3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 by Larry Wall and others
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
12 * 'It all comes from here, the stench and the peril.' --Frodo
14 * [p.719 of _The Lord of the Rings_, IV/ix: "Shelob's Lair"]
18 * This file is the lexer for Perl. It's closely linked to the
21 * The main routine is yylex(), which returns the next token.
25 =head1 Lexer interface
27 This is the lower layer of the Perl parser, managing characters and tokens.
29 =for apidoc AmU|yy_parser *|PL_parser
31 Pointer to a structure encapsulating the state of the parsing operation
32 currently in progress. The pointer can be locally changed to perform
33 a nested parse without interfering with the state of an outer parse.
34 Individual members of C<PL_parser> have their own documentation.
40 #define PERL_IN_TOKE_C
42 #include "dquote_static.c"
44 #define new_constant(a,b,c,d,e,f,g) \
45 S_new_constant(aTHX_ a,b,STR_WITH_LEN(c),d,e,f, g)
47 #define pl_yylval (PL_parser->yylval)
49 /* XXX temporary backwards compatibility */
50 #define PL_lex_brackets (PL_parser->lex_brackets)
51 #define PL_lex_allbrackets (PL_parser->lex_allbrackets)
52 #define PL_lex_fakeeof (PL_parser->lex_fakeeof)
53 #define PL_lex_brackstack (PL_parser->lex_brackstack)
54 #define PL_lex_casemods (PL_parser->lex_casemods)
55 #define PL_lex_casestack (PL_parser->lex_casestack)
56 #define PL_lex_defer (PL_parser->lex_defer)
57 #define PL_lex_dojoin (PL_parser->lex_dojoin)
58 #define PL_lex_expect (PL_parser->lex_expect)
59 #define PL_lex_formbrack (PL_parser->lex_formbrack)
60 #define PL_lex_inpat (PL_parser->lex_inpat)
61 #define PL_lex_inwhat (PL_parser->lex_inwhat)
62 #define PL_lex_op (PL_parser->lex_op)
63 #define PL_lex_repl (PL_parser->lex_repl)
64 #define PL_lex_starts (PL_parser->lex_starts)
65 #define PL_lex_stuff (PL_parser->lex_stuff)
66 #define PL_multi_start (PL_parser->multi_start)
67 #define PL_multi_open (PL_parser->multi_open)
68 #define PL_multi_close (PL_parser->multi_close)
69 #define PL_pending_ident (PL_parser->pending_ident)
70 #define PL_preambled (PL_parser->preambled)
71 #define PL_sublex_info (PL_parser->sublex_info)
72 #define PL_linestr (PL_parser->linestr)
73 #define PL_expect (PL_parser->expect)
74 #define PL_copline (PL_parser->copline)
75 #define PL_bufptr (PL_parser->bufptr)
76 #define PL_oldbufptr (PL_parser->oldbufptr)
77 #define PL_oldoldbufptr (PL_parser->oldoldbufptr)
78 #define PL_linestart (PL_parser->linestart)
79 #define PL_bufend (PL_parser->bufend)
80 #define PL_last_uni (PL_parser->last_uni)
81 #define PL_last_lop (PL_parser->last_lop)
82 #define PL_last_lop_op (PL_parser->last_lop_op)
83 #define PL_lex_state (PL_parser->lex_state)
84 #define PL_rsfp (PL_parser->rsfp)
85 #define PL_rsfp_filters (PL_parser->rsfp_filters)
86 #define PL_in_my (PL_parser->in_my)
87 #define PL_in_my_stash (PL_parser->in_my_stash)
88 #define PL_tokenbuf (PL_parser->tokenbuf)
89 #define PL_multi_end (PL_parser->multi_end)
90 #define PL_error_count (PL_parser->error_count)
93 # define PL_endwhite (PL_parser->endwhite)
94 # define PL_faketokens (PL_parser->faketokens)
95 # define PL_lasttoke (PL_parser->lasttoke)
96 # define PL_nextwhite (PL_parser->nextwhite)
97 # define PL_realtokenstart (PL_parser->realtokenstart)
98 # define PL_skipwhite (PL_parser->skipwhite)
99 # define PL_thisclose (PL_parser->thisclose)
100 # define PL_thismad (PL_parser->thismad)
101 # define PL_thisopen (PL_parser->thisopen)
102 # define PL_thisstuff (PL_parser->thisstuff)
103 # define PL_thistoken (PL_parser->thistoken)
104 # define PL_thiswhite (PL_parser->thiswhite)
105 # define PL_thiswhite (PL_parser->thiswhite)
106 # define PL_nexttoke (PL_parser->nexttoke)
107 # define PL_curforce (PL_parser->curforce)
109 # define PL_nexttoke (PL_parser->nexttoke)
110 # define PL_nexttype (PL_parser->nexttype)
111 # define PL_nextval (PL_parser->nextval)
114 /* This can't be done with embed.fnc, because struct yy_parser contains a
115 member named pending_ident, which clashes with the generated #define */
117 S_pending_ident(pTHX);
119 static const char ident_too_long[] = "Identifier too long";
122 # define CURMAD(slot,sv) if (PL_madskills) { curmad(slot,sv); sv = 0; }
123 # define NEXTVAL_NEXTTOKE PL_nexttoke[PL_curforce].next_val
125 # define CURMAD(slot,sv)
126 # define NEXTVAL_NEXTTOKE PL_nextval[PL_nexttoke]
129 #define XENUMMASK 0x3f
130 #define XFAKEEOF 0x40
131 #define XFAKEBRACK 0x80
133 #ifdef USE_UTF8_SCRIPTS
134 # define UTF (!IN_BYTES)
136 # define UTF ((PL_linestr && DO_UTF8(PL_linestr)) || (PL_hints & HINT_UTF8))
139 /* The maximum number of characters preceding the unrecognized one to display */
140 #define UNRECOGNIZED_PRECEDE_COUNT 10
142 /* In variables named $^X, these are the legal values for X.
143 * 1999-02-27 mjd-perl-patch@plover.com */
144 #define isCONTROLVAR(x) (isUPPER(x) || strchr("[\\]^_?", (x)))
146 #define SPACE_OR_TAB(c) ((c)==' '||(c)=='\t')
148 /* LEX_* are values for PL_lex_state, the state of the lexer.
149 * They are arranged oddly so that the guard on the switch statement
150 * can get by with a single comparison (if the compiler is smart enough).
153 /* #define LEX_NOTPARSING 11 is done in perl.h. */
155 #define LEX_NORMAL 10 /* normal code (ie not within "...") */
156 #define LEX_INTERPNORMAL 9 /* code within a string, eg "$foo[$x+1]" */
157 #define LEX_INTERPCASEMOD 8 /* expecting a \U, \Q or \E etc */
158 #define LEX_INTERPPUSH 7 /* starting a new sublex parse level */
159 #define LEX_INTERPSTART 6 /* expecting the start of a $var */
161 /* at end of code, eg "$x" followed by: */
162 #define LEX_INTERPEND 5 /* ... eg not one of [, { or -> */
163 #define LEX_INTERPENDMAYBE 4 /* ... eg one of [, { or -> */
165 #define LEX_INTERPCONCAT 3 /* expecting anything, eg at start of
166 string or after \E, $foo, etc */
167 #define LEX_INTERPCONST 2 /* NOT USED */
168 #define LEX_FORMLINE 1 /* expecting a format line */
169 #define LEX_KNOWNEXT 0 /* next token known; just return it */
173 static const char* const lex_state_names[] = {
192 #include "keywords.h"
194 /* CLINE is a macro that ensures PL_copline has a sane value */
199 #define CLINE (PL_copline = (CopLINE(PL_curcop) < PL_copline ? CopLINE(PL_curcop) : PL_copline))
202 # define SKIPSPACE0(s) skipspace0(s)
203 # define SKIPSPACE1(s) skipspace1(s)
204 # define SKIPSPACE2(s,tsv) skipspace2(s,&tsv)
205 # define PEEKSPACE(s) skipspace2(s,0)
207 # define SKIPSPACE0(s) skipspace(s)
208 # define SKIPSPACE1(s) skipspace(s)
209 # define SKIPSPACE2(s,tsv) skipspace(s)
210 # define PEEKSPACE(s) skipspace(s)
214 * Convenience functions to return different tokens and prime the
215 * lexer for the next token. They all take an argument.
217 * TOKEN : generic token (used for '(', DOLSHARP, etc)
218 * OPERATOR : generic operator
219 * AOPERATOR : assignment operator
220 * PREBLOCK : beginning the block after an if, while, foreach, ...
221 * PRETERMBLOCK : beginning a non-code-defining {} block (eg, hash ref)
222 * PREREF : *EXPR where EXPR is not a simple identifier
223 * TERM : expression term
224 * LOOPX : loop exiting command (goto, last, dump, etc)
225 * FTST : file test operator
226 * FUN0 : zero-argument function
227 * FUN1 : not used, except for not, which isn't a UNIOP
228 * BOop : bitwise or or xor
230 * SHop : shift operator
231 * PWop : power operator
232 * PMop : pattern-matching operator
233 * Aop : addition-level operator
234 * Mop : multiplication-level operator
235 * Eop : equality-testing operator
236 * Rop : relational operator <= != gt
238 * Also see LOP and lop() below.
241 #ifdef DEBUGGING /* Serve -DT. */
242 # define REPORT(retval) tokereport((I32)retval, &pl_yylval)
244 # define REPORT(retval) (retval)
247 #define TOKEN(retval) return ( PL_bufptr = s, REPORT(retval))
248 #define OPERATOR(retval) return (PL_expect = XTERM, PL_bufptr = s, REPORT(retval))
249 #define AOPERATOR(retval) return ao((PL_expect = XTERM, PL_bufptr = s, REPORT(retval)))
250 #define PREBLOCK(retval) return (PL_expect = XBLOCK,PL_bufptr = s, REPORT(retval))
251 #define PRETERMBLOCK(retval) return (PL_expect = XTERMBLOCK,PL_bufptr = s, REPORT(retval))
252 #define PREREF(retval) return (PL_expect = XREF,PL_bufptr = s, REPORT(retval))
253 #define TERM(retval) return (CLINE, PL_expect = XOPERATOR, PL_bufptr = s, REPORT(retval))
254 #define LOOPX(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)LOOPEX))
255 #define FTST(f) return (pl_yylval.ival=f, PL_expect=XTERMORDORDOR, PL_bufptr=s, REPORT((int)UNIOP))
256 #define FUN0(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC0))
257 #define FUN1(f) return (pl_yylval.ival=f, PL_expect=XOPERATOR, PL_bufptr=s, REPORT((int)FUNC1))
258 #define BOop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITOROP)))
259 #define BAop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)BITANDOP)))
260 #define SHop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)SHIFTOP)))
261 #define PWop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)POWOP)))
262 #define PMop(f) return(pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MATCHOP))
263 #define Aop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)ADDOP)))
264 #define Mop(f) return ao((pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)MULOP)))
265 #define Eop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)EQOP))
266 #define Rop(f) return (pl_yylval.ival=f, PL_expect=XTERM, PL_bufptr=s, REPORT((int)RELOP))
268 /* This bit of chicanery makes a unary function followed by
269 * a parenthesis into a function with one argument, highest precedence.
270 * The UNIDOR macro is for unary functions that can be followed by the //
271 * operator (such as C<shift // 0>).
273 #define UNI2(f,x) { \
274 pl_yylval.ival = f; \
277 PL_last_uni = PL_oldbufptr; \
278 PL_last_lop_op = f; \
280 return REPORT( (int)FUNC1 ); \
282 return REPORT( *s=='(' ? (int)FUNC1 : (int)UNIOP ); \
284 #define UNI(f) UNI2(f,XTERM)
285 #define UNIDOR(f) UNI2(f,XTERMORDORDOR)
287 #define UNIBRACK(f) { \
288 pl_yylval.ival = f; \
290 PL_last_uni = PL_oldbufptr; \
292 return REPORT( (int)FUNC1 ); \
294 return REPORT( (*s == '(') ? (int)FUNC1 : (int)UNIOP ); \
297 /* grandfather return to old style */
300 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC) \
301 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC; \
302 pl_yylval.ival = (f); \
310 /* how to interpret the pl_yylval associated with the token */
314 TOKENTYPE_OPNUM, /* pl_yylval.ival contains an opcode number */
320 static struct debug_tokens {
322 enum token_type type;
324 } const debug_tokens[] =
326 { ADDOP, TOKENTYPE_OPNUM, "ADDOP" },
327 { ANDAND, TOKENTYPE_NONE, "ANDAND" },
328 { ANDOP, TOKENTYPE_NONE, "ANDOP" },
329 { ANONSUB, TOKENTYPE_IVAL, "ANONSUB" },
330 { ARROW, TOKENTYPE_NONE, "ARROW" },
331 { ASSIGNOP, TOKENTYPE_OPNUM, "ASSIGNOP" },
332 { BITANDOP, TOKENTYPE_OPNUM, "BITANDOP" },
333 { BITOROP, TOKENTYPE_OPNUM, "BITOROP" },
334 { COLONATTR, TOKENTYPE_NONE, "COLONATTR" },
335 { CONTINUE, TOKENTYPE_NONE, "CONTINUE" },
336 { DEFAULT, TOKENTYPE_NONE, "DEFAULT" },
337 { DO, TOKENTYPE_NONE, "DO" },
338 { DOLSHARP, TOKENTYPE_NONE, "DOLSHARP" },
339 { DORDOR, TOKENTYPE_NONE, "DORDOR" },
340 { DOROP, TOKENTYPE_OPNUM, "DOROP" },
341 { DOTDOT, TOKENTYPE_IVAL, "DOTDOT" },
342 { ELSE, TOKENTYPE_NONE, "ELSE" },
343 { ELSIF, TOKENTYPE_IVAL, "ELSIF" },
344 { EQOP, TOKENTYPE_OPNUM, "EQOP" },
345 { FOR, TOKENTYPE_IVAL, "FOR" },
346 { FORMAT, TOKENTYPE_NONE, "FORMAT" },
347 { FUNC, TOKENTYPE_OPNUM, "FUNC" },
348 { FUNC0, TOKENTYPE_OPNUM, "FUNC0" },
349 { FUNC0SUB, TOKENTYPE_OPVAL, "FUNC0SUB" },
350 { FUNC1, TOKENTYPE_OPNUM, "FUNC1" },
351 { FUNCMETH, TOKENTYPE_OPVAL, "FUNCMETH" },
352 { GIVEN, TOKENTYPE_IVAL, "GIVEN" },
353 { HASHBRACK, TOKENTYPE_NONE, "HASHBRACK" },
354 { IF, TOKENTYPE_IVAL, "IF" },
355 { LABEL, TOKENTYPE_PVAL, "LABEL" },
356 { LOCAL, TOKENTYPE_IVAL, "LOCAL" },
357 { LOOPEX, TOKENTYPE_OPNUM, "LOOPEX" },
358 { LSTOP, TOKENTYPE_OPNUM, "LSTOP" },
359 { LSTOPSUB, TOKENTYPE_OPVAL, "LSTOPSUB" },
360 { MATCHOP, TOKENTYPE_OPNUM, "MATCHOP" },
361 { METHOD, TOKENTYPE_OPVAL, "METHOD" },
362 { MULOP, TOKENTYPE_OPNUM, "MULOP" },
363 { MY, TOKENTYPE_IVAL, "MY" },
364 { MYSUB, TOKENTYPE_NONE, "MYSUB" },
365 { NOAMP, TOKENTYPE_NONE, "NOAMP" },
366 { NOTOP, TOKENTYPE_NONE, "NOTOP" },
367 { OROP, TOKENTYPE_IVAL, "OROP" },
368 { OROR, TOKENTYPE_NONE, "OROR" },
369 { PACKAGE, TOKENTYPE_NONE, "PACKAGE" },
370 { PLUGEXPR, TOKENTYPE_OPVAL, "PLUGEXPR" },
371 { PLUGSTMT, TOKENTYPE_OPVAL, "PLUGSTMT" },
372 { PMFUNC, TOKENTYPE_OPVAL, "PMFUNC" },
373 { POSTDEC, TOKENTYPE_NONE, "POSTDEC" },
374 { POSTINC, TOKENTYPE_NONE, "POSTINC" },
375 { POWOP, TOKENTYPE_OPNUM, "POWOP" },
376 { PREDEC, TOKENTYPE_NONE, "PREDEC" },
377 { PREINC, TOKENTYPE_NONE, "PREINC" },
378 { PRIVATEREF, TOKENTYPE_OPVAL, "PRIVATEREF" },
379 { REFGEN, TOKENTYPE_NONE, "REFGEN" },
380 { RELOP, TOKENTYPE_OPNUM, "RELOP" },
381 { SHIFTOP, TOKENTYPE_OPNUM, "SHIFTOP" },
382 { SUB, TOKENTYPE_NONE, "SUB" },
383 { THING, TOKENTYPE_OPVAL, "THING" },
384 { UMINUS, TOKENTYPE_NONE, "UMINUS" },
385 { UNIOP, TOKENTYPE_OPNUM, "UNIOP" },
386 { UNIOPSUB, TOKENTYPE_OPVAL, "UNIOPSUB" },
387 { UNLESS, TOKENTYPE_IVAL, "UNLESS" },
388 { UNTIL, TOKENTYPE_IVAL, "UNTIL" },
389 { USE, TOKENTYPE_IVAL, "USE" },
390 { WHEN, TOKENTYPE_IVAL, "WHEN" },
391 { WHILE, TOKENTYPE_IVAL, "WHILE" },
392 { WORD, TOKENTYPE_OPVAL, "WORD" },
393 { YADAYADA, TOKENTYPE_IVAL, "YADAYADA" },
394 { 0, TOKENTYPE_NONE, NULL }
397 /* dump the returned token in rv, plus any optional arg in pl_yylval */
400 S_tokereport(pTHX_ I32 rv, const YYSTYPE* lvalp)
404 PERL_ARGS_ASSERT_TOKEREPORT;
407 const char *name = NULL;
408 enum token_type type = TOKENTYPE_NONE;
409 const struct debug_tokens *p;
410 SV* const report = newSVpvs("<== ");
412 for (p = debug_tokens; p->token; p++) {
413 if (p->token == (int)rv) {
420 Perl_sv_catpv(aTHX_ report, name);
421 else if ((char)rv > ' ' && (char)rv < '~')
422 Perl_sv_catpvf(aTHX_ report, "'%c'", (char)rv);
424 sv_catpvs(report, "EOF");
426 Perl_sv_catpvf(aTHX_ report, "?? %"IVdf, (IV)rv);
429 case TOKENTYPE_GVVAL: /* doesn't appear to be used */
432 Perl_sv_catpvf(aTHX_ report, "(ival=%"IVdf")", (IV)lvalp->ival);
434 case TOKENTYPE_OPNUM:
435 Perl_sv_catpvf(aTHX_ report, "(ival=op_%s)",
436 PL_op_name[lvalp->ival]);
439 Perl_sv_catpvf(aTHX_ report, "(pval=\"%s\")", lvalp->pval);
441 case TOKENTYPE_OPVAL:
443 Perl_sv_catpvf(aTHX_ report, "(opval=op_%s)",
444 PL_op_name[lvalp->opval->op_type]);
445 if (lvalp->opval->op_type == OP_CONST) {
446 Perl_sv_catpvf(aTHX_ report, " %s",
447 SvPEEK(cSVOPx_sv(lvalp->opval)));
452 sv_catpvs(report, "(opval=null)");
455 PerlIO_printf(Perl_debug_log, "### %s\n\n", SvPV_nolen_const(report));
461 /* print the buffer with suitable escapes */
464 S_printbuf(pTHX_ const char *const fmt, const char *const s)
466 SV* const tmp = newSVpvs("");
468 PERL_ARGS_ASSERT_PRINTBUF;
470 PerlIO_printf(Perl_debug_log, fmt, pv_display(tmp, s, strlen(s), 0, 60));
477 S_deprecate_commaless_var_list(pTHX) {
479 deprecate("comma-less variable list");
480 return REPORT(','); /* grandfather non-comma-format format */
486 * This subroutine detects &&=, ||=, and //= and turns an ANDAND, OROR or DORDOR
487 * into an OP_ANDASSIGN, OP_ORASSIGN, or OP_DORASSIGN
491 S_ao(pTHX_ int toketype)
494 if (*PL_bufptr == '=') {
496 if (toketype == ANDAND)
497 pl_yylval.ival = OP_ANDASSIGN;
498 else if (toketype == OROR)
499 pl_yylval.ival = OP_ORASSIGN;
500 else if (toketype == DORDOR)
501 pl_yylval.ival = OP_DORASSIGN;
509 * When Perl expects an operator and finds something else, no_op
510 * prints the warning. It always prints "<something> found where
511 * operator expected. It prints "Missing semicolon on previous line?"
512 * if the surprise occurs at the start of the line. "do you need to
513 * predeclare ..." is printed out for code like "sub bar; foo bar $x"
514 * where the compiler doesn't know if foo is a method call or a function.
515 * It prints "Missing operator before end of line" if there's nothing
516 * after the missing operator, or "... before <...>" if there is something
517 * after the missing operator.
521 S_no_op(pTHX_ const char *const what, char *s)
524 char * const oldbp = PL_bufptr;
525 const bool is_first = (PL_oldbufptr == PL_linestart);
527 PERL_ARGS_ASSERT_NO_OP;
533 yywarn(Perl_form(aTHX_ "%s found where operator expected", what));
534 if (ckWARN_d(WARN_SYNTAX)) {
536 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
537 "\t(Missing semicolon on previous line?)\n");
538 else if (PL_oldoldbufptr && isIDFIRST_lazy_if(PL_oldoldbufptr,UTF)) {
540 for (t = PL_oldoldbufptr; (isALNUM_lazy_if(t,UTF) || *t == ':'); t++)
542 if (t < PL_bufptr && isSPACE(*t))
543 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
544 "\t(Do you need to predeclare %.*s?)\n",
545 (int)(t - PL_oldoldbufptr), PL_oldoldbufptr);
549 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
550 "\t(Missing operator before %.*s?)\n", (int)(s - oldbp), oldbp);
558 * Complain about missing quote/regexp/heredoc terminator.
559 * If it's called with NULL then it cauterizes the line buffer.
560 * If we're in a delimited string and the delimiter is a control
561 * character, it's reformatted into a two-char sequence like ^C.
566 S_missingterm(pTHX_ char *s)
572 char * const nl = strrchr(s,'\n');
576 else if (isCNTRL(PL_multi_close)) {
578 tmpbuf[1] = (char)toCTRL(PL_multi_close);
583 *tmpbuf = (char)PL_multi_close;
587 q = strchr(s,'"') ? '\'' : '"';
588 Perl_croak(aTHX_ "Can't find string terminator %c%s%c anywhere before EOF",q,s,q);
592 * Check whether the named feature is enabled.
595 Perl_feature_is_enabled(pTHX_ const char *const name, STRLEN namelen)
598 HV * const hinthv = GvHV(PL_hintgv);
599 char he_name[8 + MAX_FEATURE_LEN] = "feature_";
601 PERL_ARGS_ASSERT_FEATURE_IS_ENABLED;
603 if (namelen > MAX_FEATURE_LEN)
605 memcpy(&he_name[8], name, namelen);
607 return (hinthv && hv_exists(hinthv, he_name, 8 + namelen));
611 * experimental text filters for win32 carriage-returns, utf16-to-utf8 and
612 * utf16-to-utf8-reversed.
615 #ifdef PERL_CR_FILTER
619 register const char *s = SvPVX_const(sv);
620 register const char * const e = s + SvCUR(sv);
622 PERL_ARGS_ASSERT_STRIP_RETURN;
624 /* outer loop optimized to do nothing if there are no CR-LFs */
626 if (*s++ == '\r' && *s == '\n') {
627 /* hit a CR-LF, need to copy the rest */
628 register char *d = s - 1;
631 if (*s == '\r' && s[1] == '\n')
642 S_cr_textfilter(pTHX_ int idx, SV *sv, int maxlen)
644 const I32 count = FILTER_READ(idx+1, sv, maxlen);
645 if (count > 0 && !maxlen)
652 =for apidoc Amx|void|lex_start|SV *line|PerlIO *rsfp|U32 flags
654 Creates and initialises a new lexer/parser state object, supplying
655 a context in which to lex and parse from a new source of Perl code.
656 A pointer to the new state object is placed in L</PL_parser>. An entry
657 is made on the save stack so that upon unwinding the new state object
658 will be destroyed and the former value of L</PL_parser> will be restored.
659 Nothing else need be done to clean up the parsing context.
661 The code to be parsed comes from I<line> and I<rsfp>. I<line>, if
662 non-null, provides a string (in SV form) containing code to be parsed.
663 A copy of the string is made, so subsequent modification of I<line>
664 does not affect parsing. I<rsfp>, if non-null, provides an input stream
665 from which code will be read to be parsed. If both are non-null, the
666 code in I<line> comes first and must consist of complete lines of input,
667 and I<rsfp> supplies the remainder of the source.
669 The I<flags> parameter is reserved for future use, and must always
676 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
679 const char *s = NULL;
681 yy_parser *parser, *oparser;
683 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
685 /* create and initialise a parser */
687 Newxz(parser, 1, yy_parser);
688 parser->old_parser = oparser = PL_parser;
691 parser->stack = NULL;
693 parser->stack_size = 0;
695 /* on scope exit, free this parser and restore any outer one */
697 parser->saved_curcop = PL_curcop;
699 /* initialise lexer state */
702 parser->curforce = -1;
704 parser->nexttoke = 0;
706 parser->error_count = oparser ? oparser->error_count : 0;
707 parser->copline = NOLINE;
708 parser->lex_state = LEX_NORMAL;
709 parser->expect = XSTATE;
711 parser->rsfp_filters = newAV();
713 Newx(parser->lex_brackstack, 120, char);
714 Newx(parser->lex_casestack, 12, char);
715 *parser->lex_casestack = '\0';
718 s = SvPV_const(line, len);
724 parser->linestr = newSVpvs("\n;");
726 parser->linestr = newSVpvn_flags(s, len, SvUTF8(line));
728 sv_catpvs(parser->linestr, "\n;");
730 parser->oldoldbufptr =
733 parser->linestart = SvPVX(parser->linestr);
734 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
735 parser->last_lop = parser->last_uni = NULL;
741 /* delete a parser object */
744 Perl_parser_free(pTHX_ const yy_parser *parser)
746 PERL_ARGS_ASSERT_PARSER_FREE;
748 PL_curcop = parser->saved_curcop;
749 SvREFCNT_dec(parser->linestr);
751 if (parser->rsfp == PerlIO_stdin())
752 PerlIO_clearerr(parser->rsfp);
753 else if (parser->rsfp && (!parser->old_parser ||
754 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
755 PerlIO_close(parser->rsfp);
756 SvREFCNT_dec(parser->rsfp_filters);
758 Safefree(parser->lex_brackstack);
759 Safefree(parser->lex_casestack);
760 PL_parser = parser->old_parser;
766 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
768 Buffer scalar containing the chunk currently under consideration of the
769 text currently being lexed. This is always a plain string scalar (for
770 which C<SvPOK> is true). It is not intended to be used as a scalar by
771 normal scalar means; instead refer to the buffer directly by the pointer
772 variables described below.
774 The lexer maintains various C<char*> pointers to things in the
775 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
776 reallocated, all of these pointers must be updated. Don't attempt to
777 do this manually, but rather use L</lex_grow_linestr> if you need to
778 reallocate the buffer.
780 The content of the text chunk in the buffer is commonly exactly one
781 complete line of input, up to and including a newline terminator,
782 but there are situations where it is otherwise. The octets of the
783 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
784 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
785 flag on this scalar, which may disagree with it.
787 For direct examination of the buffer, the variable
788 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
789 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
790 of these pointers is usually preferable to examination of the scalar
791 through normal scalar means.
793 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
795 Direct pointer to the end of the chunk of text currently being lexed, the
796 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
797 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
798 always located at the end of the buffer, and does not count as part of
799 the buffer's contents.
801 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
803 Points to the current position of lexing inside the lexer buffer.
804 Characters around this point may be freely examined, within
805 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
806 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
807 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
809 Lexing code (whether in the Perl core or not) moves this pointer past
810 the characters that it consumes. It is also expected to perform some
811 bookkeeping whenever a newline character is consumed. This movement
812 can be more conveniently performed by the function L</lex_read_to>,
813 which handles newlines appropriately.
815 Interpretation of the buffer's octets can be abstracted out by
816 using the slightly higher-level functions L</lex_peek_unichar> and
817 L</lex_read_unichar>.
819 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
821 Points to the start of the current line inside the lexer buffer.
822 This is useful for indicating at which column an error occurred, and
823 not much else. This must be updated by any lexing code that consumes
824 a newline; the function L</lex_read_to> handles this detail.
830 =for apidoc Amx|bool|lex_bufutf8
832 Indicates whether the octets in the lexer buffer
833 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
834 of Unicode characters. If not, they should be interpreted as Latin-1
835 characters. This is analogous to the C<SvUTF8> flag for scalars.
837 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
838 contains valid UTF-8. Lexing code must be robust in the face of invalid
841 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
842 is significant, but not the whole story regarding the input character
843 encoding. Normally, when a file is being read, the scalar contains octets
844 and its C<SvUTF8> flag is off, but the octets should be interpreted as
845 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
846 however, the scalar may have the C<SvUTF8> flag on, and in this case its
847 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
848 is in effect. This logic may change in the future; use this function
849 instead of implementing the logic yourself.
855 Perl_lex_bufutf8(pTHX)
861 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
863 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
864 at least I<len> octets (including terminating NUL). Returns a
865 pointer to the reallocated buffer. This is necessary before making
866 any direct modification of the buffer that would increase its length.
867 L</lex_stuff_pvn> provides a more convenient way to insert text into
870 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
871 this function updates all of the lexer's variables that point directly
878 Perl_lex_grow_linestr(pTHX_ STRLEN len)
882 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
883 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
884 linestr = PL_parser->linestr;
885 buf = SvPVX(linestr);
886 if (len <= SvLEN(linestr))
888 bufend_pos = PL_parser->bufend - buf;
889 bufptr_pos = PL_parser->bufptr - buf;
890 oldbufptr_pos = PL_parser->oldbufptr - buf;
891 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
892 linestart_pos = PL_parser->linestart - buf;
893 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
894 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
895 buf = sv_grow(linestr, len);
896 PL_parser->bufend = buf + bufend_pos;
897 PL_parser->bufptr = buf + bufptr_pos;
898 PL_parser->oldbufptr = buf + oldbufptr_pos;
899 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
900 PL_parser->linestart = buf + linestart_pos;
901 if (PL_parser->last_uni)
902 PL_parser->last_uni = buf + last_uni_pos;
903 if (PL_parser->last_lop)
904 PL_parser->last_lop = buf + last_lop_pos;
909 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
911 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
912 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
913 reallocating the buffer if necessary. This means that lexing code that
914 runs later will see the characters as if they had appeared in the input.
915 It is not recommended to do this as part of normal parsing, and most
916 uses of this facility run the risk of the inserted characters being
917 interpreted in an unintended manner.
919 The string to be inserted is represented by I<len> octets starting
920 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
921 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
922 The characters are recoded for the lexer buffer, according to how the
923 buffer is currently being interpreted (L</lex_bufutf8>). If a string
924 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
925 function is more convenient.
931 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
935 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
936 if (flags & ~(LEX_STUFF_UTF8))
937 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
939 if (flags & LEX_STUFF_UTF8) {
943 const char *p, *e = pv+len;
944 for (p = pv; p != e; p++)
945 highhalf += !!(((U8)*p) & 0x80);
948 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
949 bufptr = PL_parser->bufptr;
950 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
951 SvCUR_set(PL_parser->linestr,
952 SvCUR(PL_parser->linestr) + len+highhalf);
953 PL_parser->bufend += len+highhalf;
954 for (p = pv; p != e; p++) {
957 *bufptr++ = (char)(0xc0 | (c >> 6));
958 *bufptr++ = (char)(0x80 | (c & 0x3f));
965 if (flags & LEX_STUFF_UTF8) {
967 const char *p, *e = pv+len;
968 for (p = pv; p != e; p++) {
971 Perl_croak(aTHX_ "Lexing code attempted to stuff "
972 "non-Latin-1 character into Latin-1 input");
973 } else if (c >= 0xc2 && p+1 != e &&
974 (((U8)p[1]) & 0xc0) == 0x80) {
977 } else if (c >= 0x80) {
978 /* malformed UTF-8 */
980 SAVESPTR(PL_warnhook);
981 PL_warnhook = PERL_WARNHOOK_FATAL;
982 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
988 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
989 bufptr = PL_parser->bufptr;
990 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
991 SvCUR_set(PL_parser->linestr,
992 SvCUR(PL_parser->linestr) + len-highhalf);
993 PL_parser->bufend += len-highhalf;
994 for (p = pv; p != e; p++) {
997 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1000 *bufptr++ = (char)c;
1005 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1006 bufptr = PL_parser->bufptr;
1007 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1008 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1009 PL_parser->bufend += len;
1010 Copy(pv, bufptr, len, char);
1016 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1018 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1019 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1020 reallocating the buffer if necessary. This means that lexing code that
1021 runs later will see the characters as if they had appeared in the input.
1022 It is not recommended to do this as part of normal parsing, and most
1023 uses of this facility run the risk of the inserted characters being
1024 interpreted in an unintended manner.
1026 The string to be inserted is represented by octets starting at I<pv>
1027 and continuing to the first nul. These octets are interpreted as either
1028 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1029 in I<flags>. The characters are recoded for the lexer buffer, according
1030 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1031 If it is not convenient to nul-terminate a string to be inserted, the
1032 L</lex_stuff_pvn> function is more appropriate.
1038 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1040 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1041 lex_stuff_pvn(pv, strlen(pv), flags);
1045 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1047 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1048 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1049 reallocating the buffer if necessary. This means that lexing code that
1050 runs later will see the characters as if they had appeared in the input.
1051 It is not recommended to do this as part of normal parsing, and most
1052 uses of this facility run the risk of the inserted characters being
1053 interpreted in an unintended manner.
1055 The string to be inserted is the string value of I<sv>. The characters
1056 are recoded for the lexer buffer, according to how the buffer is currently
1057 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1058 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1059 need to construct a scalar.
1065 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1069 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1071 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1073 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1077 =for apidoc Amx|void|lex_unstuff|char *ptr
1079 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1080 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1081 This hides the discarded text from any lexing code that runs later,
1082 as if the text had never appeared.
1084 This is not the normal way to consume lexed text. For that, use
1091 Perl_lex_unstuff(pTHX_ char *ptr)
1095 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1096 buf = PL_parser->bufptr;
1098 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1101 bufend = PL_parser->bufend;
1103 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1104 unstuff_len = ptr - buf;
1105 Move(ptr, buf, bufend+1-ptr, char);
1106 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1107 PL_parser->bufend = bufend - unstuff_len;
1111 =for apidoc Amx|void|lex_read_to|char *ptr
1113 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1114 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1115 performing the correct bookkeeping whenever a newline character is passed.
1116 This is the normal way to consume lexed text.
1118 Interpretation of the buffer's octets can be abstracted out by
1119 using the slightly higher-level functions L</lex_peek_unichar> and
1120 L</lex_read_unichar>.
1126 Perl_lex_read_to(pTHX_ char *ptr)
1129 PERL_ARGS_ASSERT_LEX_READ_TO;
1130 s = PL_parser->bufptr;
1131 if (ptr < s || ptr > PL_parser->bufend)
1132 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1133 for (; s != ptr; s++)
1135 CopLINE_inc(PL_curcop);
1136 PL_parser->linestart = s+1;
1138 PL_parser->bufptr = ptr;
1142 =for apidoc Amx|void|lex_discard_to|char *ptr
1144 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1145 up to I<ptr>. The remaining content of the buffer will be moved, and
1146 all pointers into the buffer updated appropriately. I<ptr> must not
1147 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1148 it is not permitted to discard text that has yet to be lexed.
1150 Normally it is not necessarily to do this directly, because it suffices to
1151 use the implicit discarding behaviour of L</lex_next_chunk> and things
1152 based on it. However, if a token stretches across multiple lines,
1153 and the lexing code has kept multiple lines of text in the buffer for
1154 that purpose, then after completion of the token it would be wise to
1155 explicitly discard the now-unneeded earlier lines, to avoid future
1156 multi-line tokens growing the buffer without bound.
1162 Perl_lex_discard_to(pTHX_ char *ptr)
1166 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1167 buf = SvPVX(PL_parser->linestr);
1169 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1172 if (ptr > PL_parser->bufptr)
1173 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1174 discard_len = ptr - buf;
1175 if (PL_parser->oldbufptr < ptr)
1176 PL_parser->oldbufptr = ptr;
1177 if (PL_parser->oldoldbufptr < ptr)
1178 PL_parser->oldoldbufptr = ptr;
1179 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1180 PL_parser->last_uni = NULL;
1181 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1182 PL_parser->last_lop = NULL;
1183 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1184 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1185 PL_parser->bufend -= discard_len;
1186 PL_parser->bufptr -= discard_len;
1187 PL_parser->oldbufptr -= discard_len;
1188 PL_parser->oldoldbufptr -= discard_len;
1189 if (PL_parser->last_uni)
1190 PL_parser->last_uni -= discard_len;
1191 if (PL_parser->last_lop)
1192 PL_parser->last_lop -= discard_len;
1196 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1198 Reads in the next chunk of text to be lexed, appending it to
1199 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1200 looked to the end of the current chunk and wants to know more. It is
1201 usual, but not necessary, for lexing to have consumed the entirety of
1202 the current chunk at this time.
1204 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1205 chunk (i.e., the current chunk has been entirely consumed), normally the
1206 current chunk will be discarded at the same time that the new chunk is
1207 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1208 will not be discarded. If the current chunk has not been entirely
1209 consumed, then it will not be discarded regardless of the flag.
1211 Returns true if some new text was added to the buffer, or false if the
1212 buffer has reached the end of the input text.
1217 #define LEX_FAKE_EOF 0x80000000
1220 Perl_lex_next_chunk(pTHX_ U32 flags)
1224 STRLEN old_bufend_pos, new_bufend_pos;
1225 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1226 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1227 bool got_some_for_debugger = 0;
1229 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1230 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1231 linestr = PL_parser->linestr;
1232 buf = SvPVX(linestr);
1233 if (!(flags & LEX_KEEP_PREVIOUS) &&
1234 PL_parser->bufptr == PL_parser->bufend) {
1235 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1237 if (PL_parser->last_uni != PL_parser->bufend)
1238 PL_parser->last_uni = NULL;
1239 if (PL_parser->last_lop != PL_parser->bufend)
1240 PL_parser->last_lop = NULL;
1241 last_uni_pos = last_lop_pos = 0;
1245 old_bufend_pos = PL_parser->bufend - buf;
1246 bufptr_pos = PL_parser->bufptr - buf;
1247 oldbufptr_pos = PL_parser->oldbufptr - buf;
1248 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1249 linestart_pos = PL_parser->linestart - buf;
1250 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1251 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1253 if (flags & LEX_FAKE_EOF) {
1255 } else if (!PL_parser->rsfp) {
1257 } else if (filter_gets(linestr, old_bufend_pos)) {
1259 got_some_for_debugger = 1;
1261 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1262 sv_setpvs(linestr, "");
1264 /* End of real input. Close filehandle (unless it was STDIN),
1265 * then add implicit termination.
1267 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1268 PerlIO_clearerr(PL_parser->rsfp);
1269 else if (PL_parser->rsfp)
1270 (void)PerlIO_close(PL_parser->rsfp);
1271 PL_parser->rsfp = NULL;
1272 PL_parser->in_pod = 0;
1274 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1277 if (!PL_in_eval && PL_minus_p) {
1279 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1280 PL_minus_n = PL_minus_p = 0;
1281 } else if (!PL_in_eval && PL_minus_n) {
1282 sv_catpvs(linestr, /*{*/";}");
1285 sv_catpvs(linestr, ";");
1288 buf = SvPVX(linestr);
1289 new_bufend_pos = SvCUR(linestr);
1290 PL_parser->bufend = buf + new_bufend_pos;
1291 PL_parser->bufptr = buf + bufptr_pos;
1292 PL_parser->oldbufptr = buf + oldbufptr_pos;
1293 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1294 PL_parser->linestart = buf + linestart_pos;
1295 if (PL_parser->last_uni)
1296 PL_parser->last_uni = buf + last_uni_pos;
1297 if (PL_parser->last_lop)
1298 PL_parser->last_lop = buf + last_lop_pos;
1299 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1300 PL_curstash != PL_debstash) {
1301 /* debugger active and we're not compiling the debugger code,
1302 * so store the line into the debugger's array of lines
1304 update_debugger_info(NULL, buf+old_bufend_pos,
1305 new_bufend_pos-old_bufend_pos);
1311 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1313 Looks ahead one (Unicode) character in the text currently being lexed.
1314 Returns the codepoint (unsigned integer value) of the next character,
1315 or -1 if lexing has reached the end of the input text. To consume the
1316 peeked character, use L</lex_read_unichar>.
1318 If the next character is in (or extends into) the next chunk of input
1319 text, the next chunk will be read in. Normally the current chunk will be
1320 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1321 then the current chunk will not be discarded.
1323 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1324 is encountered, an exception is generated.
1330 Perl_lex_peek_unichar(pTHX_ U32 flags)
1334 if (flags & ~(LEX_KEEP_PREVIOUS))
1335 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1336 s = PL_parser->bufptr;
1337 bufend = PL_parser->bufend;
1343 if (!lex_next_chunk(flags))
1345 s = PL_parser->bufptr;
1346 bufend = PL_parser->bufend;
1352 len = PL_utf8skip[head];
1353 while ((STRLEN)(bufend-s) < len) {
1354 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1356 s = PL_parser->bufptr;
1357 bufend = PL_parser->bufend;
1360 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1361 if (retlen == (STRLEN)-1) {
1362 /* malformed UTF-8 */
1364 SAVESPTR(PL_warnhook);
1365 PL_warnhook = PERL_WARNHOOK_FATAL;
1366 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1372 if (!lex_next_chunk(flags))
1374 s = PL_parser->bufptr;
1381 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1383 Reads the next (Unicode) character in the text currently being lexed.
1384 Returns the codepoint (unsigned integer value) of the character read,
1385 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1386 if lexing has reached the end of the input text. To non-destructively
1387 examine the next character, use L</lex_peek_unichar> instead.
1389 If the next character is in (or extends into) the next chunk of input
1390 text, the next chunk will be read in. Normally the current chunk will be
1391 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1392 then the current chunk will not be discarded.
1394 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1395 is encountered, an exception is generated.
1401 Perl_lex_read_unichar(pTHX_ U32 flags)
1404 if (flags & ~(LEX_KEEP_PREVIOUS))
1405 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1406 c = lex_peek_unichar(flags);
1409 CopLINE_inc(PL_curcop);
1410 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1416 =for apidoc Amx|void|lex_read_space|U32 flags
1418 Reads optional spaces, in Perl style, in the text currently being
1419 lexed. The spaces may include ordinary whitespace characters and
1420 Perl-style comments. C<#line> directives are processed if encountered.
1421 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1422 at a non-space character (or the end of the input text).
1424 If spaces extend into the next chunk of input text, the next chunk will
1425 be read in. Normally the current chunk will be discarded at the same
1426 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1427 chunk will not be discarded.
1432 #define LEX_NO_NEXT_CHUNK 0x80000000
1435 Perl_lex_read_space(pTHX_ U32 flags)
1438 bool need_incline = 0;
1439 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1440 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1443 sv_free(PL_skipwhite);
1444 PL_skipwhite = NULL;
1447 PL_skipwhite = newSVpvs("");
1448 #endif /* PERL_MAD */
1449 s = PL_parser->bufptr;
1450 bufend = PL_parser->bufend;
1456 } while (!(c == '\n' || (c == 0 && s == bufend)));
1457 } else if (c == '\n') {
1459 PL_parser->linestart = s;
1464 } else if (isSPACE(c)) {
1466 } else if (c == 0 && s == bufend) {
1470 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1471 #endif /* PERL_MAD */
1472 if (flags & LEX_NO_NEXT_CHUNK)
1474 PL_parser->bufptr = s;
1475 CopLINE_inc(PL_curcop);
1476 got_more = lex_next_chunk(flags);
1477 CopLINE_dec(PL_curcop);
1478 s = PL_parser->bufptr;
1479 bufend = PL_parser->bufend;
1482 if (need_incline && PL_parser->rsfp) {
1492 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1493 #endif /* PERL_MAD */
1494 PL_parser->bufptr = s;
1499 * This subroutine has nothing to do with tilting, whether at windmills
1500 * or pinball tables. Its name is short for "increment line". It
1501 * increments the current line number in CopLINE(PL_curcop) and checks
1502 * to see whether the line starts with a comment of the form
1503 * # line 500 "foo.pm"
1504 * If so, it sets the current line number and file to the values in the comment.
1508 S_incline(pTHX_ const char *s)
1516 PERL_ARGS_ASSERT_INCLINE;
1518 CopLINE_inc(PL_curcop);
1521 while (SPACE_OR_TAB(*s))
1523 if (strnEQ(s, "line", 4))
1527 if (SPACE_OR_TAB(*s))
1531 while (SPACE_OR_TAB(*s))
1539 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1541 while (SPACE_OR_TAB(*s))
1543 if (*s == '"' && (t = strchr(s+1, '"'))) {
1549 while (!isSPACE(*t))
1553 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1555 if (*e != '\n' && *e != '\0')
1556 return; /* false alarm */
1558 line_num = atoi(n)-1;
1561 const STRLEN len = t - s;
1562 SV *const temp_sv = CopFILESV(PL_curcop);
1567 cf = SvPVX(temp_sv);
1568 tmplen = SvCUR(temp_sv);
1574 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
1575 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1576 * to *{"::_<newfilename"} */
1577 /* However, the long form of evals is only turned on by the
1578 debugger - usually they're "(eval %lu)" */
1582 STRLEN tmplen2 = len;
1583 if (tmplen + 2 <= sizeof smallbuf)
1586 Newx(tmpbuf, tmplen + 2, char);
1589 memcpy(tmpbuf + 2, cf, tmplen);
1591 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1596 if (tmplen2 + 2 <= sizeof smallbuf)
1599 Newx(tmpbuf2, tmplen2 + 2, char);
1601 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1602 /* Either they malloc'd it, or we malloc'd it,
1603 so no prefix is present in ours. */
1608 memcpy(tmpbuf2 + 2, s, tmplen2);
1611 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1613 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1614 /* adjust ${"::_<newfilename"} to store the new file name */
1615 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1616 /* The line number may differ. If that is the case,
1617 alias the saved lines that are in the array.
1618 Otherwise alias the whole array. */
1619 if (CopLINE(PL_curcop) == line_num) {
1620 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1621 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1623 else if (GvAV(*gvp)) {
1624 AV * const av = GvAV(*gvp);
1625 const I32 start = CopLINE(PL_curcop)+1;
1626 I32 items = AvFILLp(av) - start;
1628 AV * const av2 = GvAVn(gv2);
1629 SV **svp = AvARRAY(av) + start;
1630 I32 l = (I32)line_num+1;
1632 av_store(av2, l++, SvREFCNT_inc(*svp++));
1637 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1639 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1641 CopFILE_free(PL_curcop);
1642 CopFILE_setn(PL_curcop, s, len);
1644 CopLINE_set(PL_curcop, line_num);
1648 /* skip space before PL_thistoken */
1651 S_skipspace0(pTHX_ register char *s)
1653 PERL_ARGS_ASSERT_SKIPSPACE0;
1660 PL_thiswhite = newSVpvs("");
1661 sv_catsv(PL_thiswhite, PL_skipwhite);
1662 sv_free(PL_skipwhite);
1665 PL_realtokenstart = s - SvPVX(PL_linestr);
1669 /* skip space after PL_thistoken */
1672 S_skipspace1(pTHX_ register char *s)
1674 const char *start = s;
1675 I32 startoff = start - SvPVX(PL_linestr);
1677 PERL_ARGS_ASSERT_SKIPSPACE1;
1682 start = SvPVX(PL_linestr) + startoff;
1683 if (!PL_thistoken && PL_realtokenstart >= 0) {
1684 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1685 PL_thistoken = newSVpvn(tstart, start - tstart);
1687 PL_realtokenstart = -1;
1690 PL_nextwhite = newSVpvs("");
1691 sv_catsv(PL_nextwhite, PL_skipwhite);
1692 sv_free(PL_skipwhite);
1699 S_skipspace2(pTHX_ register char *s, SV **svp)
1702 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1703 const I32 startoff = s - SvPVX(PL_linestr);
1705 PERL_ARGS_ASSERT_SKIPSPACE2;
1708 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1709 if (!PL_madskills || !svp)
1711 start = SvPVX(PL_linestr) + startoff;
1712 if (!PL_thistoken && PL_realtokenstart >= 0) {
1713 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1714 PL_thistoken = newSVpvn(tstart, start - tstart);
1715 PL_realtokenstart = -1;
1719 *svp = newSVpvs("");
1720 sv_setsv(*svp, PL_skipwhite);
1721 sv_free(PL_skipwhite);
1730 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1732 AV *av = CopFILEAVx(PL_curcop);
1734 SV * const sv = newSV_type(SVt_PVMG);
1736 sv_setsv(sv, orig_sv);
1738 sv_setpvn(sv, buf, len);
1741 av_store(av, (I32)CopLINE(PL_curcop), sv);
1747 * Called to gobble the appropriate amount and type of whitespace.
1748 * Skips comments as well.
1752 S_skipspace(pTHX_ register char *s)
1756 #endif /* PERL_MAD */
1757 PERL_ARGS_ASSERT_SKIPSPACE;
1760 sv_free(PL_skipwhite);
1761 PL_skipwhite = NULL;
1763 #endif /* PERL_MAD */
1764 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1765 while (s < PL_bufend && SPACE_OR_TAB(*s))
1768 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1770 lex_read_space(LEX_KEEP_PREVIOUS |
1771 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1772 LEX_NO_NEXT_CHUNK : 0));
1774 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1775 if (PL_linestart > PL_bufptr)
1776 PL_bufptr = PL_linestart;
1781 PL_skipwhite = newSVpvn(start, s-start);
1782 #endif /* PERL_MAD */
1788 * Check the unary operators to ensure there's no ambiguity in how they're
1789 * used. An ambiguous piece of code would be:
1791 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1792 * the +5 is its argument.
1802 if (PL_oldoldbufptr != PL_last_uni)
1804 while (isSPACE(*PL_last_uni))
1807 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1809 if ((t = strchr(s, '(')) && t < PL_bufptr)
1812 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1813 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1814 (int)(s - PL_last_uni), PL_last_uni);
1818 * LOP : macro to build a list operator. Its behaviour has been replaced
1819 * with a subroutine, S_lop() for which LOP is just another name.
1822 #define LOP(f,x) return lop(f,x,s)
1826 * Build a list operator (or something that might be one). The rules:
1827 * - if we have a next token, then it's a list operator [why?]
1828 * - if the next thing is an opening paren, then it's a function
1829 * - else it's a list operator
1833 S_lop(pTHX_ I32 f, int x, char *s)
1837 PERL_ARGS_ASSERT_LOP;
1843 PL_last_lop = PL_oldbufptr;
1844 PL_last_lop_op = (OPCODE)f;
1853 return REPORT(FUNC);
1856 return REPORT(FUNC);
1859 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1860 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1861 return REPORT(LSTOP);
1868 * Sets up for an eventual force_next(). start_force(0) basically does
1869 * an unshift, while start_force(-1) does a push. yylex removes items
1874 S_start_force(pTHX_ int where)
1878 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1879 where = PL_lasttoke;
1880 assert(PL_curforce < 0 || PL_curforce == where);
1881 if (PL_curforce != where) {
1882 for (i = PL_lasttoke; i > where; --i) {
1883 PL_nexttoke[i] = PL_nexttoke[i-1];
1887 if (PL_curforce < 0) /* in case of duplicate start_force() */
1888 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1889 PL_curforce = where;
1892 curmad('^', newSVpvs(""));
1893 CURMAD('_', PL_nextwhite);
1898 S_curmad(pTHX_ char slot, SV *sv)
1904 if (PL_curforce < 0)
1905 where = &PL_thismad;
1907 where = &PL_nexttoke[PL_curforce].next_mad;
1913 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1915 else if (PL_encoding) {
1916 sv_recode_to_utf8(sv, PL_encoding);
1921 /* keep a slot open for the head of the list? */
1922 if (slot != '_' && *where && (*where)->mad_key == '^') {
1923 (*where)->mad_key = slot;
1924 sv_free(MUTABLE_SV(((*where)->mad_val)));
1925 (*where)->mad_val = (void*)sv;
1928 addmad(newMADsv(slot, sv), where, 0);
1931 # define start_force(where) NOOP
1932 # define curmad(slot, sv) NOOP
1937 * When the lexer realizes it knows the next token (for instance,
1938 * it is reordering tokens for the parser) then it can call S_force_next
1939 * to know what token to return the next time the lexer is called. Caller
1940 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1941 * and possibly PL_expect to ensure the lexer handles the token correctly.
1945 S_force_next(pTHX_ I32 type)
1950 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1951 tokereport(type, &NEXTVAL_NEXTTOKE);
1955 if (PL_curforce < 0)
1956 start_force(PL_lasttoke);
1957 PL_nexttoke[PL_curforce].next_type = type;
1958 if (PL_lex_state != LEX_KNOWNEXT)
1959 PL_lex_defer = PL_lex_state;
1960 PL_lex_state = LEX_KNOWNEXT;
1961 PL_lex_expect = PL_expect;
1964 PL_nexttype[PL_nexttoke] = type;
1966 if (PL_lex_state != LEX_KNOWNEXT) {
1967 PL_lex_defer = PL_lex_state;
1968 PL_lex_expect = PL_expect;
1969 PL_lex_state = LEX_KNOWNEXT;
1977 int yyc = PL_parser->yychar;
1978 if (yyc != YYEMPTY) {
1981 NEXTVAL_NEXTTOKE = PL_parser->yylval;
1982 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
1983 PL_lex_allbrackets--;
1985 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
1986 } else if (yyc == '('/*)*/) {
1987 PL_lex_allbrackets--;
1992 PL_parser->yychar = YYEMPTY;
1997 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
2000 SV * const sv = newSVpvn_utf8(start, len,
2003 && !is_ascii_string((const U8*)start, len)
2004 && is_utf8_string((const U8*)start, len));
2010 * When the lexer knows the next thing is a word (for instance, it has
2011 * just seen -> and it knows that the next char is a word char, then
2012 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2016 * char *start : buffer position (must be within PL_linestr)
2017 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2018 * int check_keyword : if true, Perl checks to make sure the word isn't
2019 * a keyword (do this if the word is a label, e.g. goto FOO)
2020 * int allow_pack : if true, : characters will also be allowed (require,
2021 * use, etc. do this)
2022 * int allow_initial_tick : used by the "sub" lexer only.
2026 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
2032 PERL_ARGS_ASSERT_FORCE_WORD;
2034 start = SKIPSPACE1(start);
2036 if (isIDFIRST_lazy_if(s,UTF) ||
2037 (allow_pack && *s == ':') ||
2038 (allow_initial_tick && *s == '\'') )
2040 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2041 if (check_keyword && keyword(PL_tokenbuf, len, 0))
2043 start_force(PL_curforce);
2045 curmad('X', newSVpvn(start,s-start));
2046 if (token == METHOD) {
2051 PL_expect = XOPERATOR;
2055 curmad('g', newSVpvs( "forced" ));
2056 NEXTVAL_NEXTTOKE.opval
2057 = (OP*)newSVOP(OP_CONST,0,
2058 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2059 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2067 * Called when the lexer wants $foo *foo &foo etc, but the program
2068 * text only contains the "foo" portion. The first argument is a pointer
2069 * to the "foo", and the second argument is the type symbol to prefix.
2070 * Forces the next token to be a "WORD".
2071 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2075 S_force_ident(pTHX_ register const char *s, int kind)
2079 PERL_ARGS_ASSERT_FORCE_IDENT;
2082 const STRLEN len = strlen(s);
2083 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
2084 start_force(PL_curforce);
2085 NEXTVAL_NEXTTOKE.opval = o;
2088 o->op_private = OPpCONST_ENTERED;
2089 /* XXX see note in pp_entereval() for why we forgo typo
2090 warnings if the symbol must be introduced in an eval.
2092 gv_fetchpvn_flags(s, len,
2093 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2095 kind == '$' ? SVt_PV :
2096 kind == '@' ? SVt_PVAV :
2097 kind == '%' ? SVt_PVHV :
2105 Perl_str_to_version(pTHX_ SV *sv)
2110 const char *start = SvPV_const(sv,len);
2111 const char * const end = start + len;
2112 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2114 PERL_ARGS_ASSERT_STR_TO_VERSION;
2116 while (start < end) {
2120 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2125 retval += ((NV)n)/nshift;
2134 * Forces the next token to be a version number.
2135 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2136 * and if "guessing" is TRUE, then no new token is created (and the caller
2137 * must use an alternative parsing method).
2141 S_force_version(pTHX_ char *s, int guessing)
2147 I32 startoff = s - SvPVX(PL_linestr);
2150 PERL_ARGS_ASSERT_FORCE_VERSION;
2158 while (isDIGIT(*d) || *d == '_' || *d == '.')
2162 start_force(PL_curforce);
2163 curmad('X', newSVpvn(s,d-s));
2166 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2168 #ifdef USE_LOCALE_NUMERIC
2169 char *loc = setlocale(LC_NUMERIC, "C");
2171 s = scan_num(s, &pl_yylval);
2172 #ifdef USE_LOCALE_NUMERIC
2173 setlocale(LC_NUMERIC, loc);
2175 version = pl_yylval.opval;
2176 ver = cSVOPx(version)->op_sv;
2177 if (SvPOK(ver) && !SvNIOK(ver)) {
2178 SvUPGRADE(ver, SVt_PVNV);
2179 SvNV_set(ver, str_to_version(ver));
2180 SvNOK_on(ver); /* hint that it is a version */
2183 else if (guessing) {
2186 sv_free(PL_nextwhite); /* let next token collect whitespace */
2188 s = SvPVX(PL_linestr) + startoff;
2196 if (PL_madskills && !version) {
2197 sv_free(PL_nextwhite); /* let next token collect whitespace */
2199 s = SvPVX(PL_linestr) + startoff;
2202 /* NOTE: The parser sees the package name and the VERSION swapped */
2203 start_force(PL_curforce);
2204 NEXTVAL_NEXTTOKE.opval = version;
2211 * S_force_strict_version
2212 * Forces the next token to be a version number using strict syntax rules.
2216 S_force_strict_version(pTHX_ char *s)
2221 I32 startoff = s - SvPVX(PL_linestr);
2223 const char *errstr = NULL;
2225 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2227 while (isSPACE(*s)) /* leading whitespace */
2230 if (is_STRICT_VERSION(s,&errstr)) {
2232 s = (char *)scan_version(s, ver, 0);
2233 version = newSVOP(OP_CONST, 0, ver);
2235 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2236 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2240 yyerror(errstr); /* version required */
2245 if (PL_madskills && !version) {
2246 sv_free(PL_nextwhite); /* let next token collect whitespace */
2248 s = SvPVX(PL_linestr) + startoff;
2251 /* NOTE: The parser sees the package name and the VERSION swapped */
2252 start_force(PL_curforce);
2253 NEXTVAL_NEXTTOKE.opval = version;
2261 * Tokenize a quoted string passed in as an SV. It finds the next
2262 * chunk, up to end of string or a backslash. It may make a new
2263 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2268 S_tokeq(pTHX_ SV *sv)
2272 register char *send;
2277 PERL_ARGS_ASSERT_TOKEQ;
2282 s = SvPV_force(sv, len);
2283 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2286 /* This is relying on the SV being "well formed" with a trailing '\0' */
2287 while (s < send && !(*s == '\\' && s[1] == '\\'))
2292 if ( PL_hints & HINT_NEW_STRING ) {
2293 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2297 if (s + 1 < send && (s[1] == '\\'))
2298 s++; /* all that, just for this */
2303 SvCUR_set(sv, d - SvPVX_const(sv));
2305 if ( PL_hints & HINT_NEW_STRING )
2306 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2311 * Now come three functions related to double-quote context,
2312 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2313 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2314 * interact with PL_lex_state, and create fake ( ... ) argument lists
2315 * to handle functions and concatenation.
2316 * They assume that whoever calls them will be setting up a fake
2317 * join call, because each subthing puts a ',' after it. This lets
2320 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2322 * (I'm not sure whether the spurious commas at the end of lcfirst's
2323 * arguments and join's arguments are created or not).
2328 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2330 * Pattern matching will set PL_lex_op to the pattern-matching op to
2331 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2333 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2335 * Everything else becomes a FUNC.
2337 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2338 * had an OP_CONST or OP_READLINE). This just sets us up for a
2339 * call to S_sublex_push().
2343 S_sublex_start(pTHX)
2346 register const I32 op_type = pl_yylval.ival;
2348 if (op_type == OP_NULL) {
2349 pl_yylval.opval = PL_lex_op;
2353 if (op_type == OP_CONST || op_type == OP_READLINE) {
2354 SV *sv = tokeq(PL_lex_stuff);
2356 if (SvTYPE(sv) == SVt_PVIV) {
2357 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2359 const char * const p = SvPV_const(sv, len);
2360 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2364 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2365 PL_lex_stuff = NULL;
2366 /* Allow <FH> // "foo" */
2367 if (op_type == OP_READLINE)
2368 PL_expect = XTERMORDORDOR;
2371 else if (op_type == OP_BACKTICK && PL_lex_op) {
2372 /* readpipe() vas overriden */
2373 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2374 pl_yylval.opval = PL_lex_op;
2376 PL_lex_stuff = NULL;
2380 PL_sublex_info.super_state = PL_lex_state;
2381 PL_sublex_info.sub_inwhat = (U16)op_type;
2382 PL_sublex_info.sub_op = PL_lex_op;
2383 PL_lex_state = LEX_INTERPPUSH;
2387 pl_yylval.opval = PL_lex_op;
2397 * Create a new scope to save the lexing state. The scope will be
2398 * ended in S_sublex_done. Returns a '(', starting the function arguments
2399 * to the uc, lc, etc. found before.
2400 * Sets PL_lex_state to LEX_INTERPCONCAT.
2409 PL_lex_state = PL_sublex_info.super_state;
2410 SAVEBOOL(PL_lex_dojoin);
2411 SAVEI32(PL_lex_brackets);
2412 SAVEI32(PL_lex_allbrackets);
2413 SAVEI8(PL_lex_fakeeof);
2414 SAVEI32(PL_lex_casemods);
2415 SAVEI32(PL_lex_starts);
2416 SAVEI8(PL_lex_state);
2417 SAVEVPTR(PL_lex_inpat);
2418 SAVEI16(PL_lex_inwhat);
2419 SAVECOPLINE(PL_curcop);
2420 SAVEPPTR(PL_bufptr);
2421 SAVEPPTR(PL_bufend);
2422 SAVEPPTR(PL_oldbufptr);
2423 SAVEPPTR(PL_oldoldbufptr);
2424 SAVEPPTR(PL_last_lop);
2425 SAVEPPTR(PL_last_uni);
2426 SAVEPPTR(PL_linestart);
2427 SAVESPTR(PL_linestr);
2428 SAVEGENERICPV(PL_lex_brackstack);
2429 SAVEGENERICPV(PL_lex_casestack);
2431 PL_linestr = PL_lex_stuff;
2432 PL_lex_stuff = NULL;
2434 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2435 = SvPVX(PL_linestr);
2436 PL_bufend += SvCUR(PL_linestr);
2437 PL_last_lop = PL_last_uni = NULL;
2438 SAVEFREESV(PL_linestr);
2440 PL_lex_dojoin = FALSE;
2441 PL_lex_brackets = 0;
2442 PL_lex_allbrackets = 0;
2443 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2444 Newx(PL_lex_brackstack, 120, char);
2445 Newx(PL_lex_casestack, 12, char);
2446 PL_lex_casemods = 0;
2447 *PL_lex_casestack = '\0';
2449 PL_lex_state = LEX_INTERPCONCAT;
2450 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2452 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2453 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2454 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2455 PL_lex_inpat = PL_sublex_info.sub_op;
2457 PL_lex_inpat = NULL;
2464 * Restores lexer state after a S_sublex_push.
2471 if (!PL_lex_starts++) {
2472 SV * const sv = newSVpvs("");
2473 if (SvUTF8(PL_linestr))
2475 PL_expect = XOPERATOR;
2476 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2480 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2481 PL_lex_state = LEX_INTERPCASEMOD;
2485 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2486 assert(PL_lex_inwhat != OP_TRANSR);
2487 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2488 PL_linestr = PL_lex_repl;
2490 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2491 PL_bufend += SvCUR(PL_linestr);
2492 PL_last_lop = PL_last_uni = NULL;
2493 SAVEFREESV(PL_linestr);
2494 PL_lex_dojoin = FALSE;
2495 PL_lex_brackets = 0;
2496 PL_lex_allbrackets = 0;
2497 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2498 PL_lex_casemods = 0;
2499 *PL_lex_casestack = '\0';
2501 if (SvEVALED(PL_lex_repl)) {
2502 PL_lex_state = LEX_INTERPNORMAL;
2504 /* we don't clear PL_lex_repl here, so that we can check later
2505 whether this is an evalled subst; that means we rely on the
2506 logic to ensure sublex_done() is called again only via the
2507 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2510 PL_lex_state = LEX_INTERPCONCAT;
2520 PL_endwhite = newSVpvs("");
2521 sv_catsv(PL_endwhite, PL_thiswhite);
2525 sv_setpvs(PL_thistoken,"");
2527 PL_realtokenstart = -1;
2531 PL_bufend = SvPVX(PL_linestr);
2532 PL_bufend += SvCUR(PL_linestr);
2533 PL_expect = XOPERATOR;
2534 PL_sublex_info.sub_inwhat = 0;
2542 Extracts a pattern, double-quoted string, or transliteration. This
2545 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2546 processing a pattern (PL_lex_inpat is true), a transliteration
2547 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2549 Returns a pointer to the character scanned up to. If this is
2550 advanced from the start pointer supplied (i.e. if anything was
2551 successfully parsed), will leave an OP for the substring scanned
2552 in pl_yylval. Caller must intuit reason for not parsing further
2553 by looking at the next characters herself.
2557 constants: \N{NAME} only
2558 case and quoting: \U \Q \E
2559 stops on @ and $, but not for $ as tail anchor
2561 In transliterations:
2562 characters are VERY literal, except for - not at the start or end
2563 of the string, which indicates a range. If the range is in bytes,
2564 scan_const expands the range to the full set of intermediate
2565 characters. If the range is in utf8, the hyphen is replaced with
2566 a certain range mark which will be handled by pmtrans() in op.c.
2568 In double-quoted strings:
2570 double-quoted style: \r and \n
2571 constants: \x31, etc.
2572 deprecated backrefs: \1 (in substitution replacements)
2573 case and quoting: \U \Q \E
2576 scan_const does *not* construct ops to handle interpolated strings.
2577 It stops processing as soon as it finds an embedded $ or @ variable
2578 and leaves it to the caller to work out what's going on.
2580 embedded arrays (whether in pattern or not) could be:
2581 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2583 $ in double-quoted strings must be the symbol of an embedded scalar.
2585 $ in pattern could be $foo or could be tail anchor. Assumption:
2586 it's a tail anchor if $ is the last thing in the string, or if it's
2587 followed by one of "()| \r\n\t"
2589 \1 (backreferences) are turned into $1
2591 The structure of the code is
2592 while (there's a character to process) {
2593 handle transliteration ranges
2594 skip regexp comments /(?#comment)/ and codes /(?{code})/
2595 skip #-initiated comments in //x patterns
2596 check for embedded arrays
2597 check for embedded scalars
2599 deprecate \1 in substitution replacements
2600 handle string-changing backslashes \l \U \Q \E, etc.
2601 switch (what was escaped) {
2602 handle \- in a transliteration (becomes a literal -)
2603 if a pattern and not \N{, go treat as regular character
2604 handle \132 (octal characters)
2605 handle \x15 and \x{1234} (hex characters)
2606 handle \N{name} (named characters, also \N{3,5} in a pattern)
2607 handle \cV (control characters)
2608 handle printf-style backslashes (\f, \r, \n, etc)
2611 } (end if backslash)
2612 handle regular character
2613 } (end while character to read)
2618 S_scan_const(pTHX_ char *start)
2621 register char *send = PL_bufend; /* end of the constant */
2622 SV *sv = newSV(send - start); /* sv for the constant. See
2623 note below on sizing. */
2624 register char *s = start; /* start of the constant */
2625 register char *d = SvPVX(sv); /* destination for copies */
2626 bool dorange = FALSE; /* are we in a translit range? */
2627 bool didrange = FALSE; /* did we just finish a range? */
2628 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2629 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
2630 to be UTF8? But, this can
2631 show as true when the source
2632 isn't utf8, as for example
2633 when it is entirely composed
2636 /* Note on sizing: The scanned constant is placed into sv, which is
2637 * initialized by newSV() assuming one byte of output for every byte of
2638 * input. This routine expects newSV() to allocate an extra byte for a
2639 * trailing NUL, which this routine will append if it gets to the end of
2640 * the input. There may be more bytes of input than output (eg., \N{LATIN
2641 * CAPITAL LETTER A}), or more output than input if the constant ends up
2642 * recoded to utf8, but each time a construct is found that might increase
2643 * the needed size, SvGROW() is called. Its size parameter each time is
2644 * based on the best guess estimate at the time, namely the length used so
2645 * far, plus the length the current construct will occupy, plus room for
2646 * the trailing NUL, plus one byte for every input byte still unscanned */
2650 UV literal_endpoint = 0;
2651 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2654 PERL_ARGS_ASSERT_SCAN_CONST;
2656 assert(PL_lex_inwhat != OP_TRANSR);
2657 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2658 /* If we are doing a trans and we know we want UTF8 set expectation */
2659 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2660 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2664 while (s < send || dorange) {
2666 /* get transliterations out of the way (they're most literal) */
2667 if (PL_lex_inwhat == OP_TRANS) {
2668 /* expand a range A-Z to the full set of characters. AIE! */
2670 I32 i; /* current expanded character */
2671 I32 min; /* first character in range */
2672 I32 max; /* last character in range */
2683 char * const c = (char*)utf8_hop((U8*)d, -1);
2687 *c = (char)UTF_TO_NATIVE(0xff);
2688 /* mark the range as done, and continue */
2694 i = d - SvPVX_const(sv); /* remember current offset */
2697 SvLEN(sv) + (has_utf8 ?
2698 (512 - UTF_CONTINUATION_MARK +
2701 /* How many two-byte within 0..255: 128 in UTF-8,
2702 * 96 in UTF-8-mod. */
2704 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2706 d = SvPVX(sv) + i; /* refresh d after realloc */
2710 for (j = 0; j <= 1; j++) {
2711 char * const c = (char*)utf8_hop((U8*)d, -1);
2712 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2718 max = (U8)0xff; /* only to \xff */
2719 uvmax = uv; /* \x{100} to uvmax */
2721 d = c; /* eat endpoint chars */
2726 d -= 2; /* eat the first char and the - */
2727 min = (U8)*d; /* first char in range */
2728 max = (U8)d[1]; /* last char in range */
2735 "Invalid range \"%c-%c\" in transliteration operator",
2736 (char)min, (char)max);
2740 if (literal_endpoint == 2 &&
2741 ((isLOWER(min) && isLOWER(max)) ||
2742 (isUPPER(min) && isUPPER(max)))) {
2744 for (i = min; i <= max; i++)
2746 *d++ = NATIVE_TO_NEED(has_utf8,i);
2748 for (i = min; i <= max; i++)
2750 *d++ = NATIVE_TO_NEED(has_utf8,i);
2755 for (i = min; i <= max; i++)
2758 const U8 ch = (U8)NATIVE_TO_UTF(i);
2759 if (UNI_IS_INVARIANT(ch))
2762 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2763 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2772 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2774 *d++ = (char)UTF_TO_NATIVE(0xff);
2776 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2780 /* mark the range as done, and continue */
2784 literal_endpoint = 0;
2789 /* range begins (ignore - as first or last char) */
2790 else if (*s == '-' && s+1 < send && s != start) {
2792 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2799 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2809 literal_endpoint = 0;
2810 native_range = TRUE;
2815 /* if we get here, we're not doing a transliteration */
2817 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2818 except for the last char, which will be done separately. */
2819 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2821 while (s+1 < send && *s != ')')
2822 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2824 else if (s[2] == '{' /* This should match regcomp.c */
2825 || (s[2] == '?' && s[3] == '{'))
2828 char *regparse = s + (s[2] == '{' ? 3 : 4);
2831 while (count && (c = *regparse)) {
2832 if (c == '\\' && regparse[1])
2840 if (*regparse != ')')
2841 regparse--; /* Leave one char for continuation. */
2842 while (s < regparse)
2843 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2847 /* likewise skip #-initiated comments in //x patterns */
2848 else if (*s == '#' && PL_lex_inpat &&
2849 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
2850 while (s+1 < send && *s != '\n')
2851 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2854 /* check for embedded arrays
2855 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2857 else if (*s == '@' && s[1]) {
2858 if (isALNUM_lazy_if(s+1,UTF))
2860 if (strchr(":'{$", s[1]))
2862 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2863 break; /* in regexp, neither @+ nor @- are interpolated */
2866 /* check for embedded scalars. only stop if we're sure it's a
2869 else if (*s == '$') {
2870 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2872 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2874 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2875 "Possible unintended interpolation of $\\ in regex");
2877 break; /* in regexp, $ might be tail anchor */
2881 /* End of else if chain - OP_TRANS rejoin rest */
2884 if (*s == '\\' && s+1 < send) {
2885 char* e; /* Can be used for ending '}', etc. */
2889 /* warn on \1 - \9 in substitution replacements, but note that \11
2890 * is an octal; and \19 is \1 followed by '9' */
2891 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2892 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2894 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2899 /* string-change backslash escapes */
2900 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2904 /* In a pattern, process \N, but skip any other backslash escapes.
2905 * This is because we don't want to translate an escape sequence
2906 * into a meta symbol and have the regex compiler use the meta
2907 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2908 * in spite of this, we do have to process \N here while the proper
2909 * charnames handler is in scope. See bugs #56444 and #62056.
2910 * There is a complication because \N in a pattern may also stand
2911 * for 'match a non-nl', and not mean a charname, in which case its
2912 * processing should be deferred to the regex compiler. To be a
2913 * charname it must be followed immediately by a '{', and not look
2914 * like \N followed by a curly quantifier, i.e., not something like
2915 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2917 else if (PL_lex_inpat
2920 || regcurly(s + 1)))
2922 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2923 goto default_action;
2928 /* quoted - in transliterations */
2930 if (PL_lex_inwhat == OP_TRANS) {
2937 if ((isALPHA(*s) || isDIGIT(*s)))
2938 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2939 "Unrecognized escape \\%c passed through",
2941 /* default action is to copy the quoted character */
2942 goto default_action;
2945 /* eg. \132 indicates the octal constant 0132 */
2946 case '0': case '1': case '2': case '3':
2947 case '4': case '5': case '6': case '7':
2951 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2954 goto NUM_ESCAPE_INSERT;
2956 /* eg. \o{24} indicates the octal constant \024 */
2962 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
2968 goto NUM_ESCAPE_INSERT;
2971 /* eg. \x24 indicates the hex constant 0x24 */
2975 char* const e = strchr(s, '}');
2976 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2977 PERL_SCAN_DISALLOW_PREFIX;
2982 yyerror("Missing right brace on \\x{}");
2986 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2992 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
2993 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2999 /* Insert oct or hex escaped character. There will always be
3000 * enough room in sv since such escapes will be longer than any
3001 * UTF-8 sequence they can end up as, except if they force us
3002 * to recode the rest of the string into utf8 */
3004 /* Here uv is the ordinal of the next character being added in
3005 * unicode (converted from native). */
3006 if (!UNI_IS_INVARIANT(uv)) {
3007 if (!has_utf8 && uv > 255) {
3008 /* Might need to recode whatever we have accumulated so
3009 * far if it contains any chars variant in utf8 or
3012 SvCUR_set(sv, d - SvPVX_const(sv));
3015 /* See Note on sizing above. */
3016 sv_utf8_upgrade_flags_grow(sv,
3017 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3018 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3019 d = SvPVX(sv) + SvCUR(sv);
3024 d = (char*)uvuni_to_utf8((U8*)d, uv);
3025 if (PL_lex_inwhat == OP_TRANS &&
3026 PL_sublex_info.sub_op) {
3027 PL_sublex_info.sub_op->op_private |=
3028 (PL_lex_repl ? OPpTRANS_FROM_UTF
3032 if (uv > 255 && !dorange)
3033 native_range = FALSE;
3046 /* In a non-pattern \N must be a named character, like \N{LATIN
3047 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3048 * mean to match a non-newline. For non-patterns, named
3049 * characters are converted to their string equivalents. In
3050 * patterns, named characters are not converted to their
3051 * ultimate forms for the same reasons that other escapes
3052 * aren't. Instead, they are converted to the \N{U+...} form
3053 * to get the value from the charnames that is in effect right
3054 * now, while preserving the fact that it was a named character
3055 * so that the regex compiler knows this */
3057 /* This section of code doesn't generally use the
3058 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3059 * a close examination of this macro and determined it is a
3060 * no-op except on utfebcdic variant characters. Every
3061 * character generated by this that would normally need to be
3062 * enclosed by this macro is invariant, so the macro is not
3063 * needed, and would complicate use of copy(). XXX There are
3064 * other parts of this file where the macro is used
3065 * inconsistently, but are saved by it being a no-op */
3067 /* The structure of this section of code (besides checking for
3068 * errors and upgrading to utf8) is:
3069 * Further disambiguate between the two meanings of \N, and if
3070 * not a charname, go process it elsewhere
3071 * If of form \N{U+...}, pass it through if a pattern;
3072 * otherwise convert to utf8
3073 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3074 * pattern; otherwise convert to utf8 */
3076 /* Here, s points to the 'N'; the test below is guaranteed to
3077 * succeed if we are being called on a pattern as we already
3078 * know from a test above that the next character is a '{'.
3079 * On a non-pattern \N must mean 'named sequence, which
3080 * requires braces */
3083 yyerror("Missing braces on \\N{}");
3088 /* If there is no matching '}', it is an error. */
3089 if (! (e = strchr(s, '}'))) {
3090 if (! PL_lex_inpat) {
3091 yyerror("Missing right brace on \\N{}");
3093 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3098 /* Here it looks like a named character */
3102 /* XXX This block is temporary code. \N{} implies that the
3103 * pattern is to have Unicode semantics, and therefore
3104 * currently has to be encoded in utf8. By putting it in
3105 * utf8 now, we save a whole pass in the regular expression
3106 * compiler. Once that code is changed so Unicode
3107 * semantics doesn't necessarily have to be in utf8, this
3108 * block should be removed. However, the code that parses
3109 * the output of this would have to be changed to not
3110 * necessarily expect utf8 */
3112 SvCUR_set(sv, d - SvPVX_const(sv));
3115 /* See Note on sizing above. */
3116 sv_utf8_upgrade_flags_grow(sv,
3117 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3118 /* 5 = '\N{' + cur char + NUL */
3119 (STRLEN)(send - s) + 5);
3120 d = SvPVX(sv) + SvCUR(sv);
3125 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3126 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3127 | PERL_SCAN_DISALLOW_PREFIX;
3130 /* For \N{U+...}, the '...' is a unicode value even on
3131 * EBCDIC machines */
3132 s += 2; /* Skip to next char after the 'U+' */
3134 uv = grok_hex(s, &len, &flags, NULL);
3135 if (len == 0 || len != (STRLEN)(e - s)) {
3136 yyerror("Invalid hexadecimal number in \\N{U+...}");
3143 /* On non-EBCDIC platforms, pass through to the regex
3144 * compiler unchanged. The reason we evaluated the
3145 * number above is to make sure there wasn't a syntax
3146 * error. But on EBCDIC we convert to native so
3147 * downstream code can continue to assume it's native
3149 s -= 5; /* Include the '\N{U+' */
3151 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3154 (unsigned int) UNI_TO_NATIVE(uv));
3156 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3160 else { /* Not a pattern: convert the hex to string */
3162 /* If destination is not in utf8, unconditionally
3163 * recode it to be so. This is because \N{} implies
3164 * Unicode semantics, and scalars have to be in utf8
3165 * to guarantee those semantics */
3167 SvCUR_set(sv, d - SvPVX_const(sv));
3170 /* See Note on sizing above. */
3171 sv_utf8_upgrade_flags_grow(
3173 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3174 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3175 d = SvPVX(sv) + SvCUR(sv);
3179 /* Add the string to the output */
3180 if (UNI_IS_INVARIANT(uv)) {
3183 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3186 else { /* Here is \N{NAME} but not \N{U+...}. */
3188 SV *res; /* result from charnames */
3189 const char *str; /* the string in 'res' */
3190 STRLEN len; /* its length */
3192 /* Get the value for NAME */
3193 res = newSVpvn(s, e - s);
3194 res = new_constant( NULL, 0, "charnames",
3195 /* includes all of: \N{...} */
3196 res, NULL, s - 3, e - s + 4 );
3198 /* Most likely res will be in utf8 already since the
3199 * standard charnames uses pack U, but a custom translator
3200 * can leave it otherwise, so make sure. XXX This can be
3201 * revisited to not have charnames use utf8 for characters
3202 * that don't need it when regexes don't have to be in utf8
3203 * for Unicode semantics. If doing so, remember EBCDIC */
3204 sv_utf8_upgrade(res);
3205 str = SvPV_const(res, len);
3207 /* Don't accept malformed input */
3208 if (! is_utf8_string((U8 *) str, len)) {
3209 yyerror("Malformed UTF-8 returned by \\N");
3211 else if (PL_lex_inpat) {
3213 if (! len) { /* The name resolved to an empty string */
3214 Copy("\\N{}", d, 4, char);
3218 /* In order to not lose information for the regex
3219 * compiler, pass the result in the specially made
3220 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3221 * the code points in hex of each character
3222 * returned by charnames */
3224 const char *str_end = str + len;
3225 STRLEN char_length; /* cur char's byte length */
3226 STRLEN output_length; /* and the number of bytes
3227 after this is translated
3229 const STRLEN off = d - SvPVX_const(sv);
3231 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3232 * max('U+', '.'); and 1 for NUL */
3233 char hex_string[2 * UTF8_MAXBYTES + 5];
3235 /* Get the first character of the result. */
3236 U32 uv = utf8n_to_uvuni((U8 *) str,
3241 /* The call to is_utf8_string() above hopefully
3242 * guarantees that there won't be an error. But
3243 * it's easy here to make sure. The function just
3244 * above warns and returns 0 if invalid utf8, but
3245 * it can also return 0 if the input is validly a
3246 * NUL. Disambiguate */
3247 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3248 uv = UNICODE_REPLACEMENT;
3251 /* Convert first code point to hex, including the
3252 * boiler plate before it. For all these, we
3253 * convert to native format so that downstream code
3254 * can continue to assume the input is native */
3256 my_snprintf(hex_string, sizeof(hex_string),
3258 (unsigned int) UNI_TO_NATIVE(uv));
3260 /* Make sure there is enough space to hold it */
3261 d = off + SvGROW(sv, off
3263 + (STRLEN)(send - e)
3264 + 2); /* '}' + NUL */
3266 Copy(hex_string, d, output_length, char);
3269 /* For each subsequent character, append dot and
3270 * its ordinal in hex */
3271 while ((str += char_length) < str_end) {
3272 const STRLEN off = d - SvPVX_const(sv);
3273 U32 uv = utf8n_to_uvuni((U8 *) str,
3277 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3278 uv = UNICODE_REPLACEMENT;
3282 my_snprintf(hex_string, sizeof(hex_string),
3284 (unsigned int) UNI_TO_NATIVE(uv));
3286 d = off + SvGROW(sv, off
3288 + (STRLEN)(send - e)
3289 + 2); /* '}' + NUL */
3290 Copy(hex_string, d, output_length, char);
3294 *d++ = '}'; /* Done. Add the trailing brace */
3297 else { /* Here, not in a pattern. Convert the name to a
3300 /* If destination is not in utf8, unconditionally
3301 * recode it to be so. This is because \N{} implies
3302 * Unicode semantics, and scalars have to be in utf8
3303 * to guarantee those semantics */
3305 SvCUR_set(sv, d - SvPVX_const(sv));
3308 /* See Note on sizing above. */
3309 sv_utf8_upgrade_flags_grow(sv,
3310 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3311 len + (STRLEN)(send - s) + 1);
3312 d = SvPVX(sv) + SvCUR(sv);
3314 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3316 /* See Note on sizing above. (NOTE: SvCUR() is not
3317 * set correctly here). */
3318 const STRLEN off = d - SvPVX_const(sv);
3319 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3321 Copy(str, d, len, char);
3326 /* Deprecate non-approved name syntax */
3327 if (ckWARN_d(WARN_DEPRECATED)) {
3328 bool problematic = FALSE;
3331 /* For non-ut8 input, look to see that the first
3332 * character is an alpha, then loop through the rest
3333 * checking that each is a continuation */
3335 if (! isALPHAU(*i)) problematic = TRUE;
3336 else for (i = s + 1; i < e; i++) {
3337 if (isCHARNAME_CONT(*i)) continue;
3343 /* Similarly for utf8. For invariants can check
3344 * directly. We accept anything above the latin1
3345 * range because it is immaterial to Perl if it is
3346 * correct or not, and is expensive to check. But
3347 * it is fairly easy in the latin1 range to convert
3348 * the variants into a single character and check
3350 if (UTF8_IS_INVARIANT(*i)) {
3351 if (! isALPHAU(*i)) problematic = TRUE;
3352 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3353 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*i,
3359 if (! problematic) for (i = s + UTF8SKIP(s);
3363 if (UTF8_IS_INVARIANT(*i)) {
3364 if (isCHARNAME_CONT(*i)) continue;
3365 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3367 } else if (isCHARNAME_CONT(
3369 TWO_BYTE_UTF8_TO_UNI(*i, *(i+1)))))
3378 /* The e-i passed to the final %.*s makes sure that
3379 * should the trailing NUL be missing that this
3380 * print won't run off the end of the string */
3381 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3382 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s",
3383 (int)(i - s + 1), s, (int)(e - i), i + 1);
3386 } /* End \N{NAME} */
3389 native_range = FALSE; /* \N{} is defined to be Unicode */
3391 s = e + 1; /* Point to just after the '}' */
3394 /* \c is a control character */
3398 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3401 yyerror("Missing control char name in \\c");
3405 /* printf-style backslashes, formfeeds, newlines, etc */
3407 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3410 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3413 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3416 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3419 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3422 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3425 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3431 } /* end if (backslash) */
3438 /* If we started with encoded form, or already know we want it,
3439 then encode the next character */
3440 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3444 /* One might think that it is wasted effort in the case of the
3445 * source being utf8 (this_utf8 == TRUE) to take the next character
3446 * in the source, convert it to an unsigned value, and then convert
3447 * it back again. But the source has not been validated here. The
3448 * routine that does the conversion checks for errors like
3451 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3452 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3454 SvCUR_set(sv, d - SvPVX_const(sv));
3457 /* See Note on sizing above. */
3458 sv_utf8_upgrade_flags_grow(sv,
3459 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3460 need + (STRLEN)(send - s) + 1);
3461 d = SvPVX(sv) + SvCUR(sv);
3463 } else if (need > len) {
3464 /* encoded value larger than old, may need extra space (NOTE:
3465 * SvCUR() is not set correctly here). See Note on sizing
3467 const STRLEN off = d - SvPVX_const(sv);
3468 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3472 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3474 if (uv > 255 && !dorange)
3475 native_range = FALSE;
3479 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3481 } /* while loop to process each character */
3483 /* terminate the string and set up the sv */
3485 SvCUR_set(sv, d - SvPVX_const(sv));
3486 if (SvCUR(sv) >= SvLEN(sv))
3487 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3490 if (PL_encoding && !has_utf8) {
3491 sv_recode_to_utf8(sv, PL_encoding);
3497 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3498 PL_sublex_info.sub_op->op_private |=
3499 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3503 /* shrink the sv if we allocated more than we used */
3504 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3505 SvPV_shrink_to_cur(sv);
3508 /* return the substring (via pl_yylval) only if we parsed anything */
3509 if (s > PL_bufptr) {
3510 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3511 const char *const key = PL_lex_inpat ? "qr" : "q";
3512 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3516 if (PL_lex_inwhat == OP_TRANS) {
3519 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3527 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3530 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3537 * Returns TRUE if there's more to the expression (e.g., a subscript),
3540 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3542 * ->[ and ->{ return TRUE
3543 * { and [ outside a pattern are always subscripts, so return TRUE
3544 * if we're outside a pattern and it's not { or [, then return FALSE
3545 * if we're in a pattern and the first char is a {
3546 * {4,5} (any digits around the comma) returns FALSE
3547 * if we're in a pattern and the first char is a [
3549 * [SOMETHING] has a funky algorithm to decide whether it's a
3550 * character class or not. It has to deal with things like
3551 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3552 * anything else returns TRUE
3555 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3558 S_intuit_more(pTHX_ register char *s)
3562 PERL_ARGS_ASSERT_INTUIT_MORE;
3564 if (PL_lex_brackets)
3566 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3568 if (*s != '{' && *s != '[')
3573 /* In a pattern, so maybe we have {n,m}. */
3581 /* On the other hand, maybe we have a character class */
3584 if (*s == ']' || *s == '^')
3587 /* this is terrifying, and it works */
3588 int weight = 2; /* let's weigh the evidence */
3590 unsigned char un_char = 255, last_un_char;
3591 const char * const send = strchr(s,']');
3592 char tmpbuf[sizeof PL_tokenbuf * 4];
3594 if (!send) /* has to be an expression */
3597 Zero(seen,256,char);
3600 else if (isDIGIT(*s)) {
3602 if (isDIGIT(s[1]) && s[2] == ']')
3608 for (; s < send; s++) {
3609 last_un_char = un_char;
3610 un_char = (unsigned char)*s;
3615 weight -= seen[un_char] * 10;
3616 if (isALNUM_lazy_if(s+1,UTF)) {
3618 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3619 len = (int)strlen(tmpbuf);
3620 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
3625 else if (*s == '$' && s[1] &&
3626 strchr("[#!%*<>()-=",s[1])) {
3627 if (/*{*/ strchr("])} =",s[2]))
3636 if (strchr("wds]",s[1]))
3638 else if (seen[(U8)'\''] || seen[(U8)'"'])
3640 else if (strchr("rnftbxcav",s[1]))
3642 else if (isDIGIT(s[1])) {
3644 while (s[1] && isDIGIT(s[1]))
3654 if (strchr("aA01! ",last_un_char))
3656 if (strchr("zZ79~",s[1]))
3658 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3659 weight -= 5; /* cope with negative subscript */
3662 if (!isALNUM(last_un_char)
3663 && !(last_un_char == '$' || last_un_char == '@'
3664 || last_un_char == '&')
3665 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3670 if (keyword(tmpbuf, d - tmpbuf, 0))
3673 if (un_char == last_un_char + 1)
3675 weight -= seen[un_char];
3680 if (weight >= 0) /* probably a character class */
3690 * Does all the checking to disambiguate
3692 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3693 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3695 * First argument is the stuff after the first token, e.g. "bar".
3697 * Not a method if bar is a filehandle.
3698 * Not a method if foo is a subroutine prototyped to take a filehandle.
3699 * Not a method if it's really "Foo $bar"
3700 * Method if it's "foo $bar"
3701 * Not a method if it's really "print foo $bar"
3702 * Method if it's really "foo package::" (interpreted as package->foo)
3703 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3704 * Not a method if bar is a filehandle or package, but is quoted with
3709 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3712 char *s = start + (*start == '$');
3713 char tmpbuf[sizeof PL_tokenbuf];
3720 PERL_ARGS_ASSERT_INTUIT_METHOD;
3723 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3727 const char *proto = SvPVX_const(cv);
3738 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3739 /* start is the beginning of the possible filehandle/object,
3740 * and s is the end of it
3741 * tmpbuf is a copy of it
3744 if (*start == '$') {
3745 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3746 isUPPER(*PL_tokenbuf))
3749 len = start - SvPVX(PL_linestr);
3753 start = SvPVX(PL_linestr) + len;
3757 return *s == '(' ? FUNCMETH : METHOD;
3759 if (!keyword(tmpbuf, len, 0)) {
3760 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3764 soff = s - SvPVX(PL_linestr);
3768 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
3769 if (indirgv && GvCVu(indirgv))
3771 /* filehandle or package name makes it a method */
3772 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
3774 soff = s - SvPVX(PL_linestr);
3777 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3778 return 0; /* no assumptions -- "=>" quotes bareword */
3780 start_force(PL_curforce);
3781 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3782 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3783 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3785 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3790 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3792 return *s == '(' ? FUNCMETH : METHOD;
3798 /* Encoded script support. filter_add() effectively inserts a
3799 * 'pre-processing' function into the current source input stream.
3800 * Note that the filter function only applies to the current source file
3801 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3803 * The datasv parameter (which may be NULL) can be used to pass
3804 * private data to this instance of the filter. The filter function
3805 * can recover the SV using the FILTER_DATA macro and use it to
3806 * store private buffers and state information.
3808 * The supplied datasv parameter is upgraded to a PVIO type
3809 * and the IoDIRP/IoANY field is used to store the function pointer,
3810 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3811 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3812 * private use must be set using malloc'd pointers.
3816 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3825 if (!PL_rsfp_filters)
3826 PL_rsfp_filters = newAV();
3829 SvUPGRADE(datasv, SVt_PVIO);
3830 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3831 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3832 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3833 FPTR2DPTR(void *, IoANY(datasv)),
3834 SvPV_nolen(datasv)));
3835 av_unshift(PL_rsfp_filters, 1);
3836 av_store(PL_rsfp_filters, 0, datasv) ;
3841 /* Delete most recently added instance of this filter function. */
3843 Perl_filter_del(pTHX_ filter_t funcp)
3848 PERL_ARGS_ASSERT_FILTER_DEL;
3851 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3852 FPTR2DPTR(void*, funcp)));
3854 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3856 /* if filter is on top of stack (usual case) just pop it off */
3857 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3858 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3859 sv_free(av_pop(PL_rsfp_filters));
3863 /* we need to search for the correct entry and clear it */
3864 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3868 /* Invoke the idxth filter function for the current rsfp. */
3869 /* maxlen 0 = read one text line */
3871 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3876 /* This API is bad. It should have been using unsigned int for maxlen.
3877 Not sure if we want to change the API, but if not we should sanity
3878 check the value here. */
3879 const unsigned int correct_length
3888 PERL_ARGS_ASSERT_FILTER_READ;
3890 if (!PL_parser || !PL_rsfp_filters)
3892 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3893 /* Provide a default input filter to make life easy. */
3894 /* Note that we append to the line. This is handy. */
3895 DEBUG_P(PerlIO_printf(Perl_debug_log,
3896 "filter_read %d: from rsfp\n", idx));
3897 if (correct_length) {
3900 const int old_len = SvCUR(buf_sv);
3902 /* ensure buf_sv is large enough */
3903 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3904 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3905 correct_length)) <= 0) {
3906 if (PerlIO_error(PL_rsfp))
3907 return -1; /* error */
3909 return 0 ; /* end of file */
3911 SvCUR_set(buf_sv, old_len + len) ;
3912 SvPVX(buf_sv)[old_len + len] = '\0';
3915 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3916 if (PerlIO_error(PL_rsfp))
3917 return -1; /* error */
3919 return 0 ; /* end of file */
3922 return SvCUR(buf_sv);
3924 /* Skip this filter slot if filter has been deleted */
3925 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3926 DEBUG_P(PerlIO_printf(Perl_debug_log,
3927 "filter_read %d: skipped (filter deleted)\n",
3929 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
3931 /* Get function pointer hidden within datasv */
3932 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
3933 DEBUG_P(PerlIO_printf(Perl_debug_log,
3934 "filter_read %d: via function %p (%s)\n",
3935 idx, (void*)datasv, SvPV_nolen_const(datasv)));
3936 /* Call function. The function is expected to */
3937 /* call "FILTER_READ(idx+1, buf_sv)" first. */
3938 /* Return: <0:error, =0:eof, >0:not eof */
3939 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
3943 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
3947 PERL_ARGS_ASSERT_FILTER_GETS;
3949 #ifdef PERL_CR_FILTER
3950 if (!PL_rsfp_filters) {
3951 filter_add(S_cr_textfilter,NULL);
3954 if (PL_rsfp_filters) {
3956 SvCUR_set(sv, 0); /* start with empty line */
3957 if (FILTER_READ(0, sv, 0) > 0)
3958 return ( SvPVX(sv) ) ;
3963 return (sv_gets(sv, PL_rsfp, append));
3967 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3972 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3974 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3978 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3979 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3981 return GvHV(gv); /* Foo:: */
3984 /* use constant CLASS => 'MyClass' */
3985 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3986 if (gv && GvCV(gv)) {
3987 SV * const sv = cv_const_sv(GvCV(gv));
3989 pkgname = SvPV_const(sv, len);
3992 return gv_stashpvn(pkgname, len, 0);
3996 * S_readpipe_override
3997 * Check whether readpipe() is overridden, and generates the appropriate
3998 * optree, provided sublex_start() is called afterwards.
4001 S_readpipe_override(pTHX)
4004 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4005 pl_yylval.ival = OP_BACKTICK;
4007 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4009 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4010 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4011 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4013 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4014 op_append_elem(OP_LIST,
4015 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4016 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4023 * The intent of this yylex wrapper is to minimize the changes to the
4024 * tokener when we aren't interested in collecting madprops. It remains
4025 * to be seen how successful this strategy will be...
4032 char *s = PL_bufptr;
4034 /* make sure PL_thiswhite is initialized */
4038 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
4039 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4040 return S_pending_ident(aTHX);
4042 /* previous token ate up our whitespace? */
4043 if (!PL_lasttoke && PL_nextwhite) {
4044 PL_thiswhite = PL_nextwhite;
4048 /* isolate the token, and figure out where it is without whitespace */
4049 PL_realtokenstart = -1;
4053 assert(PL_curforce < 0);
4055 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4056 if (!PL_thistoken) {
4057 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4058 PL_thistoken = newSVpvs("");
4060 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4061 PL_thistoken = newSVpvn(tstart, s - tstart);
4064 if (PL_thismad) /* install head */
4065 CURMAD('X', PL_thistoken);
4068 /* last whitespace of a sublex? */
4069 if (optype == ')' && PL_endwhite) {
4070 CURMAD('X', PL_endwhite);
4075 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4076 if (!PL_thiswhite && !PL_endwhite && !optype) {
4077 sv_free(PL_thistoken);
4082 /* put off final whitespace till peg */
4083 if (optype == ';' && !PL_rsfp) {
4084 PL_nextwhite = PL_thiswhite;
4087 else if (PL_thisopen) {
4088 CURMAD('q', PL_thisopen);
4090 sv_free(PL_thistoken);
4094 /* Store actual token text as madprop X */
4095 CURMAD('X', PL_thistoken);
4099 /* add preceding whitespace as madprop _ */
4100 CURMAD('_', PL_thiswhite);
4104 /* add quoted material as madprop = */
4105 CURMAD('=', PL_thisstuff);
4109 /* add terminating quote as madprop Q */
4110 CURMAD('Q', PL_thisclose);
4114 /* special processing based on optype */
4118 /* opval doesn't need a TOKEN since it can already store mp */
4128 if (pl_yylval.opval)
4129 append_madprops(PL_thismad, pl_yylval.opval, 0);
4137 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4146 /* remember any fake bracket that lexer is about to discard */
4147 if (PL_lex_brackets == 1 &&
4148 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4151 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4154 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4155 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4158 break; /* don't bother looking for trailing comment */
4167 /* attach a trailing comment to its statement instead of next token */
4171 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4173 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4175 if (*s == '\n' || *s == '#') {
4176 while (s < PL_bufend && *s != '\n')
4180 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4181 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4198 /* Create new token struct. Note: opvals return early above. */
4199 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4206 S_tokenize_use(pTHX_ int is_use, char *s) {
4209 PERL_ARGS_ASSERT_TOKENIZE_USE;
4211 if (PL_expect != XSTATE)
4212 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4213 is_use ? "use" : "no"));
4215 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4216 s = force_version(s, TRUE);
4217 if (*s == ';' || *s == '}'
4218 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4219 start_force(PL_curforce);
4220 NEXTVAL_NEXTTOKE.opval = NULL;
4223 else if (*s == 'v') {
4224 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4225 s = force_version(s, FALSE);
4229 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4230 s = force_version(s, FALSE);
4232 pl_yylval.ival = is_use;
4236 static const char* const exp_name[] =
4237 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4238 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4242 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4244 S_word_takes_any_delimeter(char *p, STRLEN len)
4246 return (len == 1 && strchr("msyq", p[0])) ||
4248 (p[0] == 't' && p[1] == 'r') ||
4249 (p[0] == 'q' && strchr("qwxr", p[1]))));
4255 Works out what to call the token just pulled out of the input
4256 stream. The yacc parser takes care of taking the ops we return and
4257 stitching them into a tree.
4263 if read an identifier
4264 if we're in a my declaration
4265 croak if they tried to say my($foo::bar)
4266 build the ops for a my() declaration
4267 if it's an access to a my() variable
4268 are we in a sort block?
4269 croak if my($a); $a <=> $b
4270 build ops for access to a my() variable
4271 if in a dq string, and they've said @foo and we can't find @foo
4273 build ops for a bareword
4274 if we already built the token before, use it.
4279 #pragma segment Perl_yylex
4285 register char *s = PL_bufptr;
4291 /* orig_keyword, gvp, and gv are initialized here because
4292 * jump to the label just_a_word_zero can bypass their
4293 * initialization later. */
4294 I32 orig_keyword = 0;
4299 SV* tmp = newSVpvs("");
4300 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4301 (IV)CopLINE(PL_curcop),
4302 lex_state_names[PL_lex_state],
4303 exp_name[PL_expect],
4304 pv_display(tmp, s, strlen(s), 0, 60));
4307 /* check if there's an identifier for us to look at */
4308 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4309 return REPORT(S_pending_ident(aTHX));
4311 /* no identifier pending identification */
4313 switch (PL_lex_state) {
4315 case LEX_NORMAL: /* Some compilers will produce faster */
4316 case LEX_INTERPNORMAL: /* code if we comment these out. */
4320 /* when we've already built the next token, just pull it out of the queue */
4324 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4326 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4327 PL_nexttoke[PL_lasttoke].next_mad = 0;
4328 if (PL_thismad && PL_thismad->mad_key == '_') {
4329 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4330 PL_thismad->mad_val = 0;
4331 mad_free(PL_thismad);
4336 PL_lex_state = PL_lex_defer;
4337 PL_expect = PL_lex_expect;
4338 PL_lex_defer = LEX_NORMAL;
4339 if (!PL_nexttoke[PL_lasttoke].next_type)
4344 pl_yylval = PL_nextval[PL_nexttoke];
4346 PL_lex_state = PL_lex_defer;
4347 PL_expect = PL_lex_expect;
4348 PL_lex_defer = LEX_NORMAL;
4354 next_type = PL_nexttoke[PL_lasttoke].next_type;
4356 next_type = PL_nexttype[PL_nexttoke];
4358 if (next_type & (7<<24)) {
4359 if (next_type & (1<<24)) {
4360 if (PL_lex_brackets > 100)
4361 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4362 PL_lex_brackstack[PL_lex_brackets++] =
4363 (char) ((next_type >> 16) & 0xff);
4365 if (next_type & (2<<24))
4366 PL_lex_allbrackets++;
4367 if (next_type & (4<<24))
4368 PL_lex_allbrackets--;
4369 next_type &= 0xffff;
4372 /* FIXME - can these be merged? */
4375 return REPORT(next_type);
4379 /* interpolated case modifiers like \L \U, including \Q and \E.
4380 when we get here, PL_bufptr is at the \
4382 case LEX_INTERPCASEMOD:
4384 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4385 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
4387 /* handle \E or end of string */
4388 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4390 if (PL_lex_casemods) {
4391 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4392 PL_lex_casestack[PL_lex_casemods] = '\0';
4394 if (PL_bufptr != PL_bufend
4395 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
4397 PL_lex_state = LEX_INTERPCONCAT;
4400 PL_thistoken = newSVpvs("\\E");
4403 PL_lex_allbrackets--;
4407 while (PL_bufptr != PL_bufend &&
4408 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4410 PL_thiswhite = newSVpvs("");
4411 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4415 if (PL_bufptr != PL_bufend)
4418 PL_lex_state = LEX_INTERPCONCAT;
4422 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4423 "### Saw case modifier\n"); });
4425 if (s[1] == '\\' && s[2] == 'E') {
4428 PL_thiswhite = newSVpvs("");
4429 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4432 PL_lex_state = LEX_INTERPCONCAT;
4437 if (!PL_madskills) /* when just compiling don't need correct */
4438 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4439 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4440 if ((*s == 'L' || *s == 'U') &&
4441 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4442 PL_lex_casestack[--PL_lex_casemods] = '\0';
4443 PL_lex_allbrackets--;
4446 if (PL_lex_casemods > 10)
4447 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4448 PL_lex_casestack[PL_lex_casemods++] = *s;
4449 PL_lex_casestack[PL_lex_casemods] = '\0';
4450 PL_lex_state = LEX_INTERPCONCAT;
4451 start_force(PL_curforce);
4452 NEXTVAL_NEXTTOKE.ival = 0;
4453 force_next((2<<24)|'(');
4454 start_force(PL_curforce);
4456 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4458 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4460 NEXTVAL_NEXTTOKE.ival = OP_LC;
4462 NEXTVAL_NEXTTOKE.ival = OP_UC;
4464 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4466 Perl_croak(aTHX_ "panic: yylex");
4468 SV* const tmpsv = newSVpvs("\\ ");
4469 /* replace the space with the character we want to escape
4471 SvPVX(tmpsv)[1] = *s;
4477 if (PL_lex_starts) {
4483 sv_free(PL_thistoken);
4484 PL_thistoken = newSVpvs("");
4487 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4488 if (PL_lex_casemods == 1 && PL_lex_inpat)
4497 case LEX_INTERPPUSH:
4498 return REPORT(sublex_push());
4500 case LEX_INTERPSTART:
4501 if (PL_bufptr == PL_bufend)
4502 return REPORT(sublex_done());
4503 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4504 "### Interpolated variable\n"); });
4506 PL_lex_dojoin = (*PL_bufptr == '@');
4507 PL_lex_state = LEX_INTERPNORMAL;
4508 if (PL_lex_dojoin) {
4509 start_force(PL_curforce);
4510 NEXTVAL_NEXTTOKE.ival = 0;
4512 start_force(PL_curforce);
4513 force_ident("\"", '$');
4514 start_force(PL_curforce);
4515 NEXTVAL_NEXTTOKE.ival = 0;
4517 start_force(PL_curforce);
4518 NEXTVAL_NEXTTOKE.ival = 0;
4519 force_next((2<<24)|'(');
4520 start_force(PL_curforce);
4521 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4524 if (PL_lex_starts++) {
4529 sv_free(PL_thistoken);
4530 PL_thistoken = newSVpvs("");
4533 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4534 if (!PL_lex_casemods && PL_lex_inpat)
4541 case LEX_INTERPENDMAYBE:
4542 if (intuit_more(PL_bufptr)) {
4543 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4549 if (PL_lex_dojoin) {
4550 PL_lex_dojoin = FALSE;
4551 PL_lex_state = LEX_INTERPCONCAT;
4555 sv_free(PL_thistoken);
4556 PL_thistoken = newSVpvs("");
4559 PL_lex_allbrackets--;
4562 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4563 && SvEVALED(PL_lex_repl))
4565 if (PL_bufptr != PL_bufend)
4566 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4570 case LEX_INTERPCONCAT:
4572 if (PL_lex_brackets)
4573 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4575 if (PL_bufptr == PL_bufend)
4576 return REPORT(sublex_done());
4578 if (SvIVX(PL_linestr) == '\'') {
4579 SV *sv = newSVsv(PL_linestr);
4582 else if ( PL_hints & HINT_NEW_RE )
4583 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4584 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4588 s = scan_const(PL_bufptr);
4590 PL_lex_state = LEX_INTERPCASEMOD;
4592 PL_lex_state = LEX_INTERPSTART;
4595 if (s != PL_bufptr) {
4596 start_force(PL_curforce);
4598 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4600 NEXTVAL_NEXTTOKE = pl_yylval;
4603 if (PL_lex_starts++) {
4607 sv_free(PL_thistoken);
4608 PL_thistoken = newSVpvs("");
4611 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4612 if (!PL_lex_casemods && PL_lex_inpat)
4625 PL_lex_state = LEX_NORMAL;
4626 s = scan_formline(PL_bufptr);
4627 if (!PL_lex_formbrack)
4633 PL_oldoldbufptr = PL_oldbufptr;
4639 sv_free(PL_thistoken);
4642 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4646 if (isIDFIRST_lazy_if(s,UTF))
4649 unsigned char c = *s;
4650 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4651 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4652 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4657 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4661 goto fake_eof; /* emulate EOF on ^D or ^Z */
4670 if (PL_lex_brackets &&
4671 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
4672 yyerror((const char *)
4674 ? "Format not terminated"
4675 : "Missing right curly or square bracket"));
4677 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4678 "### Tokener got EOF\n");
4682 if (s++ < PL_bufend)
4683 goto retry; /* ignore stray nulls */
4686 if (!PL_in_eval && !PL_preambled) {
4687 PL_preambled = TRUE;
4693 /* Generate a string of Perl code to load the debugger.
4694 * If PERL5DB is set, it will return the contents of that,
4695 * otherwise a compile-time require of perl5db.pl. */
4697 const char * const pdb = PerlEnv_getenv("PERL5DB");
4700 sv_setpv(PL_linestr, pdb);
4701 sv_catpvs(PL_linestr,";");
4703 SETERRNO(0,SS_NORMAL);
4704 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4707 sv_setpvs(PL_linestr,"");
4708 if (PL_preambleav) {
4709 SV **svp = AvARRAY(PL_preambleav);
4710 SV **const end = svp + AvFILLp(PL_preambleav);
4712 sv_catsv(PL_linestr, *svp);
4714 sv_catpvs(PL_linestr, ";");
4716 sv_free(MUTABLE_SV(PL_preambleav));
4717 PL_preambleav = NULL;
4720 sv_catpvs(PL_linestr,
4721 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4722 if (PL_minus_n || PL_minus_p) {
4723 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4725 sv_catpvs(PL_linestr,"chomp;");
4728 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4729 || *PL_splitstr == '"')
4730 && strchr(PL_splitstr + 1, *PL_splitstr))
4731 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4733 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4734 bytes can be used as quoting characters. :-) */
4735 const char *splits = PL_splitstr;
4736 sv_catpvs(PL_linestr, "our @F=split(q\0");
4739 if (*splits == '\\')
4740 sv_catpvn(PL_linestr, splits, 1);
4741 sv_catpvn(PL_linestr, splits, 1);
4742 } while (*splits++);
4743 /* This loop will embed the trailing NUL of
4744 PL_linestr as the last thing it does before
4746 sv_catpvs(PL_linestr, ");");
4750 sv_catpvs(PL_linestr,"our @F=split(' ');");
4753 sv_catpvs(PL_linestr, "\n");
4754 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4755 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4756 PL_last_lop = PL_last_uni = NULL;
4757 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4758 update_debugger_info(PL_linestr, NULL, 0);
4763 bof = PL_rsfp ? TRUE : FALSE;
4766 fake_eof = LEX_FAKE_EOF;
4768 PL_bufptr = PL_bufend;
4769 CopLINE_inc(PL_curcop);
4770 if (!lex_next_chunk(fake_eof)) {
4771 CopLINE_dec(PL_curcop);
4773 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4775 CopLINE_dec(PL_curcop);
4778 PL_realtokenstart = -1;
4781 /* If it looks like the start of a BOM or raw UTF-16,
4782 * check if it in fact is. */
4783 if (bof && PL_rsfp &&
4788 bof = PerlIO_tell(PL_rsfp) == (Off_t)SvCUR(PL_linestr);
4790 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4791 s = swallow_bom((U8*)s);
4794 if (PL_parser->in_pod) {
4795 /* Incest with pod. */
4798 sv_catsv(PL_thiswhite, PL_linestr);
4800 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4801 sv_setpvs(PL_linestr, "");
4802 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4803 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4804 PL_last_lop = PL_last_uni = NULL;
4805 PL_parser->in_pod = 0;
4810 } while (PL_parser->in_pod);
4811 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4812 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4813 PL_last_lop = PL_last_uni = NULL;
4814 if (CopLINE(PL_curcop) == 1) {
4815 while (s < PL_bufend && isSPACE(*s))
4817 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4821 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4825 if (*s == '#' && *(s+1) == '!')
4827 #ifdef ALTERNATE_SHEBANG
4829 static char const as[] = ALTERNATE_SHEBANG;
4830 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4831 d = s + (sizeof(as) - 1);
4833 #endif /* ALTERNATE_SHEBANG */
4842 while (*d && !isSPACE(*d))
4846 #ifdef ARG_ZERO_IS_SCRIPT
4847 if (ipathend > ipath) {
4849 * HP-UX (at least) sets argv[0] to the script name,
4850 * which makes $^X incorrect. And Digital UNIX and Linux,
4851 * at least, set argv[0] to the basename of the Perl
4852 * interpreter. So, having found "#!", we'll set it right.
4854 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4856 assert(SvPOK(x) || SvGMAGICAL(x));
4857 if (sv_eq(x, CopFILESV(PL_curcop))) {
4858 sv_setpvn(x, ipath, ipathend - ipath);
4864 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4865 const char * const lstart = SvPV_const(x,llen);
4867 bstart += blen - llen;
4868 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4869 sv_setpvn(x, ipath, ipathend - ipath);
4874 TAINT_NOT; /* $^X is always tainted, but that's OK */
4876 #endif /* ARG_ZERO_IS_SCRIPT */
4881 d = instr(s,"perl -");
4883 d = instr(s,"perl");
4885 /* avoid getting into infinite loops when shebang
4886 * line contains "Perl" rather than "perl" */
4888 for (d = ipathend-4; d >= ipath; --d) {
4889 if ((*d == 'p' || *d == 'P')
4890 && !ibcmp(d, "perl", 4))
4900 #ifdef ALTERNATE_SHEBANG
4902 * If the ALTERNATE_SHEBANG on this system starts with a
4903 * character that can be part of a Perl expression, then if
4904 * we see it but not "perl", we're probably looking at the
4905 * start of Perl code, not a request to hand off to some
4906 * other interpreter. Similarly, if "perl" is there, but
4907 * not in the first 'word' of the line, we assume the line
4908 * contains the start of the Perl program.
4910 if (d && *s != '#') {
4911 const char *c = ipath;
4912 while (*c && !strchr("; \t\r\n\f\v#", *c))
4915 d = NULL; /* "perl" not in first word; ignore */
4917 *s = '#'; /* Don't try to parse shebang line */
4919 #endif /* ALTERNATE_SHEBANG */
4924 !instr(s,"indir") &&
4925 instr(PL_origargv[0],"perl"))
4932 while (s < PL_bufend && isSPACE(*s))
4934 if (s < PL_bufend) {
4935 Newx(newargv,PL_origargc+3,char*);
4937 while (s < PL_bufend && !isSPACE(*s))
4940 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4943 newargv = PL_origargv;
4946 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4948 Perl_croak(aTHX_ "Can't exec %s", ipath);
4951 while (*d && !isSPACE(*d))
4953 while (SPACE_OR_TAB(*d))
4957 const bool switches_done = PL_doswitches;
4958 const U32 oldpdb = PL_perldb;
4959 const bool oldn = PL_minus_n;
4960 const bool oldp = PL_minus_p;
4964 bool baduni = FALSE;
4966 const char *d2 = d1 + 1;
4967 if (parse_unicode_opts((const char **)&d2)
4971 if (baduni || *d1 == 'M' || *d1 == 'm') {
4972 const char * const m = d1;
4973 while (*d1 && !isSPACE(*d1))
4975 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4978 d1 = moreswitches(d1);
4980 if (PL_doswitches && !switches_done) {
4981 int argc = PL_origargc;
4982 char **argv = PL_origargv;
4985 } while (argc && argv[0][0] == '-' && argv[0][1]);
4986 init_argv_symbols(argc,argv);
4988 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
4989 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
4990 /* if we have already added "LINE: while (<>) {",
4991 we must not do it again */
4993 sv_setpvs(PL_linestr, "");
4994 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4995 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4996 PL_last_lop = PL_last_uni = NULL;
4997 PL_preambled = FALSE;
4998 if (PERLDB_LINE || PERLDB_SAVESRC)
4999 (void)gv_fetchfile(PL_origfilename);
5006 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5008 PL_lex_state = LEX_FORMLINE;
5013 #ifdef PERL_STRICT_CR
5014 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5016 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5018 case ' ': case '\t': case '\f': case 013:
5020 PL_realtokenstart = -1;
5022 PL_thiswhite = newSVpvs("");
5023 sv_catpvn(PL_thiswhite, s, 1);
5030 PL_realtokenstart = -1;
5034 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
5035 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
5036 /* handle eval qq[#line 1 "foo"\n ...] */
5037 CopLINE_dec(PL_curcop);
5040 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
5042 if (!PL_in_eval || PL_rsfp)
5047 while (d < PL_bufend && *d != '\n')
5051 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5052 Perl_croak(aTHX_ "panic: input overflow");
5055 PL_thiswhite = newSVpvn(s, d - s);
5060 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5062 PL_lex_state = LEX_FORMLINE;
5068 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
5069 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
5072 TOKEN(PEG); /* make sure any #! line is accessible */
5077 /* if (PL_madskills && PL_lex_formbrack) { */
5079 while (d < PL_bufend && *d != '\n')
5083 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5084 Perl_croak(aTHX_ "panic: input overflow");
5085 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
5087 PL_thiswhite = newSVpvs("");
5088 if (CopLINE(PL_curcop) == 1) {
5089 sv_setpvs(PL_thiswhite, "");
5092 sv_catpvn(PL_thiswhite, s, d - s);
5106 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
5114 while (s < PL_bufend && SPACE_OR_TAB(*s))
5117 if (strnEQ(s,"=>",2)) {
5118 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
5119 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5120 OPERATOR('-'); /* unary minus */
5122 PL_last_uni = PL_oldbufptr;
5124 case 'r': ftst = OP_FTEREAD; break;
5125 case 'w': ftst = OP_FTEWRITE; break;
5126 case 'x': ftst = OP_FTEEXEC; break;
5127 case 'o': ftst = OP_FTEOWNED; break;
5128 case 'R': ftst = OP_FTRREAD; break;
5129 case 'W': ftst = OP_FTRWRITE; break;
5130 case 'X': ftst = OP_FTREXEC; break;
5131 case 'O': ftst = OP_FTROWNED; break;
5132 case 'e': ftst = OP_FTIS; break;
5133 case 'z': ftst = OP_FTZERO; break;
5134 case 's': ftst = OP_FTSIZE; break;
5135 case 'f': ftst = OP_FTFILE; break;
5136 case 'd': ftst = OP_FTDIR; break;
5137 case 'l': ftst = OP_FTLINK; break;
5138 case 'p': ftst = OP_FTPIPE; break;
5139 case 'S': ftst = OP_FTSOCK; break;
5140 case 'u': ftst = OP_FTSUID; break;
5141 case 'g': ftst = OP_FTSGID; break;
5142 case 'k': ftst = OP_FTSVTX; break;
5143 case 'b': ftst = OP_FTBLK; break;
5144 case 'c': ftst = OP_FTCHR; break;
5145 case 't': ftst = OP_FTTTY; break;
5146 case 'T': ftst = OP_FTTEXT; break;
5147 case 'B': ftst = OP_FTBINARY; break;
5148 case 'M': case 'A': case 'C':
5149 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5151 case 'M': ftst = OP_FTMTIME; break;
5152 case 'A': ftst = OP_FTATIME; break;
5153 case 'C': ftst = OP_FTCTIME; break;
5161 PL_last_lop_op = (OPCODE)ftst;
5162 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5163 "### Saw file test %c\n", (int)tmp);
5168 /* Assume it was a minus followed by a one-letter named
5169 * subroutine call (or a -bareword), then. */
5170 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5171 "### '-%c' looked like a file test but was not\n",
5178 const char tmp = *s++;
5181 if (PL_expect == XOPERATOR)
5186 else if (*s == '>') {
5189 if (isIDFIRST_lazy_if(s,UTF)) {
5190 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5198 if (PL_expect == XOPERATOR) {
5199 if (*s == '=' && !PL_lex_allbrackets &&
5200 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5207 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5209 OPERATOR('-'); /* unary minus */
5215 const char tmp = *s++;
5218 if (PL_expect == XOPERATOR)
5223 if (PL_expect == XOPERATOR) {
5224 if (*s == '=' && !PL_lex_allbrackets &&
5225 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5232 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5239 if (PL_expect != XOPERATOR) {
5240 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5241 PL_expect = XOPERATOR;
5242 force_ident(PL_tokenbuf, '*');
5250 if (*s == '=' && !PL_lex_allbrackets &&
5251 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5257 if (*s == '=' && !PL_lex_allbrackets &&
5258 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5265 if (PL_expect == XOPERATOR) {
5266 if (s[1] == '=' && !PL_lex_allbrackets &&
5267 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5272 PL_tokenbuf[0] = '%';
5273 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5274 sizeof PL_tokenbuf - 1, FALSE);
5275 if (!PL_tokenbuf[1]) {
5278 PL_pending_ident = '%';
5282 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5283 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5288 if (PL_lex_brackets > 100)
5289 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5290 PL_lex_brackstack[PL_lex_brackets++] = 0;
5291 PL_lex_allbrackets++;
5293 const char tmp = *s++;
5298 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5300 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5308 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5315 goto just_a_word_zero_gv;
5318 switch (PL_expect) {
5324 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5326 PL_bufptr = s; /* update in case we back off */
5329 "Use of := for an empty attribute list is not allowed");
5336 PL_expect = XTERMBLOCK;
5339 stuffstart = s - SvPVX(PL_linestr) - 1;
5343 while (isIDFIRST_lazy_if(s,UTF)) {
5346 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5347 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5348 if (tmp < 0) tmp = -tmp;
5363 sv = newSVpvn(s, len);
5365 d = scan_str(d,TRUE,TRUE);
5367 /* MUST advance bufptr here to avoid bogus
5368 "at end of line" context messages from yyerror().
5370 PL_bufptr = s + len;
5371 yyerror("Unterminated attribute parameter in attribute list");
5375 return REPORT(0); /* EOF indicator */
5379 sv_catsv(sv, PL_lex_stuff);
5380 attrs = op_append_elem(OP_LIST, attrs,
5381 newSVOP(OP_CONST, 0, sv));
5382 SvREFCNT_dec(PL_lex_stuff);
5383 PL_lex_stuff = NULL;
5386 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5388 if (PL_in_my == KEY_our) {
5389 deprecate(":unique");
5392 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5395 /* NOTE: any CV attrs applied here need to be part of
5396 the CVf_BUILTIN_ATTRS define in cv.h! */
5397 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5399 CvLVALUE_on(PL_compcv);
5401 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5403 deprecate(":locked");
5405 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5407 CvMETHOD_on(PL_compcv);
5409 /* After we've set the flags, it could be argued that
5410 we don't need to do the attributes.pm-based setting
5411 process, and shouldn't bother appending recognized
5412 flags. To experiment with that, uncomment the
5413 following "else". (Note that's already been
5414 uncommented. That keeps the above-applied built-in
5415 attributes from being intercepted (and possibly
5416 rejected) by a package's attribute routines, but is
5417 justified by the performance win for the common case
5418 of applying only built-in attributes.) */
5420 attrs = op_append_elem(OP_LIST, attrs,
5421 newSVOP(OP_CONST, 0,
5425 if (*s == ':' && s[1] != ':')
5428 break; /* require real whitespace or :'s */
5429 /* XXX losing whitespace on sequential attributes here */
5433 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5434 if (*s != ';' && *s != '}' && *s != tmp
5435 && (tmp != '=' || *s != ')')) {
5436 const char q = ((*s == '\'') ? '"' : '\'');
5437 /* If here for an expression, and parsed no attrs, back
5439 if (tmp == '=' && !attrs) {
5443 /* MUST advance bufptr here to avoid bogus "at end of line"
5444 context messages from yyerror().
5447 yyerror( (const char *)
5449 ? Perl_form(aTHX_ "Invalid separator character "
5450 "%c%c%c in attribute list", q, *s, q)
5451 : "Unterminated attribute list" ) );
5459 start_force(PL_curforce);
5460 NEXTVAL_NEXTTOKE.opval = attrs;
5461 CURMAD('_', PL_nextwhite);
5466 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5467 (s - SvPVX(PL_linestr)) - stuffstart);
5472 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5476 PL_lex_allbrackets--;
5480 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5481 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5485 PL_lex_allbrackets++;
5488 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5494 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5497 PL_lex_allbrackets--;
5503 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5506 if (PL_lex_brackets <= 0)
5507 yyerror("Unmatched right square bracket");
5510 PL_lex_allbrackets--;
5511 if (PL_lex_state == LEX_INTERPNORMAL) {
5512 if (PL_lex_brackets == 0) {
5513 if (*s == '-' && s[1] == '>')
5514 PL_lex_state = LEX_INTERPENDMAYBE;
5515 else if (*s != '[' && *s != '{')
5516 PL_lex_state = LEX_INTERPEND;
5523 if (PL_lex_brackets > 100) {
5524 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5526 switch (PL_expect) {
5528 if (PL_lex_formbrack) {
5532 if (PL_oldoldbufptr == PL_last_lop)
5533 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5535 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5536 PL_lex_allbrackets++;
5537 OPERATOR(HASHBRACK);
5539 while (s < PL_bufend && SPACE_OR_TAB(*s))
5542 PL_tokenbuf[0] = '\0';
5543 if (d < PL_bufend && *d == '-') {
5544 PL_tokenbuf[0] = '-';
5546 while (d < PL_bufend && SPACE_OR_TAB(*d))
5549 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5550 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5552 while (d < PL_bufend && SPACE_OR_TAB(*d))
5555 const char minus = (PL_tokenbuf[0] == '-');
5556 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5564 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5565 PL_lex_allbrackets++;
5570 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5571 PL_lex_allbrackets++;
5576 if (PL_oldoldbufptr == PL_last_lop)
5577 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5579 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5580 PL_lex_allbrackets++;
5583 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5585 /* This hack is to get the ${} in the message. */
5587 yyerror("syntax error");
5590 OPERATOR(HASHBRACK);
5592 /* This hack serves to disambiguate a pair of curlies
5593 * as being a block or an anon hash. Normally, expectation
5594 * determines that, but in cases where we're not in a
5595 * position to expect anything in particular (like inside
5596 * eval"") we have to resolve the ambiguity. This code
5597 * covers the case where the first term in the curlies is a
5598 * quoted string. Most other cases need to be explicitly
5599 * disambiguated by prepending a "+" before the opening
5600 * curly in order to force resolution as an anon hash.
5602 * XXX should probably propagate the outer expectation
5603 * into eval"" to rely less on this hack, but that could
5604 * potentially break current behavior of eval"".
5608 if (*s == '\'' || *s == '"' || *s == '`') {
5609 /* common case: get past first string, handling escapes */
5610 for (t++; t < PL_bufend && *t != *s;)
5611 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5615 else if (*s == 'q') {
5618 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5621 /* skip q//-like construct */
5623 char open, close, term;
5626 while (t < PL_bufend && isSPACE(*t))
5628 /* check for q => */
5629 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5630 OPERATOR(HASHBRACK);
5634 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5638 for (t++; t < PL_bufend; t++) {
5639 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5641 else if (*t == open)
5645 for (t++; t < PL_bufend; t++) {
5646 if (*t == '\\' && t+1 < PL_bufend)
5648 else if (*t == close && --brackets <= 0)
5650 else if (*t == open)
5657 /* skip plain q word */
5658 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5661 else if (isALNUM_lazy_if(t,UTF)) {
5663 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5666 while (t < PL_bufend && isSPACE(*t))
5668 /* if comma follows first term, call it an anon hash */
5669 /* XXX it could be a comma expression with loop modifiers */
5670 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5671 || (*t == '=' && t[1] == '>')))
5672 OPERATOR(HASHBRACK);
5673 if (PL_expect == XREF)
5676 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5682 pl_yylval.ival = CopLINE(PL_curcop);
5683 if (isSPACE(*s) || *s == '#')
5684 PL_copline = NOLINE; /* invalidate current command line number */
5687 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5691 if (PL_lex_brackets <= 0)
5692 yyerror("Unmatched right curly bracket");
5694 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5695 PL_lex_allbrackets--;
5696 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5697 PL_lex_formbrack = 0;
5698 if (PL_lex_state == LEX_INTERPNORMAL) {
5699 if (PL_lex_brackets == 0) {
5700 if (PL_expect & XFAKEBRACK) {
5701 PL_expect &= XENUMMASK;
5702 PL_lex_state = LEX_INTERPEND;
5707 PL_thiswhite = newSVpvs("");
5708 sv_catpvs(PL_thiswhite,"}");
5711 return yylex(); /* ignore fake brackets */
5713 if (*s == '-' && s[1] == '>')
5714 PL_lex_state = LEX_INTERPENDMAYBE;
5715 else if (*s != '[' && *s != '{')
5716 PL_lex_state = LEX_INTERPEND;
5719 if (PL_expect & XFAKEBRACK) {
5720 PL_expect &= XENUMMASK;
5722 return yylex(); /* ignore fake brackets */
5724 start_force(PL_curforce);
5726 curmad('X', newSVpvn(s-1,1));
5727 CURMAD('_', PL_thiswhite);
5732 PL_thistoken = newSVpvs("");
5738 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5739 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5746 if (PL_expect == XOPERATOR) {
5747 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5748 && isIDFIRST_lazy_if(s,UTF))
5750 CopLINE_dec(PL_curcop);
5751 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5752 CopLINE_inc(PL_curcop);
5754 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5755 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5762 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5764 PL_expect = XOPERATOR;
5765 force_ident(PL_tokenbuf, '&');
5769 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5775 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5776 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5783 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5784 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5792 const char tmp = *s++;
5794 if (!PL_lex_allbrackets &&
5795 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5802 if (!PL_lex_allbrackets &&
5803 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
5811 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5812 && strchr("+-*/%.^&|<",tmp))
5813 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5814 "Reversed %c= operator",(int)tmp);
5816 if (PL_expect == XSTATE && isALPHA(tmp) &&
5817 (s == PL_linestart+1 || s[-2] == '\n') )
5819 if (PL_in_eval && !PL_rsfp) {
5824 if (strnEQ(s,"=cut",4)) {
5840 PL_thiswhite = newSVpvs("");
5841 sv_catpvn(PL_thiswhite, PL_linestart,
5842 PL_bufend - PL_linestart);
5846 PL_parser->in_pod = 1;
5850 if (PL_lex_brackets < PL_lex_formbrack) {
5852 #ifdef PERL_STRICT_CR
5853 while (SPACE_OR_TAB(*t))
5855 while (SPACE_OR_TAB(*t) || *t == '\r')
5858 if (*t == '\n' || *t == '#') {
5864 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5873 const char tmp = *s++;
5875 /* was this !=~ where !~ was meant?
5876 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5878 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5879 const char *t = s+1;
5881 while (t < PL_bufend && isSPACE(*t))
5884 if (*t == '/' || *t == '?' ||
5885 ((*t == 'm' || *t == 's' || *t == 'y')
5886 && !isALNUM(t[1])) ||
5887 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5888 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5889 "!=~ should be !~");
5891 if (!PL_lex_allbrackets &&
5892 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5904 if (PL_expect != XOPERATOR) {
5905 if (s[1] != '<' && !strchr(s,'>'))
5908 s = scan_heredoc(s);
5910 s = scan_inputsymbol(s);
5911 TERM(sublex_start());
5917 if (*s == '=' && !PL_lex_allbrackets &&
5918 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5922 SHop(OP_LEFT_SHIFT);
5927 if (!PL_lex_allbrackets &&
5928 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5935 if (!PL_lex_allbrackets &&
5936 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5944 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5952 const char tmp = *s++;
5954 if (*s == '=' && !PL_lex_allbrackets &&
5955 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5959 SHop(OP_RIGHT_SHIFT);
5961 else if (tmp == '=') {
5962 if (!PL_lex_allbrackets &&
5963 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5971 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5980 if (PL_expect == XOPERATOR) {
5981 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5982 return deprecate_commaless_var_list();
5986 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
5987 PL_tokenbuf[0] = '@';
5988 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
5989 sizeof PL_tokenbuf - 1, FALSE);
5990 if (PL_expect == XOPERATOR)
5991 no_op("Array length", s);
5992 if (!PL_tokenbuf[1])
5994 PL_expect = XOPERATOR;
5995 PL_pending_ident = '#';
5999 PL_tokenbuf[0] = '$';
6000 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
6001 sizeof PL_tokenbuf - 1, FALSE);
6002 if (PL_expect == XOPERATOR)
6004 if (!PL_tokenbuf[1]) {
6006 yyerror("Final $ should be \\$ or $name");
6010 /* This kludge not intended to be bulletproof. */
6011 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
6012 pl_yylval.opval = newSVOP(OP_CONST, 0,
6013 newSViv(CopARYBASE_get(&PL_compiling)));
6014 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
6020 const char tmp = *s;
6021 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6024 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6025 && intuit_more(s)) {
6027 PL_tokenbuf[0] = '@';
6028 if (ckWARN(WARN_SYNTAX)) {
6031 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
6034 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6035 while (t < PL_bufend && *t != ']')
6037 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6038 "Multidimensional syntax %.*s not supported",
6039 (int)((t - PL_bufptr) + 1), PL_bufptr);
6043 else if (*s == '{') {
6045 PL_tokenbuf[0] = '%';
6046 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6047 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6049 char tmpbuf[sizeof PL_tokenbuf];
6052 } while (isSPACE(*t));
6053 if (isIDFIRST_lazy_if(t,UTF)) {
6055 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6059 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
6060 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6061 "You need to quote \"%s\"",
6068 PL_expect = XOPERATOR;
6069 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6070 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6071 if (!islop || PL_last_lop_op == OP_GREPSTART)
6072 PL_expect = XOPERATOR;
6073 else if (strchr("$@\"'`q", *s))
6074 PL_expect = XTERM; /* e.g. print $fh "foo" */
6075 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6076 PL_expect = XTERM; /* e.g. print $fh &sub */
6077 else if (isIDFIRST_lazy_if(s,UTF)) {
6078 char tmpbuf[sizeof PL_tokenbuf];
6080 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6081 if ((t2 = keyword(tmpbuf, len, 0))) {
6082 /* binary operators exclude handle interpretations */
6094 PL_expect = XTERM; /* e.g. print $fh length() */
6099 PL_expect = XTERM; /* e.g. print $fh subr() */
6102 else if (isDIGIT(*s))
6103 PL_expect = XTERM; /* e.g. print $fh 3 */
6104 else if (*s == '.' && isDIGIT(s[1]))
6105 PL_expect = XTERM; /* e.g. print $fh .3 */
6106 else if ((*s == '?' || *s == '-' || *s == '+')
6107 && !isSPACE(s[1]) && s[1] != '=')
6108 PL_expect = XTERM; /* e.g. print $fh -1 */
6109 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6111 PL_expect = XTERM; /* e.g. print $fh /.../
6112 XXX except DORDOR operator
6114 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6116 PL_expect = XTERM; /* print $fh <<"EOF" */
6119 PL_pending_ident = '$';
6123 if (PL_expect == XOPERATOR)
6125 PL_tokenbuf[0] = '@';
6126 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6127 if (!PL_tokenbuf[1]) {
6130 if (PL_lex_state == LEX_NORMAL)
6132 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6134 PL_tokenbuf[0] = '%';
6136 /* Warn about @ where they meant $. */
6137 if (*s == '[' || *s == '{') {
6138 if (ckWARN(WARN_SYNTAX)) {
6139 const char *t = s + 1;
6140 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
6142 if (*t == '}' || *t == ']') {
6144 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6145 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6146 "Scalar value %.*s better written as $%.*s",
6147 (int)(t-PL_bufptr), PL_bufptr,
6148 (int)(t-PL_bufptr-1), PL_bufptr+1);
6153 PL_pending_ident = '@';
6156 case '/': /* may be division, defined-or, or pattern */
6157 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
6158 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6159 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6164 case '?': /* may either be conditional or pattern */
6165 if (PL_expect == XOPERATOR) {
6168 if (!PL_lex_allbrackets &&
6169 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6173 PL_lex_allbrackets++;
6179 /* A // operator. */
6180 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6181 (*s == '=' ? LEX_FAKEEOF_ASSIGN :
6182 LEX_FAKEEOF_LOGIC)) {
6190 if (*s == '=' && !PL_lex_allbrackets &&
6191 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6200 /* Disable warning on "study /blah/" */
6201 if (PL_oldoldbufptr == PL_last_uni
6202 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6203 || memNE(PL_last_uni, "study", 5)
6204 || isALNUM_lazy_if(PL_last_uni+5,UTF)
6208 deprecate("?PATTERN? without explicit operator");
6209 s = scan_pat(s,OP_MATCH);
6210 TERM(sublex_start());
6214 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6215 #ifdef PERL_STRICT_CR
6218 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6220 && (s == PL_linestart || s[-1] == '\n') )
6222 PL_lex_formbrack = 0;
6226 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6230 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6233 if (!PL_lex_allbrackets &&
6234 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6241 pl_yylval.ival = OPf_SPECIAL;
6247 if (*s == '=' && !PL_lex_allbrackets &&
6248 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6255 case '0': case '1': case '2': case '3': case '4':
6256 case '5': case '6': case '7': case '8': case '9':
6257 s = scan_num(s, &pl_yylval);
6258 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6259 if (PL_expect == XOPERATOR)
6264 s = scan_str(s,!!PL_madskills,FALSE);
6265 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6266 if (PL_expect == XOPERATOR) {
6267 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6268 return deprecate_commaless_var_list();
6275 pl_yylval.ival = OP_CONST;
6276 TERM(sublex_start());
6279 s = scan_str(s,!!PL_madskills,FALSE);
6280 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6281 if (PL_expect == XOPERATOR) {
6282 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6283 return deprecate_commaless_var_list();
6290 pl_yylval.ival = OP_CONST;
6291 /* FIXME. I think that this can be const if char *d is replaced by
6292 more localised variables. */
6293 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6294 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6295 pl_yylval.ival = OP_STRINGIFY;
6299 TERM(sublex_start());
6302 s = scan_str(s,!!PL_madskills,FALSE);
6303 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6304 if (PL_expect == XOPERATOR)
6305 no_op("Backticks",s);
6308 readpipe_override();
6309 TERM(sublex_start());
6313 if (PL_lex_inwhat && isDIGIT(*s))
6314 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6316 if (PL_expect == XOPERATOR)
6317 no_op("Backslash",s);
6321 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6322 char *start = s + 2;
6323 while (isDIGIT(*start) || *start == '_')
6325 if (*start == '.' && isDIGIT(start[1])) {
6326 s = scan_num(s, &pl_yylval);
6329 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6330 else if (!isALPHA(*start) && (PL_expect == XTERM
6331 || PL_expect == XREF || PL_expect == XSTATE
6332 || PL_expect == XTERMORDORDOR)) {
6333 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
6335 s = scan_num(s, &pl_yylval);
6342 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6385 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6387 /* Some keywords can be followed by any delimiter, including ':' */
6388 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6390 /* x::* is just a word, unless x is "CORE" */
6391 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6395 while (d < PL_bufend && isSPACE(*d))
6396 d++; /* no comments skipped here, or s### is misparsed */
6398 /* Is this a word before a => operator? */
6399 if (*d == '=' && d[1] == '>') {
6402 = (OP*)newSVOP(OP_CONST, 0,
6403 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6404 pl_yylval.opval->op_private = OPpCONST_BARE;
6408 /* Check for plugged-in keyword */
6412 char *saved_bufptr = PL_bufptr;
6414 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6416 if (result == KEYWORD_PLUGIN_DECLINE) {
6417 /* not a plugged-in keyword */
6418 PL_bufptr = saved_bufptr;
6419 } else if (result == KEYWORD_PLUGIN_STMT) {
6420 pl_yylval.opval = o;
6423 return REPORT(PLUGSTMT);
6424 } else if (result == KEYWORD_PLUGIN_EXPR) {
6425 pl_yylval.opval = o;
6427 PL_expect = XOPERATOR;
6428 return REPORT(PLUGEXPR);
6430 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6435 /* Check for built-in keyword */
6436 tmp = keyword(PL_tokenbuf, len, 0);
6438 /* Is this a label? */
6439 if (!anydelim && PL_expect == XSTATE
6440 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6442 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
6447 if (tmp < 0) { /* second-class keyword? */
6448 GV *ogv = NULL; /* override (winner) */
6449 GV *hgv = NULL; /* hidden (loser) */
6450 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6452 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
6455 if (GvIMPORTED_CV(gv))
6457 else if (! CvMETHOD(cv))
6461 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
6462 (gv = *gvp) && isGV_with_GP(gv) &&
6463 GvCVu(gv) && GvIMPORTED_CV(gv))
6470 tmp = 0; /* overridden by import or by GLOBAL */
6473 && -tmp==KEY_lock /* XXX generalizable kludge */
6476 tmp = 0; /* any sub overrides "weak" keyword */
6478 else { /* no override */
6480 if (tmp == KEY_dump) {
6481 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6482 "dump() better written as CORE::dump()");
6486 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6487 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6488 "Ambiguous call resolved as CORE::%s(), "
6489 "qualify as such or use &",
6497 default: /* not a keyword */
6498 /* Trade off - by using this evil construction we can pull the
6499 variable gv into the block labelled keylookup. If not, then
6500 we have to give it function scope so that the goto from the
6501 earlier ':' case doesn't bypass the initialisation. */
6503 just_a_word_zero_gv:
6511 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6515 SV *nextPL_nextwhite = 0;
6519 /* Get the rest if it looks like a package qualifier */
6521 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6523 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6526 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
6527 *s == '\'' ? "'" : "::");
6532 if (PL_expect == XOPERATOR) {
6533 if (PL_bufptr == PL_linestart) {
6534 CopLINE_dec(PL_curcop);
6535 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6536 CopLINE_inc(PL_curcop);
6539 no_op("Bareword",s);
6542 /* Look for a subroutine with this name in current package,
6543 unless name is "Foo::", in which case Foo is a bareword
6544 (and a package name). */
6546 if (len > 2 && !PL_madskills &&
6547 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6549 if (ckWARN(WARN_BAREWORD)
6550 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
6551 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6552 "Bareword \"%s\" refers to nonexistent package",
6555 PL_tokenbuf[len] = '\0';
6561 /* Mustn't actually add anything to a symbol table.
6562 But also don't want to "initialise" any placeholder
6563 constants that might already be there into full
6564 blown PVGVs with attached PVCV. */
6565 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6566 GV_NOADD_NOINIT, SVt_PVCV);
6571 /* if we saw a global override before, get the right name */
6573 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
6574 len ? len : strlen(PL_tokenbuf));
6576 SV * const tmp_sv = sv;
6577 sv = newSVpvs("CORE::GLOBAL::");
6578 sv_catsv(sv, tmp_sv);
6579 SvREFCNT_dec(tmp_sv);
6583 if (PL_madskills && !PL_thistoken) {
6584 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6585 PL_thistoken = newSVpvn(start,s - start);
6586 PL_realtokenstart = s - SvPVX(PL_linestr);
6590 /* Presume this is going to be a bareword of some sort. */
6592 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6593 pl_yylval.opval->op_private = OPpCONST_BARE;
6595 /* And if "Foo::", then that's what it certainly is. */
6600 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc(sv));
6601 const_op->op_private = OPpCONST_BARE;
6602 rv2cv_op = newCVREF(0, const_op);
6604 cv = rv2cv_op_cv(rv2cv_op, 0);
6606 /* See if it's the indirect object for a list operator. */
6608 if (PL_oldoldbufptr &&
6609 PL_oldoldbufptr < PL_bufptr &&
6610 (PL_oldoldbufptr == PL_last_lop
6611 || PL_oldoldbufptr == PL_last_uni) &&
6612 /* NO SKIPSPACE BEFORE HERE! */
6613 (PL_expect == XREF ||
6614 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6616 bool immediate_paren = *s == '(';
6618 /* (Now we can afford to cross potential line boundary.) */
6619 s = SKIPSPACE2(s,nextPL_nextwhite);
6621 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6624 /* Two barewords in a row may indicate method call. */
6626 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6627 (tmp = intuit_method(s, gv, cv))) {
6629 if (tmp == METHOD && !PL_lex_allbrackets &&
6630 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6631 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6635 /* If not a declared subroutine, it's an indirect object. */
6636 /* (But it's an indir obj regardless for sort.) */
6637 /* Also, if "_" follows a filetest operator, it's a bareword */
6640 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6642 (PL_last_lop_op != OP_MAPSTART &&
6643 PL_last_lop_op != OP_GREPSTART))))
6644 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6645 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6648 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6653 PL_expect = XOPERATOR;
6656 s = SKIPSPACE2(s,nextPL_nextwhite);
6657 PL_nextwhite = nextPL_nextwhite;
6662 /* Is this a word before a => operator? */
6663 if (*s == '=' && s[1] == '>' && !pkgname) {
6666 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6667 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6668 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6672 /* If followed by a paren, it's certainly a subroutine. */
6677 while (SPACE_OR_TAB(*d))
6679 if (*d == ')' && (sv = cv_const_sv(cv))) {
6686 PL_nextwhite = PL_thiswhite;
6689 start_force(PL_curforce);
6691 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6692 PL_expect = XOPERATOR;
6695 PL_nextwhite = nextPL_nextwhite;
6696 curmad('X', PL_thistoken);
6697 PL_thistoken = newSVpvs("");
6706 /* If followed by var or block, call it a method (unless sub) */
6708 if ((*s == '$' || *s == '{') && !cv) {
6710 PL_last_lop = PL_oldbufptr;
6711 PL_last_lop_op = OP_METHOD;
6712 if (!PL_lex_allbrackets &&
6713 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6714 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6718 /* If followed by a bareword, see if it looks like indir obj. */
6721 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6722 && (tmp = intuit_method(s, gv, cv))) {
6724 if (tmp == METHOD && !PL_lex_allbrackets &&
6725 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6726 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6730 /* Not a method, so call it a subroutine (if defined) */
6733 if (lastchar == '-')
6734 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6735 "Ambiguous use of -%s resolved as -&%s()",
6736 PL_tokenbuf, PL_tokenbuf);
6737 /* Check for a constant sub */
6738 if ((sv = cv_const_sv(cv))) {
6741 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6742 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6743 pl_yylval.opval->op_private = 0;
6744 pl_yylval.opval->op_flags |= OPf_SPECIAL;
6748 op_free(pl_yylval.opval);
6749 pl_yylval.opval = rv2cv_op;
6750 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6751 PL_last_lop = PL_oldbufptr;
6752 PL_last_lop_op = OP_ENTERSUB;
6753 /* Is there a prototype? */
6761 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
6764 while (*proto == ';')
6769 *proto == '$' || *proto == '_'
6770 || *proto == '*' || *proto == '+'
6775 *proto == '\\' && proto[1] && proto[2] == '\0'
6779 if (*proto == '\\' && proto[1] == '[') {
6780 const char *p = proto + 2;
6781 while(*p && *p != ']')
6783 if(*p == ']' && !p[1]) OPERATOR(UNIOPSUB);
6785 if (*proto == '&' && *s == '{') {
6787 sv_setpvs(PL_subname, "__ANON__");
6789 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6790 if (!PL_lex_allbrackets &&
6791 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6792 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6799 PL_nextwhite = PL_thiswhite;
6802 start_force(PL_curforce);
6803 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6806 PL_nextwhite = nextPL_nextwhite;
6807 curmad('X', PL_thistoken);
6808 PL_thistoken = newSVpvs("");
6811 if (!PL_lex_allbrackets &&
6812 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6813 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6818 /* Guess harder when madskills require "best effort". */
6819 if (PL_madskills && (!gv || !GvCVu(gv))) {
6820 int probable_sub = 0;
6821 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6823 else if (isALPHA(*s)) {
6827 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6828 if (!keyword(tmpbuf, tmplen, 0))
6831 while (d < PL_bufend && isSPACE(*d))
6833 if (*d == '=' && d[1] == '>')
6838 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
6839 op_free(pl_yylval.opval);
6840 pl_yylval.opval = rv2cv_op;
6841 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6842 PL_last_lop = PL_oldbufptr;
6843 PL_last_lop_op = OP_ENTERSUB;
6844 PL_nextwhite = PL_thiswhite;
6846 start_force(PL_curforce);
6847 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6849 PL_nextwhite = nextPL_nextwhite;
6850 curmad('X', PL_thistoken);
6851 PL_thistoken = newSVpvs("");
6853 if (!PL_lex_allbrackets &&
6854 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6855 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6859 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6862 if (!PL_lex_allbrackets &&
6863 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6864 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6869 /* Call it a bare word */
6871 if (PL_hints & HINT_STRICT_SUBS)
6872 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6875 /* after "print" and similar functions (corresponding to
6876 * "F? L" in opcode.pl), whatever wasn't already parsed as
6877 * a filehandle should be subject to "strict subs".
6878 * Likewise for the optional indirect-object argument to system
6879 * or exec, which can't be a bareword */
6880 if ((PL_last_lop_op == OP_PRINT
6881 || PL_last_lop_op == OP_PRTF
6882 || PL_last_lop_op == OP_SAY
6883 || PL_last_lop_op == OP_SYSTEM
6884 || PL_last_lop_op == OP_EXEC)
6885 && (PL_hints & HINT_STRICT_SUBS))
6886 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6887 if (lastchar != '-') {
6888 if (ckWARN(WARN_RESERVED)) {
6892 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
6893 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6901 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
6902 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6903 "Operator or semicolon missing before %c%s",
6904 lastchar, PL_tokenbuf);
6905 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6906 "Ambiguous use of %c resolved as operator %c",
6907 lastchar, lastchar);
6913 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6914 newSVpv(CopFILE(PL_curcop),0));
6918 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6919 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
6922 case KEY___PACKAGE__:
6923 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6925 ? newSVhek(HvNAME_HEK(PL_curstash))
6932 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6933 const char *pname = "main";
6934 if (PL_tokenbuf[2] == 'D')
6935 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
6936 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
6940 GvIOp(gv) = newIO();
6941 IoIFP(GvIOp(gv)) = PL_rsfp;
6942 #if defined(HAS_FCNTL) && defined(F_SETFD)
6944 const int fd = PerlIO_fileno(PL_rsfp);
6945 fcntl(fd,F_SETFD,fd >= 3);
6948 /* Mark this internal pseudo-handle as clean */
6949 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6950 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6951 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6953 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6954 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6955 /* if the script was opened in binmode, we need to revert
6956 * it to text mode for compatibility; but only iff it has CRs
6957 * XXX this is a questionable hack at best. */
6958 if (PL_bufend-PL_bufptr > 2
6959 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6962 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6963 loc = PerlIO_tell(PL_rsfp);
6964 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6967 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6969 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6970 #endif /* NETWARE */
6971 #ifdef PERLIO_IS_STDIO /* really? */
6972 # if defined(__BORLANDC__)
6973 /* XXX see note in do_binmode() */
6974 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
6978 PerlIO_seek(PL_rsfp, loc, 0);
6982 #ifdef PERLIO_LAYERS
6985 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6986 else if (PL_encoding) {
6993 XPUSHs(PL_encoding);
6995 call_method("name", G_SCALAR);
6999 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7000 Perl_form(aTHX_ ":encoding(%"SVf")",
7009 if (PL_realtokenstart >= 0) {
7010 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7012 PL_endwhite = newSVpvs("");
7013 sv_catsv(PL_endwhite, PL_thiswhite);
7015 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
7016 PL_realtokenstart = -1;
7018 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
7034 if (PL_expect == XSTATE) {
7041 if (*s == ':' && s[1] == ':') {
7044 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7045 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
7046 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
7049 else if (tmp == KEY_require || tmp == KEY_do)
7050 /* that's a way to remember we saw "CORE::" */
7063 LOP(OP_ACCEPT,XTERM);
7066 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7071 LOP(OP_ATAN2,XTERM);
7077 LOP(OP_BINMODE,XTERM);
7080 LOP(OP_BLESS,XTERM);
7089 /* When 'use switch' is in effect, continue has a dual
7090 life as a control operator. */
7092 if (!FEATURE_IS_ENABLED("switch"))
7095 /* We have to disambiguate the two senses of
7096 "continue". If the next token is a '{' then
7097 treat it as the start of a continue block;
7098 otherwise treat it as a control operator.
7110 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7120 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7129 if (!PL_cryptseen) {
7130 PL_cryptseen = TRUE;
7134 LOP(OP_CRYPT,XTERM);
7137 LOP(OP_CHMOD,XTERM);
7140 LOP(OP_CHOWN,XTERM);
7143 LOP(OP_CONNECT,XTERM);
7162 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7163 if (orig_keyword == KEY_do) {
7172 PL_hints |= HINT_BLOCK_SCOPE;
7182 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7183 STR_WITH_LEN("NDBM_File::"),
7184 STR_WITH_LEN("DB_File::"),
7185 STR_WITH_LEN("GDBM_File::"),
7186 STR_WITH_LEN("SDBM_File::"),
7187 STR_WITH_LEN("ODBM_File::"),
7189 LOP(OP_DBMOPEN,XTERM);
7195 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7202 pl_yylval.ival = CopLINE(PL_curcop);
7206 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7220 if (*s == '{') { /* block eval */
7221 PL_expect = XTERMBLOCK;
7222 UNIBRACK(OP_ENTERTRY);
7224 else { /* string eval */
7226 UNIBRACK(OP_ENTEREVAL);
7241 case KEY_endhostent:
7247 case KEY_endservent:
7250 case KEY_endprotoent:
7261 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7263 pl_yylval.ival = CopLINE(PL_curcop);
7265 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7268 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
7271 if ((PL_bufend - p) >= 3 &&
7272 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7274 else if ((PL_bufend - p) >= 4 &&
7275 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7278 if (isIDFIRST_lazy_if(p,UTF)) {
7279 p = scan_ident(p, PL_bufend,
7280 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
7284 Perl_croak(aTHX_ "Missing $ on loop variable");
7286 s = SvPVX(PL_linestr) + soff;
7292 LOP(OP_FORMLINE,XTERM);
7298 LOP(OP_FCNTL,XTERM);
7304 LOP(OP_FLOCK,XTERM);
7307 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7312 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7317 LOP(OP_GREPSTART, XREF);
7320 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7335 case KEY_getpriority:
7336 LOP(OP_GETPRIORITY,XTERM);
7338 case KEY_getprotobyname:
7341 case KEY_getprotobynumber:
7342 LOP(OP_GPBYNUMBER,XTERM);
7344 case KEY_getprotoent:
7356 case KEY_getpeername:
7357 UNI(OP_GETPEERNAME);
7359 case KEY_gethostbyname:
7362 case KEY_gethostbyaddr:
7363 LOP(OP_GHBYADDR,XTERM);
7365 case KEY_gethostent:
7368 case KEY_getnetbyname:
7371 case KEY_getnetbyaddr:
7372 LOP(OP_GNBYADDR,XTERM);
7377 case KEY_getservbyname:
7378 LOP(OP_GSBYNAME,XTERM);
7380 case KEY_getservbyport:
7381 LOP(OP_GSBYPORT,XTERM);
7383 case KEY_getservent:
7386 case KEY_getsockname:
7387 UNI(OP_GETSOCKNAME);
7389 case KEY_getsockopt:
7390 LOP(OP_GSOCKOPT,XTERM);
7405 pl_yylval.ival = CopLINE(PL_curcop);
7415 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7417 pl_yylval.ival = CopLINE(PL_curcop);
7421 LOP(OP_INDEX,XTERM);
7427 LOP(OP_IOCTL,XTERM);
7439 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7456 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7461 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7475 LOP(OP_LISTEN,XTERM);
7484 s = scan_pat(s,OP_MATCH);
7485 TERM(sublex_start());
7488 LOP(OP_MAPSTART, XREF);
7491 LOP(OP_MKDIR,XTERM);
7494 LOP(OP_MSGCTL,XTERM);
7497 LOP(OP_MSGGET,XTERM);
7500 LOP(OP_MSGRCV,XTERM);
7503 LOP(OP_MSGSND,XTERM);
7508 PL_in_my = (U16)tmp;
7510 if (isIDFIRST_lazy_if(s,UTF)) {
7514 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7515 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7517 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7518 if (!PL_in_my_stash) {
7521 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7525 if (PL_madskills) { /* just add type to declarator token */
7526 sv_catsv(PL_thistoken, PL_nextwhite);
7528 sv_catpvn(PL_thistoken, start, s - start);
7536 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7540 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7545 s = tokenize_use(0, s);
7549 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7552 if (!PL_lex_allbrackets &&
7553 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7554 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7560 if (isIDFIRST_lazy_if(s,UTF)) {
7562 for (d = s; isALNUM_lazy_if(d,UTF);)
7564 for (t=d; isSPACE(*t);)
7566 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7568 && !(t[0] == '=' && t[1] == '>')
7570 int parms_len = (int)(d-s);
7571 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7572 "Precedence problem: open %.*s should be open(%.*s)",
7573 parms_len, s, parms_len, s);
7579 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7581 pl_yylval.ival = OP_OR;
7591 LOP(OP_OPEN_DIR,XTERM);
7594 checkcomma(s,PL_tokenbuf,"filehandle");
7598 checkcomma(s,PL_tokenbuf,"filehandle");
7617 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7619 s = force_strict_version(s);
7620 PL_lex_expect = XBLOCK;
7624 LOP(OP_PIPE_OP,XTERM);
7627 s = scan_str(s,!!PL_madskills,FALSE);
7630 pl_yylval.ival = OP_CONST;
7631 TERM(sublex_start());
7638 s = scan_str(s,!!PL_madskills,FALSE);
7641 PL_expect = XOPERATOR;
7642 if (SvCUR(PL_lex_stuff)) {
7644 d = SvPV_force(PL_lex_stuff, len);
7646 for (; isSPACE(*d) && len; --len, ++d)
7651 if (!warned && ckWARN(WARN_QW)) {
7652 for (; !isSPACE(*d) && len; --len, ++d) {
7654 Perl_warner(aTHX_ packWARN(WARN_QW),
7655 "Possible attempt to separate words with commas");
7658 else if (*d == '#') {
7659 Perl_warner(aTHX_ packWARN(WARN_QW),
7660 "Possible attempt to put comments in qw() list");
7666 for (; !isSPACE(*d) && len; --len, ++d)
7669 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7670 words = op_append_elem(OP_LIST, words,
7671 newSVOP(OP_CONST, 0, tokeq(sv)));
7676 words = newNULLLIST();
7678 SvREFCNT_dec(PL_lex_stuff);
7679 PL_lex_stuff = NULL;
7681 PL_expect = XOPERATOR;
7682 pl_yylval.opval = sawparens(words);
7687 s = scan_str(s,!!PL_madskills,FALSE);
7690 pl_yylval.ival = OP_STRINGIFY;
7691 if (SvIVX(PL_lex_stuff) == '\'')
7692 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
7693 TERM(sublex_start());
7696 s = scan_pat(s,OP_QR);
7697 TERM(sublex_start());
7700 s = scan_str(s,!!PL_madskills,FALSE);
7703 readpipe_override();
7704 TERM(sublex_start());
7712 s = force_version(s, FALSE);
7714 else if (*s != 'v' || !isDIGIT(s[1])
7715 || (s = force_version(s, TRUE), *s == 'v'))
7717 *PL_tokenbuf = '\0';
7718 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7719 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7720 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7722 yyerror("<> should be quotes");
7724 if (orig_keyword == KEY_require) {
7732 PL_last_uni = PL_oldbufptr;
7733 PL_last_lop_op = OP_REQUIRE;
7735 return REPORT( (int)REQUIRE );
7741 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7745 LOP(OP_RENAME,XTERM);
7754 LOP(OP_RINDEX,XTERM);
7763 UNIDOR(OP_READLINE);
7766 UNIDOR(OP_BACKTICK);
7775 LOP(OP_REVERSE,XTERM);
7778 UNIDOR(OP_READLINK);
7785 if (pl_yylval.opval)
7786 TERM(sublex_start());
7788 TOKEN(1); /* force error */
7791 checkcomma(s,PL_tokenbuf,"filehandle");
7801 LOP(OP_SELECT,XTERM);
7807 LOP(OP_SEMCTL,XTERM);
7810 LOP(OP_SEMGET,XTERM);
7813 LOP(OP_SEMOP,XTERM);
7819 LOP(OP_SETPGRP,XTERM);
7821 case KEY_setpriority:
7822 LOP(OP_SETPRIORITY,XTERM);
7824 case KEY_sethostent:
7830 case KEY_setservent:
7833 case KEY_setprotoent:
7843 LOP(OP_SEEKDIR,XTERM);
7845 case KEY_setsockopt:
7846 LOP(OP_SSOCKOPT,XTERM);
7852 LOP(OP_SHMCTL,XTERM);
7855 LOP(OP_SHMGET,XTERM);
7858 LOP(OP_SHMREAD,XTERM);
7861 LOP(OP_SHMWRITE,XTERM);
7864 LOP(OP_SHUTDOWN,XTERM);
7873 LOP(OP_SOCKET,XTERM);
7875 case KEY_socketpair:
7876 LOP(OP_SOCKPAIR,XTERM);
7879 checkcomma(s,PL_tokenbuf,"subroutine name");
7881 if (*s == ';' || *s == ')') /* probably a close */
7882 Perl_croak(aTHX_ "sort is now a reserved word");
7884 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7888 LOP(OP_SPLIT,XTERM);
7891 LOP(OP_SPRINTF,XTERM);
7894 LOP(OP_SPLICE,XTERM);
7909 LOP(OP_SUBSTR,XTERM);
7915 char tmpbuf[sizeof PL_tokenbuf];
7916 SSize_t tboffset = 0;
7917 expectation attrful;
7918 bool have_name, have_proto;
7919 const int key = tmp;
7924 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7925 SV *subtoken = newSVpvn(tstart, s - tstart);
7929 s = SKIPSPACE2(s,tmpwhite);
7934 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7935 (*s == ':' && s[1] == ':'))
7938 SV *nametoke = NULL;
7942 attrful = XATTRBLOCK;
7943 /* remember buffer pos'n for later force_word */
7944 tboffset = s - PL_oldbufptr;
7945 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7948 nametoke = newSVpvn(s, d - s);
7950 if (memchr(tmpbuf, ':', len))
7951 sv_setpvn(PL_subname, tmpbuf, len);
7953 sv_setsv(PL_subname,PL_curstname);
7954 sv_catpvs(PL_subname,"::");
7955 sv_catpvn(PL_subname,tmpbuf,len);
7962 CURMAD('X', nametoke);
7963 CURMAD('_', tmpwhite);
7964 (void) force_word(PL_oldbufptr + tboffset, WORD,
7967 s = SKIPSPACE2(d,tmpwhite);
7974 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7975 PL_expect = XTERMBLOCK;
7976 attrful = XATTRTERM;
7977 sv_setpvs(PL_subname,"?");
7981 if (key == KEY_format) {
7983 PL_lex_formbrack = PL_lex_brackets + 1;
7985 PL_thistoken = subtoken;
7989 (void) force_word(PL_oldbufptr + tboffset, WORD,
7995 /* Look for a prototype */
7998 bool bad_proto = FALSE;
7999 bool in_brackets = FALSE;
8000 char greedy_proto = ' ';
8001 bool proto_after_greedy_proto = FALSE;
8002 bool must_be_last = FALSE;
8003 bool underscore = FALSE;
8004 bool seen_underscore = FALSE;
8005 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
8007 s = scan_str(s,!!PL_madskills,FALSE);
8009 Perl_croak(aTHX_ "Prototype not terminated");
8010 /* strip spaces and check for bad characters */
8011 d = SvPVX(PL_lex_stuff);
8013 for (p = d; *p; ++p) {
8017 if (warnillegalproto) {
8019 proto_after_greedy_proto = TRUE;
8020 if (!strchr("$@%*;[]&\\_+", *p)) {
8032 else if ( *p == ']' ) {
8033 in_brackets = FALSE;
8035 else if ( (*p == '@' || *p == '%') &&
8036 ( tmp < 2 || d[tmp-2] != '\\' ) &&
8038 must_be_last = TRUE;
8041 else if ( *p == '_' ) {
8042 underscore = seen_underscore = TRUE;
8049 if (proto_after_greedy_proto)
8050 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8051 "Prototype after '%c' for %"SVf" : %s",
8052 greedy_proto, SVfARG(PL_subname), d);
8054 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8055 "Illegal character %sin prototype for %"SVf" : %s",
8056 seen_underscore ? "after '_' " : "",
8057 SVfARG(PL_subname), d);
8058 SvCUR_set(PL_lex_stuff, tmp);
8063 CURMAD('q', PL_thisopen);
8064 CURMAD('_', tmpwhite);
8065 CURMAD('=', PL_thisstuff);
8066 CURMAD('Q', PL_thisclose);
8067 NEXTVAL_NEXTTOKE.opval =
8068 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8069 PL_lex_stuff = NULL;
8072 s = SKIPSPACE2(s,tmpwhite);
8080 if (*s == ':' && s[1] != ':')
8081 PL_expect = attrful;
8082 else if (*s != '{' && key == KEY_sub) {
8084 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8085 else if (*s != ';' && *s != '}')
8086 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8093 curmad('^', newSVpvs(""));
8094 CURMAD('_', tmpwhite);
8098 PL_thistoken = subtoken;
8101 NEXTVAL_NEXTTOKE.opval =
8102 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8103 PL_lex_stuff = NULL;
8109 sv_setpvs(PL_subname, "__ANON__");
8111 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8115 (void) force_word(PL_oldbufptr + tboffset, WORD,
8124 LOP(OP_SYSTEM,XREF);
8127 LOP(OP_SYMLINK,XTERM);
8130 LOP(OP_SYSCALL,XTERM);
8133 LOP(OP_SYSOPEN,XTERM);
8136 LOP(OP_SYSSEEK,XTERM);
8139 LOP(OP_SYSREAD,XTERM);
8142 LOP(OP_SYSWRITE,XTERM);
8146 TERM(sublex_start());
8167 LOP(OP_TRUNCATE,XTERM);
8179 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8181 pl_yylval.ival = CopLINE(PL_curcop);
8185 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8187 pl_yylval.ival = CopLINE(PL_curcop);
8191 LOP(OP_UNLINK,XTERM);
8197 LOP(OP_UNPACK,XTERM);
8200 LOP(OP_UTIME,XTERM);
8206 LOP(OP_UNSHIFT,XTERM);
8209 s = tokenize_use(1, s);
8219 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8221 pl_yylval.ival = CopLINE(PL_curcop);
8225 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8227 pl_yylval.ival = CopLINE(PL_curcop);
8231 PL_hints |= HINT_BLOCK_SCOPE;
8238 LOP(OP_WAITPID,XTERM);
8247 ctl_l[0] = toCTRL('L');
8249 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
8252 /* Make sure $^L is defined */
8253 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
8258 if (PL_expect == XOPERATOR) {
8259 if (*s == '=' && !PL_lex_allbrackets &&
8260 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8268 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8270 pl_yylval.ival = OP_XOR;
8275 TERM(sublex_start());
8280 #pragma segment Main
8284 S_pending_ident(pTHX)
8289 /* pit holds the identifier we read and pending_ident is reset */
8290 char pit = PL_pending_ident;
8291 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8292 /* All routes through this function want to know if there is a colon. */
8293 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8294 PL_pending_ident = 0;
8296 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
8297 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8298 "### Pending identifier '%s'\n", PL_tokenbuf); });
8300 /* if we're in a my(), we can't allow dynamics here.
8301 $foo'bar has already been turned into $foo::bar, so
8302 just check for colons.
8304 if it's a legal name, the OP is a PADANY.
8307 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8309 yyerror(Perl_form(aTHX_ "No package name allowed for "
8310 "variable %s in \"our\"",
8312 tmp = allocmy(PL_tokenbuf, tokenbuf_len, 0);
8316 yyerror(Perl_form(aTHX_ PL_no_myglob,
8317 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
8319 pl_yylval.opval = newOP(OP_PADANY, 0);
8320 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len, 0);
8326 build the ops for accesses to a my() variable.
8328 Deny my($a) or my($b) in a sort block, *if* $a or $b is
8329 then used in a comparison. This catches most, but not
8330 all cases. For instance, it catches
8331 sort { my($a); $a <=> $b }
8333 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
8334 (although why you'd do that is anyone's guess).
8339 tmp = pad_findmy(PL_tokenbuf, tokenbuf_len, 0);
8340 if (tmp != NOT_IN_PAD) {
8341 /* might be an "our" variable" */
8342 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8343 /* build ops for a bareword */
8344 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8345 HEK * const stashname = HvNAME_HEK(stash);
8346 SV * const sym = newSVhek(stashname);
8347 sv_catpvs(sym, "::");
8348 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
8349 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8350 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8353 ? (GV_ADDMULTI | GV_ADDINEVAL)
8356 ((PL_tokenbuf[0] == '$') ? SVt_PV
8357 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8362 /* if it's a sort block and they're naming $a or $b */
8363 if (PL_last_lop_op == OP_SORT &&
8364 PL_tokenbuf[0] == '$' &&
8365 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
8368 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
8369 d < PL_bufend && *d != '\n';
8372 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
8373 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
8379 pl_yylval.opval = newOP(OP_PADANY, 0);
8380 pl_yylval.opval->op_targ = tmp;
8386 Whine if they've said @foo in a doublequoted string,
8387 and @foo isn't a variable we can find in the symbol
8390 if (ckWARN(WARN_AMBIGUOUS) &&
8391 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8392 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
8394 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8395 /* DO NOT warn for @- and @+ */
8396 && !( PL_tokenbuf[2] == '\0' &&
8397 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8400 /* Downgraded from fatal to warning 20000522 mjd */
8401 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8402 "Possible unintended interpolation of %s in string",
8407 /* build ops for a bareword */
8408 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
8410 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8411 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8412 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD,
8413 ((PL_tokenbuf[0] == '$') ? SVt_PV
8414 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8420 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8424 PERL_ARGS_ASSERT_CHECKCOMMA;
8426 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8427 if (ckWARN(WARN_SYNTAX)) {
8430 for (w = s+2; *w && level; w++) {
8438 /* the list of chars below is for end of statements or
8439 * block / parens, boolean operators (&&, ||, //) and branch
8440 * constructs (or, and, if, until, unless, while, err, for).
8441 * Not a very solid hack... */
8442 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8443 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8444 "%s (...) interpreted as function",name);
8447 while (s < PL_bufend && isSPACE(*s))
8451 while (s < PL_bufend && isSPACE(*s))
8453 if (isIDFIRST_lazy_if(s,UTF)) {
8454 const char * const w = s++;
8455 while (isALNUM_lazy_if(s,UTF))
8457 while (s < PL_bufend && isSPACE(*s))
8461 if (keyword(w, s - w, 0))
8464 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
8465 if (gv && GvCVu(gv))
8467 Perl_croak(aTHX_ "No comma allowed after %s", what);
8472 /* Either returns sv, or mortalizes sv and returns a new SV*.
8473 Best used as sv=new_constant(..., sv, ...).
8474 If s, pv are NULL, calls subroutine with one argument,
8475 and type is used with error messages only. */
8478 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
8479 SV *sv, SV *pv, const char *type, STRLEN typelen)
8482 HV * const table = GvHV(PL_hintgv); /* ^H */
8486 const char *why1 = "", *why2 = "", *why3 = "";
8488 PERL_ARGS_ASSERT_NEW_CONSTANT;
8490 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8493 why2 = (const char *)
8494 (strEQ(key,"charnames")
8495 ? "(possibly a missing \"use charnames ...\")"
8497 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
8498 (type ? type: "undef"), why2);
8500 /* This is convoluted and evil ("goto considered harmful")
8501 * but I do not understand the intricacies of all the different
8502 * failure modes of %^H in here. The goal here is to make
8503 * the most probable error message user-friendly. --jhi */
8508 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
8509 (type ? type: "undef"), why1, why2, why3);
8511 yyerror(SvPVX_const(msg));
8516 /* charnames doesn't work well if there have been errors found */
8517 if (PL_error_count > 0 && strEQ(key,"charnames"))
8518 return &PL_sv_undef;
8520 cvp = hv_fetch(table, key, keylen, FALSE);
8521 if (!cvp || !SvOK(*cvp)) {
8524 why3 = "} is not defined";
8527 sv_2mortal(sv); /* Parent created it permanently */
8530 pv = newSVpvn_flags(s, len, SVs_TEMP);
8532 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
8534 typesv = &PL_sv_undef;
8536 PUSHSTACKi(PERLSI_OVERLOAD);
8548 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
8552 /* Check the eval first */
8553 if (!PL_in_eval && SvTRUE(ERRSV)) {
8554 sv_catpvs(ERRSV, "Propagated");
8555 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
8557 res = SvREFCNT_inc_simple(sv);
8561 SvREFCNT_inc_simple_void(res);
8570 why1 = "Call to &{$^H{";
8572 why3 = "}} did not return a defined value";
8580 /* Returns a NUL terminated string, with the length of the string written to
8584 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
8587 register char *d = dest;
8588 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
8590 PERL_ARGS_ASSERT_SCAN_WORD;
8594 Perl_croak(aTHX_ ident_too_long);
8595 if (isALNUM(*s)) /* UTF handled below */
8597 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
8602 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
8606 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
8607 char *t = s + UTF8SKIP(s);
8609 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
8613 Perl_croak(aTHX_ ident_too_long);
8614 Copy(s, d, len, char);
8627 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
8630 char *bracket = NULL;
8632 register char *d = dest;
8633 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
8635 PERL_ARGS_ASSERT_SCAN_IDENT;
8640 while (isDIGIT(*s)) {
8642 Perl_croak(aTHX_ ident_too_long);
8649 Perl_croak(aTHX_ ident_too_long);
8650 if (isALNUM(*s)) /* UTF handled below */
8652 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
8657 else if (*s == ':' && s[1] == ':') {
8661 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
8662 char *t = s + UTF8SKIP(s);
8663 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
8665 if (d + (t - s) > e)
8666 Perl_croak(aTHX_ ident_too_long);
8667 Copy(s, d, t - s, char);
8678 if (PL_lex_state != LEX_NORMAL)
8679 PL_lex_state = LEX_INTERPENDMAYBE;
8682 if (*s == '$' && s[1] &&
8683 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
8696 if (*d == '^' && *s && isCONTROLVAR(*s)) {
8701 if (isSPACE(s[-1])) {
8703 const char ch = *s++;
8704 if (!SPACE_OR_TAB(ch)) {
8710 if (isIDFIRST_lazy_if(d,UTF)) {
8714 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
8715 end += UTF8SKIP(end);
8716 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
8717 end += UTF8SKIP(end);
8719 Copy(s, d, end - s, char);
8724 while ((isALNUM(*s) || *s == ':') && d < e)
8727 Perl_croak(aTHX_ ident_too_long);
8730 while (s < send && SPACE_OR_TAB(*s))
8732 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
8733 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
8734 const char * const brack =
8736 ((*s == '[') ? "[...]" : "{...}");
8737 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
8738 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8739 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
8740 funny, dest, brack, funny, dest, brack);
8743 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
8744 PL_lex_allbrackets++;
8748 /* Handle extended ${^Foo} variables
8749 * 1999-02-27 mjd-perl-patch@plover.com */
8750 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
8754 while (isALNUM(*s) && d < e) {
8758 Perl_croak(aTHX_ ident_too_long);
8763 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
8764 PL_lex_state = LEX_INTERPEND;
8767 if (PL_lex_state == LEX_NORMAL) {
8768 if (ckWARN(WARN_AMBIGUOUS) &&
8769 (keyword(dest, d - dest, 0)
8770 || get_cvn_flags(dest, d - dest, 0)))
8774 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8775 "Ambiguous use of %c{%s} resolved to %c%s",
8776 funny, dest, funny, dest);
8781 s = bracket; /* let the parser handle it */
8785 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
8786 PL_lex_state = LEX_INTERPEND;
8791 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
8793 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
8794 * the parse starting at 's', based on the subset that are valid in this
8795 * context input to this routine in 'valid_flags'. Advances s. Returns
8796 * TRUE if the input was a valid flag, so the next char may be as well;
8797 * otherwise FALSE. 'charset' should point to a NUL upon first call on the
8798 * current regex. This routine will set it to any charset modifier found.
8799 * The caller shouldn't change it. This way, another charset modifier
8800 * encountered in the parse can be detected as an error, as we have decided
8805 if (! strchr(valid_flags, c)) {
8814 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
8815 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
8816 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
8817 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
8818 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
8819 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
8820 case LOCALE_PAT_MOD:
8822 /* In 5.14, qr//lt is legal but deprecated; the 't' means they
8823 * can't be regex modifiers.
8824 * In 5.14, s///le is legal and ambiguous. Try to disambiguate as
8825 * much as easily done. s///lei, for example, has to mean regex
8826 * modifiers if it's not an error (as does any word character
8827 * following the 'e'). Otherwise, we resolve to the backwards-
8828 * compatible, but less likely 's/// le ...', i.e. as meaning
8829 * less-than-or-equal. The reason it's not likely is that s//
8830 * returns a number for code in the field (/r returns a string, but
8831 * that wasn't added until the 5.13 series), and so '<=' should be
8832 * used for comparing, not 'le'. */
8833 if (*((*s) + 1) == 't') {
8836 else if (*((*s) + 1) == 'e' && ! isALNUM(*((*s) + 2))) {
8838 /* 'e' is valid only for substitutes, s///e. If it is not
8839 * valid in the current context, then 'm//le' must mean the
8840 * comparison operator, so use the regular deprecation message.
8842 if (! strchr(valid_flags, 'e')) {
8845 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
8846 "Ambiguous use of 's//le...' resolved as 's// le...'; Rewrite as 's//el' if you meant 'use locale rules and evaluate rhs as an expression'. In Perl 5.16, it will be resolved the other way");
8850 goto multiple_charsets;
8852 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
8855 case UNICODE_PAT_MOD:
8856 /* In 5.14, qr//unless and qr//until are legal but deprecated; the
8857 * 'n' means they can't be regex modifiers */
8858 if (*((*s) + 1) == 'n') {
8862 goto multiple_charsets;
8864 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
8867 case ASCII_RESTRICT_PAT_MOD:
8868 /* In 5.14, qr//and is legal but deprecated; the 'n' means they
8869 * can't be regex modifiers */
8870 if (*((*s) + 1) == 'n') {
8873 if (*((*s) + 1) == ASCII_RESTRICT_PAT_MOD) {
8874 /* Doubled modifier implies more restricted */
8875 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
8879 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
8881 if (*charset) { /* Do this after the increment of *s in /aa, so
8882 the return advances the ptr correctly */
8883 goto multiple_charsets;
8887 case DEPENDS_PAT_MOD:
8889 goto multiple_charsets;
8891 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
8900 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX),
8901 "Having no space between pattern and following word is deprecated");
8905 if (*charset != c) {
8906 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
8909 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
8912 /* Pretend that it worked, so will continue processing before dieing */
8918 S_scan_pat(pTHX_ char *start, I32 type)
8922 char *s = scan_str(start,!!PL_madskills,FALSE);
8923 const char * const valid_flags =
8924 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
8925 char charset = '\0'; /* character set modifier */
8930 PERL_ARGS_ASSERT_SCAN_PAT;
8933 const char * const delimiter = skipspace(start);
8937 ? "Search pattern not terminated or ternary operator parsed as search pattern"
8938 : "Search pattern not terminated" ));
8941 pm = (PMOP*)newPMOP(type, 0);
8942 if (PL_multi_open == '?') {
8943 /* This is the only point in the code that sets PMf_ONCE: */
8944 pm->op_pmflags |= PMf_ONCE;
8946 /* Hence it's safe to do this bit of PMOP book-keeping here, which
8947 allows us to restrict the list needed by reset to just the ??
8949 assert(type != OP_TRANS);
8951 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
8954 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
8957 elements = mg->mg_len / sizeof(PMOP**);
8958 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
8959 ((PMOP**)mg->mg_ptr) [elements++] = pm;
8960 mg->mg_len = elements * sizeof(PMOP**);
8961 PmopSTASH_set(pm,PL_curstash);
8967 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
8969 if (PL_madskills && modstart != s) {
8970 SV* tmptoken = newSVpvn(modstart, s - modstart);
8971 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
8974 /* issue a warning if /c is specified,but /g is not */
8975 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
8977 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
8978 "Use of /c modifier is meaningless without /g" );
8981 PL_lex_op = (OP*)pm;
8982 pl_yylval.ival = OP_MATCH;
8987 S_scan_subst(pTHX_ char *start)
8994 char charset = '\0'; /* character set modifier */
8999 PERL_ARGS_ASSERT_SCAN_SUBST;
9001 pl_yylval.ival = OP_NULL;
9003 s = scan_str(start,!!PL_madskills,FALSE);
9006 Perl_croak(aTHX_ "Substitution pattern not terminated");
9008 if (s[-1] == PL_multi_open)
9012 CURMAD('q', PL_thisopen);
9013 CURMAD('_', PL_thiswhite);
9014 CURMAD('E', PL_thisstuff);
9015 CURMAD('Q', PL_thisclose);
9016 PL_realtokenstart = s - SvPVX(PL_linestr);
9020 first_start = PL_multi_start;
9021 s = scan_str(s,!!PL_madskills,FALSE);
9024 SvREFCNT_dec(PL_lex_stuff);
9025 PL_lex_stuff = NULL;
9027 Perl_croak(aTHX_ "Substitution replacement not terminated");
9029 PL_multi_start = first_start; /* so whole substitution is taken together */
9031 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9035 CURMAD('z', PL_thisopen);
9036 CURMAD('R', PL_thisstuff);
9037 CURMAD('Z', PL_thisclose);
9043 if (*s == EXEC_PAT_MOD) {
9047 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
9056 curmad('m', newSVpvn(modstart, s - modstart));
9057 append_madprops(PL_thismad, (OP*)pm, 0);
9061 if ((pm->op_pmflags & PMf_CONTINUE)) {
9062 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9066 SV * const repl = newSVpvs("");
9068 PL_sublex_info.super_bufptr = s;
9069 PL_sublex_info.super_bufend = PL_bufend;
9071 pm->op_pmflags |= PMf_EVAL;
9074 sv_catpvs(repl, "eval ");
9076 sv_catpvs(repl, "do ");
9078 sv_catpvs(repl, "{");
9079 sv_catsv(repl, PL_lex_repl);
9080 if (strchr(SvPVX(PL_lex_repl), '#'))
9081 sv_catpvs(repl, "\n");
9082 sv_catpvs(repl, "}");
9084 SvREFCNT_dec(PL_lex_repl);
9088 PL_lex_op = (OP*)pm;
9089 pl_yylval.ival = OP_SUBST;
9094 S_scan_trans(pTHX_ char *start)
9103 bool nondestruct = 0;
9108 PERL_ARGS_ASSERT_SCAN_TRANS;
9110 pl_yylval.ival = OP_NULL;
9112 s = scan_str(start,!!PL_madskills,FALSE);
9114 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9116 if (s[-1] == PL_multi_open)
9120 CURMAD('q', PL_thisopen);
9121 CURMAD('_', PL_thiswhite);
9122 CURMAD('E', PL_thisstuff);
9123 CURMAD('Q', PL_thisclose);
9124 PL_realtokenstart = s - SvPVX(PL_linestr);
9128 s = scan_str(s,!!PL_madskills,FALSE);
9131 SvREFCNT_dec(PL_lex_stuff);
9132 PL_lex_stuff = NULL;
9134 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9137 CURMAD('z', PL_thisopen);
9138 CURMAD('R', PL_thisstuff);
9139 CURMAD('Z', PL_thisclose);
9142 complement = del = squash = 0;
9149 complement = OPpTRANS_COMPLEMENT;
9152 del = OPpTRANS_DELETE;
9155 squash = OPpTRANS_SQUASH;
9167 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
9168 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)tbl);
9169 o->op_private &= ~OPpTRANS_ALL;
9170 o->op_private |= del|squash|complement|
9171 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9172 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
9175 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9180 curmad('m', newSVpvn(modstart, s - modstart));
9181 append_madprops(PL_thismad, o, 0);
9190 S_scan_heredoc(pTHX_ register char *s)
9194 I32 op_type = OP_SCALAR;
9198 const char *found_newline;
9202 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
9204 I32 stuffstart = s - SvPVX(PL_linestr);
9207 PL_realtokenstart = -1;
9210 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9214 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9218 while (SPACE_OR_TAB(*peek))
9220 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9223 s = delimcpy(d, e, s, PL_bufend, term, &len);
9233 if (!isALNUM_lazy_if(s,UTF))
9234 deprecate("bare << to mean <<\"\"");
9235 for (; isALNUM_lazy_if(s,UTF); s++) {
9240 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9241 Perl_croak(aTHX_ "Delimiter for here document is too long");
9244 len = d - PL_tokenbuf;
9248 tstart = PL_tokenbuf + !outer;
9249 PL_thisclose = newSVpvn(tstart, len - !outer);
9250 tstart = SvPVX(PL_linestr) + stuffstart;
9251 PL_thisopen = newSVpvn(tstart, s - tstart);
9252 stuffstart = s - SvPVX(PL_linestr);
9255 #ifndef PERL_STRICT_CR
9256 d = strchr(s, '\r');
9258 char * const olds = s;
9260 while (s < PL_bufend) {
9266 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9275 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9282 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
9283 herewas = newSVpvn(s,PL_bufend-s);
9287 herewas = newSVpvn(s-1,found_newline-s+1);
9290 herewas = newSVpvn(s,found_newline-s);
9295 tstart = SvPVX(PL_linestr) + stuffstart;
9297 sv_catpvn(PL_thisstuff, tstart, s - tstart);
9299 PL_thisstuff = newSVpvn(tstart, s - tstart);
9302 s += SvCUR(herewas);
9305 stuffstart = s - SvPVX(PL_linestr);
9311 tmpstr = newSV_type(SVt_PVIV);
9315 SvIV_set(tmpstr, -1);
9317 else if (term == '`') {
9318 op_type = OP_BACKTICK;
9319 SvIV_set(tmpstr, '\\');
9323 PL_multi_start = CopLINE(PL_curcop);
9324 PL_multi_open = PL_multi_close = '<';
9325 term = *PL_tokenbuf;
9326 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
9327 char * const bufptr = PL_sublex_info.super_bufptr;
9328 char * const bufend = PL_sublex_info.super_bufend;
9329 char * const olds = s - SvCUR(herewas);
9330 s = strchr(bufptr, '\n');
9334 while (s < bufend &&
9335 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9337 CopLINE_inc(PL_curcop);
9340 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9341 missingterm(PL_tokenbuf);
9343 sv_setpvn(herewas,bufptr,d-bufptr+1);
9344 sv_setpvn(tmpstr,d+1,s-d);
9346 sv_catpvn(herewas,s,bufend-s);
9347 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
9354 while (s < PL_bufend &&
9355 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9357 CopLINE_inc(PL_curcop);
9359 if (s >= PL_bufend) {
9360 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9361 missingterm(PL_tokenbuf);
9363 sv_setpvn(tmpstr,d+1,s-d);
9367 sv_catpvn(PL_thisstuff, d + 1, s - d);
9369 PL_thisstuff = newSVpvn(d + 1, s - d);
9370 stuffstart = s - SvPVX(PL_linestr);
9374 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
9376 sv_catpvn(herewas,s,PL_bufend-s);
9377 sv_setsv(PL_linestr,herewas);
9378 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
9379 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9380 PL_last_lop = PL_last_uni = NULL;
9383 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9384 while (s >= PL_bufend) { /* multiple line string? */
9387 tstart = SvPVX(PL_linestr) + stuffstart;
9389 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
9391 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
9395 CopLINE_inc(PL_curcop);
9396 if (!outer || !lex_next_chunk(0)) {
9397 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9398 missingterm(PL_tokenbuf);
9400 CopLINE_dec(PL_curcop);
9403 stuffstart = s - SvPVX(PL_linestr);
9405 CopLINE_inc(PL_curcop);
9406 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9407 PL_last_lop = PL_last_uni = NULL;
9408 #ifndef PERL_STRICT_CR
9409 if (PL_bufend - PL_linestart >= 2) {
9410 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
9411 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9413 PL_bufend[-2] = '\n';
9415 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9417 else if (PL_bufend[-1] == '\r')
9418 PL_bufend[-1] = '\n';
9420 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9421 PL_bufend[-1] = '\n';
9423 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
9424 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
9425 *(SvPVX(PL_linestr) + off ) = ' ';
9426 sv_catsv(PL_linestr,herewas);
9427 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9428 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
9432 sv_catsv(tmpstr,PL_linestr);
9437 PL_multi_end = CopLINE(PL_curcop);
9438 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9439 SvPV_shrink_to_cur(tmpstr);
9441 SvREFCNT_dec(herewas);
9443 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9445 else if (PL_encoding)
9446 sv_recode_to_utf8(tmpstr, PL_encoding);
9448 PL_lex_stuff = tmpstr;
9449 pl_yylval.ival = op_type;
9454 takes: current position in input buffer
9455 returns: new position in input buffer
9456 side-effects: pl_yylval and lex_op are set.
9461 <FH> read from filehandle
9462 <pkg::FH> read from package qualified filehandle
9463 <pkg'FH> read from package qualified filehandle
9464 <$fh> read from filehandle in $fh
9470 S_scan_inputsymbol(pTHX_ char *start)
9473 register char *s = start; /* current position in buffer */
9476 char *d = PL_tokenbuf; /* start of temp holding space */
9477 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9479 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
9481 end = strchr(s, '\n');
9484 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9486 /* die if we didn't have space for the contents of the <>,
9487 or if it didn't end, or if we see a newline
9490 if (len >= (I32)sizeof PL_tokenbuf)
9491 Perl_croak(aTHX_ "Excessively long <> operator");
9493 Perl_croak(aTHX_ "Unterminated <> operator");
9498 Remember, only scalar variables are interpreted as filehandles by
9499 this code. Anything more complex (e.g., <$fh{$num}>) will be
9500 treated as a glob() call.
9501 This code makes use of the fact that except for the $ at the front,
9502 a scalar variable and a filehandle look the same.
9504 if (*d == '$' && d[1]) d++;
9506 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9507 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9510 /* If we've tried to read what we allow filehandles to look like, and
9511 there's still text left, then it must be a glob() and not a getline.
9512 Use scan_str to pull out the stuff between the <> and treat it
9513 as nothing more than a string.
9516 if (d - PL_tokenbuf != len) {
9517 pl_yylval.ival = OP_GLOB;
9518 s = scan_str(start,!!PL_madskills,FALSE);
9520 Perl_croak(aTHX_ "Glob not terminated");
9524 bool readline_overriden = FALSE;
9527 /* we're in a filehandle read situation */
9530 /* turn <> into <ARGV> */
9532 Copy("ARGV",d,5,char);
9534 /* Check whether readline() is overriden */
9535 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
9537 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
9539 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
9540 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
9541 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
9542 readline_overriden = TRUE;
9544 /* if <$fh>, create the ops to turn the variable into a
9548 /* try to find it in the pad for this block, otherwise find
9549 add symbol table ops
9551 const PADOFFSET tmp = pad_findmy(d, len, 0);
9552 if (tmp != NOT_IN_PAD) {
9553 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
9554 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
9555 HEK * const stashname = HvNAME_HEK(stash);
9556 SV * const sym = sv_2mortal(newSVhek(stashname));
9557 sv_catpvs(sym, "::");
9563 OP * const o = newOP(OP_PADSV, 0);
9565 PL_lex_op = readline_overriden
9566 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9567 op_append_elem(OP_LIST, o,
9568 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
9569 : (OP*)newUNOP(OP_READLINE, 0, o);
9578 ? (GV_ADDMULTI | GV_ADDINEVAL)
9581 PL_lex_op = readline_overriden
9582 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9583 op_append_elem(OP_LIST,
9584 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
9585 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9586 : (OP*)newUNOP(OP_READLINE, 0,
9587 newUNOP(OP_RV2SV, 0,
9588 newGVOP(OP_GV, 0, gv)));
9590 if (!readline_overriden)
9591 PL_lex_op->op_flags |= OPf_SPECIAL;
9592 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
9593 pl_yylval.ival = OP_NULL;
9596 /* If it's none of the above, it must be a literal filehandle
9597 (<Foo::BAR> or <FOO>) so build a simple readline OP */
9599 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
9600 PL_lex_op = readline_overriden
9601 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9602 op_append_elem(OP_LIST,
9603 newGVOP(OP_GV, 0, gv),
9604 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9605 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
9606 pl_yylval.ival = OP_NULL;
9615 takes: start position in buffer
9616 keep_quoted preserve \ on the embedded delimiter(s)
9617 keep_delims preserve the delimiters around the string
9618 returns: position to continue reading from buffer
9619 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
9620 updates the read buffer.
9622 This subroutine pulls a string out of the input. It is called for:
9623 q single quotes q(literal text)
9624 ' single quotes 'literal text'
9625 qq double quotes qq(interpolate $here please)
9626 " double quotes "interpolate $here please"
9627 qx backticks qx(/bin/ls -l)
9628 ` backticks `/bin/ls -l`
9629 qw quote words @EXPORT_OK = qw( func() $spam )
9630 m// regexp match m/this/
9631 s/// regexp substitute s/this/that/
9632 tr/// string transliterate tr/this/that/
9633 y/// string transliterate y/this/that/
9634 ($*@) sub prototypes sub foo ($)
9635 (stuff) sub attr parameters sub foo : attr(stuff)
9636 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
9638 In most of these cases (all but <>, patterns and transliterate)
9639 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
9640 calls scan_str(). s/// makes yylex() call scan_subst() which calls
9641 scan_str(). tr/// and y/// make yylex() call scan_trans() which
9644 It skips whitespace before the string starts, and treats the first
9645 character as the delimiter. If the delimiter is one of ([{< then
9646 the corresponding "close" character )]}> is used as the closing
9647 delimiter. It allows quoting of delimiters, and if the string has
9648 balanced delimiters ([{<>}]) it allows nesting.
9650 On success, the SV with the resulting string is put into lex_stuff or,
9651 if that is already non-NULL, into lex_repl. The second case occurs only
9652 when parsing the RHS of the special constructs s/// and tr/// (y///).
9653 For convenience, the terminating delimiter character is stuffed into
9658 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
9661 SV *sv; /* scalar value: string */
9662 const char *tmps; /* temp string, used for delimiter matching */
9663 register char *s = start; /* current position in the buffer */
9664 register char term; /* terminating character */
9665 register char *to; /* current position in the sv's data */
9666 I32 brackets = 1; /* bracket nesting level */
9667 bool has_utf8 = FALSE; /* is there any utf8 content? */
9668 I32 termcode; /* terminating char. code */
9669 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
9670 STRLEN termlen; /* length of terminating string */
9671 int last_off = 0; /* last position for nesting bracket */
9677 PERL_ARGS_ASSERT_SCAN_STR;
9679 /* skip space before the delimiter */
9685 if (PL_realtokenstart >= 0) {
9686 stuffstart = PL_realtokenstart;
9687 PL_realtokenstart = -1;
9690 stuffstart = start - SvPVX(PL_linestr);
9692 /* mark where we are, in case we need to report errors */
9695 /* after skipping whitespace, the next character is the terminator */
9698 termcode = termstr[0] = term;
9702 termcode = utf8_to_uvchr((U8*)s, &termlen);
9703 Copy(s, termstr, termlen, U8);
9704 if (!UTF8_IS_INVARIANT(term))
9708 /* mark where we are */
9709 PL_multi_start = CopLINE(PL_curcop);
9710 PL_multi_open = term;
9712 /* find corresponding closing delimiter */
9713 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
9714 termcode = termstr[0] = term = tmps[5];
9716 PL_multi_close = term;
9718 /* create a new SV to hold the contents. 79 is the SV's initial length.
9719 What a random number. */
9720 sv = newSV_type(SVt_PVIV);
9722 SvIV_set(sv, termcode);
9723 (void)SvPOK_only(sv); /* validate pointer */
9725 /* move past delimiter and try to read a complete string */
9727 sv_catpvn(sv, s, termlen);
9730 tstart = SvPVX(PL_linestr) + stuffstart;
9731 if (!PL_thisopen && !keep_delims) {
9732 PL_thisopen = newSVpvn(tstart, s - tstart);
9733 stuffstart = s - SvPVX(PL_linestr);
9737 if (PL_encoding && !UTF) {
9741 int offset = s - SvPVX_const(PL_linestr);
9742 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
9743 &offset, (char*)termstr, termlen);
9744 const char * const ns = SvPVX_const(PL_linestr) + offset;
9745 char * const svlast = SvEND(sv) - 1;
9747 for (; s < ns; s++) {
9748 if (*s == '\n' && !PL_rsfp)
9749 CopLINE_inc(PL_curcop);
9752 goto read_more_line;
9754 /* handle quoted delimiters */
9755 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
9757 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
9759 if ((svlast-1 - t) % 2) {
9763 SvCUR_set(sv, SvCUR(sv) - 1);
9768 if (PL_multi_open == PL_multi_close) {
9774 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
9775 /* At here, all closes are "was quoted" one,
9776 so we don't check PL_multi_close. */
9778 if (!keep_quoted && *(t+1) == PL_multi_open)
9783 else if (*t == PL_multi_open)
9791 SvCUR_set(sv, w - SvPVX_const(sv));
9793 last_off = w - SvPVX(sv);
9794 if (--brackets <= 0)
9800 SvCUR_set(sv, SvCUR(sv) - 1);
9806 /* extend sv if need be */
9807 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
9808 /* set 'to' to the next character in the sv's string */
9809 to = SvPVX(sv)+SvCUR(sv);
9811 /* if open delimiter is the close delimiter read unbridle */
9812 if (PL_multi_open == PL_multi_close) {
9813 for (; s < PL_bufend; s++,to++) {
9814 /* embedded newlines increment the current line number */
9815 if (*s == '\n' && !PL_rsfp)
9816 CopLINE_inc(PL_curcop);
9817 /* handle quoted delimiters */
9818 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
9819 if (!keep_quoted && s[1] == term)
9821 /* any other quotes are simply copied straight through */
9825 /* terminate when run out of buffer (the for() condition), or
9826 have found the terminator */
9827 else if (*s == term) {
9830 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
9833 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9839 /* if the terminator isn't the same as the start character (e.g.,
9840 matched brackets), we have to allow more in the quoting, and
9841 be prepared for nested brackets.
9844 /* read until we run out of string, or we find the terminator */
9845 for (; s < PL_bufend; s++,to++) {
9846 /* embedded newlines increment the line count */
9847 if (*s == '\n' && !PL_rsfp)
9848 CopLINE_inc(PL_curcop);
9849 /* backslashes can escape the open or closing characters */
9850 if (*s == '\\' && s+1 < PL_bufend) {
9852 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
9857 /* allow nested opens and closes */
9858 else if (*s == PL_multi_close && --brackets <= 0)
9860 else if (*s == PL_multi_open)
9862 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9867 /* terminate the copied string and update the sv's end-of-string */
9869 SvCUR_set(sv, to - SvPVX_const(sv));
9872 * this next chunk reads more into the buffer if we're not done yet
9876 break; /* handle case where we are done yet :-) */
9878 #ifndef PERL_STRICT_CR
9879 if (to - SvPVX_const(sv) >= 2) {
9880 if ((to[-2] == '\r' && to[-1] == '\n') ||
9881 (to[-2] == '\n' && to[-1] == '\r'))
9885 SvCUR_set(sv, to - SvPVX_const(sv));
9887 else if (to[-1] == '\r')
9890 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
9895 /* if we're out of file, or a read fails, bail and reset the current
9896 line marker so we can report where the unterminated string began
9900 char * const tstart = SvPVX(PL_linestr) + stuffstart;
9902 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
9904 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
9907 CopLINE_inc(PL_curcop);
9908 PL_bufptr = PL_bufend;
9909 if (!lex_next_chunk(0)) {
9911 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9920 /* at this point, we have successfully read the delimited string */
9922 if (!PL_encoding || UTF) {
9925 char * const tstart = SvPVX(PL_linestr) + stuffstart;
9926 const int len = s - tstart;
9928 sv_catpvn(PL_thisstuff, tstart, len);
9930 PL_thisstuff = newSVpvn(tstart, len);
9931 if (!PL_thisclose && !keep_delims)
9932 PL_thisclose = newSVpvn(s,termlen);
9937 sv_catpvn(sv, s, termlen);
9943 char * const tstart = SvPVX(PL_linestr) + stuffstart;
9944 const int len = s - tstart - termlen;
9946 sv_catpvn(PL_thisstuff, tstart, len);
9948 PL_thisstuff = newSVpvn(tstart, len);
9949 if (!PL_thisclose && !keep_delims)
9950 PL_thisclose = newSVpvn(s - termlen,termlen);
9954 if (has_utf8 || PL_encoding)
9957 PL_multi_end = CopLINE(PL_curcop);
9959 /* if we allocated too much space, give some back */
9960 if (SvCUR(sv) + 5 < SvLEN(sv)) {
9961 SvLEN_set(sv, SvCUR(sv) + 1);
9962 SvPV_renew(sv, SvLEN(sv));
9965 /* decide whether this is the first or second quoted string we've read
9978 takes: pointer to position in buffer
9979 returns: pointer to new position in buffer
9980 side-effects: builds ops for the constant in pl_yylval.op
9982 Read a number in any of the formats that Perl accepts:
9984 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
9985 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
9988 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
9990 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
9993 If it reads a number without a decimal point or an exponent, it will
9994 try converting the number to an integer and see if it can do so
9995 without loss of precision.
9999 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10002 register const char *s = start; /* current position in buffer */
10003 register char *d; /* destination in temp buffer */
10004 register char *e; /* end of temp buffer */
10005 NV nv; /* number read, as a double */
10006 SV *sv = NULL; /* place to put the converted number */
10007 bool floatit; /* boolean: int or float? */
10008 const char *lastub = NULL; /* position of last underbar */
10009 static char const number_too_long[] = "Number too long";
10011 PERL_ARGS_ASSERT_SCAN_NUM;
10013 /* We use the first character to decide what type of number this is */
10017 Perl_croak(aTHX_ "panic: scan_num");
10019 /* if it starts with a 0, it could be an octal number, a decimal in
10020 0.13 disguise, or a hexadecimal number, or a binary number. */
10024 u holds the "number so far"
10025 shift the power of 2 of the base
10026 (hex == 4, octal == 3, binary == 1)
10027 overflowed was the number more than we can hold?
10029 Shift is used when we add a digit. It also serves as an "are
10030 we in octal/hex/binary?" indicator to disallow hex characters
10031 when in octal mode.
10036 bool overflowed = FALSE;
10037 bool just_zero = TRUE; /* just plain 0 or binary number? */
10038 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10039 static const char* const bases[5] =
10040 { "", "binary", "", "octal", "hexadecimal" };
10041 static const char* const Bases[5] =
10042 { "", "Binary", "", "Octal", "Hexadecimal" };
10043 static const char* const maxima[5] =
10045 "0b11111111111111111111111111111111",
10049 const char *base, *Base, *max;
10051 /* check for hex */
10052 if (s[1] == 'x' || s[1] == 'X') {
10056 } else if (s[1] == 'b' || s[1] == 'B') {
10061 /* check for a decimal in disguise */
10062 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
10064 /* so it must be octal */
10071 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10072 "Misplaced _ in number");
10076 base = bases[shift];
10077 Base = Bases[shift];
10078 max = maxima[shift];
10080 /* read the rest of the number */
10082 /* x is used in the overflow test,
10083 b is the digit we're adding on. */
10088 /* if we don't mention it, we're done */
10092 /* _ are ignored -- but warned about if consecutive */
10094 if (lastub && s == lastub + 1)
10095 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10096 "Misplaced _ in number");
10100 /* 8 and 9 are not octal */
10101 case '8': case '9':
10103 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10107 case '2': case '3': case '4':
10108 case '5': case '6': case '7':
10110 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10113 case '0': case '1':
10114 b = *s++ & 15; /* ASCII digit -> value of digit */
10118 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10119 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10120 /* make sure they said 0x */
10123 b = (*s++ & 7) + 9;
10125 /* Prepare to put the digit we have onto the end
10126 of the number so far. We check for overflows.
10132 x = u << shift; /* make room for the digit */
10134 if ((x >> shift) != u
10135 && !(PL_hints & HINT_NEW_BINARY)) {
10138 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10139 "Integer overflow in %s number",
10142 u = x | b; /* add the digit to the end */
10145 n *= nvshift[shift];
10146 /* If an NV has not enough bits in its
10147 * mantissa to represent an UV this summing of
10148 * small low-order numbers is a waste of time
10149 * (because the NV cannot preserve the
10150 * low-order bits anyway): we could just
10151 * remember when did we overflow and in the
10152 * end just multiply n by the right
10160 /* if we get here, we had success: make a scalar value from
10165 /* final misplaced underbar check */
10166 if (s[-1] == '_') {
10167 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10171 if (n > 4294967295.0)
10172 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10173 "%s number > %s non-portable",
10179 if (u > 0xffffffff)
10180 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10181 "%s number > %s non-portable",
10186 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10187 sv = new_constant(start, s - start, "integer",
10188 sv, NULL, NULL, 0);
10189 else if (PL_hints & HINT_NEW_BINARY)
10190 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10195 handle decimal numbers.
10196 we're also sent here when we read a 0 as the first digit
10198 case '1': case '2': case '3': case '4': case '5':
10199 case '6': case '7': case '8': case '9': case '.':
10202 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10205 /* read next group of digits and _ and copy into d */
10206 while (isDIGIT(*s) || *s == '_') {
10207 /* skip underscores, checking for misplaced ones
10211 if (lastub && s == lastub + 1)
10212 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10213 "Misplaced _ in number");
10217 /* check for end of fixed-length buffer */
10219 Perl_croak(aTHX_ number_too_long);
10220 /* if we're ok, copy the character */
10225 /* final misplaced underbar check */
10226 if (lastub && s == lastub + 1) {
10227 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10230 /* read a decimal portion if there is one. avoid
10231 3..5 being interpreted as the number 3. followed
10234 if (*s == '.' && s[1] != '.') {
10239 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10240 "Misplaced _ in number");
10244 /* copy, ignoring underbars, until we run out of digits.
10246 for (; isDIGIT(*s) || *s == '_'; s++) {
10247 /* fixed length buffer check */
10249 Perl_croak(aTHX_ number_too_long);
10251 if (lastub && s == lastub + 1)
10252 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10253 "Misplaced _ in number");
10259 /* fractional part ending in underbar? */
10260 if (s[-1] == '_') {
10261 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10262 "Misplaced _ in number");
10264 if (*s == '.' && isDIGIT(s[1])) {
10265 /* oops, it's really a v-string, but without the "v" */
10271 /* read exponent part, if present */
10272 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10276 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10277 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10279 /* stray preinitial _ */
10281 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10282 "Misplaced _ in number");
10286 /* allow positive or negative exponent */
10287 if (*s == '+' || *s == '-')
10290 /* stray initial _ */
10292 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10293 "Misplaced _ in number");
10297 /* read digits of exponent */
10298 while (isDIGIT(*s) || *s == '_') {
10301 Perl_croak(aTHX_ number_too_long);
10305 if (((lastub && s == lastub + 1) ||
10306 (!isDIGIT(s[1]) && s[1] != '_')))
10307 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10308 "Misplaced _ in number");
10316 We try to do an integer conversion first if no characters
10317 indicating "float" have been found.
10322 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10324 if (flags == IS_NUMBER_IN_UV) {
10326 sv = newSViv(uv); /* Prefer IVs over UVs. */
10329 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10330 if (uv <= (UV) IV_MIN)
10331 sv = newSViv(-(IV)uv);
10338 /* terminate the string */
10340 nv = Atof(PL_tokenbuf);
10345 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10346 const char *const key = floatit ? "float" : "integer";
10347 const STRLEN keylen = floatit ? 5 : 7;
10348 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10349 key, keylen, sv, NULL, NULL, 0);
10353 /* if it starts with a v, it could be a v-string */
10356 sv = newSV(5); /* preallocate storage space */
10357 s = scan_vstring(s, PL_bufend, sv);
10361 /* make the op for the constant and return */
10364 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10366 lvalp->opval = NULL;
10372 S_scan_formline(pTHX_ register char *s)
10375 register char *eol;
10377 SV * const stuff = newSVpvs("");
10378 bool needargs = FALSE;
10379 bool eofmt = FALSE;
10381 char *tokenstart = s;
10382 SV* savewhite = NULL;
10384 if (PL_madskills) {
10385 savewhite = PL_thiswhite;
10390 PERL_ARGS_ASSERT_SCAN_FORMLINE;
10392 while (!needargs) {
10395 #ifdef PERL_STRICT_CR
10396 while (SPACE_OR_TAB(*t))
10399 while (SPACE_OR_TAB(*t) || *t == '\r')
10402 if (*t == '\n' || t == PL_bufend) {
10407 if (PL_in_eval && !PL_rsfp) {
10408 eol = (char *) memchr(s,'\n',PL_bufend-s);
10413 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10415 for (t = s; t < eol; t++) {
10416 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10418 goto enough; /* ~~ must be first line in formline */
10420 if (*t == '@' || *t == '^')
10424 sv_catpvn(stuff, s, eol-s);
10425 #ifndef PERL_STRICT_CR
10426 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10427 char *end = SvPVX(stuff) + SvCUR(stuff);
10430 SvCUR_set(stuff, SvCUR(stuff) - 1);
10441 if (PL_madskills) {
10443 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
10445 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
10448 PL_bufptr = PL_bufend;
10449 CopLINE_inc(PL_curcop);
10450 got_some = lex_next_chunk(0);
10451 CopLINE_dec(PL_curcop);
10454 tokenstart = PL_bufptr;
10462 if (SvCUR(stuff)) {
10465 PL_lex_state = LEX_NORMAL;
10466 start_force(PL_curforce);
10467 NEXTVAL_NEXTTOKE.ival = 0;
10471 PL_lex_state = LEX_FORMLINE;
10473 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10475 else if (PL_encoding)
10476 sv_recode_to_utf8(stuff, PL_encoding);
10478 start_force(PL_curforce);
10479 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10481 start_force(PL_curforce);
10482 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
10486 SvREFCNT_dec(stuff);
10488 PL_lex_formbrack = 0;
10492 if (PL_madskills) {
10494 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
10496 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
10497 PL_thiswhite = savewhite;
10504 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10507 const I32 oldsavestack_ix = PL_savestack_ix;
10508 CV* const outsidecv = PL_compcv;
10511 assert(SvTYPE(PL_compcv) == SVt_PVCV);
10513 SAVEI32(PL_subline);
10514 save_item(PL_subname);
10515 SAVESPTR(PL_compcv);
10517 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
10518 CvFLAGS(PL_compcv) |= flags;
10520 PL_subline = CopLINE(PL_curcop);
10521 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10522 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
10523 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10525 return oldsavestack_ix;
10529 #pragma segment Perl_yylex
10532 S_yywarn(pTHX_ const char *const s)
10536 PERL_ARGS_ASSERT_YYWARN;
10538 PL_in_eval |= EVAL_WARNONLY;
10540 PL_in_eval &= ~EVAL_WARNONLY;
10545 Perl_yyerror(pTHX_ const char *const s)
10548 const char *where = NULL;
10549 const char *context = NULL;
10552 int yychar = PL_parser->yychar;
10554 PERL_ARGS_ASSERT_YYERROR;
10556 if (!yychar || (yychar == ';' && !PL_rsfp))
10558 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
10559 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
10560 PL_oldbufptr != PL_bufptr) {
10563 The code below is removed for NetWare because it abends/crashes on NetWare
10564 when the script has error such as not having the closing quotes like:
10565 if ($var eq "value)
10566 Checking of white spaces is anyway done in NetWare code.
10569 while (isSPACE(*PL_oldoldbufptr))
10572 context = PL_oldoldbufptr;
10573 contlen = PL_bufptr - PL_oldoldbufptr;
10575 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
10576 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
10579 The code below is removed for NetWare because it abends/crashes on NetWare
10580 when the script has error such as not having the closing quotes like:
10581 if ($var eq "value)
10582 Checking of white spaces is anyway done in NetWare code.
10585 while (isSPACE(*PL_oldbufptr))
10588 context = PL_oldbufptr;
10589 contlen = PL_bufptr - PL_oldbufptr;
10591 else if (yychar > 255)
10592 where = "next token ???";
10593 else if (yychar == -2) { /* YYEMPTY */
10594 if (PL_lex_state == LEX_NORMAL ||
10595 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
10596 where = "at end of line";
10597 else if (PL_lex_inpat)
10598 where = "within pattern";
10600 where = "within string";
10603 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
10605 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
10606 else if (isPRINT_LC(yychar)) {
10607 const char string = yychar;
10608 sv_catpvn(where_sv, &string, 1);
10611 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
10612 where = SvPVX_const(where_sv);
10614 msg = sv_2mortal(newSVpv(s, 0));
10615 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
10616 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
10618 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
10620 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
10621 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
10622 Perl_sv_catpvf(aTHX_ msg,
10623 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
10624 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
10627 if (PL_in_eval & EVAL_WARNONLY) {
10628 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
10632 if (PL_error_count >= 10) {
10633 if (PL_in_eval && SvCUR(ERRSV))
10634 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
10635 SVfARG(ERRSV), OutCopFILE(PL_curcop));
10637 Perl_croak(aTHX_ "%s has too many errors.\n",
10638 OutCopFILE(PL_curcop));
10641 PL_in_my_stash = NULL;
10645 #pragma segment Main
10649 S_swallow_bom(pTHX_ U8 *s)
10652 const STRLEN slen = SvCUR(PL_linestr);
10654 PERL_ARGS_ASSERT_SWALLOW_BOM;
10658 if (s[1] == 0xFE) {
10659 /* UTF-16 little-endian? (or UTF-32LE?) */
10660 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
10661 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
10662 #ifndef PERL_NO_UTF16_FILTER
10663 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
10665 if (PL_bufend > (char*)s) {
10666 s = add_utf16_textfilter(s, TRUE);
10669 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10674 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
10675 #ifndef PERL_NO_UTF16_FILTER
10676 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
10678 if (PL_bufend > (char *)s) {
10679 s = add_utf16_textfilter(s, FALSE);
10682 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10687 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
10688 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10689 s += 3; /* UTF-8 */
10695 if (s[2] == 0xFE && s[3] == 0xFF) {
10696 /* UTF-32 big-endian */
10697 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
10700 else if (s[2] == 0 && s[3] != 0) {
10703 * are a good indicator of UTF-16BE. */
10704 #ifndef PERL_NO_UTF16_FILTER
10705 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
10706 s = add_utf16_textfilter(s, FALSE);
10708 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10714 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
10715 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10716 s += 4; /* UTF-8 */
10722 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
10725 * are a good indicator of UTF-16LE. */
10726 #ifndef PERL_NO_UTF16_FILTER
10727 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
10728 s = add_utf16_textfilter(s, TRUE);
10730 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10738 #ifndef PERL_NO_UTF16_FILTER
10740 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
10743 SV *const filter = FILTER_DATA(idx);
10744 /* We re-use this each time round, throwing the contents away before we
10746 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
10747 SV *const utf8_buffer = filter;
10748 IV status = IoPAGE(filter);
10749 const bool reverse = cBOOL(IoLINES(filter));
10752 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
10754 /* As we're automatically added, at the lowest level, and hence only called
10755 from this file, we can be sure that we're not called in block mode. Hence
10756 don't bother writing code to deal with block mode. */
10758 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
10761 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
10763 DEBUG_P(PerlIO_printf(Perl_debug_log,
10764 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10765 FPTR2DPTR(void *, S_utf16_textfilter),
10766 reverse ? 'l' : 'b', idx, maxlen, status,
10767 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10774 /* First, look in our buffer of existing UTF-8 data: */
10775 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
10779 } else if (status == 0) {
10781 IoPAGE(filter) = 0;
10782 nl = SvEND(utf8_buffer);
10785 STRLEN got = nl - SvPVX(utf8_buffer);
10786 /* Did we have anything to append? */
10788 sv_catpvn(sv, SvPVX(utf8_buffer), got);
10789 /* Everything else in this code works just fine if SVp_POK isn't
10790 set. This, however, needs it, and we need it to work, else
10791 we loop infinitely because the buffer is never consumed. */
10792 sv_chop(utf8_buffer, nl);
10796 /* OK, not a complete line there, so need to read some more UTF-16.
10797 Read an extra octect if the buffer currently has an odd number. */
10801 if (SvCUR(utf16_buffer) >= 2) {
10802 /* Location of the high octet of the last complete code point.
10803 Gosh, UTF-16 is a pain. All the benefits of variable length,
10804 *coupled* with all the benefits of partial reads and
10806 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
10807 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
10809 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
10813 /* We have the first half of a surrogate. Read more. */
10814 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
10817 status = FILTER_READ(idx + 1, utf16_buffer,
10818 160 + (SvCUR(utf16_buffer) & 1));
10819 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
10820 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
10823 IoPAGE(filter) = status;
10828 chars = SvCUR(utf16_buffer) >> 1;
10829 have = SvCUR(utf8_buffer);
10830 SvGROW(utf8_buffer, have + chars * 3 + 1);
10833 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
10834 (U8*)SvPVX_const(utf8_buffer) + have,
10835 chars * 2, &newlen);
10837 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
10838 (U8*)SvPVX_const(utf8_buffer) + have,
10839 chars * 2, &newlen);
10841 SvCUR_set(utf8_buffer, have + newlen);
10844 /* No need to keep this SV "well-formed" with a '\0' after the end, as
10845 it's private to us, and utf16_to_utf8{,reversed} take a
10846 (pointer,length) pair, rather than a NUL-terminated string. */
10847 if(SvCUR(utf16_buffer) & 1) {
10848 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
10849 SvCUR_set(utf16_buffer, 1);
10851 SvCUR_set(utf16_buffer, 0);
10854 DEBUG_P(PerlIO_printf(Perl_debug_log,
10855 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10857 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10858 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
10863 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
10865 SV *filter = filter_add(S_utf16_textfilter, NULL);
10867 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
10869 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
10870 sv_setpvs(filter, "");
10871 IoLINES(filter) = reversed;
10872 IoPAGE(filter) = 1; /* Not EOF */
10874 /* Sadly, we have to return a valid pointer, come what may, so we have to
10875 ignore any error return from this. */
10876 SvCUR_set(PL_linestr, 0);
10877 if (FILTER_READ(0, PL_linestr, 0)) {
10878 SvUTF8_on(PL_linestr);
10880 SvUTF8_on(PL_linestr);
10882 PL_bufend = SvEND(PL_linestr);
10883 return (U8*)SvPVX(PL_linestr);
10888 Returns a pointer to the next character after the parsed
10889 vstring, as well as updating the passed in sv.
10891 Function must be called like
10894 s = scan_vstring(s,e,sv);
10896 where s and e are the start and end of the string.
10897 The sv should already be large enough to store the vstring
10898 passed in, for performance reasons.
10903 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
10906 const char *pos = s;
10907 const char *start = s;
10909 PERL_ARGS_ASSERT_SCAN_VSTRING;
10911 if (*pos == 'v') pos++; /* get past 'v' */
10912 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
10914 if ( *pos != '.') {
10915 /* this may not be a v-string if followed by => */
10916 const char *next = pos;
10917 while (next < e && isSPACE(*next))
10919 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
10920 /* return string not v-string */
10921 sv_setpvn(sv,(char *)s,pos-s);
10922 return (char *)pos;
10926 if (!isALPHA(*pos)) {
10927 U8 tmpbuf[UTF8_MAXBYTES+1];
10930 s++; /* get past 'v' */
10935 /* this is atoi() that tolerates underscores */
10938 const char *end = pos;
10940 while (--end >= s) {
10942 const UV orev = rev;
10943 rev += (*end - '0') * mult;
10946 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10947 "Integer overflow in decimal number");
10951 if (rev > 0x7FFFFFFF)
10952 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
10954 /* Append native character for the rev point */
10955 tmpend = uvchr_to_utf8(tmpbuf, rev);
10956 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
10957 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
10959 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
10965 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
10969 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
10976 Perl_keyword_plugin_standard(pTHX_
10977 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
10979 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
10980 PERL_UNUSED_CONTEXT;
10981 PERL_UNUSED_ARG(keyword_ptr);
10982 PERL_UNUSED_ARG(keyword_len);
10983 PERL_UNUSED_ARG(op_ptr);
10984 return KEYWORD_PLUGIN_DECLINE;
10987 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
10989 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
10991 SAVEI32(PL_lex_brackets);
10992 if (PL_lex_brackets > 100)
10993 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
10994 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
10995 SAVEI32(PL_lex_allbrackets);
10996 PL_lex_allbrackets = 0;
10997 SAVEI8(PL_lex_fakeeof);
10998 PL_lex_fakeeof = (U8)fakeeof;
10999 if(yyparse(gramtype) && !PL_parser->error_count)
11000 qerror(Perl_mess(aTHX_ "Parse error"));
11003 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11005 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11009 SAVEVPTR(PL_eval_root);
11010 PL_eval_root = NULL;
11011 parse_recdescent(gramtype, fakeeof);
11017 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11019 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11022 if (flags & ~PARSE_OPTIONAL)
11023 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11024 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11025 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11026 if (!PL_parser->error_count)
11027 qerror(Perl_mess(aTHX_ "Parse error"));
11028 exprop = newOP(OP_NULL, 0);
11034 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11036 Parse a Perl arithmetic expression. This may contain operators of precedence
11037 down to the bit shift operators. The expression must be followed (and thus
11038 terminated) either by a comparison or lower-precedence operator or by
11039 something that would normally terminate an expression such as semicolon.
11040 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11041 otherwise it is mandatory. It is up to the caller to ensure that the
11042 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11043 the source of the code to be parsed and the lexical context for the
11046 The op tree representing the expression is returned. If an optional
11047 expression is absent, a null pointer is returned, otherwise the pointer
11050 If an error occurs in parsing or compilation, in most cases a valid op
11051 tree is returned anyway. The error is reflected in the parser state,
11052 normally resulting in a single exception at the top level of parsing
11053 which covers all the compilation errors that occurred. Some compilation
11054 errors, however, will throw an exception immediately.
11060 Perl_parse_arithexpr(pTHX_ U32 flags)
11062 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11066 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11068 Parse a Perl term expression. This may contain operators of precedence
11069 down to the assignment operators. The expression must be followed (and thus
11070 terminated) either by a comma or lower-precedence operator or by
11071 something that would normally terminate an expression such as semicolon.
11072 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11073 otherwise it is mandatory. It is up to the caller to ensure that the
11074 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11075 the source of the code to be parsed and the lexical context for the
11078 The op tree representing the expression is returned. If an optional
11079 expression is absent, a null pointer is returned, otherwise the pointer
11082 If an error occurs in parsing or compilation, in most cases a valid op
11083 tree is returned anyway. The error is reflected in the parser state,
11084 normally resulting in a single exception at the top level of parsing
11085 which covers all the compilation errors that occurred. Some compilation
11086 errors, however, will throw an exception immediately.
11092 Perl_parse_termexpr(pTHX_ U32 flags)
11094 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11098 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11100 Parse a Perl list expression. This may contain operators of precedence
11101 down to the comma operator. The expression must be followed (and thus
11102 terminated) either by a low-precedence logic operator such as C<or> or by
11103 something that would normally terminate an expression such as semicolon.
11104 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11105 otherwise it is mandatory. It is up to the caller to ensure that the
11106 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11107 the source of the code to be parsed and the lexical context for the
11110 The op tree representing the expression is returned. If an optional
11111 expression is absent, a null pointer is returned, otherwise the pointer
11114 If an error occurs in parsing or compilation, in most cases a valid op
11115 tree is returned anyway. The error is reflected in the parser state,
11116 normally resulting in a single exception at the top level of parsing
11117 which covers all the compilation errors that occurred. Some compilation
11118 errors, however, will throw an exception immediately.
11124 Perl_parse_listexpr(pTHX_ U32 flags)
11126 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11130 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11132 Parse a single complete Perl expression. This allows the full
11133 expression grammar, including the lowest-precedence operators such
11134 as C<or>. The expression must be followed (and thus terminated) by a
11135 token that an expression would normally be terminated by: end-of-file,
11136 closing bracketing punctuation, semicolon, or one of the keywords that
11137 signals a postfix expression-statement modifier. If I<flags> includes
11138 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11139 mandatory. It is up to the caller to ensure that the dynamic parser
11140 state (L</PL_parser> et al) is correctly set to reflect the source of
11141 the code to be parsed and the lexical context for the expression.
11143 The op tree representing the expression is returned. If an optional
11144 expression is absent, a null pointer is returned, otherwise the pointer
11147 If an error occurs in parsing or compilation, in most cases a valid op
11148 tree is returned anyway. The error is reflected in the parser state,
11149 normally resulting in a single exception at the top level of parsing
11150 which covers all the compilation errors that occurred. Some compilation
11151 errors, however, will throw an exception immediately.
11157 Perl_parse_fullexpr(pTHX_ U32 flags)
11159 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11163 =for apidoc Amx|OP *|parse_block|U32 flags
11165 Parse a single complete Perl code block. This consists of an opening
11166 brace, a sequence of statements, and a closing brace. The block
11167 constitutes a lexical scope, so C<my> variables and various compile-time
11168 effects can be contained within it. It is up to the caller to ensure
11169 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11170 reflect the source of the code to be parsed and the lexical context for
11173 The op tree representing the code block is returned. This is always a
11174 real op, never a null pointer. It will normally be a C<lineseq> list,
11175 including C<nextstate> or equivalent ops. No ops to construct any kind
11176 of runtime scope are included by virtue of it being a block.
11178 If an error occurs in parsing or compilation, in most cases a valid op
11179 tree (most likely null) is returned anyway. The error is reflected in
11180 the parser state, normally resulting in a single exception at the top
11181 level of parsing which covers all the compilation errors that occurred.
11182 Some compilation errors, however, will throw an exception immediately.
11184 The I<flags> parameter is reserved for future use, and must always
11191 Perl_parse_block(pTHX_ U32 flags)
11194 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11195 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11199 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11201 Parse a single unadorned Perl statement. This may be a normal imperative
11202 statement or a declaration that has compile-time effect. It does not
11203 include any label or other affixture. It is up to the caller to ensure
11204 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11205 reflect the source of the code to be parsed and the lexical context for
11208 The op tree representing the statement is returned. This may be a
11209 null pointer if the statement is null, for example if it was actually
11210 a subroutine definition (which has compile-time side effects). If not
11211 null, it will be ops directly implementing the statement, suitable to
11212 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11213 equivalent op (except for those embedded in a scope contained entirely
11214 within the statement).
11216 If an error occurs in parsing or compilation, in most cases a valid op
11217 tree (most likely null) is returned anyway. The error is reflected in
11218 the parser state, normally resulting in a single exception at the top
11219 level of parsing which covers all the compilation errors that occurred.
11220 Some compilation errors, however, will throw an exception immediately.
11222 The I<flags> parameter is reserved for future use, and must always
11229 Perl_parse_barestmt(pTHX_ U32 flags)
11232 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11233 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11237 =for apidoc Amx|SV *|parse_label|U32 flags
11239 Parse a single label, possibly optional, of the type that may prefix a
11240 Perl statement. It is up to the caller to ensure that the dynamic parser
11241 state (L</PL_parser> et al) is correctly set to reflect the source of
11242 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11243 label is optional, otherwise it is mandatory.
11245 The name of the label is returned in the form of a fresh scalar. If an
11246 optional label is absent, a null pointer is returned.
11248 If an error occurs in parsing, which can only occur if the label is
11249 mandatory, a valid label is returned anyway. The error is reflected in
11250 the parser state, normally resulting in a single exception at the top
11251 level of parsing which covers all the compilation errors that occurred.
11257 Perl_parse_label(pTHX_ U32 flags)
11259 if (flags & ~PARSE_OPTIONAL)
11260 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11261 if (PL_lex_state == LEX_KNOWNEXT) {
11262 PL_parser->yychar = yylex();
11263 if (PL_parser->yychar == LABEL) {
11264 char *lpv = pl_yylval.pval;
11265 STRLEN llen = strlen(lpv);
11267 PL_parser->yychar = YYEMPTY;
11268 lsv = newSV_type(SVt_PV);
11269 SvPV_set(lsv, lpv);
11270 SvCUR_set(lsv, llen);
11271 SvLEN_set(lsv, llen+1);
11281 STRLEN wlen, bufptr_pos;
11285 if (!isIDFIRST_A(c))
11289 } while(isWORDCHAR_A(c));
11291 if (word_takes_any_delimeter(s, wlen))
11293 bufptr_pos = s - SvPVX(PL_linestr);
11295 lex_read_space(LEX_KEEP_PREVIOUS);
11297 s = SvPVX(PL_linestr) + bufptr_pos;
11298 if (t[0] == ':' && t[1] != ':') {
11299 PL_oldoldbufptr = PL_oldbufptr;
11302 return newSVpvn(s, wlen);
11306 if (flags & PARSE_OPTIONAL) {
11309 qerror(Perl_mess(aTHX_ "Parse error"));
11310 return newSVpvs("x");
11317 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11319 Parse a single complete Perl statement. This may be a normal imperative
11320 statement or a declaration that has compile-time effect, and may include
11321 optional labels. It is up to the caller to ensure that the dynamic
11322 parser state (L</PL_parser> et al) is correctly set to reflect the source
11323 of the code to be parsed and the lexical context for the statement.
11325 The op tree representing the statement is returned. This may be a
11326 null pointer if the statement is null, for example if it was actually
11327 a subroutine definition (which has compile-time side effects). If not
11328 null, it will be the result of a L</newSTATEOP> call, normally including
11329 a C<nextstate> or equivalent op.
11331 If an error occurs in parsing or compilation, in most cases a valid op
11332 tree (most likely null) is returned anyway. The error is reflected in
11333 the parser state, normally resulting in a single exception at the top
11334 level of parsing which covers all the compilation errors that occurred.
11335 Some compilation errors, however, will throw an exception immediately.
11337 The I<flags> parameter is reserved for future use, and must always
11344 Perl_parse_fullstmt(pTHX_ U32 flags)
11347 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11348 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11352 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11354 Parse a sequence of zero or more Perl statements. These may be normal
11355 imperative statements, including optional labels, or declarations
11356 that have compile-time effect, or any mixture thereof. The statement
11357 sequence ends when a closing brace or end-of-file is encountered in a
11358 place where a new statement could have validly started. It is up to
11359 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11360 is correctly set to reflect the source of the code to be parsed and the
11361 lexical context for the statements.
11363 The op tree representing the statement sequence is returned. This may
11364 be a null pointer if the statements were all null, for example if there
11365 were no statements or if there were only subroutine definitions (which
11366 have compile-time side effects). If not null, it will be a C<lineseq>
11367 list, normally including C<nextstate> or equivalent ops.
11369 If an error occurs in parsing or compilation, in most cases a valid op
11370 tree is returned anyway. The error is reflected in the parser state,
11371 normally resulting in a single exception at the top level of parsing
11372 which covers all the compilation errors that occurred. Some compilation
11373 errors, however, will throw an exception immediately.
11375 The I<flags> parameter is reserved for future use, and must always
11382 Perl_parse_stmtseq(pTHX_ U32 flags)
11387 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
11388 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
11389 c = lex_peek_unichar(0);
11390 if (c != -1 && c != /*{*/'}')
11391 qerror(Perl_mess(aTHX_ "Parse error"));
11396 Perl_munge_qwlist_to_paren_list(pTHX_ OP *qwlist)
11398 PERL_ARGS_ASSERT_MUNGE_QWLIST_TO_PAREN_LIST;
11399 deprecate("qw(...) as parentheses");
11400 force_next((4<<24)|')');
11401 if (qwlist->op_type == OP_STUB) {
11405 start_force(PL_curforce);
11406 NEXTVAL_NEXTTOKE.opval = qwlist;
11409 force_next((2<<24)|'(');
11414 * c-indentation-style: bsd
11415 * c-basic-offset: 4
11416 * indent-tabs-mode: t
11419 * ex: set ts=8 sts=4 sw=4 noet: