PL_multi_open == '?') {
9296 /* This is the only point in the code that sets PMf_ONCE: */
9297 pm->op_pmflags |= PMf_ONCE;
9299 /* Hence it's safe to do this bit of PMOP book-keeping here, which
9300 allows us to restrict the list needed by reset to just the ??
9302 assert(type != OP_TRANS);
9304 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
9307 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
9310 elements = mg->mg_len / sizeof(PMOP**);
9311 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
9312 ((PMOP**)mg->mg_ptr) [elements++] = pm;
9313 mg->mg_len = elements * sizeof(PMOP**);
9314 PmopSTASH_set(pm,PL_curstash);
9321 /* if qr/...(?{..}).../, then need to parse the pattern within a new
9322 * anon CV. False positives like qr/[(?{]/ are harmless */
9324 if (type == OP_QR) {
9326 char *e, *p = SvPV(PL_lex_stuff, len);
9328 for (; p < e; p++) {
9329 if (p[0] == '(' && p[1] == '?'
9330 && (p[2] == '{' || (p[2] == '?' && p[3] == '{')))
9332 pm->op_pmflags |= PMf_HAS_CV;
9336 pm->op_pmflags |= PMf_IS_QR;
9339 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
9341 if (PL_madskills && modstart != s) {
9342 SV* tmptoken = newSVpvn(modstart, s - modstart);
9343 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
9346 /* issue a warning if /c is specified,but /g is not */
9347 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
9349 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
9350 "Use of /c modifier is meaningless without /g" );
9353 PL_lex_op = (OP*)pm;
9354 pl_yylval.ival = OP_MATCH;
9359 S_scan_subst(pTHX_ char *start)
9366 char charset = '\0'; /* character set modifier */
9371 PERL_ARGS_ASSERT_SCAN_SUBST;
9373 pl_yylval.ival = OP_NULL;
9375 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
9378 Perl_croak(aTHX_ "Substitution pattern not terminated");
9380 if (s[-1] == PL_multi_open)
9384 CURMAD('q', PL_thisopen);
9385 CURMAD('_', PL_thiswhite);
9386 CURMAD('E', PL_thisstuff);
9387 CURMAD('Q', PL_thisclose);
9388 PL_realtokenstart = s - SvPVX(PL_linestr);
9392 first_start = PL_multi_start;
9393 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
9396 SvREFCNT_dec(PL_lex_stuff);
9397 PL_lex_stuff = NULL;
9399 Perl_croak(aTHX_ "Substitution replacement not terminated");
9401 PL_multi_start = first_start; /* so whole substitution is taken together */
9403 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9407 CURMAD('z', PL_thisopen);
9408 CURMAD('R', PL_thisstuff);
9409 CURMAD('Z', PL_thisclose);
9415 if (*s == EXEC_PAT_MOD) {
9419 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
9428 curmad('m', newSVpvn(modstart, s - modstart));
9429 append_madprops(PL_thismad, (OP*)pm, 0);
9433 if ((pm->op_pmflags & PMf_CONTINUE)) {
9434 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9438 SV * const repl = newSVpvs("");
9441 pm->op_pmflags |= PMf_EVAL;
9444 sv_catpvs(repl, "eval ");
9446 sv_catpvs(repl, "do ");
9448 sv_catpvs(repl, "{");
9449 sv_catsv(repl, PL_sublex_info.repl);
9450 sv_catpvs(repl, "}");
9452 SvREFCNT_dec(PL_sublex_info.repl);
9453 PL_sublex_info.repl = repl;
9456 PL_lex_op = (OP*)pm;
9457 pl_yylval.ival = OP_SUBST;
9462 S_scan_trans(pTHX_ char *start)
9470 bool nondestruct = 0;
9475 PERL_ARGS_ASSERT_SCAN_TRANS;
9477 pl_yylval.ival = OP_NULL;
9479 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
9481 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9483 if (s[-1] == PL_multi_open)
9487 CURMAD('q', PL_thisopen);
9488 CURMAD('_', PL_thiswhite);
9489 CURMAD('E', PL_thisstuff);
9490 CURMAD('Q', PL_thisclose);
9491 PL_realtokenstart = s - SvPVX(PL_linestr);
9495 s = scan_str(s,!!PL_madskills,FALSE,FALSE);
9498 SvREFCNT_dec(PL_lex_stuff);
9499 PL_lex_stuff = NULL;
9501 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9504 CURMAD('z', PL_thisopen);
9505 CURMAD('R', PL_thisstuff);
9506 CURMAD('Z', PL_thisclose);
9509 complement = del = squash = 0;
9516 complement = OPpTRANS_COMPLEMENT;
9519 del = OPpTRANS_DELETE;
9522 squash = OPpTRANS_SQUASH;
9534 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)NULL);
9535 o->op_private &= ~OPpTRANS_ALL;
9536 o->op_private |= del|squash|complement|
9537 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9538 (DO_UTF8(PL_sublex_info.repl) ? OPpTRANS_TO_UTF : 0);
9541 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9546 curmad('m', newSVpvn(modstart, s - modstart));
9547 append_madprops(PL_thismad, o, 0);
9556 Takes a pointer to the first < in <<FOO.
9557 Returns a pointer to the byte following <<FOO.
9559 This function scans a heredoc, which involves different methods
9560 depending on whether we are in a string eval, quoted construct, etc.
9561 This is because PL_linestr could containing a single line of input, or
9562 a whole string being evalled, or the contents of the current quote-
9565 The two basic methods are:
9566 - Steal lines from the input stream
9567 - Scan the heredoc in PL_linestr and remove it therefrom
9569 In a file scope or filtered eval, the first method is used; in a
9570 string eval, the second.
9572 In a quote-like operator, we have to choose between the two,
9573 depending on where we can find a newline. We peek into outer lex-
9574 ing scopes until we find one with a newline in it. If we reach the
9575 outermost lexing scope and it is a file, we use the stream method.
9576 Otherwise it is treated as an eval.
9580 S_scan_heredoc(pTHX_ register char *s)
9583 I32 op_type = OP_SCALAR;
9590 const bool infile = PL_rsfp || PL_parser->filtered;
9591 LEXSHARED *shared = PL_parser->lex_shared;
9593 I32 stuffstart = s - SvPVX(PL_linestr);
9596 PL_realtokenstart = -1;
9599 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9602 d = PL_tokenbuf + 1;
9603 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9604 *PL_tokenbuf = '\n';
9606 while (SPACE_OR_TAB(*peek))
9608 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9611 s = delimcpy(d, e, s, PL_bufend, term, &len);
9613 Perl_croak(aTHX_ "Unterminated delimiter for here document");
9619 /* <<\FOO is equivalent to <<'FOO' */
9623 if (!isALNUM_lazy_if(s,UTF))
9624 deprecate("bare << to mean <<\"\"");
9625 for (; isALNUM_lazy_if(s,UTF); s++) {
9630 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9631 Perl_croak(aTHX_ "Delimiter for here document is too long");
9634 len = d - PL_tokenbuf;
9638 tstart = PL_tokenbuf + 1;
9639 PL_thisclose = newSVpvn(tstart, len - 1);
9640 tstart = SvPVX(PL_linestr) + stuffstart;
9641 PL_thisopen = newSVpvn(tstart, s - tstart);
9642 stuffstart = s - SvPVX(PL_linestr);
9645 #ifndef PERL_STRICT_CR
9646 d = strchr(s, '\r');
9648 char * const olds = s;
9650 while (s < PL_bufend) {
9656 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9665 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9671 tstart = SvPVX(PL_linestr) + stuffstart;
9673 sv_catpvn(PL_thisstuff, tstart, s - tstart);
9675 PL_thisstuff = newSVpvn(tstart, s - tstart);
9678 stuffstart = s - SvPVX(PL_linestr);
9681 tmpstr = newSV_type(SVt_PVIV);
9685 SvIV_set(tmpstr, -1);
9687 else if (term == '`') {
9688 op_type = OP_BACKTICK;
9689 SvIV_set(tmpstr, '\\');
9692 PL_multi_start = CopLINE(PL_curcop) + 1;
9693 PL_multi_open = PL_multi_close = '<';
9694 /* inside a string eval or quote-like operator */
9695 if (!infile || PL_lex_inwhat) {
9698 char * const olds = s;
9699 PERL_CONTEXT * const cx = &cxstack[cxstack_ix];
9700 /* These two fields are not set until an inner lexing scope is
9701 entered. But we need them set here. */
9702 shared->ls_bufptr = s;
9703 shared->ls_linestr = PL_linestr;
9705 /* Look for a newline. If the current buffer does not have one,
9706 peek into the line buffer of the parent lexing scope, going
9707 up as many levels as necessary to find one with a newline
9710 while (!(s = (char *)memchr(
9711 (void *)shared->ls_bufptr, '\n',
9712 SvEND(shared->ls_linestr)-shared->ls_bufptr
9714 shared = shared->ls_prev;
9715 /* shared is only null if we have gone beyond the outermost
9716 lexing scope. In a file, we will have broken out of the
9717 loop in the previous iteration. In an eval, the string buf-
9718 fer ends with "\n;", so the while condition below will have
9719 evaluated to false. So shared can never be null. */
9721 /* A LEXSHARED struct with a null ls_prev pointer is the outer-
9722 most lexing scope. In a file, shared->ls_linestr at that
9723 level is just one line, so there is no body to steal. */
9724 if (infile && !shared->ls_prev) {
9730 s = (char*)memchr((void*)s, '\n', PL_bufend - s);
9733 linestr = shared->ls_linestr;
9734 bufend = SvEND(linestr);
9736 while (s < bufend &&
9737 (*s != '\n' || memNE(s,PL_tokenbuf,len)) ) {
9739 ++shared->herelines;
9744 sv_setpvn(tmpstr,d+1,s-d);
9748 sv_catpvn(PL_thisstuff, d + 1, s - d);
9750 PL_thisstuff = newSVpvn(d + 1, s - d);
9751 stuffstart = s - SvPVX(PL_linestr);
9755 /* the preceding stmt passes a newline */
9756 shared->herelines++;
9758 /* s now points to the newline after the heredoc terminator.
9759 d points to the newline before the body of the heredoc.
9762 /* We are going to modify linestr in place here, so set
9763 aside copies of the string if necessary for re-evals or
9765 /* See the Paranoia note in case LEX_INTERPEND in yylex, for why we
9766 check shared->re_eval_str. */
9767 if (shared->re_eval_start || shared->re_eval_str) {
9768 /* Set aside the rest of the regexp */
9769 if (!shared->re_eval_str)
9770 shared->re_eval_str =
9771 newSVpvn(shared->re_eval_start,
9772 bufend - shared->re_eval_start);
9773 shared->re_eval_start -= s-d;
9775 if (CxTYPE(cx) == CXt_EVAL && CxOLD_OP_TYPE(cx) == OP_ENTEREVAL
9776 && cx->blk_eval.cur_text == linestr) {
9777 cx->blk_eval.cur_text = newSVsv(linestr);
9778 SvSCREAM_on(cx->blk_eval.cur_text);
9780 /* Copy everything from s onwards back to d. */
9781 Move(s,d,bufend-s + 1,char);
9782 SvCUR_set(linestr, SvCUR(linestr) - (s-d));
9783 /* Setting PL_bufend only applies when we have not dug deeper
9784 into other scopes, because sublex_done sets PL_bufend to
9785 SvEND(PL_linestr). */
9786 if (shared == PL_parser->lex_shared) PL_bufend = SvEND(linestr);
9793 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9794 term = PL_tokenbuf[1];
9796 linestr_save = PL_linestr; /* must restore this afterwards */
9797 d = s; /* and this */
9798 PL_linestr = newSVpvs("");
9799 PL_bufend = SvPVX(PL_linestr);
9803 tstart = SvPVX(PL_linestr) + stuffstart;
9805 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
9807 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
9810 PL_bufptr = PL_bufend;
9811 CopLINE_set(PL_curcop,
9812 PL_multi_start + shared->herelines);
9813 if (!lex_next_chunk(LEX_NO_TERM)
9814 && (!SvCUR(tmpstr) || SvEND(tmpstr)[-1] != '\n')) {
9815 SvREFCNT_dec(linestr_save);
9818 CopLINE_set(PL_curcop, (line_t)PL_multi_start - 1);
9819 if (!SvCUR(PL_linestr) || PL_bufend[-1] != '\n') {
9820 lex_grow_linestr(SvCUR(PL_linestr) + 2);
9821 sv_catpvs(PL_linestr, "\n\0");
9825 stuffstart = s - SvPVX(PL_linestr);
9827 shared->herelines++;
9828 PL_last_lop = PL_last_uni = NULL;
9829 #ifndef PERL_STRICT_CR
9830 if (PL_bufend - PL_linestart >= 2) {
9831 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
9832 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9834 PL_bufend[-2] = '\n';
9836 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9838 else if (PL_bufend[-1] == '\r')
9839 PL_bufend[-1] = '\n';
9841 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9842 PL_bufend[-1] = '\n';
9844 if (*s == term && memEQ(s,PL_tokenbuf + 1,len)) {
9845 SvREFCNT_dec(PL_linestr);
9846 PL_linestr = linestr_save;
9847 PL_linestart = SvPVX(linestr_save);
9848 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9853 sv_catsv(tmpstr,PL_linestr);
9857 PL_multi_end = CopLINE(PL_curcop);
9858 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9859 SvPV_shrink_to_cur(tmpstr);
9862 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9864 else if (PL_encoding)
9865 sv_recode_to_utf8(tmpstr, PL_encoding);
9867 PL_lex_stuff = tmpstr;
9868 pl_yylval.ival = op_type;
9872 SvREFCNT_dec(tmpstr);
9873 CopLINE_set(PL_curcop, (line_t)PL_multi_start - 1);
9874 missingterm(PL_tokenbuf + 1);
9878 takes: current position in input buffer
9879 returns: new position in input buffer
9880 side-effects: pl_yylval and lex_op are set.
9885 <FH> read from filehandle
9886 <pkg::FH> read from package qualified filehandle
9887 <pkg'FH> read from package qualified filehandle
9888 <$fh> read from filehandle in $fh
9894 S_scan_inputsymbol(pTHX_ char *start)
9897 char *s = start; /* current position in buffer */
9900 char *d = PL_tokenbuf; /* start of temp holding space */
9901 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9903 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
9905 end = strchr(s, '\n');
9908 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9910 /* die if we didn't have space for the contents of the <>,
9911 or if it didn't end, or if we see a newline
9914 if (len >= (I32)sizeof PL_tokenbuf)
9915 Perl_croak(aTHX_ "Excessively long <> operator");
9917 Perl_croak(aTHX_ "Unterminated <> operator");
9922 Remember, only scalar variables are interpreted as filehandles by
9923 this code. Anything more complex (e.g., <$fh{$num}>) will be
9924 treated as a glob() call.
9925 This code makes use of the fact that except for the $ at the front,
9926 a scalar variable and a filehandle look the same.
9928 if (*d == '$' && d[1]) d++;
9930 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9931 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9932 d += UTF ? UTF8SKIP(d) : 1;
9934 /* If we've tried to read what we allow filehandles to look like, and
9935 there's still text left, then it must be a glob() and not a getline.
9936 Use scan_str to pull out the stuff between the <> and treat it
9937 as nothing more than a string.
9940 if (d - PL_tokenbuf != len) {
9941 pl_yylval.ival = OP_GLOB;
9942 s = scan_str(start,!!PL_madskills,FALSE,FALSE);
9944 Perl_croak(aTHX_ "Glob not terminated");
9948 bool readline_overriden = FALSE;
9951 /* we're in a filehandle read situation */
9954 /* turn <> into <ARGV> */
9956 Copy("ARGV",d,5,char);
9958 /* Check whether readline() is overriden */
9959 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
9961 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
9963 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
9964 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
9965 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
9966 readline_overriden = TRUE;
9968 /* if <$fh>, create the ops to turn the variable into a
9972 /* try to find it in the pad for this block, otherwise find
9973 add symbol table ops
9975 const PADOFFSET tmp = pad_findmy_pvn(d, len, UTF ? SVf_UTF8 : 0);
9976 if (tmp != NOT_IN_PAD) {
9977 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
9978 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
9979 HEK * const stashname = HvNAME_HEK(stash);
9980 SV * const sym = sv_2mortal(newSVhek(stashname));
9981 sv_catpvs(sym, "::");
9987 OP * const o = newOP(OP_PADSV, 0);
9989 PL_lex_op = readline_overriden
9990 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9991 op_append_elem(OP_LIST, o,
9992 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
9993 : (OP*)newUNOP(OP_READLINE, 0, o);
10002 ? (GV_ADDMULTI | GV_ADDINEVAL)
10003 : GV_ADDMULTI) | ( UTF ? SVf_UTF8 : 0 ),
10005 PL_lex_op = readline_overriden
10006 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10007 op_append_elem(OP_LIST,
10008 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
10009 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
10010 : (OP*)newUNOP(OP_READLINE, 0,
10011 newUNOP(OP_RV2SV, 0,
10012 newGVOP(OP_GV, 0, gv)));
10014 if (!readline_overriden)
10015 PL_lex_op->op_flags |= OPf_SPECIAL;
10016 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
10017 pl_yylval.ival = OP_NULL;
10020 /* If it's none of the above, it must be a literal filehandle
10021 (<Foo::BAR> or <FOO>) so build a simple readline OP */
10023 GV * const gv = gv_fetchpv(d, GV_ADD | ( UTF ? SVf_UTF8 : 0 ), SVt_PVIO);
10024 PL_lex_op = readline_overriden
10025 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
10026 op_append_elem(OP_LIST,
10027 newGVOP(OP_GV, 0, gv),
10028 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
10029 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
10030 pl_yylval.ival = OP_NULL;
10039 takes: start position in buffer
10040 keep_quoted preserve \ on the embedded delimiter(s)
10041 keep_delims preserve the delimiters around the string
10042 re_reparse compiling a run-time /(?{})/:
10043 collapse // to /, and skip encoding src
10044 returns: position to continue reading from buffer
10045 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
10046 updates the read buffer.
10048 This subroutine pulls a string out of the input. It is called for:
10049 q single quotes q(literal text)
10050 ' single quotes 'literal text'
10051 qq double quotes qq(interpolate $here please)
10052 " double quotes "interpolate $here please"
10053 qx backticks qx(/bin/ls -l)
10054 ` backticks `/bin/ls -l`
10055 qw quote words @EXPORT_OK = qw( func() $spam )
10056 m// regexp match m/this/
10057 s/// regexp substitute s/this/that/
10058 tr/// string transliterate tr/this/that/
10059 y/// string transliterate y/this/that/
10060 ($*@) sub prototypes sub foo ($)
10061 (stuff) sub attr parameters sub foo : attr(stuff)
10062 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
10064 In most of these cases (all but <>, patterns and transliterate)
10065 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
10066 calls scan_str(). s/// makes yylex() call scan_subst() which calls
10067 scan_str(). tr/// and y/// make yylex() call scan_trans() which
10070 It skips whitespace before the string starts, and treats the first
10071 character as the delimiter. If the delimiter is one of ([{< then
10072 the corresponding "close" character )]}> is used as the closing
10073 delimiter. It allows quoting of delimiters, and if the string has
10074 balanced delimiters ([{<>}]) it allows nesting.
10076 On success, the SV with the resulting string is put into lex_stuff or,
10077 if that is already non-NULL, into lex_repl. The second case occurs only
10078 when parsing the RHS of the special constructs s/// and tr/// (y///).
10079 For convenience, the terminating delimiter character is stuffed into
10084 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims, int re_reparse)
10087 SV *sv; /* scalar value: string */
10088 const char *tmps; /* temp string, used for delimiter matching */
10089 char *s = start; /* current position in the buffer */
10090 char term; /* terminating character */
10091 char *to; /* current position in the sv's data */
10092 I32 brackets = 1; /* bracket nesting level */
10093 bool has_utf8 = FALSE; /* is there any utf8 content? */
10094 I32 termcode; /* terminating char. code */
10095 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
10096 STRLEN termlen; /* length of terminating string */
10097 int last_off = 0; /* last position for nesting bracket */
10103 PERL_ARGS_ASSERT_SCAN_STR;
10105 /* skip space before the delimiter */
10111 if (PL_realtokenstart >= 0) {
10112 stuffstart = PL_realtokenstart;
10113 PL_realtokenstart = -1;
10116 stuffstart = start - SvPVX(PL_linestr);
10118 /* mark where we are, in case we need to report errors */
10121 /* after skipping whitespace, the next character is the terminator */
10124 termcode = termstr[0] = term;
10128 termcode = utf8_to_uvchr_buf((U8*)s, (U8*)PL_bufend, &termlen);
10129 Copy(s, termstr, termlen, U8);
10130 if (!UTF8_IS_INVARIANT(term))
10134 /* mark where we are */
10135 PL_multi_start = CopLINE(PL_curcop);
10136 PL_multi_open = term;
10138 /* find corresponding closing delimiter */
10139 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
10140 termcode = termstr[0] = term = tmps[5];
10142 PL_multi_close = term;
10144 /* create a new SV to hold the contents. 79 is the SV's initial length.
10145 What a random number. */
10146 sv = newSV_type(SVt_PVIV);
10148 SvIV_set(sv, termcode);
10149 (void)SvPOK_only(sv); /* validate pointer */
10151 /* move past delimiter and try to read a complete string */
10153 sv_catpvn(sv, s, termlen);
10156 tstart = SvPVX(PL_linestr) + stuffstart;
10157 if (!PL_thisopen && !keep_delims) {
10158 PL_thisopen = newSVpvn(tstart, s - tstart);
10159 stuffstart = s - SvPVX(PL_linestr);
10163 if (PL_encoding && !UTF && !re_reparse) {
10167 int offset = s - SvPVX_const(PL_linestr);
10168 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
10169 &offset, (char*)termstr, termlen);
10170 const char * const ns = SvPVX_const(PL_linestr) + offset;
10171 char * const svlast = SvEND(sv) - 1;
10173 for (; s < ns; s++) {
10174 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10175 COPLINE_INC_WITH_HERELINES;
10178 goto read_more_line;
10180 /* handle quoted delimiters */
10181 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
10183 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
10185 if ((svlast-1 - t) % 2) {
10186 if (!keep_quoted) {
10187 *(svlast-1) = term;
10189 SvCUR_set(sv, SvCUR(sv) - 1);
10194 if (PL_multi_open == PL_multi_close) {
10200 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
10201 /* At here, all closes are "was quoted" one,
10202 so we don't check PL_multi_close. */
10204 if (!keep_quoted && *(t+1) == PL_multi_open)
10209 else if (*t == PL_multi_open)
10217 SvCUR_set(sv, w - SvPVX_const(sv));
10219 last_off = w - SvPVX(sv);
10220 if (--brackets <= 0)
10225 if (!keep_delims) {
10226 SvCUR_set(sv, SvCUR(sv) - 1);
10232 /* extend sv if need be */
10233 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
10234 /* set 'to' to the next character in the sv's string */
10235 to = SvPVX(sv)+SvCUR(sv);
10237 /* if open delimiter is the close delimiter read unbridle */
10238 if (PL_multi_open == PL_multi_close) {
10239 for (; s < PL_bufend; s++,to++) {
10240 /* embedded newlines increment the current line number */
10241 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10242 COPLINE_INC_WITH_HERELINES;
10243 /* handle quoted delimiters */
10244 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
10247 || (re_reparse && s[1] == '\\'))
10250 /* any other quotes are simply copied straight through */
10254 /* terminate when run out of buffer (the for() condition), or
10255 have found the terminator */
10256 else if (*s == term) {
10259 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
10262 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10268 /* if the terminator isn't the same as the start character (e.g.,
10269 matched brackets), we have to allow more in the quoting, and
10270 be prepared for nested brackets.
10273 /* read until we run out of string, or we find the terminator */
10274 for (; s < PL_bufend; s++,to++) {
10275 /* embedded newlines increment the line count */
10276 if (*s == '\n' && !PL_rsfp && !PL_parser->filtered)
10277 COPLINE_INC_WITH_HERELINES;
10278 /* backslashes can escape the open or closing characters */
10279 if (*s == '\\' && s+1 < PL_bufend) {
10280 if (!keep_quoted &&
10281 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
10286 /* allow nested opens and closes */
10287 else if (*s == PL_multi_close && --brackets <= 0)
10289 else if (*s == PL_multi_open)
10291 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
10296 /* terminate the copied string and update the sv's end-of-string */
10298 SvCUR_set(sv, to - SvPVX_const(sv));
10301 * this next chunk reads more into the buffer if we're not done yet
10305 break; /* handle case where we are done yet :-) */
10307 #ifndef PERL_STRICT_CR
10308 if (to - SvPVX_const(sv) >= 2) {
10309 if ((to[-2] == '\r' && to[-1] == '\n') ||
10310 (to[-2] == '\n' && to[-1] == '\r'))
10314 SvCUR_set(sv, to - SvPVX_const(sv));
10316 else if (to[-1] == '\r')
10319 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
10324 /* if we're out of file, or a read fails, bail and reset the current
10325 line marker so we can report where the unterminated string began
10328 if (PL_madskills) {
10329 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10331 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
10333 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
10336 COPLINE_INC_WITH_HERELINES;
10337 PL_bufptr = PL_bufend;
10338 if (!lex_next_chunk(0)) {
10340 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
10349 /* at this point, we have successfully read the delimited string */
10351 if (!PL_encoding || UTF || re_reparse) {
10353 if (PL_madskills) {
10354 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10355 const int len = s - tstart;
10357 sv_catpvn(PL_thisstuff, tstart, len);
10359 PL_thisstuff = newSVpvn(tstart, len);
10360 if (!PL_thisclose && !keep_delims)
10361 PL_thisclose = newSVpvn(s,termlen);
10366 sv_catpvn(sv, s, termlen);
10371 if (PL_madskills) {
10372 char * const tstart = SvPVX(PL_linestr) + stuffstart;
10373 const int len = s - tstart - termlen;
10375 sv_catpvn(PL_thisstuff, tstart, len);
10377 PL_thisstuff = newSVpvn(tstart, len);
10378 if (!PL_thisclose && !keep_delims)
10379 PL_thisclose = newSVpvn(s - termlen,termlen);
10383 if (has_utf8 || (PL_encoding && !re_reparse))
10386 PL_multi_end = CopLINE(PL_curcop);
10388 /* if we allocated too much space, give some back */
10389 if (SvCUR(sv) + 5 < SvLEN(sv)) {
10390 SvLEN_set(sv, SvCUR(sv) + 1);
10391 SvPV_renew(sv, SvLEN(sv));
10394 /* decide whether this is the first or second quoted string we've read
10399 PL_sublex_info.repl = sv;
10407 takes: pointer to position in buffer
10408 returns: pointer to new position in buffer
10409 side-effects: builds ops for the constant in pl_yylval.op
10411 Read a number in any of the formats that Perl accepts:
10413 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
10414 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
10417 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
10419 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
10422 If it reads a number without a decimal point or an exponent, it will
10423 try converting the number to an integer and see if it can do so
10424 without loss of precision.
10428 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10431 const char *s = start; /* current position in buffer */
10432 char *d; /* destination in temp buffer */
10433 char *e; /* end of temp buffer */
10434 NV nv; /* number read, as a double */
10435 SV *sv = NULL; /* place to put the converted number */
10436 bool floatit; /* boolean: int or float? */
10437 const char *lastub = NULL; /* position of last underbar */
10438 static char const number_too_long[] = "Number too long";
10440 PERL_ARGS_ASSERT_SCAN_NUM;
10442 /* We use the first character to decide what type of number this is */
10446 Perl_croak(aTHX_ "panic: scan_num, *s=%d", *s);
10448 /* if it starts with a 0, it could be an octal number, a decimal in
10449 0.13 disguise, or a hexadecimal number, or a binary number. */
10453 u holds the "number so far"
10454 shift the power of 2 of the base
10455 (hex == 4, octal == 3, binary == 1)
10456 overflowed was the number more than we can hold?
10458 Shift is used when we add a digit. It also serves as an "are
10459 we in octal/hex/binary?" indicator to disallow hex characters
10460 when in octal mode.
10465 bool overflowed = FALSE;
10466 bool just_zero = TRUE; /* just plain 0 or binary number? */
10467 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10468 static const char* const bases[5] =
10469 { "", "binary", "", "octal", "hexadecimal" };
10470 static const char* const Bases[5] =
10471 { "", "Binary", "", "Octal", "Hexadecimal" };
10472 static const char* const maxima[5] =
10474 "0b11111111111111111111111111111111",
10478 const char *base, *Base, *max;
10480 /* check for hex */
10481 if (s[1] == 'x' || s[1] == 'X') {
10485 } else if (s[1] == 'b' || s[1] == 'B') {
10490 /* check for a decimal in disguise */
10491 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
10493 /* so it must be octal */
10500 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10501 "Misplaced _ in number");
10505 base = bases[shift];
10506 Base = Bases[shift];
10507 max = maxima[shift];
10509 /* read the rest of the number */
10511 /* x is used in the overflow test,
10512 b is the digit we're adding on. */
10517 /* if we don't mention it, we're done */
10521 /* _ are ignored -- but warned about if consecutive */
10523 if (lastub && s == lastub + 1)
10524 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10525 "Misplaced _ in number");
10529 /* 8 and 9 are not octal */
10530 case '8': case '9':
10532 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10536 case '2': case '3': case '4':
10537 case '5': case '6': case '7':
10539 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10542 case '0': case '1':
10543 b = *s++ & 15; /* ASCII digit -> value of digit */
10547 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10548 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10549 /* make sure they said 0x */
10552 b = (*s++ & 7) + 9;
10554 /* Prepare to put the digit we have onto the end
10555 of the number so far. We check for overflows.
10561 x = u << shift; /* make room for the digit */
10563 if ((x >> shift) != u
10564 && !(PL_hints & HINT_NEW_BINARY)) {
10567 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10568 "Integer overflow in %s number",
10571 u = x | b; /* add the digit to the end */
10574 n *= nvshift[shift];
10575 /* If an NV has not enough bits in its
10576 * mantissa to represent an UV this summing of
10577 * small low-order numbers is a waste of time
10578 * (because the NV cannot preserve the
10579 * low-order bits anyway): we could just
10580 * remember when did we overflow and in the
10581 * end just multiply n by the right
10589 /* if we get here, we had success: make a scalar value from
10594 /* final misplaced underbar check */
10595 if (s[-1] == '_') {
10596 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10600 if (n > 4294967295.0)
10601 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10602 "%s number > %s non-portable",
10608 if (u > 0xffffffff)
10609 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10610 "%s number > %s non-portable",
10615 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10616 sv = new_constant(start, s - start, "integer",
10617 sv, NULL, NULL, 0);
10618 else if (PL_hints & HINT_NEW_BINARY)
10619 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10624 handle decimal numbers.
10625 we're also sent here when we read a 0 as the first digit
10627 case '1': case '2': case '3': case '4': case '5':
10628 case '6': case '7': case '8': case '9': case '.':
10631 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10634 /* read next group of digits and _ and copy into d */
10635 while (isDIGIT(*s) || *s == '_') {
10636 /* skip underscores, checking for misplaced ones
10640 if (lastub && s == lastub + 1)
10641 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10642 "Misplaced _ in number");
10646 /* check for end of fixed-length buffer */
10648 Perl_croak(aTHX_ number_too_long);
10649 /* if we're ok, copy the character */
10654 /* final misplaced underbar check */
10655 if (lastub && s == lastub + 1) {
10656 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10659 /* read a decimal portion if there is one. avoid
10660 3..5 being interpreted as the number 3. followed
10663 if (*s == '.' && s[1] != '.') {
10668 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10669 "Misplaced _ in number");
10673 /* copy, ignoring underbars, until we run out of digits.
10675 for (; isDIGIT(*s) || *s == '_'; s++) {
10676 /* fixed length buffer check */
10678 Perl_croak(aTHX_ number_too_long);
10680 if (lastub && s == lastub + 1)
10681 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10682 "Misplaced _ in number");
10688 /* fractional part ending in underbar? */
10689 if (s[-1] == '_') {
10690 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10691 "Misplaced _ in number");
10693 if (*s == '.' && isDIGIT(s[1])) {
10694 /* oops, it's really a v-string, but without the "v" */
10700 /* read exponent part, if present */
10701 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10705 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10706 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10708 /* stray preinitial _ */
10710 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10711 "Misplaced _ in number");
10715 /* allow positive or negative exponent */
10716 if (*s == '+' || *s == '-')
10719 /* stray initial _ */
10721 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10722 "Misplaced _ in number");
10726 /* read digits of exponent */
10727 while (isDIGIT(*s) || *s == '_') {
10730 Perl_croak(aTHX_ number_too_long);
10734 if (((lastub && s == lastub + 1) ||
10735 (!isDIGIT(s[1]) && s[1] != '_')))
10736 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10737 "Misplaced _ in number");
10745 We try to do an integer conversion first if no characters
10746 indicating "float" have been found.
10751 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10753 if (flags == IS_NUMBER_IN_UV) {
10755 sv = newSViv(uv); /* Prefer IVs over UVs. */
10758 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10759 if (uv <= (UV) IV_MIN)
10760 sv = newSViv(-(IV)uv);
10767 /* terminate the string */
10769 nv = Atof(PL_tokenbuf);
10774 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10775 const char *const key = floatit ? "float" : "integer";
10776 const STRLEN keylen = floatit ? 5 : 7;
10777 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10778 key, keylen, sv, NULL, NULL, 0);
10782 /* if it starts with a v, it could be a v-string */
10785 sv = newSV(5); /* preallocate storage space */
10786 s = scan_vstring(s, PL_bufend, sv);
10790 /* make the op for the constant and return */
10793 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10795 lvalp->opval = NULL;
10801 S_scan_formline(pTHX_ register char *s)
10806 SV * const stuff = newSVpvs("");
10807 bool needargs = FALSE;
10808 bool eofmt = FALSE;
10810 char *tokenstart = s;
10811 SV* savewhite = NULL;
10813 if (PL_madskills) {
10814 savewhite = PL_thiswhite;
10819 PERL_ARGS_ASSERT_SCAN_FORMLINE;
10821 while (!needargs) {
10824 #ifdef PERL_STRICT_CR
10825 while (SPACE_OR_TAB(*t))
10828 while (SPACE_OR_TAB(*t) || *t == '\r')
10831 if (*t == '\n' || t == PL_bufend) {
10836 eol = (char *) memchr(s,'\n',PL_bufend-s);
10840 for (t = s; t < eol; t++) {
10841 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10843 goto enough; /* ~~ must be first line in formline */
10845 if (*t == '@' || *t == '^')
10849 sv_catpvn(stuff, s, eol-s);
10850 #ifndef PERL_STRICT_CR
10851 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10852 char *end = SvPVX(stuff) + SvCUR(stuff);
10855 SvCUR_set(stuff, SvCUR(stuff) - 1);
10863 if ((PL_rsfp || PL_parser->filtered)
10864 && PL_parser->form_lex_state == LEX_NORMAL) {
10867 if (PL_madskills) {
10869 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
10871 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
10874 PL_bufptr = PL_bufend;
10875 COPLINE_INC_WITH_HERELINES;
10876 got_some = lex_next_chunk(0);
10877 CopLINE_dec(PL_curcop);
10880 tokenstart = PL_bufptr;
10888 if (!SvCUR(stuff) || needargs)
10889 PL_lex_state = PL_parser->form_lex_state;
10890 if (SvCUR(stuff)) {
10891 PL_expect = XSTATE;
10893 start_force(PL_curforce);
10894 NEXTVAL_NEXTTOKE.ival = 0;
10895 force_next(FORMLBRACK);
10898 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10900 else if (PL_encoding)
10901 sv_recode_to_utf8(stuff, PL_encoding);
10903 start_force(PL_curforce);
10904 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10908 SvREFCNT_dec(stuff);
10910 PL_lex_formbrack = 0;
10913 if (PL_madskills) {
10915 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
10917 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
10918 PL_thiswhite = savewhite;
10925 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10928 const I32 oldsavestack_ix = PL_savestack_ix;
10929 CV* const outsidecv = PL_compcv;
10931 SAVEI32(PL_subline);
10932 save_item(PL_subname);
10933 SAVESPTR(PL_compcv);
10935 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
10936 CvFLAGS(PL_compcv) |= flags;
10938 PL_subline = CopLINE(PL_curcop);
10939 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10940 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
10941 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10942 if (outsidecv && CvPADLIST(outsidecv))
10943 CvPADLIST(PL_compcv)->xpadl_outid = CvPADLIST(outsidecv)->xpadl_id;
10945 return oldsavestack_ix;
10949 #pragma segment Perl_yylex
10952 S_yywarn(pTHX_ const char *const s, U32 flags)
10956 PERL_ARGS_ASSERT_YYWARN;
10958 PL_in_eval |= EVAL_WARNONLY;
10959 yyerror_pv(s, flags);
10960 PL_in_eval &= ~EVAL_WARNONLY;
10965 Perl_yyerror(pTHX_ const char *const s)
10967 PERL_ARGS_ASSERT_YYERROR;
10968 return yyerror_pvn(s, strlen(s), 0);
10972 Perl_yyerror_pv(pTHX_ const char *const s, U32 flags)
10974 PERL_ARGS_ASSERT_YYERROR_PV;
10975 return yyerror_pvn(s, strlen(s), flags);
10979 Perl_yyerror_pvn(pTHX_ const char *const s, STRLEN len, U32 flags)
10982 const char *context = NULL;
10985 SV * const where_sv = newSVpvs_flags("", SVs_TEMP);
10986 int yychar = PL_parser->yychar;
10987 U32 is_utf8 = flags & SVf_UTF8;
10989 PERL_ARGS_ASSERT_YYERROR_PVN;
10991 if (!yychar || (yychar == ';' && !PL_rsfp))
10992 sv_catpvs(where_sv, "at EOF");
10993 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
10994 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
10995 PL_oldbufptr != PL_bufptr) {
10998 The code below is removed for NetWare because it abends/crashes on NetWare
10999 when the script has error such as not having the closing quotes like:
11000 if ($var eq "value)
11001 Checking of white spaces is anyway done in NetWare code.
11004 while (isSPACE(*PL_oldoldbufptr))
11007 context = PL_oldoldbufptr;
11008 contlen = PL_bufptr - PL_oldoldbufptr;
11010 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
11011 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
11014 The code below is removed for NetWare because it abends/crashes on NetWare
11015 when the script has error such as not having the closing quotes like:
11016 if ($var eq "value)
11017 Checking of white spaces is anyway done in NetWare code.
11020 while (isSPACE(*PL_oldbufptr))
11023 context = PL_oldbufptr;
11024 contlen = PL_bufptr - PL_oldbufptr;
11026 else if (yychar > 255)
11027 sv_catpvs(where_sv, "next token ???");
11028 else if (yychar == -2) { /* YYEMPTY */
11029 if (PL_lex_state == LEX_NORMAL ||
11030 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
11031 sv_catpvs(where_sv, "at end of line");
11032 else if (PL_lex_inpat)
11033 sv_catpvs(where_sv, "within pattern");
11035 sv_catpvs(where_sv, "within string");
11038 sv_catpvs(where_sv, "next char ");
11040 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
11041 else if (isPRINT_LC(yychar)) {
11042 const char string = yychar;
11043 sv_catpvn(where_sv, &string, 1);
11046 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
11048 msg = sv_2mortal(newSVpvn_flags(s, len, is_utf8));
11049 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
11050 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
11052 Perl_sv_catpvf(aTHX_ msg, "near \"%"SVf"\"\n",
11053 SVfARG(newSVpvn_flags(context, contlen,
11054 SVs_TEMP | (UTF ? SVf_UTF8 : 0))));
11056 Perl_sv_catpvf(aTHX_ msg, "%"SVf"\n", SVfARG(where_sv));
11057 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
11058 Perl_sv_catpvf(aTHX_ msg,
11059 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
11060 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
11063 if (PL_in_eval & EVAL_WARNONLY) {
11064 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
11068 if (PL_error_count >= 10) {
11069 if (PL_in_eval && SvCUR(ERRSV))
11070 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
11071 SVfARG(ERRSV), OutCopFILE(PL_curcop));
11073 Perl_croak(aTHX_ "%s has too many errors.\n",
11074 OutCopFILE(PL_curcop));
11077 PL_in_my_stash = NULL;
11081 #pragma segment Main
11085 S_swallow_bom(pTHX_ U8 *s)
11088 const STRLEN slen = SvCUR(PL_linestr);
11090 PERL_ARGS_ASSERT_SWALLOW_BOM;
11094 if (s[1] == 0xFE) {
11095 /* UTF-16 little-endian? (or UTF-32LE?) */
11096 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
11097 /* diag_listed_as: Unsupported script encoding %s */
11098 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
11099 #ifndef PERL_NO_UTF16_FILTER
11100 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
11102 if (PL_bufend > (char*)s) {
11103 s = add_utf16_textfilter(s, TRUE);
11106 /* diag_listed_as: Unsupported script encoding %s */
11107 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11112 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
11113 #ifndef PERL_NO_UTF16_FILTER
11114 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
11116 if (PL_bufend > (char *)s) {
11117 s = add_utf16_textfilter(s, FALSE);
11120 /* diag_listed_as: Unsupported script encoding %s */
11121 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11126 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
11127 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
11128 s += 3; /* UTF-8 */
11134 if (s[2] == 0xFE && s[3] == 0xFF) {
11135 /* UTF-32 big-endian */
11136 /* diag_listed_as: Unsupported script encoding %s */
11137 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
11140 else if (s[2] == 0 && s[3] != 0) {
11143 * are a good indicator of UTF-16BE. */
11144 #ifndef PERL_NO_UTF16_FILTER
11145 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
11146 s = add_utf16_textfilter(s, FALSE);
11148 /* diag_listed_as: Unsupported script encoding %s */
11149 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
11155 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
11156 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
11157 s += 4; /* UTF-8 */
11163 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
11166 * are a good indicator of UTF-16LE. */
11167 #ifndef PERL_NO_UTF16_FILTER
11168 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
11169 s = add_utf16_textfilter(s, TRUE);
11171 /* diag_listed_as: Unsupported script encoding %s */
11172 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
11180 #ifndef PERL_NO_UTF16_FILTER
11182 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
11185 SV *const filter = FILTER_DATA(idx);
11186 /* We re-use this each time round, throwing the contents away before we
11188 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
11189 SV *const utf8_buffer = filter;
11190 IV status = IoPAGE(filter);
11191 const bool reverse = cBOOL(IoLINES(filter));
11194 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
11196 /* As we're automatically added, at the lowest level, and hence only called
11197 from this file, we can be sure that we're not called in block mode. Hence
11198 don't bother writing code to deal with block mode. */
11200 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
11203 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
11205 DEBUG_P(PerlIO_printf(Perl_debug_log,
11206 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11207 FPTR2DPTR(void *, S_utf16_textfilter),
11208 reverse ? 'l' : 'b', idx, maxlen, status,
11209 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11216 /* First, look in our buffer of existing UTF-8 data: */
11217 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
11221 } else if (status == 0) {
11223 IoPAGE(filter) = 0;
11224 nl = SvEND(utf8_buffer);
11227 STRLEN got = nl - SvPVX(utf8_buffer);
11228 /* Did we have anything to append? */
11230 sv_catpvn(sv, SvPVX(utf8_buffer), got);
11231 /* Everything else in this code works just fine if SVp_POK isn't
11232 set. This, however, needs it, and we need it to work, else
11233 we loop infinitely because the buffer is never consumed. */
11234 sv_chop(utf8_buffer, nl);
11238 /* OK, not a complete line there, so need to read some more UTF-16.
11239 Read an extra octect if the buffer currently has an odd number. */
11243 if (SvCUR(utf16_buffer) >= 2) {
11244 /* Location of the high octet of the last complete code point.
11245 Gosh, UTF-16 is a pain. All the benefits of variable length,
11246 *coupled* with all the benefits of partial reads and
11248 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
11249 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
11251 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
11255 /* We have the first half of a surrogate. Read more. */
11256 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
11259 status = FILTER_READ(idx + 1, utf16_buffer,
11260 160 + (SvCUR(utf16_buffer) & 1));
11261 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
11262 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
11265 IoPAGE(filter) = status;
11270 chars = SvCUR(utf16_buffer) >> 1;
11271 have = SvCUR(utf8_buffer);
11272 SvGROW(utf8_buffer, have + chars * 3 + 1);
11275 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
11276 (U8*)SvPVX_const(utf8_buffer) + have,
11277 chars * 2, &newlen);
11279 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
11280 (U8*)SvPVX_const(utf8_buffer) + have,
11281 chars * 2, &newlen);
11283 SvCUR_set(utf8_buffer, have + newlen);
11286 /* No need to keep this SV "well-formed" with a '\0' after the end, as
11287 it's private to us, and utf16_to_utf8{,reversed} take a
11288 (pointer,length) pair, rather than a NUL-terminated string. */
11289 if(SvCUR(utf16_buffer) & 1) {
11290 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
11291 SvCUR_set(utf16_buffer, 1);
11293 SvCUR_set(utf16_buffer, 0);
11296 DEBUG_P(PerlIO_printf(Perl_debug_log,
11297 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
11299 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
11300 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
11305 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
11307 SV *filter = filter_add(S_utf16_textfilter, NULL);
11309 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
11311 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
11312 sv_setpvs(filter, "");
11313 IoLINES(filter) = reversed;
11314 IoPAGE(filter) = 1; /* Not EOF */
11316 /* Sadly, we have to return a valid pointer, come what may, so we have to
11317 ignore any error return from this. */
11318 SvCUR_set(PL_linestr, 0);
11319 if (FILTER_READ(0, PL_linestr, 0)) {
11320 SvUTF8_on(PL_linestr);
11322 SvUTF8_on(PL_linestr);
11324 PL_bufend = SvEND(PL_linestr);
11325 return (U8*)SvPVX(PL_linestr);
11330 Returns a pointer to the next character after the parsed
11331 vstring, as well as updating the passed in sv.
11333 Function must be called like
11336 s = scan_vstring(s,e,sv);
11338 where s and e are the start and end of the string.
11339 The sv should already be large enough to store the vstring
11340 passed in, for performance reasons.
11345 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
11348 const char *pos = s;
11349 const char *start = s;
11351 PERL_ARGS_ASSERT_SCAN_VSTRING;
11353 if (*pos == 'v') pos++; /* get past 'v' */
11354 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11356 if ( *pos != '.') {
11357 /* this may not be a v-string if followed by => */
11358 const char *next = pos;
11359 while (next < e && isSPACE(*next))
11361 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
11362 /* return string not v-string */
11363 sv_setpvn(sv,(char *)s,pos-s);
11364 return (char *)pos;
11368 if (!isALPHA(*pos)) {
11369 U8 tmpbuf[UTF8_MAXBYTES+1];
11372 s++; /* get past 'v' */
11377 /* this is atoi() that tolerates underscores */
11380 const char *end = pos;
11382 while (--end >= s) {
11384 const UV orev = rev;
11385 rev += (*end - '0') * mult;
11388 /* diag_listed_as: Integer overflow in %s number */
11389 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
11390 "Integer overflow in decimal number");
11394 if (rev > 0x7FFFFFFF)
11395 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
11397 /* Append native character for the rev point */
11398 tmpend = uvchr_to_utf8(tmpbuf, rev);
11399 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
11400 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
11402 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
11408 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
11412 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
11419 Perl_keyword_plugin_standard(pTHX_
11420 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
11422 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
11423 PERL_UNUSED_CONTEXT;
11424 PERL_UNUSED_ARG(keyword_ptr);
11425 PERL_UNUSED_ARG(keyword_len);
11426 PERL_UNUSED_ARG(op_ptr);
11427 return KEYWORD_PLUGIN_DECLINE;
11430 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
11432 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
11434 SAVEI32(PL_lex_brackets);
11435 if (PL_lex_brackets > 100)
11436 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
11437 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
11438 SAVEI32(PL_lex_allbrackets);
11439 PL_lex_allbrackets = 0;
11440 SAVEI8(PL_lex_fakeeof);
11441 PL_lex_fakeeof = (U8)fakeeof;
11442 if(yyparse(gramtype) && !PL_parser->error_count)
11443 qerror(Perl_mess(aTHX_ "Parse error"));
11446 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11448 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11452 SAVEVPTR(PL_eval_root);
11453 PL_eval_root = NULL;
11454 parse_recdescent(gramtype, fakeeof);
11460 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11462 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11465 if (flags & ~PARSE_OPTIONAL)
11466 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11467 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11468 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11469 if (!PL_parser->error_count)
11470 qerror(Perl_mess(aTHX_ "Parse error"));
11471 exprop = newOP(OP_NULL, 0);
11477 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11479 Parse a Perl arithmetic expression. This may contain operators of precedence
11480 down to the bit shift operators. The expression must be followed (and thus
11481 terminated) either by a comparison or lower-precedence operator or by
11482 something that would normally terminate an expression such as semicolon.
11483 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11484 otherwise it is mandatory. It is up to the caller to ensure that the
11485 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11486 the source of the code to be parsed and the lexical context for the
11489 The op tree representing the expression is returned. If an optional
11490 expression is absent, a null pointer is returned, otherwise the pointer
11493 If an error occurs in parsing or compilation, in most cases a valid op
11494 tree is returned anyway. The error is reflected in the parser state,
11495 normally resulting in a single exception at the top level of parsing
11496 which covers all the compilation errors that occurred. Some compilation
11497 errors, however, will throw an exception immediately.
11503 Perl_parse_arithexpr(pTHX_ U32 flags)
11505 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11509 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11511 Parse a Perl term expression. This may contain operators of precedence
11512 down to the assignment operators. The expression must be followed (and thus
11513 terminated) either by a comma or lower-precedence operator or by
11514 something that would normally terminate an expression such as semicolon.
11515 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11516 otherwise it is mandatory. It is up to the caller to ensure that the
11517 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11518 the source of the code to be parsed and the lexical context for the
11521 The op tree representing the expression is returned. If an optional
11522 expression is absent, a null pointer is returned, otherwise the pointer
11525 If an error occurs in parsing or compilation, in most cases a valid op
11526 tree is returned anyway. The error is reflected in the parser state,
11527 normally resulting in a single exception at the top level of parsing
11528 which covers all the compilation errors that occurred. Some compilation
11529 errors, however, will throw an exception immediately.
11535 Perl_parse_termexpr(pTHX_ U32 flags)
11537 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11541 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11543 Parse a Perl list expression. This may contain operators of precedence
11544 down to the comma operator. The expression must be followed (and thus
11545 terminated) either by a low-precedence logic operator such as C<or> or by
11546 something that would normally terminate an expression such as semicolon.
11547 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11548 otherwise it is mandatory. It is up to the caller to ensure that the
11549 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11550 the source of the code to be parsed and the lexical context for the
11553 The op tree representing the expression is returned. If an optional
11554 expression is absent, a null pointer is returned, otherwise the pointer
11557 If an error occurs in parsing or compilation, in most cases a valid op
11558 tree is returned anyway. The error is reflected in the parser state,
11559 normally resulting in a single exception at the top level of parsing
11560 which covers all the compilation errors that occurred. Some compilation
11561 errors, however, will throw an exception immediately.
11567 Perl_parse_listexpr(pTHX_ U32 flags)
11569 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11573 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11575 Parse a single complete Perl expression. This allows the full
11576 expression grammar, including the lowest-precedence operators such
11577 as C<or>. The expression must be followed (and thus terminated) by a
11578 token that an expression would normally be terminated by: end-of-file,
11579 closing bracketing punctuation, semicolon, or one of the keywords that
11580 signals a postfix expression-statement modifier. If I<flags> includes
11581 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11582 mandatory. It is up to the caller to ensure that the dynamic parser
11583 state (L</PL_parser> et al) is correctly set to reflect the source of
11584 the code to be parsed and the lexical context for the expression.
11586 The op tree representing the expression is returned. If an optional
11587 expression is absent, a null pointer is returned, otherwise the pointer
11590 If an error occurs in parsing or compilation, in most cases a valid op
11591 tree is returned anyway. The error is reflected in the parser state,
11592 normally resulting in a single exception at the top level of parsing
11593 which covers all the compilation errors that occurred. Some compilation
11594 errors, however, will throw an exception immediately.
11600 Perl_parse_fullexpr(pTHX_ U32 flags)
11602 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11606 =for apidoc Amx|OP *|parse_block|U32 flags
11608 Parse a single complete Perl code block. This consists of an opening
11609 brace, a sequence of statements, and a closing brace. The block
11610 constitutes a lexical scope, so C<my> variables and various compile-time
11611 effects can be contained within it. It is up to the caller to ensure
11612 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11613 reflect the source of the code to be parsed and the lexical context for
11616 The op tree representing the code block is returned. This is always a
11617 real op, never a null pointer. It will normally be a C<lineseq> list,
11618 including C<nextstate> or equivalent ops. No ops to construct any kind
11619 of runtime scope are included by virtue of it being a block.
11621 If an error occurs in parsing or compilation, in most cases a valid op
11622 tree (most likely null) is returned anyway. The error is reflected in
11623 the parser state, normally resulting in a single exception at the top
11624 level of parsing which covers all the compilation errors that occurred.
11625 Some compilation errors, however, will throw an exception immediately.
11627 The I<flags> parameter is reserved for future use, and must always
11634 Perl_parse_block(pTHX_ U32 flags)
11637 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11638 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11642 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11644 Parse a single unadorned Perl statement. This may be a normal imperative
11645 statement or a declaration that has compile-time effect. It does not
11646 include any label or other affixture. It is up to the caller to ensure
11647 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11648 reflect the source of the code to be parsed and the lexical context for
11651 The op tree representing the statement is returned. This may be a
11652 null pointer if the statement is null, for example if it was actually
11653 a subroutine definition (which has compile-time side effects). If not
11654 null, it will be ops directly implementing the statement, suitable to
11655 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11656 equivalent op (except for those embedded in a scope contained entirely
11657 within the statement).
11659 If an error occurs in parsing or compilation, in most cases a valid op
11660 tree (most likely null) is returned anyway. The error is reflected in
11661 the parser state, normally resulting in a single exception at the top
11662 level of parsing which covers all the compilation errors that occurred.
11663 Some compilation errors, however, will throw an exception immediately.
11665 The I<flags> parameter is reserved for future use, and must always
11672 Perl_parse_barestmt(pTHX_ U32 flags)
11675 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11676 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11680 =for apidoc Amx|SV *|parse_label|U32 flags
11682 Parse a single label, possibly optional, of the type that may prefix a
11683 Perl statement. It is up to the caller to ensure that the dynamic parser
11684 state (L</PL_parser> et al) is correctly set to reflect the source of
11685 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11686 label is optional, otherwise it is mandatory.
11688 The name of the label is returned in the form of a fresh scalar. If an
11689 optional label is absent, a null pointer is returned.
11691 If an error occurs in parsing, which can only occur if the label is
11692 mandatory, a valid label is returned anyway. The error is reflected in
11693 the parser state, normally resulting in a single exception at the top
11694 level of parsing which covers all the compilation errors that occurred.
11700 Perl_parse_label(pTHX_ U32 flags)
11702 if (flags & ~PARSE_OPTIONAL)
11703 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11704 if (PL_lex_state == LEX_KNOWNEXT) {
11705 PL_parser->yychar = yylex();
11706 if (PL_parser->yychar == LABEL) {
11708 PL_parser->yychar = YYEMPTY;
11709 lsv = newSV_type(SVt_PV);
11710 sv_copypv(lsv, cSVOPx(pl_yylval.opval)->op_sv);
11718 STRLEN wlen, bufptr_pos;
11721 if (!isIDFIRST_lazy_if(s, UTF))
11723 t = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &wlen);
11724 if (word_takes_any_delimeter(s, wlen))
11726 bufptr_pos = s - SvPVX(PL_linestr);
11728 lex_read_space(LEX_KEEP_PREVIOUS);
11730 s = SvPVX(PL_linestr) + bufptr_pos;
11731 if (t[0] == ':' && t[1] != ':') {
11732 PL_oldoldbufptr = PL_oldbufptr;
11735 return newSVpvn_flags(s, wlen, UTF ? SVf_UTF8 : 0);
11739 if (flags & PARSE_OPTIONAL) {
11742 qerror(Perl_mess(aTHX_ "Parse error"));
11743 return newSVpvs("x");
11750 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11752 Parse a single complete Perl statement. This may be a normal imperative
11753 statement or a declaration that has compile-time effect, and may include
11754 optional labels. It is up to the caller to ensure that the dynamic
11755 parser state (L</PL_parser> et al) is correctly set to reflect the source
11756 of the code to be parsed and the lexical context for the statement.
11758 The op tree representing the statement is returned. This may be a
11759 null pointer if the statement is null, for example if it was actually
11760 a subroutine definition (which has compile-time side effects). If not
11761 null, it will be the result of a L</newSTATEOP> call, normally including
11762 a C<nextstate> or equivalent op.
11764 If an error occurs in parsing or compilation, in most cases a valid op
11765 tree (most likely null) is returned anyway. The error is reflected in
11766 the parser state, normally resulting in a single exception at the top
11767 level of parsing which covers all the compilation errors that occurred.
11768 Some compilation errors, however, will throw an exception immediately.
11770 The I<flags> parameter is reserved for future use, and must always
11777 Perl_parse_fullstmt(pTHX_ U32 flags)
11780 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11781 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11785 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11787 Parse a sequence of zero or more Perl statements. These may be normal
11788 imperative statements, including optional labels, or declarations
11789 that have compile-time effect, or any mixture thereof. The statement
11790 sequence ends when a closing brace or end-of-file is encountered in a
11791 place where a new statement could have validly started. It is up to
11792 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11793 is correctly set to reflect the source of the code to be parsed and the
11794 lexical context for the statements.
11796 The op tree representing the statement sequence is returned. This may
11797 be a null pointer if the statements were all null, for example if there
11798 were no statements or if there were only subroutine definitions (which
11799 have compile-time side effects). If not null, it will be a C<lineseq>
11800 list, normally including C<nextstate> or equivalent ops.
11802 If an error occurs in parsing or compilation, in most cases a valid op
11803 tree is returned anyway. The error is reflected in the parser state,
11804 normally resulting in a single exception at the top level of parsing
11805 which covers all the compilation errors that occurred. Some compilation
11806 errors, however, will throw an exception immediately.
11808 The I<flags> parameter is reserved for future use, and must always
11815 Perl_parse_stmtseq(pTHX_ U32 flags)
11820 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
11821 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
11822 c = lex_peek_unichar(0);
11823 if (c != -1 && c != /*{*/'}')
11824 qerror(Perl_mess(aTHX_ "Parse error"));
11830 * c-indentation-style: bsd
11831 * c-basic-offset: 4
11832 * indent-tabs-mode: nil
11835 * ex: set ts=8 sts=4 sw=4 et: