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
670 be zero, except for one flag that is currently reserved for perl's internal
676 /* LEX_START_SAME_FILTER indicates that this is not a new file, so it
677 can share filters with the current parser. */
680 Perl_lex_start(pTHX_ SV *line, PerlIO *rsfp, U32 flags)
683 const char *s = NULL;
685 yy_parser *parser, *oparser;
686 if (flags && flags != LEX_START_SAME_FILTER)
687 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_start");
689 /* create and initialise a parser */
691 Newxz(parser, 1, yy_parser);
692 parser->old_parser = oparser = PL_parser;
695 parser->stack = NULL;
697 parser->stack_size = 0;
699 /* on scope exit, free this parser and restore any outer one */
701 parser->saved_curcop = PL_curcop;
703 /* initialise lexer state */
706 parser->curforce = -1;
708 parser->nexttoke = 0;
710 parser->error_count = oparser ? oparser->error_count : 0;
711 parser->copline = NOLINE;
712 parser->lex_state = LEX_NORMAL;
713 parser->expect = XSTATE;
715 parser->rsfp_filters =
716 !(flags & LEX_START_SAME_FILTER) || !oparser
718 : MUTABLE_AV(SvREFCNT_inc(oparser->rsfp_filters));
720 Newx(parser->lex_brackstack, 120, char);
721 Newx(parser->lex_casestack, 12, char);
722 *parser->lex_casestack = '\0';
725 s = SvPV_const(line, len);
731 parser->linestr = newSVpvs("\n;");
733 parser->linestr = newSVpvn_flags(s, len, SvUTF8(line));
735 sv_catpvs(parser->linestr, "\n;");
737 parser->oldoldbufptr =
740 parser->linestart = SvPVX(parser->linestr);
741 parser->bufend = parser->bufptr + SvCUR(parser->linestr);
742 parser->last_lop = parser->last_uni = NULL;
748 /* delete a parser object */
751 Perl_parser_free(pTHX_ const yy_parser *parser)
753 PERL_ARGS_ASSERT_PARSER_FREE;
755 PL_curcop = parser->saved_curcop;
756 SvREFCNT_dec(parser->linestr);
758 if (parser->rsfp == PerlIO_stdin())
759 PerlIO_clearerr(parser->rsfp);
760 else if (parser->rsfp && (!parser->old_parser ||
761 (parser->old_parser && parser->rsfp != parser->old_parser->rsfp)))
762 PerlIO_close(parser->rsfp);
763 SvREFCNT_dec(parser->rsfp_filters);
765 Safefree(parser->lex_brackstack);
766 Safefree(parser->lex_casestack);
767 PL_parser = parser->old_parser;
773 =for apidoc AmxU|SV *|PL_parser-E<gt>linestr
775 Buffer scalar containing the chunk currently under consideration of the
776 text currently being lexed. This is always a plain string scalar (for
777 which C<SvPOK> is true). It is not intended to be used as a scalar by
778 normal scalar means; instead refer to the buffer directly by the pointer
779 variables described below.
781 The lexer maintains various C<char*> pointers to things in the
782 C<PL_parser-E<gt>linestr> buffer. If C<PL_parser-E<gt>linestr> is ever
783 reallocated, all of these pointers must be updated. Don't attempt to
784 do this manually, but rather use L</lex_grow_linestr> if you need to
785 reallocate the buffer.
787 The content of the text chunk in the buffer is commonly exactly one
788 complete line of input, up to and including a newline terminator,
789 but there are situations where it is otherwise. The octets of the
790 buffer may be intended to be interpreted as either UTF-8 or Latin-1.
791 The function L</lex_bufutf8> tells you which. Do not use the C<SvUTF8>
792 flag on this scalar, which may disagree with it.
794 For direct examination of the buffer, the variable
795 L</PL_parser-E<gt>bufend> points to the end of the buffer. The current
796 lexing position is pointed to by L</PL_parser-E<gt>bufptr>. Direct use
797 of these pointers is usually preferable to examination of the scalar
798 through normal scalar means.
800 =for apidoc AmxU|char *|PL_parser-E<gt>bufend
802 Direct pointer to the end of the chunk of text currently being lexed, the
803 end of the lexer buffer. This is equal to C<SvPVX(PL_parser-E<gt>linestr)
804 + SvCUR(PL_parser-E<gt>linestr)>. A NUL character (zero octet) is
805 always located at the end of the buffer, and does not count as part of
806 the buffer's contents.
808 =for apidoc AmxU|char *|PL_parser-E<gt>bufptr
810 Points to the current position of lexing inside the lexer buffer.
811 Characters around this point may be freely examined, within
812 the range delimited by C<SvPVX(L</PL_parser-E<gt>linestr>)> and
813 L</PL_parser-E<gt>bufend>. The octets of the buffer may be intended to be
814 interpreted as either UTF-8 or Latin-1, as indicated by L</lex_bufutf8>.
816 Lexing code (whether in the Perl core or not) moves this pointer past
817 the characters that it consumes. It is also expected to perform some
818 bookkeeping whenever a newline character is consumed. This movement
819 can be more conveniently performed by the function L</lex_read_to>,
820 which handles newlines appropriately.
822 Interpretation of the buffer's octets can be abstracted out by
823 using the slightly higher-level functions L</lex_peek_unichar> and
824 L</lex_read_unichar>.
826 =for apidoc AmxU|char *|PL_parser-E<gt>linestart
828 Points to the start of the current line inside the lexer buffer.
829 This is useful for indicating at which column an error occurred, and
830 not much else. This must be updated by any lexing code that consumes
831 a newline; the function L</lex_read_to> handles this detail.
837 =for apidoc Amx|bool|lex_bufutf8
839 Indicates whether the octets in the lexer buffer
840 (L</PL_parser-E<gt>linestr>) should be interpreted as the UTF-8 encoding
841 of Unicode characters. If not, they should be interpreted as Latin-1
842 characters. This is analogous to the C<SvUTF8> flag for scalars.
844 In UTF-8 mode, it is not guaranteed that the lexer buffer actually
845 contains valid UTF-8. Lexing code must be robust in the face of invalid
848 The actual C<SvUTF8> flag of the L</PL_parser-E<gt>linestr> scalar
849 is significant, but not the whole story regarding the input character
850 encoding. Normally, when a file is being read, the scalar contains octets
851 and its C<SvUTF8> flag is off, but the octets should be interpreted as
852 UTF-8 if the C<use utf8> pragma is in effect. During a string eval,
853 however, the scalar may have the C<SvUTF8> flag on, and in this case its
854 octets should be interpreted as UTF-8 unless the C<use bytes> pragma
855 is in effect. This logic may change in the future; use this function
856 instead of implementing the logic yourself.
862 Perl_lex_bufutf8(pTHX)
868 =for apidoc Amx|char *|lex_grow_linestr|STRLEN len
870 Reallocates the lexer buffer (L</PL_parser-E<gt>linestr>) to accommodate
871 at least I<len> octets (including terminating NUL). Returns a
872 pointer to the reallocated buffer. This is necessary before making
873 any direct modification of the buffer that would increase its length.
874 L</lex_stuff_pvn> provides a more convenient way to insert text into
877 Do not use C<SvGROW> or C<sv_grow> directly on C<PL_parser-E<gt>linestr>;
878 this function updates all of the lexer's variables that point directly
885 Perl_lex_grow_linestr(pTHX_ STRLEN len)
889 STRLEN bufend_pos, bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
890 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
891 linestr = PL_parser->linestr;
892 buf = SvPVX(linestr);
893 if (len <= SvLEN(linestr))
895 bufend_pos = PL_parser->bufend - buf;
896 bufptr_pos = PL_parser->bufptr - buf;
897 oldbufptr_pos = PL_parser->oldbufptr - buf;
898 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
899 linestart_pos = PL_parser->linestart - buf;
900 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
901 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
902 buf = sv_grow(linestr, len);
903 PL_parser->bufend = buf + bufend_pos;
904 PL_parser->bufptr = buf + bufptr_pos;
905 PL_parser->oldbufptr = buf + oldbufptr_pos;
906 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
907 PL_parser->linestart = buf + linestart_pos;
908 if (PL_parser->last_uni)
909 PL_parser->last_uni = buf + last_uni_pos;
910 if (PL_parser->last_lop)
911 PL_parser->last_lop = buf + last_lop_pos;
916 =for apidoc Amx|void|lex_stuff_pvn|const char *pv|STRLEN len|U32 flags
918 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
919 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
920 reallocating the buffer if necessary. This means that lexing code that
921 runs later will see the characters as if they had appeared in the input.
922 It is not recommended to do this as part of normal parsing, and most
923 uses of this facility run the risk of the inserted characters being
924 interpreted in an unintended manner.
926 The string to be inserted is represented by I<len> octets starting
927 at I<pv>. These octets are interpreted as either UTF-8 or Latin-1,
928 according to whether the C<LEX_STUFF_UTF8> flag is set in I<flags>.
929 The characters are recoded for the lexer buffer, according to how the
930 buffer is currently being interpreted (L</lex_bufutf8>). If a string
931 to be inserted is available as a Perl scalar, the L</lex_stuff_sv>
932 function is more convenient.
938 Perl_lex_stuff_pvn(pTHX_ const char *pv, STRLEN len, U32 flags)
942 PERL_ARGS_ASSERT_LEX_STUFF_PVN;
943 if (flags & ~(LEX_STUFF_UTF8))
944 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_pvn");
946 if (flags & LEX_STUFF_UTF8) {
950 const char *p, *e = pv+len;
951 for (p = pv; p != e; p++)
952 highhalf += !!(((U8)*p) & 0x80);
955 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len+highhalf);
956 bufptr = PL_parser->bufptr;
957 Move(bufptr, bufptr+len+highhalf, PL_parser->bufend+1-bufptr, char);
958 SvCUR_set(PL_parser->linestr,
959 SvCUR(PL_parser->linestr) + len+highhalf);
960 PL_parser->bufend += len+highhalf;
961 for (p = pv; p != e; p++) {
964 *bufptr++ = (char)(0xc0 | (c >> 6));
965 *bufptr++ = (char)(0x80 | (c & 0x3f));
972 if (flags & LEX_STUFF_UTF8) {
974 const char *p, *e = pv+len;
975 for (p = pv; p != e; p++) {
978 Perl_croak(aTHX_ "Lexing code attempted to stuff "
979 "non-Latin-1 character into Latin-1 input");
980 } else if (c >= 0xc2 && p+1 != e &&
981 (((U8)p[1]) & 0xc0) == 0x80) {
984 } else if (c >= 0x80) {
985 /* malformed UTF-8 */
987 SAVESPTR(PL_warnhook);
988 PL_warnhook = PERL_WARNHOOK_FATAL;
989 utf8n_to_uvuni((U8*)p, e-p, NULL, 0);
995 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len-highhalf);
996 bufptr = PL_parser->bufptr;
997 Move(bufptr, bufptr+len-highhalf, PL_parser->bufend+1-bufptr, char);
998 SvCUR_set(PL_parser->linestr,
999 SvCUR(PL_parser->linestr) + len-highhalf);
1000 PL_parser->bufend += len-highhalf;
1001 for (p = pv; p != e; p++) {
1004 *bufptr++ = (char)(((c & 0x3) << 6) | (p[1] & 0x3f));
1007 *bufptr++ = (char)c;
1012 lex_grow_linestr(SvCUR(PL_parser->linestr)+1+len);
1013 bufptr = PL_parser->bufptr;
1014 Move(bufptr, bufptr+len, PL_parser->bufend+1-bufptr, char);
1015 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) + len);
1016 PL_parser->bufend += len;
1017 Copy(pv, bufptr, len, char);
1023 =for apidoc Amx|void|lex_stuff_pv|const char *pv|U32 flags
1025 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1026 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1027 reallocating the buffer if necessary. This means that lexing code that
1028 runs later will see the characters as if they had appeared in the input.
1029 It is not recommended to do this as part of normal parsing, and most
1030 uses of this facility run the risk of the inserted characters being
1031 interpreted in an unintended manner.
1033 The string to be inserted is represented by octets starting at I<pv>
1034 and continuing to the first nul. These octets are interpreted as either
1035 UTF-8 or Latin-1, according to whether the C<LEX_STUFF_UTF8> flag is set
1036 in I<flags>. The characters are recoded for the lexer buffer, according
1037 to how the buffer is currently being interpreted (L</lex_bufutf8>).
1038 If it is not convenient to nul-terminate a string to be inserted, the
1039 L</lex_stuff_pvn> function is more appropriate.
1045 Perl_lex_stuff_pv(pTHX_ const char *pv, U32 flags)
1047 PERL_ARGS_ASSERT_LEX_STUFF_PV;
1048 lex_stuff_pvn(pv, strlen(pv), flags);
1052 =for apidoc Amx|void|lex_stuff_sv|SV *sv|U32 flags
1054 Insert characters into the lexer buffer (L</PL_parser-E<gt>linestr>),
1055 immediately after the current lexing point (L</PL_parser-E<gt>bufptr>),
1056 reallocating the buffer if necessary. This means that lexing code that
1057 runs later will see the characters as if they had appeared in the input.
1058 It is not recommended to do this as part of normal parsing, and most
1059 uses of this facility run the risk of the inserted characters being
1060 interpreted in an unintended manner.
1062 The string to be inserted is the string value of I<sv>. The characters
1063 are recoded for the lexer buffer, according to how the buffer is currently
1064 being interpreted (L</lex_bufutf8>). If a string to be inserted is
1065 not already a Perl scalar, the L</lex_stuff_pvn> function avoids the
1066 need to construct a scalar.
1072 Perl_lex_stuff_sv(pTHX_ SV *sv, U32 flags)
1076 PERL_ARGS_ASSERT_LEX_STUFF_SV;
1078 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_stuff_sv");
1080 lex_stuff_pvn(pv, len, flags | (SvUTF8(sv) ? LEX_STUFF_UTF8 : 0));
1084 =for apidoc Amx|void|lex_unstuff|char *ptr
1086 Discards text about to be lexed, from L</PL_parser-E<gt>bufptr> up to
1087 I<ptr>. Text following I<ptr> will be moved, and the buffer shortened.
1088 This hides the discarded text from any lexing code that runs later,
1089 as if the text had never appeared.
1091 This is not the normal way to consume lexed text. For that, use
1098 Perl_lex_unstuff(pTHX_ char *ptr)
1102 PERL_ARGS_ASSERT_LEX_UNSTUFF;
1103 buf = PL_parser->bufptr;
1105 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1108 bufend = PL_parser->bufend;
1110 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_unstuff");
1111 unstuff_len = ptr - buf;
1112 Move(ptr, buf, bufend+1-ptr, char);
1113 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - unstuff_len);
1114 PL_parser->bufend = bufend - unstuff_len;
1118 =for apidoc Amx|void|lex_read_to|char *ptr
1120 Consume text in the lexer buffer, from L</PL_parser-E<gt>bufptr> up
1121 to I<ptr>. This advances L</PL_parser-E<gt>bufptr> to match I<ptr>,
1122 performing the correct bookkeeping whenever a newline character is passed.
1123 This is the normal way to consume lexed text.
1125 Interpretation of the buffer's octets can be abstracted out by
1126 using the slightly higher-level functions L</lex_peek_unichar> and
1127 L</lex_read_unichar>.
1133 Perl_lex_read_to(pTHX_ char *ptr)
1136 PERL_ARGS_ASSERT_LEX_READ_TO;
1137 s = PL_parser->bufptr;
1138 if (ptr < s || ptr > PL_parser->bufend)
1139 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_to");
1140 for (; s != ptr; s++)
1142 CopLINE_inc(PL_curcop);
1143 PL_parser->linestart = s+1;
1145 PL_parser->bufptr = ptr;
1149 =for apidoc Amx|void|lex_discard_to|char *ptr
1151 Discards the first part of the L</PL_parser-E<gt>linestr> buffer,
1152 up to I<ptr>. The remaining content of the buffer will be moved, and
1153 all pointers into the buffer updated appropriately. I<ptr> must not
1154 be later in the buffer than the position of L</PL_parser-E<gt>bufptr>:
1155 it is not permitted to discard text that has yet to be lexed.
1157 Normally it is not necessarily to do this directly, because it suffices to
1158 use the implicit discarding behaviour of L</lex_next_chunk> and things
1159 based on it. However, if a token stretches across multiple lines,
1160 and the lexing code has kept multiple lines of text in the buffer for
1161 that purpose, then after completion of the token it would be wise to
1162 explicitly discard the now-unneeded earlier lines, to avoid future
1163 multi-line tokens growing the buffer without bound.
1169 Perl_lex_discard_to(pTHX_ char *ptr)
1173 PERL_ARGS_ASSERT_LEX_DISCARD_TO;
1174 buf = SvPVX(PL_parser->linestr);
1176 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1179 if (ptr > PL_parser->bufptr)
1180 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_discard_to");
1181 discard_len = ptr - buf;
1182 if (PL_parser->oldbufptr < ptr)
1183 PL_parser->oldbufptr = ptr;
1184 if (PL_parser->oldoldbufptr < ptr)
1185 PL_parser->oldoldbufptr = ptr;
1186 if (PL_parser->last_uni && PL_parser->last_uni < ptr)
1187 PL_parser->last_uni = NULL;
1188 if (PL_parser->last_lop && PL_parser->last_lop < ptr)
1189 PL_parser->last_lop = NULL;
1190 Move(ptr, buf, PL_parser->bufend+1-ptr, char);
1191 SvCUR_set(PL_parser->linestr, SvCUR(PL_parser->linestr) - discard_len);
1192 PL_parser->bufend -= discard_len;
1193 PL_parser->bufptr -= discard_len;
1194 PL_parser->oldbufptr -= discard_len;
1195 PL_parser->oldoldbufptr -= discard_len;
1196 if (PL_parser->last_uni)
1197 PL_parser->last_uni -= discard_len;
1198 if (PL_parser->last_lop)
1199 PL_parser->last_lop -= discard_len;
1203 =for apidoc Amx|bool|lex_next_chunk|U32 flags
1205 Reads in the next chunk of text to be lexed, appending it to
1206 L</PL_parser-E<gt>linestr>. This should be called when lexing code has
1207 looked to the end of the current chunk and wants to know more. It is
1208 usual, but not necessary, for lexing to have consumed the entirety of
1209 the current chunk at this time.
1211 If L</PL_parser-E<gt>bufptr> is pointing to the very end of the current
1212 chunk (i.e., the current chunk has been entirely consumed), normally the
1213 current chunk will be discarded at the same time that the new chunk is
1214 read in. If I<flags> includes C<LEX_KEEP_PREVIOUS>, the current chunk
1215 will not be discarded. If the current chunk has not been entirely
1216 consumed, then it will not be discarded regardless of the flag.
1218 Returns true if some new text was added to the buffer, or false if the
1219 buffer has reached the end of the input text.
1224 #define LEX_FAKE_EOF 0x80000000
1227 Perl_lex_next_chunk(pTHX_ U32 flags)
1231 STRLEN old_bufend_pos, new_bufend_pos;
1232 STRLEN bufptr_pos, oldbufptr_pos, oldoldbufptr_pos;
1233 STRLEN linestart_pos, last_uni_pos, last_lop_pos;
1234 bool got_some_for_debugger = 0;
1236 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_FAKE_EOF))
1237 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_next_chunk");
1238 linestr = PL_parser->linestr;
1239 buf = SvPVX(linestr);
1240 if (!(flags & LEX_KEEP_PREVIOUS) &&
1241 PL_parser->bufptr == PL_parser->bufend) {
1242 old_bufend_pos = bufptr_pos = oldbufptr_pos = oldoldbufptr_pos = 0;
1244 if (PL_parser->last_uni != PL_parser->bufend)
1245 PL_parser->last_uni = NULL;
1246 if (PL_parser->last_lop != PL_parser->bufend)
1247 PL_parser->last_lop = NULL;
1248 last_uni_pos = last_lop_pos = 0;
1252 old_bufend_pos = PL_parser->bufend - buf;
1253 bufptr_pos = PL_parser->bufptr - buf;
1254 oldbufptr_pos = PL_parser->oldbufptr - buf;
1255 oldoldbufptr_pos = PL_parser->oldoldbufptr - buf;
1256 linestart_pos = PL_parser->linestart - buf;
1257 last_uni_pos = PL_parser->last_uni ? PL_parser->last_uni - buf : 0;
1258 last_lop_pos = PL_parser->last_lop ? PL_parser->last_lop - buf : 0;
1260 if (flags & LEX_FAKE_EOF) {
1262 } else if (!PL_parser->rsfp) {
1264 } else if (filter_gets(linestr, old_bufend_pos)) {
1266 got_some_for_debugger = 1;
1268 if (!SvPOK(linestr)) /* can get undefined by filter_gets */
1269 sv_setpvs(linestr, "");
1271 /* End of real input. Close filehandle (unless it was STDIN),
1272 * then add implicit termination.
1274 if ((PerlIO*)PL_parser->rsfp == PerlIO_stdin())
1275 PerlIO_clearerr(PL_parser->rsfp);
1276 else if (PL_parser->rsfp)
1277 (void)PerlIO_close(PL_parser->rsfp);
1278 PL_parser->rsfp = NULL;
1279 PL_parser->in_pod = 0;
1281 if (PL_madskills && !PL_in_eval && (PL_minus_p || PL_minus_n))
1284 if (!PL_in_eval && PL_minus_p) {
1286 /*{*/";}continue{print or die qq(-p destination: $!\\n);}");
1287 PL_minus_n = PL_minus_p = 0;
1288 } else if (!PL_in_eval && PL_minus_n) {
1289 sv_catpvs(linestr, /*{*/";}");
1292 sv_catpvs(linestr, ";");
1295 buf = SvPVX(linestr);
1296 new_bufend_pos = SvCUR(linestr);
1297 PL_parser->bufend = buf + new_bufend_pos;
1298 PL_parser->bufptr = buf + bufptr_pos;
1299 PL_parser->oldbufptr = buf + oldbufptr_pos;
1300 PL_parser->oldoldbufptr = buf + oldoldbufptr_pos;
1301 PL_parser->linestart = buf + linestart_pos;
1302 if (PL_parser->last_uni)
1303 PL_parser->last_uni = buf + last_uni_pos;
1304 if (PL_parser->last_lop)
1305 PL_parser->last_lop = buf + last_lop_pos;
1306 if (got_some_for_debugger && (PERLDB_LINE || PERLDB_SAVESRC) &&
1307 PL_curstash != PL_debstash) {
1308 /* debugger active and we're not compiling the debugger code,
1309 * so store the line into the debugger's array of lines
1311 update_debugger_info(NULL, buf+old_bufend_pos,
1312 new_bufend_pos-old_bufend_pos);
1318 =for apidoc Amx|I32|lex_peek_unichar|U32 flags
1320 Looks ahead one (Unicode) character in the text currently being lexed.
1321 Returns the codepoint (unsigned integer value) of the next character,
1322 or -1 if lexing has reached the end of the input text. To consume the
1323 peeked character, use L</lex_read_unichar>.
1325 If the next character is in (or extends into) the next chunk of input
1326 text, the next chunk will be read in. Normally the current chunk will be
1327 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1328 then the current chunk will not be discarded.
1330 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1331 is encountered, an exception is generated.
1337 Perl_lex_peek_unichar(pTHX_ U32 flags)
1341 if (flags & ~(LEX_KEEP_PREVIOUS))
1342 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_peek_unichar");
1343 s = PL_parser->bufptr;
1344 bufend = PL_parser->bufend;
1350 if (!lex_next_chunk(flags))
1352 s = PL_parser->bufptr;
1353 bufend = PL_parser->bufend;
1359 len = PL_utf8skip[head];
1360 while ((STRLEN)(bufend-s) < len) {
1361 if (!lex_next_chunk(flags | LEX_KEEP_PREVIOUS))
1363 s = PL_parser->bufptr;
1364 bufend = PL_parser->bufend;
1367 unichar = utf8n_to_uvuni((U8*)s, bufend-s, &retlen, UTF8_CHECK_ONLY);
1368 if (retlen == (STRLEN)-1) {
1369 /* malformed UTF-8 */
1371 SAVESPTR(PL_warnhook);
1372 PL_warnhook = PERL_WARNHOOK_FATAL;
1373 utf8n_to_uvuni((U8*)s, bufend-s, NULL, 0);
1379 if (!lex_next_chunk(flags))
1381 s = PL_parser->bufptr;
1388 =for apidoc Amx|I32|lex_read_unichar|U32 flags
1390 Reads the next (Unicode) character in the text currently being lexed.
1391 Returns the codepoint (unsigned integer value) of the character read,
1392 and moves L</PL_parser-E<gt>bufptr> past the character, or returns -1
1393 if lexing has reached the end of the input text. To non-destructively
1394 examine the next character, use L</lex_peek_unichar> instead.
1396 If the next character is in (or extends into) the next chunk of input
1397 text, the next chunk will be read in. Normally the current chunk will be
1398 discarded at the same time, but if I<flags> includes C<LEX_KEEP_PREVIOUS>
1399 then the current chunk will not be discarded.
1401 If the input is being interpreted as UTF-8 and a UTF-8 encoding error
1402 is encountered, an exception is generated.
1408 Perl_lex_read_unichar(pTHX_ U32 flags)
1411 if (flags & ~(LEX_KEEP_PREVIOUS))
1412 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_unichar");
1413 c = lex_peek_unichar(flags);
1416 CopLINE_inc(PL_curcop);
1417 PL_parser->bufptr += UTF8SKIP(PL_parser->bufptr);
1423 =for apidoc Amx|void|lex_read_space|U32 flags
1425 Reads optional spaces, in Perl style, in the text currently being
1426 lexed. The spaces may include ordinary whitespace characters and
1427 Perl-style comments. C<#line> directives are processed if encountered.
1428 L</PL_parser-E<gt>bufptr> is moved past the spaces, so that it points
1429 at a non-space character (or the end of the input text).
1431 If spaces extend into the next chunk of input text, the next chunk will
1432 be read in. Normally the current chunk will be discarded at the same
1433 time, but if I<flags> includes C<LEX_KEEP_PREVIOUS> then the current
1434 chunk will not be discarded.
1439 #define LEX_NO_NEXT_CHUNK 0x80000000
1442 Perl_lex_read_space(pTHX_ U32 flags)
1445 bool need_incline = 0;
1446 if (flags & ~(LEX_KEEP_PREVIOUS|LEX_NO_NEXT_CHUNK))
1447 Perl_croak(aTHX_ "Lexing code internal error (%s)", "lex_read_space");
1450 sv_free(PL_skipwhite);
1451 PL_skipwhite = NULL;
1454 PL_skipwhite = newSVpvs("");
1455 #endif /* PERL_MAD */
1456 s = PL_parser->bufptr;
1457 bufend = PL_parser->bufend;
1463 } while (!(c == '\n' || (c == 0 && s == bufend)));
1464 } else if (c == '\n') {
1466 PL_parser->linestart = s;
1471 } else if (isSPACE(c)) {
1473 } else if (c == 0 && s == bufend) {
1477 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1478 #endif /* PERL_MAD */
1479 if (flags & LEX_NO_NEXT_CHUNK)
1481 PL_parser->bufptr = s;
1482 CopLINE_inc(PL_curcop);
1483 got_more = lex_next_chunk(flags);
1484 CopLINE_dec(PL_curcop);
1485 s = PL_parser->bufptr;
1486 bufend = PL_parser->bufend;
1489 if (need_incline && PL_parser->rsfp) {
1499 sv_catpvn(PL_skipwhite, PL_parser->bufptr, s-PL_parser->bufptr);
1500 #endif /* PERL_MAD */
1501 PL_parser->bufptr = s;
1506 * This subroutine has nothing to do with tilting, whether at windmills
1507 * or pinball tables. Its name is short for "increment line". It
1508 * increments the current line number in CopLINE(PL_curcop) and checks
1509 * to see whether the line starts with a comment of the form
1510 * # line 500 "foo.pm"
1511 * If so, it sets the current line number and file to the values in the comment.
1515 S_incline(pTHX_ const char *s)
1523 PERL_ARGS_ASSERT_INCLINE;
1525 CopLINE_inc(PL_curcop);
1528 while (SPACE_OR_TAB(*s))
1530 if (strnEQ(s, "line", 4))
1534 if (SPACE_OR_TAB(*s))
1538 while (SPACE_OR_TAB(*s))
1546 if (!SPACE_OR_TAB(*s) && *s != '\r' && *s != '\n' && *s != '\0')
1548 while (SPACE_OR_TAB(*s))
1550 if (*s == '"' && (t = strchr(s+1, '"'))) {
1556 while (!isSPACE(*t))
1560 while (SPACE_OR_TAB(*e) || *e == '\r' || *e == '\f')
1562 if (*e != '\n' && *e != '\0')
1563 return; /* false alarm */
1565 line_num = atoi(n)-1;
1568 const STRLEN len = t - s;
1569 SV *const temp_sv = CopFILESV(PL_curcop);
1574 cf = SvPVX(temp_sv);
1575 tmplen = SvCUR(temp_sv);
1581 if (tmplen > 7 && strnEQ(cf, "(eval ", 6)) {
1582 /* must copy *{"::_<(eval N)[oldfilename:L]"}
1583 * to *{"::_<newfilename"} */
1584 /* However, the long form of evals is only turned on by the
1585 debugger - usually they're "(eval %lu)" */
1589 STRLEN tmplen2 = len;
1590 if (tmplen + 2 <= sizeof smallbuf)
1593 Newx(tmpbuf, tmplen + 2, char);
1596 memcpy(tmpbuf + 2, cf, tmplen);
1598 gvp = (GV**)hv_fetch(PL_defstash, tmpbuf, tmplen, FALSE);
1603 if (tmplen2 + 2 <= sizeof smallbuf)
1606 Newx(tmpbuf2, tmplen2 + 2, char);
1608 if (tmpbuf2 != smallbuf || tmpbuf != smallbuf) {
1609 /* Either they malloc'd it, or we malloc'd it,
1610 so no prefix is present in ours. */
1615 memcpy(tmpbuf2 + 2, s, tmplen2);
1618 gv2 = *(GV**)hv_fetch(PL_defstash, tmpbuf2, tmplen2, TRUE);
1620 gv_init(gv2, PL_defstash, tmpbuf2, tmplen2, FALSE);
1621 /* adjust ${"::_<newfilename"} to store the new file name */
1622 GvSV(gv2) = newSVpvn(tmpbuf2 + 2, tmplen2 - 2);
1623 /* The line number may differ. If that is the case,
1624 alias the saved lines that are in the array.
1625 Otherwise alias the whole array. */
1626 if (CopLINE(PL_curcop) == line_num) {
1627 GvHV(gv2) = MUTABLE_HV(SvREFCNT_inc(GvHV(*gvp)));
1628 GvAV(gv2) = MUTABLE_AV(SvREFCNT_inc(GvAV(*gvp)));
1630 else if (GvAV(*gvp)) {
1631 AV * const av = GvAV(*gvp);
1632 const I32 start = CopLINE(PL_curcop)+1;
1633 I32 items = AvFILLp(av) - start;
1635 AV * const av2 = GvAVn(gv2);
1636 SV **svp = AvARRAY(av) + start;
1637 I32 l = (I32)line_num+1;
1639 av_store(av2, l++, SvREFCNT_inc(*svp++));
1644 if (tmpbuf2 != smallbuf) Safefree(tmpbuf2);
1646 if (tmpbuf != smallbuf) Safefree(tmpbuf);
1648 CopFILE_free(PL_curcop);
1649 CopFILE_setn(PL_curcop, s, len);
1651 CopLINE_set(PL_curcop, line_num);
1655 /* skip space before PL_thistoken */
1658 S_skipspace0(pTHX_ register char *s)
1660 PERL_ARGS_ASSERT_SKIPSPACE0;
1667 PL_thiswhite = newSVpvs("");
1668 sv_catsv(PL_thiswhite, PL_skipwhite);
1669 sv_free(PL_skipwhite);
1672 PL_realtokenstart = s - SvPVX(PL_linestr);
1676 /* skip space after PL_thistoken */
1679 S_skipspace1(pTHX_ register char *s)
1681 const char *start = s;
1682 I32 startoff = start - SvPVX(PL_linestr);
1684 PERL_ARGS_ASSERT_SKIPSPACE1;
1689 start = SvPVX(PL_linestr) + startoff;
1690 if (!PL_thistoken && PL_realtokenstart >= 0) {
1691 const char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1692 PL_thistoken = newSVpvn(tstart, start - tstart);
1694 PL_realtokenstart = -1;
1697 PL_nextwhite = newSVpvs("");
1698 sv_catsv(PL_nextwhite, PL_skipwhite);
1699 sv_free(PL_skipwhite);
1706 S_skipspace2(pTHX_ register char *s, SV **svp)
1709 const I32 bufptroff = PL_bufptr - SvPVX(PL_linestr);
1710 const I32 startoff = s - SvPVX(PL_linestr);
1712 PERL_ARGS_ASSERT_SKIPSPACE2;
1715 PL_bufptr = SvPVX(PL_linestr) + bufptroff;
1716 if (!PL_madskills || !svp)
1718 start = SvPVX(PL_linestr) + startoff;
1719 if (!PL_thistoken && PL_realtokenstart >= 0) {
1720 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
1721 PL_thistoken = newSVpvn(tstart, start - tstart);
1722 PL_realtokenstart = -1;
1726 *svp = newSVpvs("");
1727 sv_setsv(*svp, PL_skipwhite);
1728 sv_free(PL_skipwhite);
1737 S_update_debugger_info(pTHX_ SV *orig_sv, const char *const buf, STRLEN len)
1739 AV *av = CopFILEAVx(PL_curcop);
1741 SV * const sv = newSV_type(SVt_PVMG);
1743 sv_setsv(sv, orig_sv);
1745 sv_setpvn(sv, buf, len);
1748 av_store(av, (I32)CopLINE(PL_curcop), sv);
1754 * Called to gobble the appropriate amount and type of whitespace.
1755 * Skips comments as well.
1759 S_skipspace(pTHX_ register char *s)
1763 #endif /* PERL_MAD */
1764 PERL_ARGS_ASSERT_SKIPSPACE;
1767 sv_free(PL_skipwhite);
1768 PL_skipwhite = NULL;
1770 #endif /* PERL_MAD */
1771 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
1772 while (s < PL_bufend && SPACE_OR_TAB(*s))
1775 STRLEN bufptr_pos = PL_bufptr - SvPVX(PL_linestr);
1777 lex_read_space(LEX_KEEP_PREVIOUS |
1778 (PL_sublex_info.sub_inwhat || PL_lex_state == LEX_FORMLINE ?
1779 LEX_NO_NEXT_CHUNK : 0));
1781 PL_bufptr = SvPVX(PL_linestr) + bufptr_pos;
1782 if (PL_linestart > PL_bufptr)
1783 PL_bufptr = PL_linestart;
1788 PL_skipwhite = newSVpvn(start, s-start);
1789 #endif /* PERL_MAD */
1795 * Check the unary operators to ensure there's no ambiguity in how they're
1796 * used. An ambiguous piece of code would be:
1798 * This doesn't mean rand() + 5. Because rand() is a unary operator,
1799 * the +5 is its argument.
1809 if (PL_oldoldbufptr != PL_last_uni)
1811 while (isSPACE(*PL_last_uni))
1814 while (isALNUM_lazy_if(s,UTF) || *s == '-')
1816 if ((t = strchr(s, '(')) && t < PL_bufptr)
1819 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
1820 "Warning: Use of \"%.*s\" without parentheses is ambiguous",
1821 (int)(s - PL_last_uni), PL_last_uni);
1825 * LOP : macro to build a list operator. Its behaviour has been replaced
1826 * with a subroutine, S_lop() for which LOP is just another name.
1829 #define LOP(f,x) return lop(f,x,s)
1833 * Build a list operator (or something that might be one). The rules:
1834 * - if we have a next token, then it's a list operator [why?]
1835 * - if the next thing is an opening paren, then it's a function
1836 * - else it's a list operator
1840 S_lop(pTHX_ I32 f, int x, char *s)
1844 PERL_ARGS_ASSERT_LOP;
1850 PL_last_lop = PL_oldbufptr;
1851 PL_last_lop_op = (OPCODE)f;
1860 return REPORT(FUNC);
1863 return REPORT(FUNC);
1866 if (!PL_lex_allbrackets && PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
1867 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
1868 return REPORT(LSTOP);
1875 * Sets up for an eventual force_next(). start_force(0) basically does
1876 * an unshift, while start_force(-1) does a push. yylex removes items
1881 S_start_force(pTHX_ int where)
1885 if (where < 0) /* so people can duplicate start_force(PL_curforce) */
1886 where = PL_lasttoke;
1887 assert(PL_curforce < 0 || PL_curforce == where);
1888 if (PL_curforce != where) {
1889 for (i = PL_lasttoke; i > where; --i) {
1890 PL_nexttoke[i] = PL_nexttoke[i-1];
1894 if (PL_curforce < 0) /* in case of duplicate start_force() */
1895 Zero(&PL_nexttoke[where], 1, NEXTTOKE);
1896 PL_curforce = where;
1899 curmad('^', newSVpvs(""));
1900 CURMAD('_', PL_nextwhite);
1905 S_curmad(pTHX_ char slot, SV *sv)
1911 if (PL_curforce < 0)
1912 where = &PL_thismad;
1914 where = &PL_nexttoke[PL_curforce].next_mad;
1920 if (UTF && is_utf8_string((U8*)SvPVX(sv), SvCUR(sv)))
1922 else if (PL_encoding) {
1923 sv_recode_to_utf8(sv, PL_encoding);
1928 /* keep a slot open for the head of the list? */
1929 if (slot != '_' && *where && (*where)->mad_key == '^') {
1930 (*where)->mad_key = slot;
1931 sv_free(MUTABLE_SV(((*where)->mad_val)));
1932 (*where)->mad_val = (void*)sv;
1935 addmad(newMADsv(slot, sv), where, 0);
1938 # define start_force(where) NOOP
1939 # define curmad(slot, sv) NOOP
1944 * When the lexer realizes it knows the next token (for instance,
1945 * it is reordering tokens for the parser) then it can call S_force_next
1946 * to know what token to return the next time the lexer is called. Caller
1947 * will need to set PL_nextval[] (or PL_nexttoke[].next_val with PERL_MAD),
1948 * and possibly PL_expect to ensure the lexer handles the token correctly.
1952 S_force_next(pTHX_ I32 type)
1957 PerlIO_printf(Perl_debug_log, "### forced token:\n");
1958 tokereport(type, &NEXTVAL_NEXTTOKE);
1962 if (PL_curforce < 0)
1963 start_force(PL_lasttoke);
1964 PL_nexttoke[PL_curforce].next_type = type;
1965 if (PL_lex_state != LEX_KNOWNEXT)
1966 PL_lex_defer = PL_lex_state;
1967 PL_lex_state = LEX_KNOWNEXT;
1968 PL_lex_expect = PL_expect;
1971 PL_nexttype[PL_nexttoke] = type;
1973 if (PL_lex_state != LEX_KNOWNEXT) {
1974 PL_lex_defer = PL_lex_state;
1975 PL_lex_expect = PL_expect;
1976 PL_lex_state = LEX_KNOWNEXT;
1984 int yyc = PL_parser->yychar;
1985 if (yyc != YYEMPTY) {
1988 NEXTVAL_NEXTTOKE = PL_parser->yylval;
1989 if (yyc == '{'/*}*/ || yyc == HASHBRACK || yyc == '['/*]*/) {
1990 PL_lex_allbrackets--;
1992 yyc |= (3<<24) | (PL_lex_brackstack[PL_lex_brackets] << 16);
1993 } else if (yyc == '('/*)*/) {
1994 PL_lex_allbrackets--;
1999 PL_parser->yychar = YYEMPTY;
2004 S_newSV_maybe_utf8(pTHX_ const char *const start, STRLEN len)
2007 SV * const sv = newSVpvn_utf8(start, len,
2010 && !is_ascii_string((const U8*)start, len)
2011 && is_utf8_string((const U8*)start, len));
2017 * When the lexer knows the next thing is a word (for instance, it has
2018 * just seen -> and it knows that the next char is a word char, then
2019 * it calls S_force_word to stick the next word into the PL_nexttoke/val
2023 * char *start : buffer position (must be within PL_linestr)
2024 * int token : PL_next* will be this type of bare word (e.g., METHOD,WORD)
2025 * int check_keyword : if true, Perl checks to make sure the word isn't
2026 * a keyword (do this if the word is a label, e.g. goto FOO)
2027 * int allow_pack : if true, : characters will also be allowed (require,
2028 * use, etc. do this)
2029 * int allow_initial_tick : used by the "sub" lexer only.
2033 S_force_word(pTHX_ register char *start, int token, int check_keyword, int allow_pack, int allow_initial_tick)
2039 PERL_ARGS_ASSERT_FORCE_WORD;
2041 start = SKIPSPACE1(start);
2043 if (isIDFIRST_lazy_if(s,UTF) ||
2044 (allow_pack && *s == ':') ||
2045 (allow_initial_tick && *s == '\'') )
2047 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, allow_pack, &len);
2048 if (check_keyword && keyword(PL_tokenbuf, len, 0))
2050 start_force(PL_curforce);
2052 curmad('X', newSVpvn(start,s-start));
2053 if (token == METHOD) {
2058 PL_expect = XOPERATOR;
2062 curmad('g', newSVpvs( "forced" ));
2063 NEXTVAL_NEXTTOKE.opval
2064 = (OP*)newSVOP(OP_CONST,0,
2065 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
2066 NEXTVAL_NEXTTOKE.opval->op_private |= OPpCONST_BARE;
2074 * Called when the lexer wants $foo *foo &foo etc, but the program
2075 * text only contains the "foo" portion. The first argument is a pointer
2076 * to the "foo", and the second argument is the type symbol to prefix.
2077 * Forces the next token to be a "WORD".
2078 * Creates the symbol if it didn't already exist (via gv_fetchpv()).
2082 S_force_ident(pTHX_ register const char *s, int kind)
2086 PERL_ARGS_ASSERT_FORCE_IDENT;
2089 const STRLEN len = strlen(s);
2090 OP* const o = (OP*)newSVOP(OP_CONST, 0, newSVpvn(s, len));
2091 start_force(PL_curforce);
2092 NEXTVAL_NEXTTOKE.opval = o;
2095 o->op_private = OPpCONST_ENTERED;
2096 /* XXX see note in pp_entereval() for why we forgo typo
2097 warnings if the symbol must be introduced in an eval.
2099 gv_fetchpvn_flags(s, len,
2100 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL)
2102 kind == '$' ? SVt_PV :
2103 kind == '@' ? SVt_PVAV :
2104 kind == '%' ? SVt_PVHV :
2112 Perl_str_to_version(pTHX_ SV *sv)
2117 const char *start = SvPV_const(sv,len);
2118 const char * const end = start + len;
2119 const bool utf = SvUTF8(sv) ? TRUE : FALSE;
2121 PERL_ARGS_ASSERT_STR_TO_VERSION;
2123 while (start < end) {
2127 n = utf8n_to_uvchr((U8*)start, len, &skip, 0);
2132 retval += ((NV)n)/nshift;
2141 * Forces the next token to be a version number.
2142 * If the next token appears to be an invalid version number, (e.g. "v2b"),
2143 * and if "guessing" is TRUE, then no new token is created (and the caller
2144 * must use an alternative parsing method).
2148 S_force_version(pTHX_ char *s, int guessing)
2154 I32 startoff = s - SvPVX(PL_linestr);
2157 PERL_ARGS_ASSERT_FORCE_VERSION;
2165 while (isDIGIT(*d) || *d == '_' || *d == '.')
2169 start_force(PL_curforce);
2170 curmad('X', newSVpvn(s,d-s));
2173 if (*d == ';' || isSPACE(*d) || *d == '{' || *d == '}' || !*d) {
2175 #ifdef USE_LOCALE_NUMERIC
2176 char *loc = setlocale(LC_NUMERIC, "C");
2178 s = scan_num(s, &pl_yylval);
2179 #ifdef USE_LOCALE_NUMERIC
2180 setlocale(LC_NUMERIC, loc);
2182 version = pl_yylval.opval;
2183 ver = cSVOPx(version)->op_sv;
2184 if (SvPOK(ver) && !SvNIOK(ver)) {
2185 SvUPGRADE(ver, SVt_PVNV);
2186 SvNV_set(ver, str_to_version(ver));
2187 SvNOK_on(ver); /* hint that it is a version */
2190 else if (guessing) {
2193 sv_free(PL_nextwhite); /* let next token collect whitespace */
2195 s = SvPVX(PL_linestr) + startoff;
2203 if (PL_madskills && !version) {
2204 sv_free(PL_nextwhite); /* let next token collect whitespace */
2206 s = SvPVX(PL_linestr) + startoff;
2209 /* NOTE: The parser sees the package name and the VERSION swapped */
2210 start_force(PL_curforce);
2211 NEXTVAL_NEXTTOKE.opval = version;
2218 * S_force_strict_version
2219 * Forces the next token to be a version number using strict syntax rules.
2223 S_force_strict_version(pTHX_ char *s)
2228 I32 startoff = s - SvPVX(PL_linestr);
2230 const char *errstr = NULL;
2232 PERL_ARGS_ASSERT_FORCE_STRICT_VERSION;
2234 while (isSPACE(*s)) /* leading whitespace */
2237 if (is_STRICT_VERSION(s,&errstr)) {
2239 s = (char *)scan_version(s, ver, 0);
2240 version = newSVOP(OP_CONST, 0, ver);
2242 else if ( (*s != ';' && *s != '{' && *s != '}' ) &&
2243 (s = SKIPSPACE1(s), (*s != ';' && *s != '{' && *s != '}' )))
2247 yyerror(errstr); /* version required */
2252 if (PL_madskills && !version) {
2253 sv_free(PL_nextwhite); /* let next token collect whitespace */
2255 s = SvPVX(PL_linestr) + startoff;
2258 /* NOTE: The parser sees the package name and the VERSION swapped */
2259 start_force(PL_curforce);
2260 NEXTVAL_NEXTTOKE.opval = version;
2268 * Tokenize a quoted string passed in as an SV. It finds the next
2269 * chunk, up to end of string or a backslash. It may make a new
2270 * SV containing that chunk (if HINT_NEW_STRING is on). It also
2275 S_tokeq(pTHX_ SV *sv)
2279 register char *send;
2284 PERL_ARGS_ASSERT_TOKEQ;
2289 s = SvPV_force(sv, len);
2290 if (SvTYPE(sv) >= SVt_PVIV && SvIVX(sv) == -1)
2293 /* This is relying on the SV being "well formed" with a trailing '\0' */
2294 while (s < send && !(*s == '\\' && s[1] == '\\'))
2299 if ( PL_hints & HINT_NEW_STRING ) {
2300 pv = newSVpvn_flags(SvPVX_const(pv), len, SVs_TEMP | SvUTF8(sv));
2304 if (s + 1 < send && (s[1] == '\\'))
2305 s++; /* all that, just for this */
2310 SvCUR_set(sv, d - SvPVX_const(sv));
2312 if ( PL_hints & HINT_NEW_STRING )
2313 return new_constant(NULL, 0, "q", sv, pv, "q", 1);
2318 * Now come three functions related to double-quote context,
2319 * S_sublex_start, S_sublex_push, and S_sublex_done. They're used when
2320 * converting things like "\u\Lgnat" into ucfirst(lc("gnat")). They
2321 * interact with PL_lex_state, and create fake ( ... ) argument lists
2322 * to handle functions and concatenation.
2323 * They assume that whoever calls them will be setting up a fake
2324 * join call, because each subthing puts a ',' after it. This lets
2327 * join($, , 'lower ', lcfirst( 'uPpEr', ) ,)
2329 * (I'm not sure whether the spurious commas at the end of lcfirst's
2330 * arguments and join's arguments are created or not).
2335 * Assumes that pl_yylval.ival is the op we're creating (e.g. OP_LCFIRST).
2337 * Pattern matching will set PL_lex_op to the pattern-matching op to
2338 * make (we return THING if pl_yylval.ival is OP_NULL, PMFUNC otherwise).
2340 * OP_CONST and OP_READLINE are easy--just make the new op and return.
2342 * Everything else becomes a FUNC.
2344 * Sets PL_lex_state to LEX_INTERPPUSH unless (ival was OP_NULL or we
2345 * had an OP_CONST or OP_READLINE). This just sets us up for a
2346 * call to S_sublex_push().
2350 S_sublex_start(pTHX)
2353 register const I32 op_type = pl_yylval.ival;
2355 if (op_type == OP_NULL) {
2356 pl_yylval.opval = PL_lex_op;
2360 if (op_type == OP_CONST || op_type == OP_READLINE) {
2361 SV *sv = tokeq(PL_lex_stuff);
2363 if (SvTYPE(sv) == SVt_PVIV) {
2364 /* Overloaded constants, nothing fancy: Convert to SVt_PV: */
2366 const char * const p = SvPV_const(sv, len);
2367 SV * const nsv = newSVpvn_flags(p, len, SvUTF8(sv));
2371 pl_yylval.opval = (OP*)newSVOP(op_type, 0, sv);
2372 PL_lex_stuff = NULL;
2373 /* Allow <FH> // "foo" */
2374 if (op_type == OP_READLINE)
2375 PL_expect = XTERMORDORDOR;
2378 else if (op_type == OP_BACKTICK && PL_lex_op) {
2379 /* readpipe() vas overriden */
2380 cSVOPx(cLISTOPx(cUNOPx(PL_lex_op)->op_first)->op_first->op_sibling)->op_sv = tokeq(PL_lex_stuff);
2381 pl_yylval.opval = PL_lex_op;
2383 PL_lex_stuff = NULL;
2387 PL_sublex_info.super_state = PL_lex_state;
2388 PL_sublex_info.sub_inwhat = (U16)op_type;
2389 PL_sublex_info.sub_op = PL_lex_op;
2390 PL_lex_state = LEX_INTERPPUSH;
2394 pl_yylval.opval = PL_lex_op;
2404 * Create a new scope to save the lexing state. The scope will be
2405 * ended in S_sublex_done. Returns a '(', starting the function arguments
2406 * to the uc, lc, etc. found before.
2407 * Sets PL_lex_state to LEX_INTERPCONCAT.
2416 PL_lex_state = PL_sublex_info.super_state;
2417 SAVEBOOL(PL_lex_dojoin);
2418 SAVEI32(PL_lex_brackets);
2419 SAVEI32(PL_lex_allbrackets);
2420 SAVEI8(PL_lex_fakeeof);
2421 SAVEI32(PL_lex_casemods);
2422 SAVEI32(PL_lex_starts);
2423 SAVEI8(PL_lex_state);
2424 SAVEVPTR(PL_lex_inpat);
2425 SAVEI16(PL_lex_inwhat);
2426 SAVECOPLINE(PL_curcop);
2427 SAVEPPTR(PL_bufptr);
2428 SAVEPPTR(PL_bufend);
2429 SAVEPPTR(PL_oldbufptr);
2430 SAVEPPTR(PL_oldoldbufptr);
2431 SAVEPPTR(PL_last_lop);
2432 SAVEPPTR(PL_last_uni);
2433 SAVEPPTR(PL_linestart);
2434 SAVESPTR(PL_linestr);
2435 SAVEGENERICPV(PL_lex_brackstack);
2436 SAVEGENERICPV(PL_lex_casestack);
2438 PL_linestr = PL_lex_stuff;
2439 PL_lex_stuff = NULL;
2441 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart
2442 = SvPVX(PL_linestr);
2443 PL_bufend += SvCUR(PL_linestr);
2444 PL_last_lop = PL_last_uni = NULL;
2445 SAVEFREESV(PL_linestr);
2447 PL_lex_dojoin = FALSE;
2448 PL_lex_brackets = 0;
2449 PL_lex_allbrackets = 0;
2450 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2451 Newx(PL_lex_brackstack, 120, char);
2452 Newx(PL_lex_casestack, 12, char);
2453 PL_lex_casemods = 0;
2454 *PL_lex_casestack = '\0';
2456 PL_lex_state = LEX_INTERPCONCAT;
2457 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
2459 PL_lex_inwhat = PL_sublex_info.sub_inwhat;
2460 if (PL_lex_inwhat == OP_TRANSR) PL_lex_inwhat = OP_TRANS;
2461 if (PL_lex_inwhat == OP_MATCH || PL_lex_inwhat == OP_QR || PL_lex_inwhat == OP_SUBST)
2462 PL_lex_inpat = PL_sublex_info.sub_op;
2464 PL_lex_inpat = NULL;
2471 * Restores lexer state after a S_sublex_push.
2478 if (!PL_lex_starts++) {
2479 SV * const sv = newSVpvs("");
2480 if (SvUTF8(PL_linestr))
2482 PL_expect = XOPERATOR;
2483 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
2487 if (PL_lex_casemods) { /* oops, we've got some unbalanced parens */
2488 PL_lex_state = LEX_INTERPCASEMOD;
2492 /* Is there a right-hand side to take care of? (s//RHS/ or tr//RHS/) */
2493 assert(PL_lex_inwhat != OP_TRANSR);
2494 if (PL_lex_repl && (PL_lex_inwhat == OP_SUBST || PL_lex_inwhat == OP_TRANS)) {
2495 PL_linestr = PL_lex_repl;
2497 PL_bufend = PL_bufptr = PL_oldbufptr = PL_oldoldbufptr = PL_linestart = SvPVX(PL_linestr);
2498 PL_bufend += SvCUR(PL_linestr);
2499 PL_last_lop = PL_last_uni = NULL;
2500 SAVEFREESV(PL_linestr);
2501 PL_lex_dojoin = FALSE;
2502 PL_lex_brackets = 0;
2503 PL_lex_allbrackets = 0;
2504 PL_lex_fakeeof = LEX_FAKEEOF_NEVER;
2505 PL_lex_casemods = 0;
2506 *PL_lex_casestack = '\0';
2508 if (SvEVALED(PL_lex_repl)) {
2509 PL_lex_state = LEX_INTERPNORMAL;
2511 /* we don't clear PL_lex_repl here, so that we can check later
2512 whether this is an evalled subst; that means we rely on the
2513 logic to ensure sublex_done() is called again only via the
2514 branch (in yylex()) that clears PL_lex_repl, else we'll loop */
2517 PL_lex_state = LEX_INTERPCONCAT;
2527 PL_endwhite = newSVpvs("");
2528 sv_catsv(PL_endwhite, PL_thiswhite);
2532 sv_setpvs(PL_thistoken,"");
2534 PL_realtokenstart = -1;
2538 PL_bufend = SvPVX(PL_linestr);
2539 PL_bufend += SvCUR(PL_linestr);
2540 PL_expect = XOPERATOR;
2541 PL_sublex_info.sub_inwhat = 0;
2549 Extracts a pattern, double-quoted string, or transliteration. This
2552 It looks at PL_lex_inwhat and PL_lex_inpat to find out whether it's
2553 processing a pattern (PL_lex_inpat is true), a transliteration
2554 (PL_lex_inwhat == OP_TRANS is true), or a double-quoted string.
2556 Returns a pointer to the character scanned up to. If this is
2557 advanced from the start pointer supplied (i.e. if anything was
2558 successfully parsed), will leave an OP for the substring scanned
2559 in pl_yylval. Caller must intuit reason for not parsing further
2560 by looking at the next characters herself.
2564 constants: \N{NAME} only
2565 case and quoting: \U \Q \E
2566 stops on @ and $, but not for $ as tail anchor
2568 In transliterations:
2569 characters are VERY literal, except for - not at the start or end
2570 of the string, which indicates a range. If the range is in bytes,
2571 scan_const expands the range to the full set of intermediate
2572 characters. If the range is in utf8, the hyphen is replaced with
2573 a certain range mark which will be handled by pmtrans() in op.c.
2575 In double-quoted strings:
2577 double-quoted style: \r and \n
2578 constants: \x31, etc.
2579 deprecated backrefs: \1 (in substitution replacements)
2580 case and quoting: \U \Q \E
2583 scan_const does *not* construct ops to handle interpolated strings.
2584 It stops processing as soon as it finds an embedded $ or @ variable
2585 and leaves it to the caller to work out what's going on.
2587 embedded arrays (whether in pattern or not) could be:
2588 @foo, @::foo, @'foo, @{foo}, @$foo, @+, @-.
2590 $ in double-quoted strings must be the symbol of an embedded scalar.
2592 $ in pattern could be $foo or could be tail anchor. Assumption:
2593 it's a tail anchor if $ is the last thing in the string, or if it's
2594 followed by one of "()| \r\n\t"
2596 \1 (backreferences) are turned into $1
2598 The structure of the code is
2599 while (there's a character to process) {
2600 handle transliteration ranges
2601 skip regexp comments /(?#comment)/ and codes /(?{code})/
2602 skip #-initiated comments in //x patterns
2603 check for embedded arrays
2604 check for embedded scalars
2606 deprecate \1 in substitution replacements
2607 handle string-changing backslashes \l \U \Q \E, etc.
2608 switch (what was escaped) {
2609 handle \- in a transliteration (becomes a literal -)
2610 if a pattern and not \N{, go treat as regular character
2611 handle \132 (octal characters)
2612 handle \x15 and \x{1234} (hex characters)
2613 handle \N{name} (named characters, also \N{3,5} in a pattern)
2614 handle \cV (control characters)
2615 handle printf-style backslashes (\f, \r, \n, etc)
2618 } (end if backslash)
2619 handle regular character
2620 } (end while character to read)
2625 S_scan_const(pTHX_ char *start)
2628 register char *send = PL_bufend; /* end of the constant */
2629 SV *sv = newSV(send - start); /* sv for the constant. See
2630 note below on sizing. */
2631 register char *s = start; /* start of the constant */
2632 register char *d = SvPVX(sv); /* destination for copies */
2633 bool dorange = FALSE; /* are we in a translit range? */
2634 bool didrange = FALSE; /* did we just finish a range? */
2635 bool has_utf8 = FALSE; /* Output constant is UTF8 */
2636 bool this_utf8 = cBOOL(UTF); /* Is the source string assumed
2637 to be UTF8? But, this can
2638 show as true when the source
2639 isn't utf8, as for example
2640 when it is entirely composed
2643 /* Note on sizing: The scanned constant is placed into sv, which is
2644 * initialized by newSV() assuming one byte of output for every byte of
2645 * input. This routine expects newSV() to allocate an extra byte for a
2646 * trailing NUL, which this routine will append if it gets to the end of
2647 * the input. There may be more bytes of input than output (eg., \N{LATIN
2648 * CAPITAL LETTER A}), or more output than input if the constant ends up
2649 * recoded to utf8, but each time a construct is found that might increase
2650 * the needed size, SvGROW() is called. Its size parameter each time is
2651 * based on the best guess estimate at the time, namely the length used so
2652 * far, plus the length the current construct will occupy, plus room for
2653 * the trailing NUL, plus one byte for every input byte still unscanned */
2657 UV literal_endpoint = 0;
2658 bool native_range = TRUE; /* turned to FALSE if the first endpoint is Unicode. */
2661 PERL_ARGS_ASSERT_SCAN_CONST;
2663 assert(PL_lex_inwhat != OP_TRANSR);
2664 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
2665 /* If we are doing a trans and we know we want UTF8 set expectation */
2666 has_utf8 = PL_sublex_info.sub_op->op_private & (OPpTRANS_FROM_UTF|OPpTRANS_TO_UTF);
2667 this_utf8 = PL_sublex_info.sub_op->op_private & (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
2671 while (s < send || dorange) {
2673 /* get transliterations out of the way (they're most literal) */
2674 if (PL_lex_inwhat == OP_TRANS) {
2675 /* expand a range A-Z to the full set of characters. AIE! */
2677 I32 i; /* current expanded character */
2678 I32 min; /* first character in range */
2679 I32 max; /* last character in range */
2690 char * const c = (char*)utf8_hop((U8*)d, -1);
2694 *c = (char)UTF_TO_NATIVE(0xff);
2695 /* mark the range as done, and continue */
2701 i = d - SvPVX_const(sv); /* remember current offset */
2704 SvLEN(sv) + (has_utf8 ?
2705 (512 - UTF_CONTINUATION_MARK +
2708 /* How many two-byte within 0..255: 128 in UTF-8,
2709 * 96 in UTF-8-mod. */
2711 SvGROW(sv, SvLEN(sv) + 256); /* never more than 256 chars in a range */
2713 d = SvPVX(sv) + i; /* refresh d after realloc */
2717 for (j = 0; j <= 1; j++) {
2718 char * const c = (char*)utf8_hop((U8*)d, -1);
2719 const UV uv = utf8n_to_uvchr((U8*)c, d - c, NULL, 0);
2725 max = (U8)0xff; /* only to \xff */
2726 uvmax = uv; /* \x{100} to uvmax */
2728 d = c; /* eat endpoint chars */
2733 d -= 2; /* eat the first char and the - */
2734 min = (U8)*d; /* first char in range */
2735 max = (U8)d[1]; /* last char in range */
2742 "Invalid range \"%c-%c\" in transliteration operator",
2743 (char)min, (char)max);
2747 if (literal_endpoint == 2 &&
2748 ((isLOWER(min) && isLOWER(max)) ||
2749 (isUPPER(min) && isUPPER(max)))) {
2751 for (i = min; i <= max; i++)
2753 *d++ = NATIVE_TO_NEED(has_utf8,i);
2755 for (i = min; i <= max; i++)
2757 *d++ = NATIVE_TO_NEED(has_utf8,i);
2762 for (i = min; i <= max; i++)
2765 const U8 ch = (U8)NATIVE_TO_UTF(i);
2766 if (UNI_IS_INVARIANT(ch))
2769 *d++ = (U8)UTF8_EIGHT_BIT_HI(ch);
2770 *d++ = (U8)UTF8_EIGHT_BIT_LO(ch);
2779 d = (char*)uvchr_to_utf8((U8*)d, 0x100);
2781 *d++ = (char)UTF_TO_NATIVE(0xff);
2783 d = (char*)uvchr_to_utf8((U8*)d, uvmax);
2787 /* mark the range as done, and continue */
2791 literal_endpoint = 0;
2796 /* range begins (ignore - as first or last char) */
2797 else if (*s == '-' && s+1 < send && s != start) {
2799 Perl_croak(aTHX_ "Ambiguous range in transliteration operator");
2806 *d++ = (char)UTF_TO_NATIVE(0xff); /* use illegal utf8 byte--see pmtrans */
2816 literal_endpoint = 0;
2817 native_range = TRUE;
2822 /* if we get here, we're not doing a transliteration */
2824 /* skip for regexp comments /(?#comment)/ and code /(?{code})/,
2825 except for the last char, which will be done separately. */
2826 else if (*s == '(' && PL_lex_inpat && s[1] == '?') {
2828 while (s+1 < send && *s != ')')
2829 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2831 else if (s[2] == '{' /* This should match regcomp.c */
2832 || (s[2] == '?' && s[3] == '{'))
2835 char *regparse = s + (s[2] == '{' ? 3 : 4);
2838 while (count && (c = *regparse)) {
2839 if (c == '\\' && regparse[1])
2847 if (*regparse != ')')
2848 regparse--; /* Leave one char for continuation. */
2849 while (s < regparse)
2850 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2854 /* likewise skip #-initiated comments in //x patterns */
2855 else if (*s == '#' && PL_lex_inpat &&
2856 ((PMOP*)PL_lex_inpat)->op_pmflags & RXf_PMf_EXTENDED) {
2857 while (s+1 < send && *s != '\n')
2858 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
2861 /* check for embedded arrays
2862 (@foo, @::foo, @'foo, @{foo}, @$foo, @+, @-)
2864 else if (*s == '@' && s[1]) {
2865 if (isALNUM_lazy_if(s+1,UTF))
2867 if (strchr(":'{$", s[1]))
2869 if (!PL_lex_inpat && (s[1] == '+' || s[1] == '-'))
2870 break; /* in regexp, neither @+ nor @- are interpolated */
2873 /* check for embedded scalars. only stop if we're sure it's a
2876 else if (*s == '$') {
2877 if (!PL_lex_inpat) /* not a regexp, so $ must be var */
2879 if (s + 1 < send && !strchr("()| \r\n\t", s[1])) {
2881 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
2882 "Possible unintended interpolation of $\\ in regex");
2884 break; /* in regexp, $ might be tail anchor */
2888 /* End of else if chain - OP_TRANS rejoin rest */
2891 if (*s == '\\' && s+1 < send) {
2892 char* e; /* Can be used for ending '}', etc. */
2896 /* warn on \1 - \9 in substitution replacements, but note that \11
2897 * is an octal; and \19 is \1 followed by '9' */
2898 if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat &&
2899 isDIGIT(*s) && *s != '0' && !isDIGIT(s[1]))
2901 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "\\%c better written as $%c", *s, *s);
2906 /* string-change backslash escapes */
2907 if (PL_lex_inwhat != OP_TRANS && *s && strchr("lLuUEQ", *s)) {
2911 /* In a pattern, process \N, but skip any other backslash escapes.
2912 * This is because we don't want to translate an escape sequence
2913 * into a meta symbol and have the regex compiler use the meta
2914 * symbol meaning, e.g. \x{2E} would be confused with a dot. But
2915 * in spite of this, we do have to process \N here while the proper
2916 * charnames handler is in scope. See bugs #56444 and #62056.
2917 * There is a complication because \N in a pattern may also stand
2918 * for 'match a non-nl', and not mean a charname, in which case its
2919 * processing should be deferred to the regex compiler. To be a
2920 * charname it must be followed immediately by a '{', and not look
2921 * like \N followed by a curly quantifier, i.e., not something like
2922 * \N{3,}. regcurly returns a boolean indicating if it is a legal
2924 else if (PL_lex_inpat
2927 || regcurly(s + 1)))
2929 *d++ = NATIVE_TO_NEED(has_utf8,'\\');
2930 goto default_action;
2935 /* quoted - in transliterations */
2937 if (PL_lex_inwhat == OP_TRANS) {
2944 if ((isALPHA(*s) || isDIGIT(*s)))
2945 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
2946 "Unrecognized escape \\%c passed through",
2948 /* default action is to copy the quoted character */
2949 goto default_action;
2952 /* eg. \132 indicates the octal constant 0132 */
2953 case '0': case '1': case '2': case '3':
2954 case '4': case '5': case '6': case '7':
2958 uv = NATIVE_TO_UNI(grok_oct(s, &len, &flags, NULL));
2961 goto NUM_ESCAPE_INSERT;
2963 /* eg. \o{24} indicates the octal constant \024 */
2969 bool valid = grok_bslash_o(s, &uv, &len, &error, 1);
2975 goto NUM_ESCAPE_INSERT;
2978 /* eg. \x24 indicates the hex constant 0x24 */
2982 char* const e = strchr(s, '}');
2983 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES |
2984 PERL_SCAN_DISALLOW_PREFIX;
2989 yyerror("Missing right brace on \\x{}");
2993 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
2999 I32 flags = PERL_SCAN_DISALLOW_PREFIX;
3000 uv = NATIVE_TO_UNI(grok_hex(s, &len, &flags, NULL));
3006 /* Insert oct or hex escaped character. There will always be
3007 * enough room in sv since such escapes will be longer than any
3008 * UTF-8 sequence they can end up as, except if they force us
3009 * to recode the rest of the string into utf8 */
3011 /* Here uv is the ordinal of the next character being added in
3012 * unicode (converted from native). */
3013 if (!UNI_IS_INVARIANT(uv)) {
3014 if (!has_utf8 && uv > 255) {
3015 /* Might need to recode whatever we have accumulated so
3016 * far if it contains any chars variant in utf8 or
3019 SvCUR_set(sv, d - SvPVX_const(sv));
3022 /* See Note on sizing above. */
3023 sv_utf8_upgrade_flags_grow(sv,
3024 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3025 UNISKIP(uv) + (STRLEN)(send - s) + 1);
3026 d = SvPVX(sv) + SvCUR(sv);
3031 d = (char*)uvuni_to_utf8((U8*)d, uv);
3032 if (PL_lex_inwhat == OP_TRANS &&
3033 PL_sublex_info.sub_op) {
3034 PL_sublex_info.sub_op->op_private |=
3035 (PL_lex_repl ? OPpTRANS_FROM_UTF
3039 if (uv > 255 && !dorange)
3040 native_range = FALSE;
3053 /* In a non-pattern \N must be a named character, like \N{LATIN
3054 * SMALL LETTER A} or \N{U+0041}. For patterns, it also can
3055 * mean to match a non-newline. For non-patterns, named
3056 * characters are converted to their string equivalents. In
3057 * patterns, named characters are not converted to their
3058 * ultimate forms for the same reasons that other escapes
3059 * aren't. Instead, they are converted to the \N{U+...} form
3060 * to get the value from the charnames that is in effect right
3061 * now, while preserving the fact that it was a named character
3062 * so that the regex compiler knows this */
3064 /* This section of code doesn't generally use the
3065 * NATIVE_TO_NEED() macro to transform the input. I (khw) did
3066 * a close examination of this macro and determined it is a
3067 * no-op except on utfebcdic variant characters. Every
3068 * character generated by this that would normally need to be
3069 * enclosed by this macro is invariant, so the macro is not
3070 * needed, and would complicate use of copy(). XXX There are
3071 * other parts of this file where the macro is used
3072 * inconsistently, but are saved by it being a no-op */
3074 /* The structure of this section of code (besides checking for
3075 * errors and upgrading to utf8) is:
3076 * Further disambiguate between the two meanings of \N, and if
3077 * not a charname, go process it elsewhere
3078 * If of form \N{U+...}, pass it through if a pattern;
3079 * otherwise convert to utf8
3080 * Otherwise must be \N{NAME}: convert to \N{U+c1.c2...} if a
3081 * pattern; otherwise convert to utf8 */
3083 /* Here, s points to the 'N'; the test below is guaranteed to
3084 * succeed if we are being called on a pattern as we already
3085 * know from a test above that the next character is a '{'.
3086 * On a non-pattern \N must mean 'named sequence, which
3087 * requires braces */
3090 yyerror("Missing braces on \\N{}");
3095 /* If there is no matching '}', it is an error. */
3096 if (! (e = strchr(s, '}'))) {
3097 if (! PL_lex_inpat) {
3098 yyerror("Missing right brace on \\N{}");
3100 yyerror("Missing right brace on \\N{} or unescaped left brace after \\N.");
3105 /* Here it looks like a named character */
3109 /* XXX This block is temporary code. \N{} implies that the
3110 * pattern is to have Unicode semantics, and therefore
3111 * currently has to be encoded in utf8. By putting it in
3112 * utf8 now, we save a whole pass in the regular expression
3113 * compiler. Once that code is changed so Unicode
3114 * semantics doesn't necessarily have to be in utf8, this
3115 * block should be removed. However, the code that parses
3116 * the output of this would have to be changed to not
3117 * necessarily expect utf8 */
3119 SvCUR_set(sv, d - SvPVX_const(sv));
3122 /* See Note on sizing above. */
3123 sv_utf8_upgrade_flags_grow(sv,
3124 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3125 /* 5 = '\N{' + cur char + NUL */
3126 (STRLEN)(send - s) + 5);
3127 d = SvPVX(sv) + SvCUR(sv);
3132 if (*s == 'U' && s[1] == '+') { /* \N{U+...} */
3133 I32 flags = PERL_SCAN_ALLOW_UNDERSCORES
3134 | PERL_SCAN_DISALLOW_PREFIX;
3137 /* For \N{U+...}, the '...' is a unicode value even on
3138 * EBCDIC machines */
3139 s += 2; /* Skip to next char after the 'U+' */
3141 uv = grok_hex(s, &len, &flags, NULL);
3142 if (len == 0 || len != (STRLEN)(e - s)) {
3143 yyerror("Invalid hexadecimal number in \\N{U+...}");
3150 /* On non-EBCDIC platforms, pass through to the regex
3151 * compiler unchanged. The reason we evaluated the
3152 * number above is to make sure there wasn't a syntax
3153 * error. But on EBCDIC we convert to native so
3154 * downstream code can continue to assume it's native
3156 s -= 5; /* Include the '\N{U+' */
3158 d += my_snprintf(d, e - s + 1 + 1, /* includes the }
3161 (unsigned int) UNI_TO_NATIVE(uv));
3163 Copy(s, d, e - s + 1, char); /* 1 = include the } */
3167 else { /* Not a pattern: convert the hex to string */
3169 /* If destination is not in utf8, unconditionally
3170 * recode it to be so. This is because \N{} implies
3171 * Unicode semantics, and scalars have to be in utf8
3172 * to guarantee those semantics */
3174 SvCUR_set(sv, d - SvPVX_const(sv));
3177 /* See Note on sizing above. */
3178 sv_utf8_upgrade_flags_grow(
3180 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3181 UNISKIP(uv) + (STRLEN)(send - e) + 1);
3182 d = SvPVX(sv) + SvCUR(sv);
3186 /* Add the string to the output */
3187 if (UNI_IS_INVARIANT(uv)) {
3190 else d = (char*)uvuni_to_utf8((U8*)d, uv);
3193 else { /* Here is \N{NAME} but not \N{U+...}. */
3195 SV *res; /* result from charnames */
3196 const char *str; /* the string in 'res' */
3197 STRLEN len; /* its length */
3199 /* Get the value for NAME */
3200 res = newSVpvn(s, e - s);
3201 res = new_constant( NULL, 0, "charnames",
3202 /* includes all of: \N{...} */
3203 res, NULL, s - 3, e - s + 4 );
3205 /* Most likely res will be in utf8 already since the
3206 * standard charnames uses pack U, but a custom translator
3207 * can leave it otherwise, so make sure. XXX This can be
3208 * revisited to not have charnames use utf8 for characters
3209 * that don't need it when regexes don't have to be in utf8
3210 * for Unicode semantics. If doing so, remember EBCDIC */
3211 sv_utf8_upgrade(res);
3212 str = SvPV_const(res, len);
3214 /* Don't accept malformed input */
3215 if (! is_utf8_string((U8 *) str, len)) {
3216 yyerror("Malformed UTF-8 returned by \\N");
3218 else if (PL_lex_inpat) {
3220 if (! len) { /* The name resolved to an empty string */
3221 Copy("\\N{}", d, 4, char);
3225 /* In order to not lose information for the regex
3226 * compiler, pass the result in the specially made
3227 * syntax: \N{U+c1.c2.c3...}, where c1 etc. are
3228 * the code points in hex of each character
3229 * returned by charnames */
3231 const char *str_end = str + len;
3232 STRLEN char_length; /* cur char's byte length */
3233 STRLEN output_length; /* and the number of bytes
3234 after this is translated
3236 const STRLEN off = d - SvPVX_const(sv);
3238 /* 2 hex per byte; 2 chars for '\N'; 2 chars for
3239 * max('U+', '.'); and 1 for NUL */
3240 char hex_string[2 * UTF8_MAXBYTES + 5];
3242 /* Get the first character of the result. */
3243 U32 uv = utf8n_to_uvuni((U8 *) str,
3248 /* The call to is_utf8_string() above hopefully
3249 * guarantees that there won't be an error. But
3250 * it's easy here to make sure. The function just
3251 * above warns and returns 0 if invalid utf8, but
3252 * it can also return 0 if the input is validly a
3253 * NUL. Disambiguate */
3254 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3255 uv = UNICODE_REPLACEMENT;
3258 /* Convert first code point to hex, including the
3259 * boiler plate before it. For all these, we
3260 * convert to native format so that downstream code
3261 * can continue to assume the input is native */
3263 my_snprintf(hex_string, sizeof(hex_string),
3265 (unsigned int) UNI_TO_NATIVE(uv));
3267 /* Make sure there is enough space to hold it */
3268 d = off + SvGROW(sv, off
3270 + (STRLEN)(send - e)
3271 + 2); /* '}' + NUL */
3273 Copy(hex_string, d, output_length, char);
3276 /* For each subsequent character, append dot and
3277 * its ordinal in hex */
3278 while ((str += char_length) < str_end) {
3279 const STRLEN off = d - SvPVX_const(sv);
3280 U32 uv = utf8n_to_uvuni((U8 *) str,
3284 if (uv == 0 && NATIVE_TO_ASCII(*str) != '\0') {
3285 uv = UNICODE_REPLACEMENT;
3289 my_snprintf(hex_string, sizeof(hex_string),
3291 (unsigned int) UNI_TO_NATIVE(uv));
3293 d = off + SvGROW(sv, off
3295 + (STRLEN)(send - e)
3296 + 2); /* '}' + NUL */
3297 Copy(hex_string, d, output_length, char);
3301 *d++ = '}'; /* Done. Add the trailing brace */
3304 else { /* Here, not in a pattern. Convert the name to a
3307 /* If destination is not in utf8, unconditionally
3308 * recode it to be so. This is because \N{} implies
3309 * Unicode semantics, and scalars have to be in utf8
3310 * to guarantee those semantics */
3312 SvCUR_set(sv, d - SvPVX_const(sv));
3315 /* See Note on sizing above. */
3316 sv_utf8_upgrade_flags_grow(sv,
3317 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3318 len + (STRLEN)(send - s) + 1);
3319 d = SvPVX(sv) + SvCUR(sv);
3321 } else if (len > (STRLEN)(e - s + 4)) { /* I _guess_ 4 is \N{} --jhi */
3323 /* See Note on sizing above. (NOTE: SvCUR() is not
3324 * set correctly here). */
3325 const STRLEN off = d - SvPVX_const(sv);
3326 d = off + SvGROW(sv, off + len + (STRLEN)(send - s) + 1);
3328 Copy(str, d, len, char);
3333 /* Deprecate non-approved name syntax */
3334 if (ckWARN_d(WARN_DEPRECATED)) {
3335 bool problematic = FALSE;
3338 /* For non-ut8 input, look to see that the first
3339 * character is an alpha, then loop through the rest
3340 * checking that each is a continuation */
3342 if (! isALPHAU(*i)) problematic = TRUE;
3343 else for (i = s + 1; i < e; i++) {
3344 if (isCHARNAME_CONT(*i)) continue;
3350 /* Similarly for utf8. For invariants can check
3351 * directly. We accept anything above the latin1
3352 * range because it is immaterial to Perl if it is
3353 * correct or not, and is expensive to check. But
3354 * it is fairly easy in the latin1 range to convert
3355 * the variants into a single character and check
3357 if (UTF8_IS_INVARIANT(*i)) {
3358 if (! isALPHAU(*i)) problematic = TRUE;
3359 } else if (UTF8_IS_DOWNGRADEABLE_START(*i)) {
3360 if (! isALPHAU(UNI_TO_NATIVE(TWO_BYTE_UTF8_TO_UNI(*i,
3366 if (! problematic) for (i = s + UTF8SKIP(s);
3370 if (UTF8_IS_INVARIANT(*i)) {
3371 if (isCHARNAME_CONT(*i)) continue;
3372 } else if (! UTF8_IS_DOWNGRADEABLE_START(*i)) {
3374 } else if (isCHARNAME_CONT(
3376 TWO_BYTE_UTF8_TO_UNI(*i, *(i+1)))))
3385 /* The e-i passed to the final %.*s makes sure that
3386 * should the trailing NUL be missing that this
3387 * print won't run off the end of the string */
3388 Perl_warner(aTHX_ packWARN(WARN_DEPRECATED),
3389 "Deprecated character in \\N{...}; marked by <-- HERE in \\N{%.*s<-- HERE %.*s",
3390 (int)(i - s + 1), s, (int)(e - i), i + 1);
3393 } /* End \N{NAME} */
3396 native_range = FALSE; /* \N{} is defined to be Unicode */
3398 s = e + 1; /* Point to just after the '}' */
3401 /* \c is a control character */
3405 *d++ = grok_bslash_c(*s++, has_utf8, 1);
3408 yyerror("Missing control char name in \\c");
3412 /* printf-style backslashes, formfeeds, newlines, etc */
3414 *d++ = NATIVE_TO_NEED(has_utf8,'\b');
3417 *d++ = NATIVE_TO_NEED(has_utf8,'\n');
3420 *d++ = NATIVE_TO_NEED(has_utf8,'\r');
3423 *d++ = NATIVE_TO_NEED(has_utf8,'\f');
3426 *d++ = NATIVE_TO_NEED(has_utf8,'\t');
3429 *d++ = ASCII_TO_NEED(has_utf8,'\033');
3432 *d++ = ASCII_TO_NEED(has_utf8,'\007');
3438 } /* end if (backslash) */
3445 /* If we started with encoded form, or already know we want it,
3446 then encode the next character */
3447 if (! NATIVE_IS_INVARIANT((U8)(*s)) && (this_utf8 || has_utf8)) {
3451 /* One might think that it is wasted effort in the case of the
3452 * source being utf8 (this_utf8 == TRUE) to take the next character
3453 * in the source, convert it to an unsigned value, and then convert
3454 * it back again. But the source has not been validated here. The
3455 * routine that does the conversion checks for errors like
3458 const UV nextuv = (this_utf8) ? utf8n_to_uvchr((U8*)s, send - s, &len, 0) : (UV) ((U8) *s);
3459 const STRLEN need = UNISKIP(NATIVE_TO_UNI(nextuv));
3461 SvCUR_set(sv, d - SvPVX_const(sv));
3464 /* See Note on sizing above. */
3465 sv_utf8_upgrade_flags_grow(sv,
3466 SV_GMAGIC|SV_FORCE_UTF8_UPGRADE,
3467 need + (STRLEN)(send - s) + 1);
3468 d = SvPVX(sv) + SvCUR(sv);
3470 } else if (need > len) {
3471 /* encoded value larger than old, may need extra space (NOTE:
3472 * SvCUR() is not set correctly here). See Note on sizing
3474 const STRLEN off = d - SvPVX_const(sv);
3475 d = SvGROW(sv, off + need + (STRLEN)(send - s) + 1) + off;
3479 d = (char*)uvchr_to_utf8((U8*)d, nextuv);
3481 if (uv > 255 && !dorange)
3482 native_range = FALSE;
3486 *d++ = NATIVE_TO_NEED(has_utf8,*s++);
3488 } /* while loop to process each character */
3490 /* terminate the string and set up the sv */
3492 SvCUR_set(sv, d - SvPVX_const(sv));
3493 if (SvCUR(sv) >= SvLEN(sv))
3494 Perl_croak(aTHX_ "panic: constant overflowed allocated space");
3497 if (PL_encoding && !has_utf8) {
3498 sv_recode_to_utf8(sv, PL_encoding);
3504 if (PL_lex_inwhat == OP_TRANS && PL_sublex_info.sub_op) {
3505 PL_sublex_info.sub_op->op_private |=
3506 (PL_lex_repl ? OPpTRANS_FROM_UTF : OPpTRANS_TO_UTF);
3510 /* shrink the sv if we allocated more than we used */
3511 if (SvCUR(sv) + 5 < SvLEN(sv)) {
3512 SvPV_shrink_to_cur(sv);
3515 /* return the substring (via pl_yylval) only if we parsed anything */
3516 if (s > PL_bufptr) {
3517 if ( PL_hints & ( PL_lex_inpat ? HINT_NEW_RE : HINT_NEW_STRING ) ) {
3518 const char *const key = PL_lex_inpat ? "qr" : "q";
3519 const STRLEN keylen = PL_lex_inpat ? 2 : 1;
3523 if (PL_lex_inwhat == OP_TRANS) {
3526 } else if (PL_lex_inwhat == OP_SUBST && !PL_lex_inpat) {
3534 sv = S_new_constant(aTHX_ start, s - start, key, keylen, sv, NULL,
3537 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
3544 * Returns TRUE if there's more to the expression (e.g., a subscript),
3547 * It deals with "$foo[3]" and /$foo[3]/ and /$foo[0123456789$]+/
3549 * ->[ and ->{ return TRUE
3550 * { and [ outside a pattern are always subscripts, so return TRUE
3551 * if we're outside a pattern and it's not { or [, then return FALSE
3552 * if we're in a pattern and the first char is a {
3553 * {4,5} (any digits around the comma) returns FALSE
3554 * if we're in a pattern and the first char is a [
3556 * [SOMETHING] has a funky algorithm to decide whether it's a
3557 * character class or not. It has to deal with things like
3558 * /$foo[-3]/ and /$foo[$bar]/ as well as /$foo[$\d]+/
3559 * anything else returns TRUE
3562 /* This is the one truly awful dwimmer necessary to conflate C and sed. */
3565 S_intuit_more(pTHX_ register char *s)
3569 PERL_ARGS_ASSERT_INTUIT_MORE;
3571 if (PL_lex_brackets)
3573 if (*s == '-' && s[1] == '>' && (s[2] == '[' || s[2] == '{'))
3575 if (*s != '{' && *s != '[')
3580 /* In a pattern, so maybe we have {n,m}. */
3588 /* On the other hand, maybe we have a character class */
3591 if (*s == ']' || *s == '^')
3594 /* this is terrifying, and it works */
3595 int weight = 2; /* let's weigh the evidence */
3597 unsigned char un_char = 255, last_un_char;
3598 const char * const send = strchr(s,']');
3599 char tmpbuf[sizeof PL_tokenbuf * 4];
3601 if (!send) /* has to be an expression */
3604 Zero(seen,256,char);
3607 else if (isDIGIT(*s)) {
3609 if (isDIGIT(s[1]) && s[2] == ']')
3615 for (; s < send; s++) {
3616 last_un_char = un_char;
3617 un_char = (unsigned char)*s;
3622 weight -= seen[un_char] * 10;
3623 if (isALNUM_lazy_if(s+1,UTF)) {
3625 scan_ident(s, send, tmpbuf, sizeof tmpbuf, FALSE);
3626 len = (int)strlen(tmpbuf);
3627 if (len > 1 && gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PV))
3632 else if (*s == '$' && s[1] &&
3633 strchr("[#!%*<>()-=",s[1])) {
3634 if (/*{*/ strchr("])} =",s[2]))
3643 if (strchr("wds]",s[1]))
3645 else if (seen[(U8)'\''] || seen[(U8)'"'])
3647 else if (strchr("rnftbxcav",s[1]))
3649 else if (isDIGIT(s[1])) {
3651 while (s[1] && isDIGIT(s[1]))
3661 if (strchr("aA01! ",last_un_char))
3663 if (strchr("zZ79~",s[1]))
3665 if (last_un_char == 255 && (isDIGIT(s[1]) || s[1] == '$'))
3666 weight -= 5; /* cope with negative subscript */
3669 if (!isALNUM(last_un_char)
3670 && !(last_un_char == '$' || last_un_char == '@'
3671 || last_un_char == '&')
3672 && isALPHA(*s) && s[1] && isALPHA(s[1])) {
3677 if (keyword(tmpbuf, d - tmpbuf, 0))
3680 if (un_char == last_un_char + 1)
3682 weight -= seen[un_char];
3687 if (weight >= 0) /* probably a character class */
3697 * Does all the checking to disambiguate
3699 * between foo(bar) and bar->foo. Returns 0 if not a method, otherwise
3700 * FUNCMETH (bar->foo(args)) or METHOD (bar->foo args).
3702 * First argument is the stuff after the first token, e.g. "bar".
3704 * Not a method if bar is a filehandle.
3705 * Not a method if foo is a subroutine prototyped to take a filehandle.
3706 * Not a method if it's really "Foo $bar"
3707 * Method if it's "foo $bar"
3708 * Not a method if it's really "print foo $bar"
3709 * Method if it's really "foo package::" (interpreted as package->foo)
3710 * Not a method if bar is known to be a subroutine ("sub bar; foo bar")
3711 * Not a method if bar is a filehandle or package, but is quoted with
3716 S_intuit_method(pTHX_ char *start, GV *gv, CV *cv)
3719 char *s = start + (*start == '$');
3720 char tmpbuf[sizeof PL_tokenbuf];
3727 PERL_ARGS_ASSERT_INTUIT_METHOD;
3730 if (SvTYPE(gv) == SVt_PVGV && GvIO(gv))
3734 const char *proto = SvPVX_const(cv);
3745 s = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
3746 /* start is the beginning of the possible filehandle/object,
3747 * and s is the end of it
3748 * tmpbuf is a copy of it
3751 if (*start == '$') {
3752 if (gv || PL_last_lop_op == OP_PRINT || PL_last_lop_op == OP_SAY ||
3753 isUPPER(*PL_tokenbuf))
3756 len = start - SvPVX(PL_linestr);
3760 start = SvPVX(PL_linestr) + len;
3764 return *s == '(' ? FUNCMETH : METHOD;
3766 if (!keyword(tmpbuf, len, 0)) {
3767 if (len > 2 && tmpbuf[len - 2] == ':' && tmpbuf[len - 1] == ':') {
3771 soff = s - SvPVX(PL_linestr);
3775 indirgv = gv_fetchpvn_flags(tmpbuf, len, 0, SVt_PVCV);
3776 if (indirgv && GvCVu(indirgv))
3778 /* filehandle or package name makes it a method */
3779 if (!gv || GvIO(indirgv) || gv_stashpvn(tmpbuf, len, 0)) {
3781 soff = s - SvPVX(PL_linestr);
3784 if ((PL_bufend - s) >= 2 && *s == '=' && *(s+1) == '>')
3785 return 0; /* no assumptions -- "=>" quotes bareword */
3787 start_force(PL_curforce);
3788 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0,
3789 S_newSV_maybe_utf8(aTHX_ tmpbuf, len));
3790 NEXTVAL_NEXTTOKE.opval->op_private = OPpCONST_BARE;
3792 curmad('X', newSVpvn(start,SvPVX(PL_linestr) + soff - start));
3797 PL_bufptr = SvPVX(PL_linestr) + soff; /* restart before space */
3799 return *s == '(' ? FUNCMETH : METHOD;
3805 /* Encoded script support. filter_add() effectively inserts a
3806 * 'pre-processing' function into the current source input stream.
3807 * Note that the filter function only applies to the current source file
3808 * (e.g., it will not affect files 'require'd or 'use'd by this one).
3810 * The datasv parameter (which may be NULL) can be used to pass
3811 * private data to this instance of the filter. The filter function
3812 * can recover the SV using the FILTER_DATA macro and use it to
3813 * store private buffers and state information.
3815 * The supplied datasv parameter is upgraded to a PVIO type
3816 * and the IoDIRP/IoANY field is used to store the function pointer,
3817 * and IOf_FAKE_DIRP is enabled on datasv to mark this as such.
3818 * Note that IoTOP_NAME, IoFMT_NAME, IoBOTTOM_NAME, if set for
3819 * private use must be set using malloc'd pointers.
3823 Perl_filter_add(pTHX_ filter_t funcp, SV *datasv)
3832 if (!PL_rsfp_filters)
3833 PL_rsfp_filters = newAV();
3836 SvUPGRADE(datasv, SVt_PVIO);
3837 IoANY(datasv) = FPTR2DPTR(void *, funcp); /* stash funcp into spare field */
3838 IoFLAGS(datasv) |= IOf_FAKE_DIRP;
3839 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_add func %p (%s)\n",
3840 FPTR2DPTR(void *, IoANY(datasv)),
3841 SvPV_nolen(datasv)));
3842 av_unshift(PL_rsfp_filters, 1);
3843 av_store(PL_rsfp_filters, 0, datasv) ;
3848 /* Delete most recently added instance of this filter function. */
3850 Perl_filter_del(pTHX_ filter_t funcp)
3855 PERL_ARGS_ASSERT_FILTER_DEL;
3858 DEBUG_P(PerlIO_printf(Perl_debug_log, "filter_del func %p",
3859 FPTR2DPTR(void*, funcp)));
3861 if (!PL_parser || !PL_rsfp_filters || AvFILLp(PL_rsfp_filters)<0)
3863 /* if filter is on top of stack (usual case) just pop it off */
3864 datasv = FILTER_DATA(AvFILLp(PL_rsfp_filters));
3865 if (IoANY(datasv) == FPTR2DPTR(void *, funcp)) {
3866 sv_free(av_pop(PL_rsfp_filters));
3870 /* we need to search for the correct entry and clear it */
3871 Perl_die(aTHX_ "filter_del can only delete in reverse order (currently)");
3875 /* Invoke the idxth filter function for the current rsfp. */
3876 /* maxlen 0 = read one text line */
3878 Perl_filter_read(pTHX_ int idx, SV *buf_sv, int maxlen)
3883 /* This API is bad. It should have been using unsigned int for maxlen.
3884 Not sure if we want to change the API, but if not we should sanity
3885 check the value here. */
3886 const unsigned int correct_length
3895 PERL_ARGS_ASSERT_FILTER_READ;
3897 if (!PL_parser || !PL_rsfp_filters)
3899 if (idx > AvFILLp(PL_rsfp_filters)) { /* Any more filters? */
3900 /* Provide a default input filter to make life easy. */
3901 /* Note that we append to the line. This is handy. */
3902 DEBUG_P(PerlIO_printf(Perl_debug_log,
3903 "filter_read %d: from rsfp\n", idx));
3904 if (correct_length) {
3907 const int old_len = SvCUR(buf_sv);
3909 /* ensure buf_sv is large enough */
3910 SvGROW(buf_sv, (STRLEN)(old_len + correct_length + 1)) ;
3911 if ((len = PerlIO_read(PL_rsfp, SvPVX(buf_sv) + old_len,
3912 correct_length)) <= 0) {
3913 if (PerlIO_error(PL_rsfp))
3914 return -1; /* error */
3916 return 0 ; /* end of file */
3918 SvCUR_set(buf_sv, old_len + len) ;
3919 SvPVX(buf_sv)[old_len + len] = '\0';
3922 if (sv_gets(buf_sv, PL_rsfp, SvCUR(buf_sv)) == NULL) {
3923 if (PerlIO_error(PL_rsfp))
3924 return -1; /* error */
3926 return 0 ; /* end of file */
3929 return SvCUR(buf_sv);
3931 /* Skip this filter slot if filter has been deleted */
3932 if ( (datasv = FILTER_DATA(idx)) == &PL_sv_undef) {
3933 DEBUG_P(PerlIO_printf(Perl_debug_log,
3934 "filter_read %d: skipped (filter deleted)\n",
3936 return FILTER_READ(idx+1, buf_sv, correct_length); /* recurse */
3938 /* Get function pointer hidden within datasv */
3939 funcp = DPTR2FPTR(filter_t, IoANY(datasv));
3940 DEBUG_P(PerlIO_printf(Perl_debug_log,
3941 "filter_read %d: via function %p (%s)\n",
3942 idx, (void*)datasv, SvPV_nolen_const(datasv)));
3943 /* Call function. The function is expected to */
3944 /* call "FILTER_READ(idx+1, buf_sv)" first. */
3945 /* Return: <0:error, =0:eof, >0:not eof */
3946 return (*funcp)(aTHX_ idx, buf_sv, correct_length);
3950 S_filter_gets(pTHX_ register SV *sv, STRLEN append)
3954 PERL_ARGS_ASSERT_FILTER_GETS;
3956 #ifdef PERL_CR_FILTER
3957 if (!PL_rsfp_filters) {
3958 filter_add(S_cr_textfilter,NULL);
3961 if (PL_rsfp_filters) {
3963 SvCUR_set(sv, 0); /* start with empty line */
3964 if (FILTER_READ(0, sv, 0) > 0)
3965 return ( SvPVX(sv) ) ;
3970 return (sv_gets(sv, PL_rsfp, append));
3974 S_find_in_my_stash(pTHX_ const char *pkgname, STRLEN len)
3979 PERL_ARGS_ASSERT_FIND_IN_MY_STASH;
3981 if (len == 11 && *pkgname == '_' && strEQ(pkgname, "__PACKAGE__"))
3985 (pkgname[len - 2] == ':' && pkgname[len - 1] == ':') &&
3986 (gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVHV)))
3988 return GvHV(gv); /* Foo:: */
3991 /* use constant CLASS => 'MyClass' */
3992 gv = gv_fetchpvn_flags(pkgname, len, 0, SVt_PVCV);
3993 if (gv && GvCV(gv)) {
3994 SV * const sv = cv_const_sv(GvCV(gv));
3996 pkgname = SvPV_const(sv, len);
3999 return gv_stashpvn(pkgname, len, 0);
4003 * S_readpipe_override
4004 * Check whether readpipe() is overridden, and generates the appropriate
4005 * optree, provided sublex_start() is called afterwards.
4008 S_readpipe_override(pTHX)
4011 GV *gv_readpipe = gv_fetchpvs("readpipe", GV_NOTQUAL, SVt_PVCV);
4012 pl_yylval.ival = OP_BACKTICK;
4014 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe))
4016 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readpipe", FALSE))
4017 && (gv_readpipe = *gvp) && isGV_with_GP(gv_readpipe)
4018 && GvCVu(gv_readpipe) && GvIMPORTED_CV(gv_readpipe)))
4020 PL_lex_op = (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
4021 op_append_elem(OP_LIST,
4022 newSVOP(OP_CONST, 0, &PL_sv_undef), /* value will be read later */
4023 newCVREF(0, newGVOP(OP_GV, 0, gv_readpipe))));
4030 * The intent of this yylex wrapper is to minimize the changes to the
4031 * tokener when we aren't interested in collecting madprops. It remains
4032 * to be seen how successful this strategy will be...
4039 char *s = PL_bufptr;
4041 /* make sure PL_thiswhite is initialized */
4045 /* just do what yylex would do on pending identifier; leave PL_thiswhite alone */
4046 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4047 return S_pending_ident(aTHX);
4049 /* previous token ate up our whitespace? */
4050 if (!PL_lasttoke && PL_nextwhite) {
4051 PL_thiswhite = PL_nextwhite;
4055 /* isolate the token, and figure out where it is without whitespace */
4056 PL_realtokenstart = -1;
4060 assert(PL_curforce < 0);
4062 if (!PL_thismad || PL_thismad->mad_key == '^') { /* not forced already? */
4063 if (!PL_thistoken) {
4064 if (PL_realtokenstart < 0 || !CopLINE(PL_curcop))
4065 PL_thistoken = newSVpvs("");
4067 char * const tstart = SvPVX(PL_linestr) + PL_realtokenstart;
4068 PL_thistoken = newSVpvn(tstart, s - tstart);
4071 if (PL_thismad) /* install head */
4072 CURMAD('X', PL_thistoken);
4075 /* last whitespace of a sublex? */
4076 if (optype == ')' && PL_endwhite) {
4077 CURMAD('X', PL_endwhite);
4082 /* if no whitespace and we're at EOF, bail. Otherwise fake EOF below. */
4083 if (!PL_thiswhite && !PL_endwhite && !optype) {
4084 sv_free(PL_thistoken);
4089 /* put off final whitespace till peg */
4090 if (optype == ';' && !PL_rsfp) {
4091 PL_nextwhite = PL_thiswhite;
4094 else if (PL_thisopen) {
4095 CURMAD('q', PL_thisopen);
4097 sv_free(PL_thistoken);
4101 /* Store actual token text as madprop X */
4102 CURMAD('X', PL_thistoken);
4106 /* add preceding whitespace as madprop _ */
4107 CURMAD('_', PL_thiswhite);
4111 /* add quoted material as madprop = */
4112 CURMAD('=', PL_thisstuff);
4116 /* add terminating quote as madprop Q */
4117 CURMAD('Q', PL_thisclose);
4121 /* special processing based on optype */
4125 /* opval doesn't need a TOKEN since it can already store mp */
4135 if (pl_yylval.opval)
4136 append_madprops(PL_thismad, pl_yylval.opval, 0);
4144 addmad(newMADsv('p', PL_endwhite), &PL_thismad, 0);
4153 /* remember any fake bracket that lexer is about to discard */
4154 if (PL_lex_brackets == 1 &&
4155 ((expectation)PL_lex_brackstack[0] & XFAKEBRACK))
4158 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4161 PL_thiswhite = newSVpvn(PL_bufptr, ++s - PL_bufptr);
4162 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4165 break; /* don't bother looking for trailing comment */
4174 /* attach a trailing comment to its statement instead of next token */
4178 if (PL_bufptr > PL_oldbufptr && PL_bufptr[-1] == optype) {
4180 while (s < PL_bufend && (*s == ' ' || *s == '\t'))
4182 if (*s == '\n' || *s == '#') {
4183 while (s < PL_bufend && *s != '\n')
4187 PL_thiswhite = newSVpvn(PL_bufptr, s - PL_bufptr);
4188 addmad(newMADsv('#', PL_thiswhite), &PL_thismad, 0);
4205 /* Create new token struct. Note: opvals return early above. */
4206 pl_yylval.tkval = newTOKEN(optype, pl_yylval, PL_thismad);
4213 S_tokenize_use(pTHX_ int is_use, char *s) {
4216 PERL_ARGS_ASSERT_TOKENIZE_USE;
4218 if (PL_expect != XSTATE)
4219 yyerror(Perl_form(aTHX_ "\"%s\" not allowed in expression",
4220 is_use ? "use" : "no"));
4222 if (isDIGIT(*s) || (*s == 'v' && isDIGIT(s[1]))) {
4223 s = force_version(s, TRUE);
4224 if (*s == ';' || *s == '}'
4225 || (s = SKIPSPACE1(s), (*s == ';' || *s == '}'))) {
4226 start_force(PL_curforce);
4227 NEXTVAL_NEXTTOKE.opval = NULL;
4230 else if (*s == 'v') {
4231 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4232 s = force_version(s, FALSE);
4236 s = force_word(s,WORD,FALSE,TRUE,FALSE);
4237 s = force_version(s, FALSE);
4239 pl_yylval.ival = is_use;
4243 static const char* const exp_name[] =
4244 { "OPERATOR", "TERM", "REF", "STATE", "BLOCK", "ATTRBLOCK",
4245 "ATTRTERM", "TERMBLOCK", "TERMORDORDOR"
4249 #define word_takes_any_delimeter(p,l) S_word_takes_any_delimeter(p,l)
4251 S_word_takes_any_delimeter(char *p, STRLEN len)
4253 return (len == 1 && strchr("msyq", p[0])) ||
4255 (p[0] == 't' && p[1] == 'r') ||
4256 (p[0] == 'q' && strchr("qwxr", p[1]))));
4262 Works out what to call the token just pulled out of the input
4263 stream. The yacc parser takes care of taking the ops we return and
4264 stitching them into a tree.
4270 if read an identifier
4271 if we're in a my declaration
4272 croak if they tried to say my($foo::bar)
4273 build the ops for a my() declaration
4274 if it's an access to a my() variable
4275 are we in a sort block?
4276 croak if my($a); $a <=> $b
4277 build ops for access to a my() variable
4278 if in a dq string, and they've said @foo and we can't find @foo
4280 build ops for a bareword
4281 if we already built the token before, use it.
4286 #pragma segment Perl_yylex
4292 register char *s = PL_bufptr;
4298 /* orig_keyword, gvp, and gv are initialized here because
4299 * jump to the label just_a_word_zero can bypass their
4300 * initialization later. */
4301 I32 orig_keyword = 0;
4306 SV* tmp = newSVpvs("");
4307 PerlIO_printf(Perl_debug_log, "### %"IVdf":LEX_%s/X%s %s\n",
4308 (IV)CopLINE(PL_curcop),
4309 lex_state_names[PL_lex_state],
4310 exp_name[PL_expect],
4311 pv_display(tmp, s, strlen(s), 0, 60));
4314 /* check if there's an identifier for us to look at */
4315 if (PL_lex_state != LEX_KNOWNEXT && PL_pending_ident)
4316 return REPORT(S_pending_ident(aTHX));
4318 /* no identifier pending identification */
4320 switch (PL_lex_state) {
4322 case LEX_NORMAL: /* Some compilers will produce faster */
4323 case LEX_INTERPNORMAL: /* code if we comment these out. */
4327 /* when we've already built the next token, just pull it out of the queue */
4331 pl_yylval = PL_nexttoke[PL_lasttoke].next_val;
4333 PL_thismad = PL_nexttoke[PL_lasttoke].next_mad;
4334 PL_nexttoke[PL_lasttoke].next_mad = 0;
4335 if (PL_thismad && PL_thismad->mad_key == '_') {
4336 PL_thiswhite = MUTABLE_SV(PL_thismad->mad_val);
4337 PL_thismad->mad_val = 0;
4338 mad_free(PL_thismad);
4343 PL_lex_state = PL_lex_defer;
4344 PL_expect = PL_lex_expect;
4345 PL_lex_defer = LEX_NORMAL;
4346 if (!PL_nexttoke[PL_lasttoke].next_type)
4351 pl_yylval = PL_nextval[PL_nexttoke];
4353 PL_lex_state = PL_lex_defer;
4354 PL_expect = PL_lex_expect;
4355 PL_lex_defer = LEX_NORMAL;
4361 next_type = PL_nexttoke[PL_lasttoke].next_type;
4363 next_type = PL_nexttype[PL_nexttoke];
4365 if (next_type & (7<<24)) {
4366 if (next_type & (1<<24)) {
4367 if (PL_lex_brackets > 100)
4368 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
4369 PL_lex_brackstack[PL_lex_brackets++] =
4370 (char) ((next_type >> 16) & 0xff);
4372 if (next_type & (2<<24))
4373 PL_lex_allbrackets++;
4374 if (next_type & (4<<24))
4375 PL_lex_allbrackets--;
4376 next_type &= 0xffff;
4379 /* FIXME - can these be merged? */
4382 return REPORT(next_type);
4386 /* interpolated case modifiers like \L \U, including \Q and \E.
4387 when we get here, PL_bufptr is at the \
4389 case LEX_INTERPCASEMOD:
4391 if (PL_bufptr != PL_bufend && *PL_bufptr != '\\')
4392 Perl_croak(aTHX_ "panic: INTERPCASEMOD");
4394 /* handle \E or end of string */
4395 if (PL_bufptr == PL_bufend || PL_bufptr[1] == 'E') {
4397 if (PL_lex_casemods) {
4398 const char oldmod = PL_lex_casestack[--PL_lex_casemods];
4399 PL_lex_casestack[PL_lex_casemods] = '\0';
4401 if (PL_bufptr != PL_bufend
4402 && (oldmod == 'L' || oldmod == 'U' || oldmod == 'Q')) {
4404 PL_lex_state = LEX_INTERPCONCAT;
4407 PL_thistoken = newSVpvs("\\E");
4410 PL_lex_allbrackets--;
4414 while (PL_bufptr != PL_bufend &&
4415 PL_bufptr[0] == '\\' && PL_bufptr[1] == 'E') {
4417 PL_thiswhite = newSVpvs("");
4418 sv_catpvn(PL_thiswhite, PL_bufptr, 2);
4422 if (PL_bufptr != PL_bufend)
4425 PL_lex_state = LEX_INTERPCONCAT;
4429 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4430 "### Saw case modifier\n"); });
4432 if (s[1] == '\\' && s[2] == 'E') {
4435 PL_thiswhite = newSVpvs("");
4436 sv_catpvn(PL_thiswhite, PL_bufptr, 4);
4439 PL_lex_state = LEX_INTERPCONCAT;
4444 if (!PL_madskills) /* when just compiling don't need correct */
4445 if (strnEQ(s, "L\\u", 3) || strnEQ(s, "U\\l", 3))
4446 tmp = *s, *s = s[2], s[2] = (char)tmp; /* misordered... */
4447 if ((*s == 'L' || *s == 'U') &&
4448 (strchr(PL_lex_casestack, 'L') || strchr(PL_lex_casestack, 'U'))) {
4449 PL_lex_casestack[--PL_lex_casemods] = '\0';
4450 PL_lex_allbrackets--;
4453 if (PL_lex_casemods > 10)
4454 Renew(PL_lex_casestack, PL_lex_casemods + 2, char);
4455 PL_lex_casestack[PL_lex_casemods++] = *s;
4456 PL_lex_casestack[PL_lex_casemods] = '\0';
4457 PL_lex_state = LEX_INTERPCONCAT;
4458 start_force(PL_curforce);
4459 NEXTVAL_NEXTTOKE.ival = 0;
4460 force_next((2<<24)|'(');
4461 start_force(PL_curforce);
4463 NEXTVAL_NEXTTOKE.ival = OP_LCFIRST;
4465 NEXTVAL_NEXTTOKE.ival = OP_UCFIRST;
4467 NEXTVAL_NEXTTOKE.ival = OP_LC;
4469 NEXTVAL_NEXTTOKE.ival = OP_UC;
4471 NEXTVAL_NEXTTOKE.ival = OP_QUOTEMETA;
4473 Perl_croak(aTHX_ "panic: yylex");
4475 SV* const tmpsv = newSVpvs("\\ ");
4476 /* replace the space with the character we want to escape
4478 SvPVX(tmpsv)[1] = *s;
4484 if (PL_lex_starts) {
4490 sv_free(PL_thistoken);
4491 PL_thistoken = newSVpvs("");
4494 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4495 if (PL_lex_casemods == 1 && PL_lex_inpat)
4504 case LEX_INTERPPUSH:
4505 return REPORT(sublex_push());
4507 case LEX_INTERPSTART:
4508 if (PL_bufptr == PL_bufend)
4509 return REPORT(sublex_done());
4510 DEBUG_T({ PerlIO_printf(Perl_debug_log,
4511 "### Interpolated variable\n"); });
4513 PL_lex_dojoin = (*PL_bufptr == '@');
4514 PL_lex_state = LEX_INTERPNORMAL;
4515 if (PL_lex_dojoin) {
4516 start_force(PL_curforce);
4517 NEXTVAL_NEXTTOKE.ival = 0;
4519 start_force(PL_curforce);
4520 force_ident("\"", '$');
4521 start_force(PL_curforce);
4522 NEXTVAL_NEXTTOKE.ival = 0;
4524 start_force(PL_curforce);
4525 NEXTVAL_NEXTTOKE.ival = 0;
4526 force_next((2<<24)|'(');
4527 start_force(PL_curforce);
4528 NEXTVAL_NEXTTOKE.ival = OP_JOIN; /* emulate join($", ...) */
4531 if (PL_lex_starts++) {
4536 sv_free(PL_thistoken);
4537 PL_thistoken = newSVpvs("");
4540 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4541 if (!PL_lex_casemods && PL_lex_inpat)
4548 case LEX_INTERPENDMAYBE:
4549 if (intuit_more(PL_bufptr)) {
4550 PL_lex_state = LEX_INTERPNORMAL; /* false alarm, more expr */
4556 if (PL_lex_dojoin) {
4557 PL_lex_dojoin = FALSE;
4558 PL_lex_state = LEX_INTERPCONCAT;
4562 sv_free(PL_thistoken);
4563 PL_thistoken = newSVpvs("");
4566 PL_lex_allbrackets--;
4569 if (PL_lex_inwhat == OP_SUBST && PL_linestr == PL_lex_repl
4570 && SvEVALED(PL_lex_repl))
4572 if (PL_bufptr != PL_bufend)
4573 Perl_croak(aTHX_ "Bad evalled substitution pattern");
4577 case LEX_INTERPCONCAT:
4579 if (PL_lex_brackets)
4580 Perl_croak(aTHX_ "panic: INTERPCONCAT");
4582 if (PL_bufptr == PL_bufend)
4583 return REPORT(sublex_done());
4585 if (SvIVX(PL_linestr) == '\'') {
4586 SV *sv = newSVsv(PL_linestr);
4589 else if ( PL_hints & HINT_NEW_RE )
4590 sv = new_constant(NULL, 0, "qr", sv, sv, "q", 1);
4591 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
4595 s = scan_const(PL_bufptr);
4597 PL_lex_state = LEX_INTERPCASEMOD;
4599 PL_lex_state = LEX_INTERPSTART;
4602 if (s != PL_bufptr) {
4603 start_force(PL_curforce);
4605 curmad('X', newSVpvn(PL_bufptr,s-PL_bufptr));
4607 NEXTVAL_NEXTTOKE = pl_yylval;
4610 if (PL_lex_starts++) {
4614 sv_free(PL_thistoken);
4615 PL_thistoken = newSVpvs("");
4618 /* commas only at base level: /$a\Ub$c/ => ($a,uc(b.$c)) */
4619 if (!PL_lex_casemods && PL_lex_inpat)
4632 PL_lex_state = LEX_NORMAL;
4633 s = scan_formline(PL_bufptr);
4634 if (!PL_lex_formbrack)
4640 PL_oldoldbufptr = PL_oldbufptr;
4646 sv_free(PL_thistoken);
4649 PL_realtokenstart = s - SvPVX(PL_linestr); /* assume but undo on ws */
4653 if (isIDFIRST_lazy_if(s,UTF))
4656 unsigned char c = *s;
4657 len = UTF ? Perl_utf8_length(aTHX_ (U8 *) PL_linestart, (U8 *) s) : (STRLEN) (s - PL_linestart);
4658 if (len > UNRECOGNIZED_PRECEDE_COUNT) {
4659 d = UTF ? (char *) Perl_utf8_hop(aTHX_ (U8 *) s, -UNRECOGNIZED_PRECEDE_COUNT) : s - UNRECOGNIZED_PRECEDE_COUNT;
4664 Perl_croak(aTHX_ "Unrecognized character \\x%02X; marked by <-- HERE after %s<-- HERE near column %d", c, d, (int) len + 1);
4668 goto fake_eof; /* emulate EOF on ^D or ^Z */
4677 if (PL_lex_brackets &&
4678 PL_lex_brackstack[PL_lex_brackets-1] != XFAKEEOF) {
4679 yyerror((const char *)
4681 ? "Format not terminated"
4682 : "Missing right curly or square bracket"));
4684 DEBUG_T( { PerlIO_printf(Perl_debug_log,
4685 "### Tokener got EOF\n");
4689 if (s++ < PL_bufend)
4690 goto retry; /* ignore stray nulls */
4693 if (!PL_in_eval && !PL_preambled) {
4694 PL_preambled = TRUE;
4700 /* Generate a string of Perl code to load the debugger.
4701 * If PERL5DB is set, it will return the contents of that,
4702 * otherwise a compile-time require of perl5db.pl. */
4704 const char * const pdb = PerlEnv_getenv("PERL5DB");
4707 sv_setpv(PL_linestr, pdb);
4708 sv_catpvs(PL_linestr,";");
4710 SETERRNO(0,SS_NORMAL);
4711 sv_setpvs(PL_linestr, "BEGIN { require 'perl5db.pl' };");
4714 sv_setpvs(PL_linestr,"");
4715 if (PL_preambleav) {
4716 SV **svp = AvARRAY(PL_preambleav);
4717 SV **const end = svp + AvFILLp(PL_preambleav);
4719 sv_catsv(PL_linestr, *svp);
4721 sv_catpvs(PL_linestr, ";");
4723 sv_free(MUTABLE_SV(PL_preambleav));
4724 PL_preambleav = NULL;
4727 sv_catpvs(PL_linestr,
4728 "use feature ':5." STRINGIFY(PERL_VERSION) "';");
4729 if (PL_minus_n || PL_minus_p) {
4730 sv_catpvs(PL_linestr, "LINE: while (<>) {"/*}*/);
4732 sv_catpvs(PL_linestr,"chomp;");
4735 if ((*PL_splitstr == '/' || *PL_splitstr == '\''
4736 || *PL_splitstr == '"')
4737 && strchr(PL_splitstr + 1, *PL_splitstr))
4738 Perl_sv_catpvf(aTHX_ PL_linestr, "our @F=split(%s);", PL_splitstr);
4740 /* "q\0${splitstr}\0" is legal perl. Yes, even NUL
4741 bytes can be used as quoting characters. :-) */
4742 const char *splits = PL_splitstr;
4743 sv_catpvs(PL_linestr, "our @F=split(q\0");
4746 if (*splits == '\\')
4747 sv_catpvn(PL_linestr, splits, 1);
4748 sv_catpvn(PL_linestr, splits, 1);
4749 } while (*splits++);
4750 /* This loop will embed the trailing NUL of
4751 PL_linestr as the last thing it does before
4753 sv_catpvs(PL_linestr, ");");
4757 sv_catpvs(PL_linestr,"our @F=split(' ');");
4760 sv_catpvs(PL_linestr, "\n");
4761 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4762 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4763 PL_last_lop = PL_last_uni = NULL;
4764 if ((PERLDB_LINE || PERLDB_SAVESRC) && PL_curstash != PL_debstash)
4765 update_debugger_info(PL_linestr, NULL, 0);
4770 bof = PL_rsfp ? TRUE : FALSE;
4773 fake_eof = LEX_FAKE_EOF;
4775 PL_bufptr = PL_bufend;
4776 CopLINE_inc(PL_curcop);
4777 if (!lex_next_chunk(fake_eof)) {
4778 CopLINE_dec(PL_curcop);
4780 TOKEN(';'); /* not infinite loop because rsfp is NULL now */
4782 CopLINE_dec(PL_curcop);
4785 PL_realtokenstart = -1;
4788 /* If it looks like the start of a BOM or raw UTF-16,
4789 * check if it in fact is. */
4790 if (bof && PL_rsfp &&
4795 IV offset = (IV)PerlIO_tell(PL_rsfp);
4796 bof = (offset == SvCUR(PL_linestr));
4797 #if defined(PERLIO_USING_CRLF) && defined(PERL_TEXTMODE_SCRIPTS)
4798 /* offset may include swallowed CR */
4800 bof = (offset == SvCUR(PL_linestr)+1);
4803 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4804 s = swallow_bom((U8*)s);
4807 if (PL_parser->in_pod) {
4808 /* Incest with pod. */
4811 sv_catsv(PL_thiswhite, PL_linestr);
4813 if (*s == '=' && strnEQ(s, "=cut", 4) && !isALPHA(s[4])) {
4814 sv_setpvs(PL_linestr, "");
4815 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
4816 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4817 PL_last_lop = PL_last_uni = NULL;
4818 PL_parser->in_pod = 0;
4823 } while (PL_parser->in_pod);
4824 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = PL_linestart = s;
4825 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
4826 PL_last_lop = PL_last_uni = NULL;
4827 if (CopLINE(PL_curcop) == 1) {
4828 while (s < PL_bufend && isSPACE(*s))
4830 if (*s == ':' && s[1] != ':') /* for csh execing sh scripts */
4834 PL_thiswhite = newSVpvn(PL_linestart, s - PL_linestart);
4838 if (*s == '#' && *(s+1) == '!')
4840 #ifdef ALTERNATE_SHEBANG
4842 static char const as[] = ALTERNATE_SHEBANG;
4843 if (*s == as[0] && strnEQ(s, as, sizeof(as) - 1))
4844 d = s + (sizeof(as) - 1);
4846 #endif /* ALTERNATE_SHEBANG */
4855 while (*d && !isSPACE(*d))
4859 #ifdef ARG_ZERO_IS_SCRIPT
4860 if (ipathend > ipath) {
4862 * HP-UX (at least) sets argv[0] to the script name,
4863 * which makes $^X incorrect. And Digital UNIX and Linux,
4864 * at least, set argv[0] to the basename of the Perl
4865 * interpreter. So, having found "#!", we'll set it right.
4867 SV * const x = GvSV(gv_fetchpvs("\030", GV_ADD|GV_NOTQUAL,
4869 assert(SvPOK(x) || SvGMAGICAL(x));
4870 if (sv_eq(x, CopFILESV(PL_curcop))) {
4871 sv_setpvn(x, ipath, ipathend - ipath);
4877 const char *bstart = SvPV_const(CopFILESV(PL_curcop),blen);
4878 const char * const lstart = SvPV_const(x,llen);
4880 bstart += blen - llen;
4881 if (strnEQ(bstart, lstart, llen) && bstart[-1] == '/') {
4882 sv_setpvn(x, ipath, ipathend - ipath);
4887 TAINT_NOT; /* $^X is always tainted, but that's OK */
4889 #endif /* ARG_ZERO_IS_SCRIPT */
4894 d = instr(s,"perl -");
4896 d = instr(s,"perl");
4898 /* avoid getting into infinite loops when shebang
4899 * line contains "Perl" rather than "perl" */
4901 for (d = ipathend-4; d >= ipath; --d) {
4902 if ((*d == 'p' || *d == 'P')
4903 && !ibcmp(d, "perl", 4))
4913 #ifdef ALTERNATE_SHEBANG
4915 * If the ALTERNATE_SHEBANG on this system starts with a
4916 * character that can be part of a Perl expression, then if
4917 * we see it but not "perl", we're probably looking at the
4918 * start of Perl code, not a request to hand off to some
4919 * other interpreter. Similarly, if "perl" is there, but
4920 * not in the first 'word' of the line, we assume the line
4921 * contains the start of the Perl program.
4923 if (d && *s != '#') {
4924 const char *c = ipath;
4925 while (*c && !strchr("; \t\r\n\f\v#", *c))
4928 d = NULL; /* "perl" not in first word; ignore */
4930 *s = '#'; /* Don't try to parse shebang line */
4932 #endif /* ALTERNATE_SHEBANG */
4937 !instr(s,"indir") &&
4938 instr(PL_origargv[0],"perl"))
4945 while (s < PL_bufend && isSPACE(*s))
4947 if (s < PL_bufend) {
4948 Newx(newargv,PL_origargc+3,char*);
4950 while (s < PL_bufend && !isSPACE(*s))
4953 Copy(PL_origargv+1, newargv+2, PL_origargc+1, char*);
4956 newargv = PL_origargv;
4959 PerlProc_execv(ipath, EXEC_ARGV_CAST(newargv));
4961 Perl_croak(aTHX_ "Can't exec %s", ipath);
4964 while (*d && !isSPACE(*d))
4966 while (SPACE_OR_TAB(*d))
4970 const bool switches_done = PL_doswitches;
4971 const U32 oldpdb = PL_perldb;
4972 const bool oldn = PL_minus_n;
4973 const bool oldp = PL_minus_p;
4977 bool baduni = FALSE;
4979 const char *d2 = d1 + 1;
4980 if (parse_unicode_opts((const char **)&d2)
4984 if (baduni || *d1 == 'M' || *d1 == 'm') {
4985 const char * const m = d1;
4986 while (*d1 && !isSPACE(*d1))
4988 Perl_croak(aTHX_ "Too late for \"-%.*s\" option",
4991 d1 = moreswitches(d1);
4993 if (PL_doswitches && !switches_done) {
4994 int argc = PL_origargc;
4995 char **argv = PL_origargv;
4998 } while (argc && argv[0][0] == '-' && argv[0][1]);
4999 init_argv_symbols(argc,argv);
5001 if (((PERLDB_LINE || PERLDB_SAVESRC) && !oldpdb) ||
5002 ((PL_minus_n || PL_minus_p) && !(oldn || oldp)))
5003 /* if we have already added "LINE: while (<>) {",
5004 we must not do it again */
5006 sv_setpvs(PL_linestr, "");
5007 PL_oldoldbufptr = PL_oldbufptr = s = PL_linestart = SvPVX(PL_linestr);
5008 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
5009 PL_last_lop = PL_last_uni = NULL;
5010 PL_preambled = FALSE;
5011 if (PERLDB_LINE || PERLDB_SAVESRC)
5012 (void)gv_fetchfile(PL_origfilename);
5019 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5021 PL_lex_state = LEX_FORMLINE;
5026 #ifdef PERL_STRICT_CR
5027 Perl_warn(aTHX_ "Illegal character \\%03o (carriage return)", '\r');
5029 "\t(Maybe you didn't strip carriage returns after a network transfer?)\n");
5031 case ' ': case '\t': case '\f': case 013:
5033 PL_realtokenstart = -1;
5035 PL_thiswhite = newSVpvs("");
5036 sv_catpvn(PL_thiswhite, s, 1);
5043 PL_realtokenstart = -1;
5047 if (PL_lex_state != LEX_NORMAL || (PL_in_eval && !PL_rsfp)) {
5048 if (*s == '#' && s == PL_linestart && PL_in_eval && !PL_rsfp) {
5049 /* handle eval qq[#line 1 "foo"\n ...] */
5050 CopLINE_dec(PL_curcop);
5053 if (PL_madskills && !PL_lex_formbrack && !PL_in_eval) {
5055 if (!PL_in_eval || PL_rsfp)
5060 while (d < PL_bufend && *d != '\n')
5064 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5065 Perl_croak(aTHX_ "panic: input overflow");
5068 PL_thiswhite = newSVpvn(s, d - s);
5073 if (PL_lex_formbrack && PL_lex_brackets <= PL_lex_formbrack) {
5075 PL_lex_state = LEX_FORMLINE;
5081 if (PL_madskills && CopLINE(PL_curcop) >= 1 && !PL_lex_formbrack) {
5082 if (CopLINE(PL_curcop) == 1 && s[0] == '#' && s[1] == '!') {
5085 TOKEN(PEG); /* make sure any #! line is accessible */
5090 /* if (PL_madskills && PL_lex_formbrack) { */
5092 while (d < PL_bufend && *d != '\n')
5096 else if (d > PL_bufend) /* Found by Ilya: feed random input to Perl. */
5097 Perl_croak(aTHX_ "panic: input overflow");
5098 if (PL_madskills && CopLINE(PL_curcop) >= 1) {
5100 PL_thiswhite = newSVpvs("");
5101 if (CopLINE(PL_curcop) == 1) {
5102 sv_setpvs(PL_thiswhite, "");
5105 sv_catpvn(PL_thiswhite, s, d - s);
5119 if (s[1] && isALPHA(s[1]) && !isALNUM(s[2])) {
5127 while (s < PL_bufend && SPACE_OR_TAB(*s))
5130 if (strnEQ(s,"=>",2)) {
5131 s = force_word(PL_bufptr,WORD,FALSE,FALSE,FALSE);
5132 DEBUG_T( { printbuf("### Saw unary minus before =>, forcing word %s\n", s); } );
5133 OPERATOR('-'); /* unary minus */
5135 PL_last_uni = PL_oldbufptr;
5137 case 'r': ftst = OP_FTEREAD; break;
5138 case 'w': ftst = OP_FTEWRITE; break;
5139 case 'x': ftst = OP_FTEEXEC; break;
5140 case 'o': ftst = OP_FTEOWNED; break;
5141 case 'R': ftst = OP_FTRREAD; break;
5142 case 'W': ftst = OP_FTRWRITE; break;
5143 case 'X': ftst = OP_FTREXEC; break;
5144 case 'O': ftst = OP_FTROWNED; break;
5145 case 'e': ftst = OP_FTIS; break;
5146 case 'z': ftst = OP_FTZERO; break;
5147 case 's': ftst = OP_FTSIZE; break;
5148 case 'f': ftst = OP_FTFILE; break;
5149 case 'd': ftst = OP_FTDIR; break;
5150 case 'l': ftst = OP_FTLINK; break;
5151 case 'p': ftst = OP_FTPIPE; break;
5152 case 'S': ftst = OP_FTSOCK; break;
5153 case 'u': ftst = OP_FTSUID; break;
5154 case 'g': ftst = OP_FTSGID; break;
5155 case 'k': ftst = OP_FTSVTX; break;
5156 case 'b': ftst = OP_FTBLK; break;
5157 case 'c': ftst = OP_FTCHR; break;
5158 case 't': ftst = OP_FTTTY; break;
5159 case 'T': ftst = OP_FTTEXT; break;
5160 case 'B': ftst = OP_FTBINARY; break;
5161 case 'M': case 'A': case 'C':
5162 gv_fetchpvs("\024", GV_ADD|GV_NOTQUAL, SVt_PV);
5164 case 'M': ftst = OP_FTMTIME; break;
5165 case 'A': ftst = OP_FTATIME; break;
5166 case 'C': ftst = OP_FTCTIME; break;
5174 PL_last_lop_op = (OPCODE)ftst;
5175 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5176 "### Saw file test %c\n", (int)tmp);
5181 /* Assume it was a minus followed by a one-letter named
5182 * subroutine call (or a -bareword), then. */
5183 DEBUG_T( { PerlIO_printf(Perl_debug_log,
5184 "### '-%c' looked like a file test but was not\n",
5191 const char tmp = *s++;
5194 if (PL_expect == XOPERATOR)
5199 else if (*s == '>') {
5202 if (isIDFIRST_lazy_if(s,UTF)) {
5203 s = force_word(s,METHOD,FALSE,TRUE,FALSE);
5211 if (PL_expect == XOPERATOR) {
5212 if (*s == '=' && !PL_lex_allbrackets &&
5213 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5220 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5222 OPERATOR('-'); /* unary minus */
5228 const char tmp = *s++;
5231 if (PL_expect == XOPERATOR)
5236 if (PL_expect == XOPERATOR) {
5237 if (*s == '=' && !PL_lex_allbrackets &&
5238 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5245 if (isSPACE(*s) || !isSPACE(*PL_bufptr))
5252 if (PL_expect != XOPERATOR) {
5253 s = scan_ident(s, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5254 PL_expect = XOPERATOR;
5255 force_ident(PL_tokenbuf, '*');
5263 if (*s == '=' && !PL_lex_allbrackets &&
5264 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5270 if (*s == '=' && !PL_lex_allbrackets &&
5271 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5278 if (PL_expect == XOPERATOR) {
5279 if (s[1] == '=' && !PL_lex_allbrackets &&
5280 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
5285 PL_tokenbuf[0] = '%';
5286 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
5287 sizeof PL_tokenbuf - 1, FALSE);
5288 if (!PL_tokenbuf[1]) {
5291 PL_pending_ident = '%';
5295 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5296 (s[1] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE))
5301 if (PL_lex_brackets > 100)
5302 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5303 PL_lex_brackstack[PL_lex_brackets++] = 0;
5304 PL_lex_allbrackets++;
5306 const char tmp = *s++;
5311 && (PL_expect == XOPERATOR || PL_expect == XTERMORDORDOR))
5313 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
5321 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMMA)
5328 goto just_a_word_zero_gv;
5331 switch (PL_expect) {
5337 if (!PL_in_my || PL_lex_state != LEX_NORMAL)
5339 PL_bufptr = s; /* update in case we back off */
5342 "Use of := for an empty attribute list is not allowed");
5349 PL_expect = XTERMBLOCK;
5352 stuffstart = s - SvPVX(PL_linestr) - 1;
5356 while (isIDFIRST_lazy_if(s,UTF)) {
5359 d = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
5360 if (isLOWER(*s) && (tmp = keyword(PL_tokenbuf, len, 0))) {
5361 if (tmp < 0) tmp = -tmp;
5376 sv = newSVpvn(s, len);
5378 d = scan_str(d,TRUE,TRUE);
5380 /* MUST advance bufptr here to avoid bogus
5381 "at end of line" context messages from yyerror().
5383 PL_bufptr = s + len;
5384 yyerror("Unterminated attribute parameter in attribute list");
5388 return REPORT(0); /* EOF indicator */
5392 sv_catsv(sv, PL_lex_stuff);
5393 attrs = op_append_elem(OP_LIST, attrs,
5394 newSVOP(OP_CONST, 0, sv));
5395 SvREFCNT_dec(PL_lex_stuff);
5396 PL_lex_stuff = NULL;
5399 if (len == 6 && strnEQ(SvPVX(sv), "unique", len)) {
5401 if (PL_in_my == KEY_our) {
5402 deprecate(":unique");
5405 Perl_croak(aTHX_ "The 'unique' attribute may only be applied to 'our' variables");
5408 /* NOTE: any CV attrs applied here need to be part of
5409 the CVf_BUILTIN_ATTRS define in cv.h! */
5410 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "lvalue", len)) {
5412 CvLVALUE_on(PL_compcv);
5414 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "locked", len)) {
5416 deprecate(":locked");
5418 else if (!PL_in_my && len == 6 && strnEQ(SvPVX(sv), "method", len)) {
5420 CvMETHOD_on(PL_compcv);
5422 /* After we've set the flags, it could be argued that
5423 we don't need to do the attributes.pm-based setting
5424 process, and shouldn't bother appending recognized
5425 flags. To experiment with that, uncomment the
5426 following "else". (Note that's already been
5427 uncommented. That keeps the above-applied built-in
5428 attributes from being intercepted (and possibly
5429 rejected) by a package's attribute routines, but is
5430 justified by the performance win for the common case
5431 of applying only built-in attributes.) */
5433 attrs = op_append_elem(OP_LIST, attrs,
5434 newSVOP(OP_CONST, 0,
5438 if (*s == ':' && s[1] != ':')
5441 break; /* require real whitespace or :'s */
5442 /* XXX losing whitespace on sequential attributes here */
5446 = (PL_expect == XOPERATOR ? '=' : '{'); /*'}(' for vi */
5447 if (*s != ';' && *s != '}' && *s != tmp
5448 && (tmp != '=' || *s != ')')) {
5449 const char q = ((*s == '\'') ? '"' : '\'');
5450 /* If here for an expression, and parsed no attrs, back
5452 if (tmp == '=' && !attrs) {
5456 /* MUST advance bufptr here to avoid bogus "at end of line"
5457 context messages from yyerror().
5460 yyerror( (const char *)
5462 ? Perl_form(aTHX_ "Invalid separator character "
5463 "%c%c%c in attribute list", q, *s, q)
5464 : "Unterminated attribute list" ) );
5472 start_force(PL_curforce);
5473 NEXTVAL_NEXTTOKE.opval = attrs;
5474 CURMAD('_', PL_nextwhite);
5479 PL_thistoken = newSVpvn(SvPVX(PL_linestr) + stuffstart,
5480 (s - SvPVX(PL_linestr)) - stuffstart);
5485 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING) {
5489 PL_lex_allbrackets--;
5493 if (PL_last_lop == PL_oldoldbufptr || PL_last_uni == PL_oldoldbufptr)
5494 PL_oldbufptr = PL_oldoldbufptr; /* allow print(STDOUT 123) */
5498 PL_lex_allbrackets++;
5501 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
5507 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_CLOSING)
5510 PL_lex_allbrackets--;
5516 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5519 if (PL_lex_brackets <= 0)
5520 yyerror("Unmatched right square bracket");
5523 PL_lex_allbrackets--;
5524 if (PL_lex_state == LEX_INTERPNORMAL) {
5525 if (PL_lex_brackets == 0) {
5526 if (*s == '-' && s[1] == '>')
5527 PL_lex_state = LEX_INTERPENDMAYBE;
5528 else if (*s != '[' && *s != '{')
5529 PL_lex_state = LEX_INTERPEND;
5536 if (PL_lex_brackets > 100) {
5537 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
5539 switch (PL_expect) {
5541 if (PL_lex_formbrack) {
5545 if (PL_oldoldbufptr == PL_last_lop)
5546 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5548 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5549 PL_lex_allbrackets++;
5550 OPERATOR(HASHBRACK);
5552 while (s < PL_bufend && SPACE_OR_TAB(*s))
5555 PL_tokenbuf[0] = '\0';
5556 if (d < PL_bufend && *d == '-') {
5557 PL_tokenbuf[0] = '-';
5559 while (d < PL_bufend && SPACE_OR_TAB(*d))
5562 if (d < PL_bufend && isIDFIRST_lazy_if(d,UTF)) {
5563 d = scan_word(d, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1,
5565 while (d < PL_bufend && SPACE_OR_TAB(*d))
5568 const char minus = (PL_tokenbuf[0] == '-');
5569 s = force_word(s + minus, WORD, FALSE, TRUE, FALSE);
5577 PL_lex_brackstack[PL_lex_brackets++] = XSTATE;
5578 PL_lex_allbrackets++;
5583 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5584 PL_lex_allbrackets++;
5589 if (PL_oldoldbufptr == PL_last_lop)
5590 PL_lex_brackstack[PL_lex_brackets++] = XTERM;
5592 PL_lex_brackstack[PL_lex_brackets++] = XOPERATOR;
5593 PL_lex_allbrackets++;
5596 if (PL_expect == XREF && PL_lex_state == LEX_INTERPNORMAL) {
5598 /* This hack is to get the ${} in the message. */
5600 yyerror("syntax error");
5603 OPERATOR(HASHBRACK);
5605 /* This hack serves to disambiguate a pair of curlies
5606 * as being a block or an anon hash. Normally, expectation
5607 * determines that, but in cases where we're not in a
5608 * position to expect anything in particular (like inside
5609 * eval"") we have to resolve the ambiguity. This code
5610 * covers the case where the first term in the curlies is a
5611 * quoted string. Most other cases need to be explicitly
5612 * disambiguated by prepending a "+" before the opening
5613 * curly in order to force resolution as an anon hash.
5615 * XXX should probably propagate the outer expectation
5616 * into eval"" to rely less on this hack, but that could
5617 * potentially break current behavior of eval"".
5621 if (*s == '\'' || *s == '"' || *s == '`') {
5622 /* common case: get past first string, handling escapes */
5623 for (t++; t < PL_bufend && *t != *s;)
5624 if (*t++ == '\\' && (*t == '\\' || *t == *s))
5628 else if (*s == 'q') {
5631 || ((*t == 'q' || *t == 'x') && ++t < PL_bufend
5634 /* skip q//-like construct */
5636 char open, close, term;
5639 while (t < PL_bufend && isSPACE(*t))
5641 /* check for q => */
5642 if (t+1 < PL_bufend && t[0] == '=' && t[1] == '>') {
5643 OPERATOR(HASHBRACK);
5647 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
5651 for (t++; t < PL_bufend; t++) {
5652 if (*t == '\\' && t+1 < PL_bufend && open != '\\')
5654 else if (*t == open)
5658 for (t++; t < PL_bufend; t++) {
5659 if (*t == '\\' && t+1 < PL_bufend)
5661 else if (*t == close && --brackets <= 0)
5663 else if (*t == open)
5670 /* skip plain q word */
5671 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5674 else if (isALNUM_lazy_if(t,UTF)) {
5676 while (t < PL_bufend && isALNUM_lazy_if(t,UTF))
5679 while (t < PL_bufend && isSPACE(*t))
5681 /* if comma follows first term, call it an anon hash */
5682 /* XXX it could be a comma expression with loop modifiers */
5683 if (t < PL_bufend && ((*t == ',' && (*s == 'q' || !isLOWER(*s)))
5684 || (*t == '=' && t[1] == '>')))
5685 OPERATOR(HASHBRACK);
5686 if (PL_expect == XREF)
5689 PL_lex_brackstack[PL_lex_brackets-1] = XSTATE;
5695 pl_yylval.ival = CopLINE(PL_curcop);
5696 if (isSPACE(*s) || *s == '#')
5697 PL_copline = NOLINE; /* invalidate current command line number */
5700 if (PL_lex_brackets && PL_lex_brackstack[PL_lex_brackets-1] == XFAKEEOF)
5704 if (PL_lex_brackets <= 0)
5705 yyerror("Unmatched right curly bracket");
5707 PL_expect = (expectation)PL_lex_brackstack[--PL_lex_brackets];
5708 PL_lex_allbrackets--;
5709 if (PL_lex_brackets < PL_lex_formbrack && PL_lex_state != LEX_INTERPNORMAL)
5710 PL_lex_formbrack = 0;
5711 if (PL_lex_state == LEX_INTERPNORMAL) {
5712 if (PL_lex_brackets == 0) {
5713 if (PL_expect & XFAKEBRACK) {
5714 PL_expect &= XENUMMASK;
5715 PL_lex_state = LEX_INTERPEND;
5720 PL_thiswhite = newSVpvs("");
5721 sv_catpvs(PL_thiswhite,"}");
5724 return yylex(); /* ignore fake brackets */
5726 if (*s == '-' && s[1] == '>')
5727 PL_lex_state = LEX_INTERPENDMAYBE;
5728 else if (*s != '[' && *s != '{')
5729 PL_lex_state = LEX_INTERPEND;
5732 if (PL_expect & XFAKEBRACK) {
5733 PL_expect &= XENUMMASK;
5735 return yylex(); /* ignore fake brackets */
5737 start_force(PL_curforce);
5739 curmad('X', newSVpvn(s-1,1));
5740 CURMAD('_', PL_thiswhite);
5745 PL_thistoken = newSVpvs("");
5751 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5752 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5759 if (PL_expect == XOPERATOR) {
5760 if (PL_bufptr == PL_linestart && ckWARN(WARN_SEMICOLON)
5761 && isIDFIRST_lazy_if(s,UTF))
5763 CopLINE_dec(PL_curcop);
5764 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
5765 CopLINE_inc(PL_curcop);
5767 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5768 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5775 s = scan_ident(s - 1, PL_bufend, PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
5777 PL_expect = XOPERATOR;
5778 force_ident(PL_tokenbuf, '&');
5782 pl_yylval.ival = (OPpENTERSUB_AMPER<<8);
5788 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5789 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC)) {
5796 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
5797 (*s == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_BITWISE)) {
5805 const char tmp = *s++;
5807 if (!PL_lex_allbrackets &&
5808 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5815 if (!PL_lex_allbrackets &&
5816 PL_lex_fakeeof >= LEX_FAKEEOF_COMMA) {
5824 if (tmp && isSPACE(*s) && ckWARN(WARN_SYNTAX)
5825 && strchr("+-*/%.^&|<",tmp))
5826 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5827 "Reversed %c= operator",(int)tmp);
5829 if (PL_expect == XSTATE && isALPHA(tmp) &&
5830 (s == PL_linestart+1 || s[-2] == '\n') )
5832 if (PL_in_eval && !PL_rsfp) {
5837 if (strnEQ(s,"=cut",4)) {
5853 PL_thiswhite = newSVpvs("");
5854 sv_catpvn(PL_thiswhite, PL_linestart,
5855 PL_bufend - PL_linestart);
5859 PL_parser->in_pod = 1;
5863 if (PL_lex_brackets < PL_lex_formbrack) {
5865 #ifdef PERL_STRICT_CR
5866 while (SPACE_OR_TAB(*t))
5868 while (SPACE_OR_TAB(*t) || *t == '\r')
5871 if (*t == '\n' || *t == '#') {
5877 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5886 const char tmp = *s++;
5888 /* was this !=~ where !~ was meant?
5889 * warn on m:!=~\s+([/?]|[msy]\W|tr\W): */
5891 if (*s == '~' && ckWARN(WARN_SYNTAX)) {
5892 const char *t = s+1;
5894 while (t < PL_bufend && isSPACE(*t))
5897 if (*t == '/' || *t == '?' ||
5898 ((*t == 'm' || *t == 's' || *t == 'y')
5899 && !isALNUM(t[1])) ||
5900 (*t == 't' && t[1] == 'r' && !isALNUM(t[2])))
5901 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
5902 "!=~ should be !~");
5904 if (!PL_lex_allbrackets &&
5905 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5917 if (PL_expect != XOPERATOR) {
5918 if (s[1] != '<' && !strchr(s,'>'))
5921 s = scan_heredoc(s);
5923 s = scan_inputsymbol(s);
5924 TERM(sublex_start());
5930 if (*s == '=' && !PL_lex_allbrackets &&
5931 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5935 SHop(OP_LEFT_SHIFT);
5940 if (!PL_lex_allbrackets &&
5941 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5948 if (!PL_lex_allbrackets &&
5949 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5957 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5965 const char tmp = *s++;
5967 if (*s == '=' && !PL_lex_allbrackets &&
5968 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
5972 SHop(OP_RIGHT_SHIFT);
5974 else if (tmp == '=') {
5975 if (!PL_lex_allbrackets &&
5976 PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5984 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE) {
5993 if (PL_expect == XOPERATOR) {
5994 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
5995 return deprecate_commaless_var_list();
5999 if (s[1] == '#' && (isIDFIRST_lazy_if(s+2,UTF) || strchr("{$:+-@", s[2]))) {
6000 PL_tokenbuf[0] = '@';
6001 s = scan_ident(s + 1, PL_bufend, PL_tokenbuf + 1,
6002 sizeof PL_tokenbuf - 1, FALSE);
6003 if (PL_expect == XOPERATOR)
6004 no_op("Array length", s);
6005 if (!PL_tokenbuf[1])
6007 PL_expect = XOPERATOR;
6008 PL_pending_ident = '#';
6012 PL_tokenbuf[0] = '$';
6013 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1,
6014 sizeof PL_tokenbuf - 1, FALSE);
6015 if (PL_expect == XOPERATOR)
6017 if (!PL_tokenbuf[1]) {
6019 yyerror("Final $ should be \\$ or $name");
6023 /* This kludge not intended to be bulletproof. */
6024 if (PL_tokenbuf[1] == '[' && !PL_tokenbuf[2]) {
6025 pl_yylval.opval = newSVOP(OP_CONST, 0,
6026 newSViv(CopARYBASE_get(&PL_compiling)));
6027 pl_yylval.opval->op_private = OPpCONST_ARYBASE;
6033 const char tmp = *s;
6034 if (PL_lex_state == LEX_NORMAL || PL_lex_brackets)
6037 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop)
6038 && intuit_more(s)) {
6040 PL_tokenbuf[0] = '@';
6041 if (ckWARN(WARN_SYNTAX)) {
6044 while (isSPACE(*t) || isALNUM_lazy_if(t,UTF) || *t == '$')
6047 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6048 while (t < PL_bufend && *t != ']')
6050 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6051 "Multidimensional syntax %.*s not supported",
6052 (int)((t - PL_bufptr) + 1), PL_bufptr);
6056 else if (*s == '{') {
6058 PL_tokenbuf[0] = '%';
6059 if (strEQ(PL_tokenbuf+1, "SIG") && ckWARN(WARN_SYNTAX)
6060 && (t = strchr(s, '}')) && (t = strchr(t, '=')))
6062 char tmpbuf[sizeof PL_tokenbuf];
6065 } while (isSPACE(*t));
6066 if (isIDFIRST_lazy_if(t,UTF)) {
6068 t = scan_word(t, tmpbuf, sizeof tmpbuf, TRUE,
6072 if (*t == ';' && get_cvn_flags(tmpbuf, len, 0))
6073 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6074 "You need to quote \"%s\"",
6081 PL_expect = XOPERATOR;
6082 if (PL_lex_state == LEX_NORMAL && isSPACE((char)tmp)) {
6083 const bool islop = (PL_last_lop == PL_oldoldbufptr);
6084 if (!islop || PL_last_lop_op == OP_GREPSTART)
6085 PL_expect = XOPERATOR;
6086 else if (strchr("$@\"'`q", *s))
6087 PL_expect = XTERM; /* e.g. print $fh "foo" */
6088 else if (strchr("&*<%", *s) && isIDFIRST_lazy_if(s+1,UTF))
6089 PL_expect = XTERM; /* e.g. print $fh &sub */
6090 else if (isIDFIRST_lazy_if(s,UTF)) {
6091 char tmpbuf[sizeof PL_tokenbuf];
6093 scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
6094 if ((t2 = keyword(tmpbuf, len, 0))) {
6095 /* binary operators exclude handle interpretations */
6107 PL_expect = XTERM; /* e.g. print $fh length() */
6112 PL_expect = XTERM; /* e.g. print $fh subr() */
6115 else if (isDIGIT(*s))
6116 PL_expect = XTERM; /* e.g. print $fh 3 */
6117 else if (*s == '.' && isDIGIT(s[1]))
6118 PL_expect = XTERM; /* e.g. print $fh .3 */
6119 else if ((*s == '?' || *s == '-' || *s == '+')
6120 && !isSPACE(s[1]) && s[1] != '=')
6121 PL_expect = XTERM; /* e.g. print $fh -1 */
6122 else if (*s == '/' && !isSPACE(s[1]) && s[1] != '='
6124 PL_expect = XTERM; /* e.g. print $fh /.../
6125 XXX except DORDOR operator
6127 else if (*s == '<' && s[1] == '<' && !isSPACE(s[2])
6129 PL_expect = XTERM; /* print $fh <<"EOF" */
6132 PL_pending_ident = '$';
6136 if (PL_expect == XOPERATOR)
6138 PL_tokenbuf[0] = '@';
6139 s = scan_ident(s, PL_bufend, PL_tokenbuf + 1, sizeof PL_tokenbuf - 1, FALSE);
6140 if (!PL_tokenbuf[1]) {
6143 if (PL_lex_state == LEX_NORMAL)
6145 if ((PL_expect != XREF || PL_oldoldbufptr == PL_last_lop) && intuit_more(s)) {
6147 PL_tokenbuf[0] = '%';
6149 /* Warn about @ where they meant $. */
6150 if (*s == '[' || *s == '{') {
6151 if (ckWARN(WARN_SYNTAX)) {
6152 const char *t = s + 1;
6153 while (*t && (isALNUM_lazy_if(t,UTF) || strchr(" \t$#+-'\"", *t)))
6155 if (*t == '}' || *t == ']') {
6157 PL_bufptr = PEEKSPACE(PL_bufptr); /* XXX can realloc */
6158 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
6159 "Scalar value %.*s better written as $%.*s",
6160 (int)(t-PL_bufptr), PL_bufptr,
6161 (int)(t-PL_bufptr-1), PL_bufptr+1);
6166 PL_pending_ident = '@';
6169 case '/': /* may be division, defined-or, or pattern */
6170 if (PL_expect == XTERMORDORDOR && s[1] == '/') {
6171 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6172 (s[2] == '=' ? LEX_FAKEEOF_ASSIGN : LEX_FAKEEOF_LOGIC))
6177 case '?': /* may either be conditional or pattern */
6178 if (PL_expect == XOPERATOR) {
6181 if (!PL_lex_allbrackets &&
6182 PL_lex_fakeeof >= LEX_FAKEEOF_IFELSE) {
6186 PL_lex_allbrackets++;
6192 /* A // operator. */
6193 if (!PL_lex_allbrackets && PL_lex_fakeeof >=
6194 (*s == '=' ? LEX_FAKEEOF_ASSIGN :
6195 LEX_FAKEEOF_LOGIC)) {
6203 if (*s == '=' && !PL_lex_allbrackets &&
6204 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6213 /* Disable warning on "study /blah/" */
6214 if (PL_oldoldbufptr == PL_last_uni
6215 && (*PL_last_uni != 's' || s - PL_last_uni < 5
6216 || memNE(PL_last_uni, "study", 5)
6217 || isALNUM_lazy_if(PL_last_uni+5,UTF)
6221 deprecate("?PATTERN? without explicit operator");
6222 s = scan_pat(s,OP_MATCH);
6223 TERM(sublex_start());
6227 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack
6228 #ifdef PERL_STRICT_CR
6231 && (s[1] == '\n' || (s[1] == '\r' && s[2] == '\n'))
6233 && (s == PL_linestart || s[-1] == '\n') )
6235 PL_lex_formbrack = 0;
6239 if (PL_expect == XSTATE && s[1] == '.' && s[2] == '.') {
6243 if (PL_expect == XOPERATOR || !isDIGIT(s[1])) {
6246 if (!PL_lex_allbrackets &&
6247 PL_lex_fakeeof >= LEX_FAKEEOF_RANGE) {
6254 pl_yylval.ival = OPf_SPECIAL;
6260 if (*s == '=' && !PL_lex_allbrackets &&
6261 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN) {
6268 case '0': case '1': case '2': case '3': case '4':
6269 case '5': case '6': case '7': case '8': case '9':
6270 s = scan_num(s, &pl_yylval);
6271 DEBUG_T( { printbuf("### Saw number in %s\n", s); } );
6272 if (PL_expect == XOPERATOR)
6277 s = scan_str(s,!!PL_madskills,FALSE);
6278 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6279 if (PL_expect == XOPERATOR) {
6280 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6281 return deprecate_commaless_var_list();
6288 pl_yylval.ival = OP_CONST;
6289 TERM(sublex_start());
6292 s = scan_str(s,!!PL_madskills,FALSE);
6293 DEBUG_T( { printbuf("### Saw string before %s\n", s); } );
6294 if (PL_expect == XOPERATOR) {
6295 if (PL_lex_formbrack && PL_lex_brackets == PL_lex_formbrack) {
6296 return deprecate_commaless_var_list();
6303 pl_yylval.ival = OP_CONST;
6304 /* FIXME. I think that this can be const if char *d is replaced by
6305 more localised variables. */
6306 for (d = SvPV(PL_lex_stuff, len); len; len--, d++) {
6307 if (*d == '$' || *d == '@' || *d == '\\' || !UTF8_IS_INVARIANT((U8)*d)) {
6308 pl_yylval.ival = OP_STRINGIFY;
6312 TERM(sublex_start());
6315 s = scan_str(s,!!PL_madskills,FALSE);
6316 DEBUG_T( { printbuf("### Saw backtick string before %s\n", s); } );
6317 if (PL_expect == XOPERATOR)
6318 no_op("Backticks",s);
6321 readpipe_override();
6322 TERM(sublex_start());
6326 if (PL_lex_inwhat && isDIGIT(*s))
6327 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),"Can't use \\%c to mean $%c in expression",
6329 if (PL_expect == XOPERATOR)
6330 no_op("Backslash",s);
6334 if (isDIGIT(s[1]) && PL_expect != XOPERATOR) {
6335 char *start = s + 2;
6336 while (isDIGIT(*start) || *start == '_')
6338 if (*start == '.' && isDIGIT(start[1])) {
6339 s = scan_num(s, &pl_yylval);
6342 /* avoid v123abc() or $h{v1}, allow C<print v10;> */
6343 else if (!isALPHA(*start) && (PL_expect == XTERM
6344 || PL_expect == XREF || PL_expect == XSTATE
6345 || PL_expect == XTERMORDORDOR)) {
6346 GV *const gv = gv_fetchpvn_flags(s, start - s, 0, SVt_PVCV);
6348 s = scan_num(s, &pl_yylval);
6355 if (isDIGIT(s[1]) && PL_expect == XOPERATOR) {
6398 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
6400 /* Some keywords can be followed by any delimiter, including ':' */
6401 anydelim = word_takes_any_delimeter(PL_tokenbuf, len);
6403 /* x::* is just a word, unless x is "CORE" */
6404 if (!anydelim && *s == ':' && s[1] == ':' && strNE(PL_tokenbuf, "CORE"))
6408 while (d < PL_bufend && isSPACE(*d))
6409 d++; /* no comments skipped here, or s### is misparsed */
6411 /* Is this a word before a => operator? */
6412 if (*d == '=' && d[1] == '>') {
6415 = (OP*)newSVOP(OP_CONST, 0,
6416 S_newSV_maybe_utf8(aTHX_ PL_tokenbuf, len));
6417 pl_yylval.opval->op_private = OPpCONST_BARE;
6421 /* Check for plugged-in keyword */
6425 char *saved_bufptr = PL_bufptr;
6427 result = PL_keyword_plugin(aTHX_ PL_tokenbuf, len, &o);
6429 if (result == KEYWORD_PLUGIN_DECLINE) {
6430 /* not a plugged-in keyword */
6431 PL_bufptr = saved_bufptr;
6432 } else if (result == KEYWORD_PLUGIN_STMT) {
6433 pl_yylval.opval = o;
6436 return REPORT(PLUGSTMT);
6437 } else if (result == KEYWORD_PLUGIN_EXPR) {
6438 pl_yylval.opval = o;
6440 PL_expect = XOPERATOR;
6441 return REPORT(PLUGEXPR);
6443 Perl_croak(aTHX_ "Bad plugin affecting keyword '%s'",
6448 /* Check for built-in keyword */
6449 tmp = keyword(PL_tokenbuf, len, 0);
6451 /* Is this a label? */
6452 if (!anydelim && PL_expect == XSTATE
6453 && d < PL_bufend && *d == ':' && *(d + 1) != ':') {
6455 pl_yylval.pval = CopLABEL_alloc(PL_tokenbuf);
6460 if (tmp < 0) { /* second-class keyword? */
6461 GV *ogv = NULL; /* override (winner) */
6462 GV *hgv = NULL; /* hidden (loser) */
6463 if (PL_expect != XOPERATOR && (*s != ':' || s[1] != ':')) {
6465 if ((gv = gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVCV)) &&
6468 if (GvIMPORTED_CV(gv))
6470 else if (! CvMETHOD(cv))
6474 (gvp = (GV**)hv_fetch(PL_globalstash,PL_tokenbuf,len,FALSE)) &&
6475 (gv = *gvp) && isGV_with_GP(gv) &&
6476 GvCVu(gv) && GvIMPORTED_CV(gv))
6483 tmp = 0; /* overridden by import or by GLOBAL */
6486 && -tmp==KEY_lock /* XXX generalizable kludge */
6489 tmp = 0; /* any sub overrides "weak" keyword */
6491 else { /* no override */
6493 if (tmp == KEY_dump) {
6494 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
6495 "dump() better written as CORE::dump()");
6499 if (hgv && tmp != KEY_x && tmp != KEY_CORE) /* never ambiguous */
6500 Perl_ck_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
6501 "Ambiguous call resolved as CORE::%s(), "
6502 "qualify as such or use &",
6510 default: /* not a keyword */
6511 /* Trade off - by using this evil construction we can pull the
6512 variable gv into the block labelled keylookup. If not, then
6513 we have to give it function scope so that the goto from the
6514 earlier ':' case doesn't bypass the initialisation. */
6516 just_a_word_zero_gv:
6524 const char lastchar = (PL_bufptr == PL_oldoldbufptr ? 0 : PL_bufptr[-1]);
6528 SV *nextPL_nextwhite = 0;
6532 /* Get the rest if it looks like a package qualifier */
6534 if (*s == '\'' || (*s == ':' && s[1] == ':')) {
6536 s = scan_word(s, PL_tokenbuf + len, sizeof PL_tokenbuf - len,
6539 Perl_croak(aTHX_ "Bad name after %s%s", PL_tokenbuf,
6540 *s == '\'' ? "'" : "::");
6545 if (PL_expect == XOPERATOR) {
6546 if (PL_bufptr == PL_linestart) {
6547 CopLINE_dec(PL_curcop);
6548 Perl_warner(aTHX_ packWARN(WARN_SEMICOLON), "%s", PL_warn_nosemi);
6549 CopLINE_inc(PL_curcop);
6552 no_op("Bareword",s);
6555 /* Look for a subroutine with this name in current package,
6556 unless name is "Foo::", in which case Foo is a bareword
6557 (and a package name). */
6559 if (len > 2 && !PL_madskills &&
6560 PL_tokenbuf[len - 2] == ':' && PL_tokenbuf[len - 1] == ':')
6562 if (ckWARN(WARN_BAREWORD)
6563 && ! gv_fetchpvn_flags(PL_tokenbuf, len, 0, SVt_PVHV))
6564 Perl_warner(aTHX_ packWARN(WARN_BAREWORD),
6565 "Bareword \"%s\" refers to nonexistent package",
6568 PL_tokenbuf[len] = '\0';
6574 /* Mustn't actually add anything to a symbol table.
6575 But also don't want to "initialise" any placeholder
6576 constants that might already be there into full
6577 blown PVGVs with attached PVCV. */
6578 gv = gv_fetchpvn_flags(PL_tokenbuf, len,
6579 GV_NOADD_NOINIT, SVt_PVCV);
6584 /* if we saw a global override before, get the right name */
6586 sv = S_newSV_maybe_utf8(aTHX_ PL_tokenbuf,
6587 len ? len : strlen(PL_tokenbuf));
6589 SV * const tmp_sv = sv;
6590 sv = newSVpvs("CORE::GLOBAL::");
6591 sv_catsv(sv, tmp_sv);
6592 SvREFCNT_dec(tmp_sv);
6596 if (PL_madskills && !PL_thistoken) {
6597 char *start = SvPVX(PL_linestr) + PL_realtokenstart;
6598 PL_thistoken = newSVpvn(start,s - start);
6599 PL_realtokenstart = s - SvPVX(PL_linestr);
6603 /* Presume this is going to be a bareword of some sort. */
6605 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sv);
6606 pl_yylval.opval->op_private = OPpCONST_BARE;
6608 /* And if "Foo::", then that's what it certainly is. */
6613 OP *const_op = newSVOP(OP_CONST, 0, SvREFCNT_inc_NN(sv));
6614 const_op->op_private = OPpCONST_BARE;
6615 rv2cv_op = newCVREF(0, const_op);
6617 cv = rv2cv_op_cv(rv2cv_op, 0);
6619 /* See if it's the indirect object for a list operator. */
6621 if (PL_oldoldbufptr &&
6622 PL_oldoldbufptr < PL_bufptr &&
6623 (PL_oldoldbufptr == PL_last_lop
6624 || PL_oldoldbufptr == PL_last_uni) &&
6625 /* NO SKIPSPACE BEFORE HERE! */
6626 (PL_expect == XREF ||
6627 ((PL_opargs[PL_last_lop_op] >> OASHIFT)& 7) == OA_FILEREF))
6629 bool immediate_paren = *s == '(';
6631 /* (Now we can afford to cross potential line boundary.) */
6632 s = SKIPSPACE2(s,nextPL_nextwhite);
6634 PL_nextwhite = nextPL_nextwhite; /* assume no & deception */
6637 /* Two barewords in a row may indicate method call. */
6639 if ((isIDFIRST_lazy_if(s,UTF) || *s == '$') &&
6640 (tmp = intuit_method(s, gv, cv))) {
6642 if (tmp == METHOD && !PL_lex_allbrackets &&
6643 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6644 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6648 /* If not a declared subroutine, it's an indirect object. */
6649 /* (But it's an indir obj regardless for sort.) */
6650 /* Also, if "_" follows a filetest operator, it's a bareword */
6653 ( !immediate_paren && (PL_last_lop_op == OP_SORT ||
6655 (PL_last_lop_op != OP_MAPSTART &&
6656 PL_last_lop_op != OP_GREPSTART))))
6657 || (PL_tokenbuf[0] == '_' && PL_tokenbuf[1] == '\0'
6658 && ((PL_opargs[PL_last_lop_op] & OA_CLASS_MASK) == OA_FILESTATOP))
6661 PL_expect = (PL_last_lop == PL_oldoldbufptr) ? XTERM : XOPERATOR;
6666 PL_expect = XOPERATOR;
6669 s = SKIPSPACE2(s,nextPL_nextwhite);
6670 PL_nextwhite = nextPL_nextwhite;
6675 /* Is this a word before a => operator? */
6676 if (*s == '=' && s[1] == '>' && !pkgname) {
6679 sv_setpv(((SVOP*)pl_yylval.opval)->op_sv, PL_tokenbuf);
6680 if (UTF && !IN_BYTES && is_utf8_string((U8*)PL_tokenbuf, len))
6681 SvUTF8_on(((SVOP*)pl_yylval.opval)->op_sv);
6685 /* If followed by a paren, it's certainly a subroutine. */
6690 while (SPACE_OR_TAB(*d))
6692 if (*d == ')' && (sv = cv_const_sv(cv))) {
6699 PL_nextwhite = PL_thiswhite;
6702 start_force(PL_curforce);
6704 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6705 PL_expect = XOPERATOR;
6708 PL_nextwhite = nextPL_nextwhite;
6709 curmad('X', PL_thistoken);
6710 PL_thistoken = newSVpvs("");
6719 /* If followed by var or block, call it a method (unless sub) */
6721 if ((*s == '$' || *s == '{') && !cv) {
6723 PL_last_lop = PL_oldbufptr;
6724 PL_last_lop_op = OP_METHOD;
6725 if (!PL_lex_allbrackets &&
6726 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6727 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6731 /* If followed by a bareword, see if it looks like indir obj. */
6734 && (isIDFIRST_lazy_if(s,UTF) || *s == '$')
6735 && (tmp = intuit_method(s, gv, cv))) {
6737 if (tmp == METHOD && !PL_lex_allbrackets &&
6738 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6739 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6743 /* Not a method, so call it a subroutine (if defined) */
6746 if (lastchar == '-')
6747 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6748 "Ambiguous use of -%s resolved as -&%s()",
6749 PL_tokenbuf, PL_tokenbuf);
6750 /* Check for a constant sub */
6751 if ((sv = cv_const_sv(cv))) {
6754 SvREFCNT_dec(((SVOP*)pl_yylval.opval)->op_sv);
6755 ((SVOP*)pl_yylval.opval)->op_sv = SvREFCNT_inc_simple(sv);
6756 pl_yylval.opval->op_private = 0;
6757 pl_yylval.opval->op_flags |= OPf_SPECIAL;
6761 op_free(pl_yylval.opval);
6762 pl_yylval.opval = rv2cv_op;
6763 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6764 PL_last_lop = PL_oldbufptr;
6765 PL_last_lop_op = OP_ENTERSUB;
6766 /* Is there a prototype? */
6774 const char *proto = SvPV_const(MUTABLE_SV(cv), protolen);
6777 while (*proto == ';')
6782 *proto == '$' || *proto == '_'
6783 || *proto == '*' || *proto == '+'
6788 *proto == '\\' && proto[1] && proto[2] == '\0'
6792 if (*proto == '\\' && proto[1] == '[') {
6793 const char *p = proto + 2;
6794 while(*p && *p != ']')
6796 if(*p == ']' && !p[1]) OPERATOR(UNIOPSUB);
6798 if (*proto == '&' && *s == '{') {
6800 sv_setpvs(PL_subname, "__ANON__");
6802 sv_setpvs(PL_subname, "__ANON__::__ANON__");
6803 if (!PL_lex_allbrackets &&
6804 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6805 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6812 PL_nextwhite = PL_thiswhite;
6815 start_force(PL_curforce);
6816 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6819 PL_nextwhite = nextPL_nextwhite;
6820 curmad('X', PL_thistoken);
6821 PL_thistoken = newSVpvs("");
6824 if (!PL_lex_allbrackets &&
6825 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6826 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6831 /* Guess harder when madskills require "best effort". */
6832 if (PL_madskills && (!gv || !GvCVu(gv))) {
6833 int probable_sub = 0;
6834 if (strchr("\"'`$@%0123456789!*+{[<", *s))
6836 else if (isALPHA(*s)) {
6840 d = scan_word(d, tmpbuf, sizeof tmpbuf, TRUE, &tmplen);
6841 if (!keyword(tmpbuf, tmplen, 0))
6844 while (d < PL_bufend && isSPACE(*d))
6846 if (*d == '=' && d[1] == '>')
6851 gv = gv_fetchpv(PL_tokenbuf, GV_ADD, SVt_PVCV);
6852 op_free(pl_yylval.opval);
6853 pl_yylval.opval = rv2cv_op;
6854 pl_yylval.opval->op_private |= OPpENTERSUB_NOPAREN;
6855 PL_last_lop = PL_oldbufptr;
6856 PL_last_lop_op = OP_ENTERSUB;
6857 PL_nextwhite = PL_thiswhite;
6859 start_force(PL_curforce);
6860 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6862 PL_nextwhite = nextPL_nextwhite;
6863 curmad('X', PL_thistoken);
6864 PL_thistoken = newSVpvs("");
6866 if (!PL_lex_allbrackets &&
6867 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6868 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6872 NEXTVAL_NEXTTOKE.opval = pl_yylval.opval;
6875 if (!PL_lex_allbrackets &&
6876 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
6877 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
6882 /* Call it a bare word */
6884 if (PL_hints & HINT_STRICT_SUBS)
6885 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6888 /* after "print" and similar functions (corresponding to
6889 * "F? L" in opcode.pl), whatever wasn't already parsed as
6890 * a filehandle should be subject to "strict subs".
6891 * Likewise for the optional indirect-object argument to system
6892 * or exec, which can't be a bareword */
6893 if ((PL_last_lop_op == OP_PRINT
6894 || PL_last_lop_op == OP_PRTF
6895 || PL_last_lop_op == OP_SAY
6896 || PL_last_lop_op == OP_SYSTEM
6897 || PL_last_lop_op == OP_EXEC)
6898 && (PL_hints & HINT_STRICT_SUBS))
6899 pl_yylval.opval->op_private |= OPpCONST_STRICT;
6900 if (lastchar != '-') {
6901 if (ckWARN(WARN_RESERVED)) {
6905 if (!*d && !gv_stashpv(PL_tokenbuf, 0))
6906 Perl_warner(aTHX_ packWARN(WARN_RESERVED), PL_warn_reserved,
6914 if ((lastchar == '*' || lastchar == '%' || lastchar == '&')) {
6915 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6916 "Operator or semicolon missing before %c%s",
6917 lastchar, PL_tokenbuf);
6918 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
6919 "Ambiguous use of %c resolved as operator %c",
6920 lastchar, lastchar);
6926 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6927 newSVpv(CopFILE(PL_curcop),0));
6931 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6932 Perl_newSVpvf(aTHX_ "%"IVdf, (IV)CopLINE(PL_curcop)));
6935 case KEY___PACKAGE__:
6936 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0,
6938 ? newSVhek(HvNAME_HEK(PL_curstash))
6945 if (PL_rsfp && (!PL_in_eval || PL_tokenbuf[2] == 'D')) {
6946 const char *pname = "main";
6947 if (PL_tokenbuf[2] == 'D')
6948 pname = HvNAME_get(PL_curstash ? PL_curstash : PL_defstash);
6949 gv = gv_fetchpv(Perl_form(aTHX_ "%s::DATA", pname), GV_ADD,
6953 GvIOp(gv) = newIO();
6954 IoIFP(GvIOp(gv)) = PL_rsfp;
6955 #if defined(HAS_FCNTL) && defined(F_SETFD)
6957 const int fd = PerlIO_fileno(PL_rsfp);
6958 fcntl(fd,F_SETFD,fd >= 3);
6961 /* Mark this internal pseudo-handle as clean */
6962 IoFLAGS(GvIOp(gv)) |= IOf_UNTAINT;
6963 if ((PerlIO*)PL_rsfp == PerlIO_stdin())
6964 IoTYPE(GvIOp(gv)) = IoTYPE_STD;
6966 IoTYPE(GvIOp(gv)) = IoTYPE_RDONLY;
6967 #if defined(WIN32) && !defined(PERL_TEXTMODE_SCRIPTS)
6968 /* if the script was opened in binmode, we need to revert
6969 * it to text mode for compatibility; but only iff it has CRs
6970 * XXX this is a questionable hack at best. */
6971 if (PL_bufend-PL_bufptr > 2
6972 && PL_bufend[-1] == '\n' && PL_bufend[-2] == '\r')
6975 if (IoTYPE(GvIOp(gv)) == IoTYPE_RDONLY) {
6976 loc = PerlIO_tell(PL_rsfp);
6977 (void)PerlIO_seek(PL_rsfp, 0L, 0);
6980 if (PerlLIO_setmode(PL_rsfp, O_TEXT) != -1) {
6982 if (PerlLIO_setmode(PerlIO_fileno(PL_rsfp), O_TEXT) != -1) {
6983 #endif /* NETWARE */
6984 #ifdef PERLIO_IS_STDIO /* really? */
6985 # if defined(__BORLANDC__)
6986 /* XXX see note in do_binmode() */
6987 ((FILE*)PL_rsfp)->flags &= ~_F_BIN;
6991 PerlIO_seek(PL_rsfp, loc, 0);
6995 #ifdef PERLIO_LAYERS
6998 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL, ":utf8");
6999 else if (PL_encoding) {
7006 XPUSHs(PL_encoding);
7008 call_method("name", G_SCALAR);
7012 PerlIO_apply_layers(aTHX_ PL_rsfp, NULL,
7013 Perl_form(aTHX_ ":encoding(%"SVf")",
7022 if (PL_realtokenstart >= 0) {
7023 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7025 PL_endwhite = newSVpvs("");
7026 sv_catsv(PL_endwhite, PL_thiswhite);
7028 sv_catpvn(PL_endwhite, tstart, PL_bufend - tstart);
7029 PL_realtokenstart = -1;
7031 while ((s = filter_gets(PL_endwhite, SvCUR(PL_endwhite)))
7047 if (PL_expect == XSTATE) {
7054 if (*s == ':' && s[1] == ':') {
7057 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, FALSE, &len);
7058 if (!(tmp = keyword(PL_tokenbuf, len, 0)))
7059 Perl_croak(aTHX_ "CORE::%s is not a keyword", PL_tokenbuf);
7062 else if (tmp == KEY_require || tmp == KEY_do)
7063 /* that's a way to remember we saw "CORE::" */
7076 LOP(OP_ACCEPT,XTERM);
7079 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7084 LOP(OP_ATAN2,XTERM);
7090 LOP(OP_BINMODE,XTERM);
7093 LOP(OP_BLESS,XTERM);
7102 /* When 'use switch' is in effect, continue has a dual
7103 life as a control operator. */
7105 if (!FEATURE_IS_ENABLED("switch"))
7108 /* We have to disambiguate the two senses of
7109 "continue". If the next token is a '{' then
7110 treat it as the start of a continue block;
7111 otherwise treat it as a control operator.
7123 (void)gv_fetchpvs("ENV", GV_ADD|GV_NOTQUAL, SVt_PVHV);
7133 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7142 if (!PL_cryptseen) {
7143 PL_cryptseen = TRUE;
7147 LOP(OP_CRYPT,XTERM);
7150 LOP(OP_CHMOD,XTERM);
7153 LOP(OP_CHOWN,XTERM);
7156 LOP(OP_CONNECT,XTERM);
7175 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7176 if (orig_keyword == KEY_do) {
7185 PL_hints |= HINT_BLOCK_SCOPE;
7195 Perl_populate_isa(aTHX_ STR_WITH_LEN("AnyDBM_File::ISA"),
7196 STR_WITH_LEN("NDBM_File::"),
7197 STR_WITH_LEN("DB_File::"),
7198 STR_WITH_LEN("GDBM_File::"),
7199 STR_WITH_LEN("SDBM_File::"),
7200 STR_WITH_LEN("ODBM_File::"),
7202 LOP(OP_DBMOPEN,XTERM);
7208 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7215 pl_yylval.ival = CopLINE(PL_curcop);
7219 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7233 if (*s == '{') { /* block eval */
7234 PL_expect = XTERMBLOCK;
7235 UNIBRACK(OP_ENTERTRY);
7237 else { /* string eval */
7239 UNIBRACK(OP_ENTEREVAL);
7254 case KEY_endhostent:
7260 case KEY_endservent:
7263 case KEY_endprotoent:
7274 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7276 pl_yylval.ival = CopLINE(PL_curcop);
7278 if (PL_expect == XSTATE && isIDFIRST_lazy_if(s,UTF)) {
7281 int soff = s - SvPVX(PL_linestr); /* for skipspace realloc */
7284 if ((PL_bufend - p) >= 3 &&
7285 strnEQ(p, "my", 2) && isSPACE(*(p + 2)))
7287 else if ((PL_bufend - p) >= 4 &&
7288 strnEQ(p, "our", 3) && isSPACE(*(p + 3)))
7291 if (isIDFIRST_lazy_if(p,UTF)) {
7292 p = scan_ident(p, PL_bufend,
7293 PL_tokenbuf, sizeof PL_tokenbuf, TRUE);
7297 Perl_croak(aTHX_ "Missing $ on loop variable");
7299 s = SvPVX(PL_linestr) + soff;
7305 LOP(OP_FORMLINE,XTERM);
7311 LOP(OP_FCNTL,XTERM);
7317 LOP(OP_FLOCK,XTERM);
7320 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7325 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7330 LOP(OP_GREPSTART, XREF);
7333 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7348 case KEY_getpriority:
7349 LOP(OP_GETPRIORITY,XTERM);
7351 case KEY_getprotobyname:
7354 case KEY_getprotobynumber:
7355 LOP(OP_GPBYNUMBER,XTERM);
7357 case KEY_getprotoent:
7369 case KEY_getpeername:
7370 UNI(OP_GETPEERNAME);
7372 case KEY_gethostbyname:
7375 case KEY_gethostbyaddr:
7376 LOP(OP_GHBYADDR,XTERM);
7378 case KEY_gethostent:
7381 case KEY_getnetbyname:
7384 case KEY_getnetbyaddr:
7385 LOP(OP_GNBYADDR,XTERM);
7390 case KEY_getservbyname:
7391 LOP(OP_GSBYNAME,XTERM);
7393 case KEY_getservbyport:
7394 LOP(OP_GSBYPORT,XTERM);
7396 case KEY_getservent:
7399 case KEY_getsockname:
7400 UNI(OP_GETSOCKNAME);
7402 case KEY_getsockopt:
7403 LOP(OP_GSOCKOPT,XTERM);
7418 pl_yylval.ival = CopLINE(PL_curcop);
7428 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
7430 pl_yylval.ival = CopLINE(PL_curcop);
7434 LOP(OP_INDEX,XTERM);
7440 LOP(OP_IOCTL,XTERM);
7452 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7469 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7474 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7488 LOP(OP_LISTEN,XTERM);
7497 s = scan_pat(s,OP_MATCH);
7498 TERM(sublex_start());
7501 LOP(OP_MAPSTART, XREF);
7504 LOP(OP_MKDIR,XTERM);
7507 LOP(OP_MSGCTL,XTERM);
7510 LOP(OP_MSGGET,XTERM);
7513 LOP(OP_MSGRCV,XTERM);
7516 LOP(OP_MSGSND,XTERM);
7521 PL_in_my = (U16)tmp;
7523 if (isIDFIRST_lazy_if(s,UTF)) {
7527 s = scan_word(s, PL_tokenbuf, sizeof PL_tokenbuf, TRUE, &len);
7528 if (len == 3 && strnEQ(PL_tokenbuf, "sub", 3))
7530 PL_in_my_stash = find_in_my_stash(PL_tokenbuf, len);
7531 if (!PL_in_my_stash) {
7534 my_snprintf(tmpbuf, sizeof(tmpbuf), "No such class %.1000s", PL_tokenbuf);
7538 if (PL_madskills) { /* just add type to declarator token */
7539 sv_catsv(PL_thistoken, PL_nextwhite);
7541 sv_catpvn(PL_thistoken, start, s - start);
7549 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7553 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_COMPARE)
7558 s = tokenize_use(0, s);
7562 if (*s == '(' || (s = SKIPSPACE1(s), *s == '('))
7565 if (!PL_lex_allbrackets &&
7566 PL_lex_fakeeof > LEX_FAKEEOF_LOWLOGIC)
7567 PL_lex_fakeeof = LEX_FAKEEOF_LOWLOGIC;
7573 if (isIDFIRST_lazy_if(s,UTF)) {
7575 for (d = s; isALNUM_lazy_if(d,UTF);)
7577 for (t=d; isSPACE(*t);)
7579 if ( *t && strchr("|&*+-=!?:.", *t) && ckWARN_d(WARN_PRECEDENCE)
7581 && !(t[0] == '=' && t[1] == '>')
7583 int parms_len = (int)(d-s);
7584 Perl_warner(aTHX_ packWARN(WARN_PRECEDENCE),
7585 "Precedence problem: open %.*s should be open(%.*s)",
7586 parms_len, s, parms_len, s);
7592 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
7594 pl_yylval.ival = OP_OR;
7604 LOP(OP_OPEN_DIR,XTERM);
7607 checkcomma(s,PL_tokenbuf,"filehandle");
7611 checkcomma(s,PL_tokenbuf,"filehandle");
7630 s = force_word(s,WORD,FALSE,TRUE,FALSE);
7632 s = force_strict_version(s);
7633 PL_lex_expect = XBLOCK;
7637 LOP(OP_PIPE_OP,XTERM);
7640 s = scan_str(s,!!PL_madskills,FALSE);
7643 pl_yylval.ival = OP_CONST;
7644 TERM(sublex_start());
7651 s = scan_str(s,!!PL_madskills,FALSE);
7654 PL_expect = XOPERATOR;
7655 if (SvCUR(PL_lex_stuff)) {
7657 d = SvPV_force(PL_lex_stuff, len);
7659 for (; isSPACE(*d) && len; --len, ++d)
7664 if (!warned && ckWARN(WARN_QW)) {
7665 for (; !isSPACE(*d) && len; --len, ++d) {
7667 Perl_warner(aTHX_ packWARN(WARN_QW),
7668 "Possible attempt to separate words with commas");
7671 else if (*d == '#') {
7672 Perl_warner(aTHX_ packWARN(WARN_QW),
7673 "Possible attempt to put comments in qw() list");
7679 for (; !isSPACE(*d) && len; --len, ++d)
7682 sv = newSVpvn_utf8(b, d-b, DO_UTF8(PL_lex_stuff));
7683 words = op_append_elem(OP_LIST, words,
7684 newSVOP(OP_CONST, 0, tokeq(sv)));
7689 words = newNULLLIST();
7691 SvREFCNT_dec(PL_lex_stuff);
7692 PL_lex_stuff = NULL;
7694 PL_expect = XOPERATOR;
7695 pl_yylval.opval = sawparens(words);
7700 s = scan_str(s,!!PL_madskills,FALSE);
7703 pl_yylval.ival = OP_STRINGIFY;
7704 if (SvIVX(PL_lex_stuff) == '\'')
7705 SvIV_set(PL_lex_stuff, 0); /* qq'$foo' should interpolate */
7706 TERM(sublex_start());
7709 s = scan_pat(s,OP_QR);
7710 TERM(sublex_start());
7713 s = scan_str(s,!!PL_madskills,FALSE);
7716 readpipe_override();
7717 TERM(sublex_start());
7725 s = force_version(s, FALSE);
7727 else if (*s != 'v' || !isDIGIT(s[1])
7728 || (s = force_version(s, TRUE), *s == 'v'))
7730 *PL_tokenbuf = '\0';
7731 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7732 if (isIDFIRST_lazy_if(PL_tokenbuf,UTF))
7733 gv_stashpvn(PL_tokenbuf, strlen(PL_tokenbuf), GV_ADD);
7735 yyerror("<> should be quotes");
7737 if (orig_keyword == KEY_require) {
7745 PL_last_uni = PL_oldbufptr;
7746 PL_last_lop_op = OP_REQUIRE;
7748 return REPORT( (int)REQUIRE );
7754 s = force_word(s,WORD,TRUE,FALSE,FALSE);
7758 LOP(OP_RENAME,XTERM);
7767 LOP(OP_RINDEX,XTERM);
7776 UNIDOR(OP_READLINE);
7779 UNIDOR(OP_BACKTICK);
7788 LOP(OP_REVERSE,XTERM);
7791 UNIDOR(OP_READLINK);
7798 if (pl_yylval.opval)
7799 TERM(sublex_start());
7801 TOKEN(1); /* force error */
7804 checkcomma(s,PL_tokenbuf,"filehandle");
7814 LOP(OP_SELECT,XTERM);
7820 LOP(OP_SEMCTL,XTERM);
7823 LOP(OP_SEMGET,XTERM);
7826 LOP(OP_SEMOP,XTERM);
7832 LOP(OP_SETPGRP,XTERM);
7834 case KEY_setpriority:
7835 LOP(OP_SETPRIORITY,XTERM);
7837 case KEY_sethostent:
7843 case KEY_setservent:
7846 case KEY_setprotoent:
7856 LOP(OP_SEEKDIR,XTERM);
7858 case KEY_setsockopt:
7859 LOP(OP_SSOCKOPT,XTERM);
7865 LOP(OP_SHMCTL,XTERM);
7868 LOP(OP_SHMGET,XTERM);
7871 LOP(OP_SHMREAD,XTERM);
7874 LOP(OP_SHMWRITE,XTERM);
7877 LOP(OP_SHUTDOWN,XTERM);
7886 LOP(OP_SOCKET,XTERM);
7888 case KEY_socketpair:
7889 LOP(OP_SOCKPAIR,XTERM);
7892 checkcomma(s,PL_tokenbuf,"subroutine name");
7894 if (*s == ';' || *s == ')') /* probably a close */
7895 Perl_croak(aTHX_ "sort is now a reserved word");
7897 s = force_word(s,WORD,TRUE,TRUE,FALSE);
7901 LOP(OP_SPLIT,XTERM);
7904 LOP(OP_SPRINTF,XTERM);
7907 LOP(OP_SPLICE,XTERM);
7922 LOP(OP_SUBSTR,XTERM);
7928 char tmpbuf[sizeof PL_tokenbuf];
7929 SSize_t tboffset = 0;
7930 expectation attrful;
7931 bool have_name, have_proto;
7932 const int key = tmp;
7937 char *tstart = SvPVX(PL_linestr) + PL_realtokenstart;
7938 SV *subtoken = newSVpvn(tstart, s - tstart);
7942 s = SKIPSPACE2(s,tmpwhite);
7947 if (isIDFIRST_lazy_if(s,UTF) || *s == '\'' ||
7948 (*s == ':' && s[1] == ':'))
7951 SV *nametoke = NULL;
7955 attrful = XATTRBLOCK;
7956 /* remember buffer pos'n for later force_word */
7957 tboffset = s - PL_oldbufptr;
7958 d = scan_word(s, tmpbuf, sizeof tmpbuf, TRUE, &len);
7961 nametoke = newSVpvn(s, d - s);
7963 if (memchr(tmpbuf, ':', len))
7964 sv_setpvn(PL_subname, tmpbuf, len);
7966 sv_setsv(PL_subname,PL_curstname);
7967 sv_catpvs(PL_subname,"::");
7968 sv_catpvn(PL_subname,tmpbuf,len);
7975 CURMAD('X', nametoke);
7976 CURMAD('_', tmpwhite);
7977 (void) force_word(PL_oldbufptr + tboffset, WORD,
7980 s = SKIPSPACE2(d,tmpwhite);
7987 Perl_croak(aTHX_ "Missing name in \"my sub\"");
7988 PL_expect = XTERMBLOCK;
7989 attrful = XATTRTERM;
7990 sv_setpvs(PL_subname,"?");
7994 if (key == KEY_format) {
7996 PL_lex_formbrack = PL_lex_brackets + 1;
7998 PL_thistoken = subtoken;
8002 (void) force_word(PL_oldbufptr + tboffset, WORD,
8008 /* Look for a prototype */
8011 bool bad_proto = FALSE;
8012 bool in_brackets = FALSE;
8013 char greedy_proto = ' ';
8014 bool proto_after_greedy_proto = FALSE;
8015 bool must_be_last = FALSE;
8016 bool underscore = FALSE;
8017 bool seen_underscore = FALSE;
8018 const bool warnillegalproto = ckWARN(WARN_ILLEGALPROTO);
8020 s = scan_str(s,!!PL_madskills,FALSE);
8022 Perl_croak(aTHX_ "Prototype not terminated");
8023 /* strip spaces and check for bad characters */
8024 d = SvPVX(PL_lex_stuff);
8026 for (p = d; *p; ++p) {
8030 if (warnillegalproto) {
8032 proto_after_greedy_proto = TRUE;
8033 if (!strchr("$@%*;[]&\\_+", *p)) {
8045 else if ( *p == ']' ) {
8046 in_brackets = FALSE;
8048 else if ( (*p == '@' || *p == '%') &&
8049 ( tmp < 2 || d[tmp-2] != '\\' ) &&
8051 must_be_last = TRUE;
8054 else if ( *p == '_' ) {
8055 underscore = seen_underscore = TRUE;
8062 if (proto_after_greedy_proto)
8063 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8064 "Prototype after '%c' for %"SVf" : %s",
8065 greedy_proto, SVfARG(PL_subname), d);
8067 Perl_warner(aTHX_ packWARN(WARN_ILLEGALPROTO),
8068 "Illegal character %sin prototype for %"SVf" : %s",
8069 seen_underscore ? "after '_' " : "",
8070 SVfARG(PL_subname), d);
8071 SvCUR_set(PL_lex_stuff, tmp);
8076 CURMAD('q', PL_thisopen);
8077 CURMAD('_', tmpwhite);
8078 CURMAD('=', PL_thisstuff);
8079 CURMAD('Q', PL_thisclose);
8080 NEXTVAL_NEXTTOKE.opval =
8081 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8082 PL_lex_stuff = NULL;
8085 s = SKIPSPACE2(s,tmpwhite);
8093 if (*s == ':' && s[1] != ':')
8094 PL_expect = attrful;
8095 else if (*s != '{' && key == KEY_sub) {
8097 Perl_croak(aTHX_ "Illegal declaration of anonymous subroutine");
8098 else if (*s != ';' && *s != '}')
8099 Perl_croak(aTHX_ "Illegal declaration of subroutine %"SVf, SVfARG(PL_subname));
8106 curmad('^', newSVpvs(""));
8107 CURMAD('_', tmpwhite);
8111 PL_thistoken = subtoken;
8114 NEXTVAL_NEXTTOKE.opval =
8115 (OP*)newSVOP(OP_CONST, 0, PL_lex_stuff);
8116 PL_lex_stuff = NULL;
8122 sv_setpvs(PL_subname, "__ANON__");
8124 sv_setpvs(PL_subname, "__ANON__::__ANON__");
8128 (void) force_word(PL_oldbufptr + tboffset, WORD,
8137 LOP(OP_SYSTEM,XREF);
8140 LOP(OP_SYMLINK,XTERM);
8143 LOP(OP_SYSCALL,XTERM);
8146 LOP(OP_SYSOPEN,XTERM);
8149 LOP(OP_SYSSEEK,XTERM);
8152 LOP(OP_SYSREAD,XTERM);
8155 LOP(OP_SYSWRITE,XTERM);
8159 TERM(sublex_start());
8180 LOP(OP_TRUNCATE,XTERM);
8192 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8194 pl_yylval.ival = CopLINE(PL_curcop);
8198 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8200 pl_yylval.ival = CopLINE(PL_curcop);
8204 LOP(OP_UNLINK,XTERM);
8210 LOP(OP_UNPACK,XTERM);
8213 LOP(OP_UTIME,XTERM);
8219 LOP(OP_UNSHIFT,XTERM);
8222 s = tokenize_use(1, s);
8232 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8234 pl_yylval.ival = CopLINE(PL_curcop);
8238 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_NONEXPR)
8240 pl_yylval.ival = CopLINE(PL_curcop);
8244 PL_hints |= HINT_BLOCK_SCOPE;
8251 LOP(OP_WAITPID,XTERM);
8260 ctl_l[0] = toCTRL('L');
8262 gv_fetchpvn_flags(ctl_l, 1, GV_ADD|GV_NOTQUAL, SVt_PV);
8265 /* Make sure $^L is defined */
8266 gv_fetchpvs("\f", GV_ADD|GV_NOTQUAL, SVt_PV);
8271 if (PL_expect == XOPERATOR) {
8272 if (*s == '=' && !PL_lex_allbrackets &&
8273 PL_lex_fakeeof >= LEX_FAKEEOF_ASSIGN)
8281 if (!PL_lex_allbrackets && PL_lex_fakeeof >= LEX_FAKEEOF_LOWLOGIC)
8283 pl_yylval.ival = OP_XOR;
8288 TERM(sublex_start());
8293 #pragma segment Main
8297 S_pending_ident(pTHX)
8302 /* pit holds the identifier we read and pending_ident is reset */
8303 char pit = PL_pending_ident;
8304 const STRLEN tokenbuf_len = strlen(PL_tokenbuf);
8305 /* All routes through this function want to know if there is a colon. */
8306 const char *const has_colon = (const char*) memchr (PL_tokenbuf, ':', tokenbuf_len);
8307 PL_pending_ident = 0;
8309 /* PL_realtokenstart = realtokenend = PL_bufptr - SvPVX(PL_linestr); */
8310 DEBUG_T({ PerlIO_printf(Perl_debug_log,
8311 "### Pending identifier '%s'\n", PL_tokenbuf); });
8313 /* if we're in a my(), we can't allow dynamics here.
8314 $foo'bar has already been turned into $foo::bar, so
8315 just check for colons.
8317 if it's a legal name, the OP is a PADANY.
8320 if (PL_in_my == KEY_our) { /* "our" is merely analogous to "my" */
8322 yyerror(Perl_form(aTHX_ "No package name allowed for "
8323 "variable %s in \"our\"",
8325 tmp = allocmy(PL_tokenbuf, tokenbuf_len, 0);
8329 yyerror(Perl_form(aTHX_ PL_no_myglob,
8330 PL_in_my == KEY_my ? "my" : "state", PL_tokenbuf));
8332 pl_yylval.opval = newOP(OP_PADANY, 0);
8333 pl_yylval.opval->op_targ = allocmy(PL_tokenbuf, tokenbuf_len, 0);
8339 build the ops for accesses to a my() variable.
8341 Deny my($a) or my($b) in a sort block, *if* $a or $b is
8342 then used in a comparison. This catches most, but not
8343 all cases. For instance, it catches
8344 sort { my($a); $a <=> $b }
8346 sort { my($a); $a < $b ? -1 : $a == $b ? 0 : 1; }
8347 (although why you'd do that is anyone's guess).
8352 tmp = pad_findmy(PL_tokenbuf, tokenbuf_len, 0);
8353 if (tmp != NOT_IN_PAD) {
8354 /* might be an "our" variable" */
8355 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
8356 /* build ops for a bareword */
8357 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
8358 HEK * const stashname = HvNAME_HEK(stash);
8359 SV * const sym = newSVhek(stashname);
8360 sv_catpvs(sym, "::");
8361 sv_catpvn(sym, PL_tokenbuf+1, tokenbuf_len - 1);
8362 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, sym);
8363 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8366 ? (GV_ADDMULTI | GV_ADDINEVAL)
8369 ((PL_tokenbuf[0] == '$') ? SVt_PV
8370 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8375 /* if it's a sort block and they're naming $a or $b */
8376 if (PL_last_lop_op == OP_SORT &&
8377 PL_tokenbuf[0] == '$' &&
8378 (PL_tokenbuf[1] == 'a' || PL_tokenbuf[1] == 'b')
8381 for (d = PL_in_eval ? PL_oldoldbufptr : PL_linestart;
8382 d < PL_bufend && *d != '\n';
8385 if (strnEQ(d,"<=>",3) || strnEQ(d,"cmp",3)) {
8386 Perl_croak(aTHX_ "Can't use \"my %s\" in sort comparison",
8392 pl_yylval.opval = newOP(OP_PADANY, 0);
8393 pl_yylval.opval->op_targ = tmp;
8399 Whine if they've said @foo in a doublequoted string,
8400 and @foo isn't a variable we can find in the symbol
8403 if (ckWARN(WARN_AMBIGUOUS) &&
8404 pit == '@' && PL_lex_state != LEX_NORMAL && !PL_lex_brackets) {
8405 GV *const gv = gv_fetchpvn_flags(PL_tokenbuf + 1, tokenbuf_len - 1, 0,
8407 if ((!gv || ((PL_tokenbuf[0] == '@') ? !GvAV(gv) : !GvHV(gv)))
8408 /* DO NOT warn for @- and @+ */
8409 && !( PL_tokenbuf[2] == '\0' &&
8410 ( PL_tokenbuf[1] == '-' || PL_tokenbuf[1] == '+' ))
8413 /* Downgraded from fatal to warning 20000522 mjd */
8414 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8415 "Possible unintended interpolation of %s in string",
8420 /* build ops for a bareword */
8421 pl_yylval.opval = (OP*)newSVOP(OP_CONST, 0, newSVpvn(PL_tokenbuf + 1,
8423 pl_yylval.opval->op_private = OPpCONST_ENTERED;
8424 gv_fetchpvn_flags(PL_tokenbuf+1, tokenbuf_len - 1,
8425 PL_in_eval ? (GV_ADDMULTI | GV_ADDINEVAL) : GV_ADD,
8426 ((PL_tokenbuf[0] == '$') ? SVt_PV
8427 : (PL_tokenbuf[0] == '@') ? SVt_PVAV
8433 S_checkcomma(pTHX_ const char *s, const char *name, const char *what)
8437 PERL_ARGS_ASSERT_CHECKCOMMA;
8439 if (*s == ' ' && s[1] == '(') { /* XXX gotta be a better way */
8440 if (ckWARN(WARN_SYNTAX)) {
8443 for (w = s+2; *w && level; w++) {
8451 /* the list of chars below is for end of statements or
8452 * block / parens, boolean operators (&&, ||, //) and branch
8453 * constructs (or, and, if, until, unless, while, err, for).
8454 * Not a very solid hack... */
8455 if (!*w || !strchr(";&/|})]oaiuwef!=", *w))
8456 Perl_warner(aTHX_ packWARN(WARN_SYNTAX),
8457 "%s (...) interpreted as function",name);
8460 while (s < PL_bufend && isSPACE(*s))
8464 while (s < PL_bufend && isSPACE(*s))
8466 if (isIDFIRST_lazy_if(s,UTF)) {
8467 const char * const w = s++;
8468 while (isALNUM_lazy_if(s,UTF))
8470 while (s < PL_bufend && isSPACE(*s))
8474 if (keyword(w, s - w, 0))
8477 gv = gv_fetchpvn_flags(w, s - w, 0, SVt_PVCV);
8478 if (gv && GvCVu(gv))
8480 Perl_croak(aTHX_ "No comma allowed after %s", what);
8485 /* Either returns sv, or mortalizes sv and returns a new SV*.
8486 Best used as sv=new_constant(..., sv, ...).
8487 If s, pv are NULL, calls subroutine with one argument,
8488 and type is used with error messages only. */
8491 S_new_constant(pTHX_ const char *s, STRLEN len, const char *key, STRLEN keylen,
8492 SV *sv, SV *pv, const char *type, STRLEN typelen)
8495 HV * const table = GvHV(PL_hintgv); /* ^H */
8499 const char *why1 = "", *why2 = "", *why3 = "";
8501 PERL_ARGS_ASSERT_NEW_CONSTANT;
8503 if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
8506 why2 = (const char *)
8507 (strEQ(key,"charnames")
8508 ? "(possibly a missing \"use charnames ...\")"
8510 msg = Perl_newSVpvf(aTHX_ "Constant(%s) unknown: %s",
8511 (type ? type: "undef"), why2);
8513 /* This is convoluted and evil ("goto considered harmful")
8514 * but I do not understand the intricacies of all the different
8515 * failure modes of %^H in here. The goal here is to make
8516 * the most probable error message user-friendly. --jhi */
8521 msg = Perl_newSVpvf(aTHX_ "Constant(%s): %s%s%s",
8522 (type ? type: "undef"), why1, why2, why3);
8524 yyerror(SvPVX_const(msg));
8529 /* charnames doesn't work well if there have been errors found */
8530 if (PL_error_count > 0 && strEQ(key,"charnames"))
8531 return &PL_sv_undef;
8533 cvp = hv_fetch(table, key, keylen, FALSE);
8534 if (!cvp || !SvOK(*cvp)) {
8537 why3 = "} is not defined";
8540 sv_2mortal(sv); /* Parent created it permanently */
8543 pv = newSVpvn_flags(s, len, SVs_TEMP);
8545 typesv = newSVpvn_flags(type, typelen, SVs_TEMP);
8547 typesv = &PL_sv_undef;
8549 PUSHSTACKi(PERLSI_OVERLOAD);
8561 call_sv(cv, G_SCALAR | ( PL_in_eval ? 0 : G_EVAL));
8565 /* Check the eval first */
8566 if (!PL_in_eval && SvTRUE(ERRSV)) {
8567 sv_catpvs(ERRSV, "Propagated");
8568 yyerror(SvPV_nolen_const(ERRSV)); /* Duplicates the message inside eval */
8570 res = SvREFCNT_inc_simple(sv);
8574 SvREFCNT_inc_simple_void(res);
8583 why1 = "Call to &{$^H{";
8585 why3 = "}} did not return a defined value";
8593 /* Returns a NUL terminated string, with the length of the string written to
8597 S_scan_word(pTHX_ register char *s, char *dest, STRLEN destlen, int allow_package, STRLEN *slp)
8600 register char *d = dest;
8601 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
8603 PERL_ARGS_ASSERT_SCAN_WORD;
8607 Perl_croak(aTHX_ ident_too_long);
8608 if (isALNUM(*s)) /* UTF handled below */
8610 else if (allow_package && (*s == '\'') && isIDFIRST_lazy_if(s+1,UTF)) {
8615 else if (allow_package && (s[0] == ':') && (s[1] == ':') && (s[2] != '$')) {
8619 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
8620 char *t = s + UTF8SKIP(s);
8622 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
8626 Perl_croak(aTHX_ ident_too_long);
8627 Copy(s, d, len, char);
8640 S_scan_ident(pTHX_ register char *s, register const char *send, char *dest, STRLEN destlen, I32 ck_uni)
8643 char *bracket = NULL;
8645 register char *d = dest;
8646 register char * const e = d + destlen - 3; /* two-character token, ending NUL */
8648 PERL_ARGS_ASSERT_SCAN_IDENT;
8653 while (isDIGIT(*s)) {
8655 Perl_croak(aTHX_ ident_too_long);
8662 Perl_croak(aTHX_ ident_too_long);
8663 if (isALNUM(*s)) /* UTF handled below */
8665 else if (*s == '\'' && isIDFIRST_lazy_if(s+1,UTF)) {
8670 else if (*s == ':' && s[1] == ':') {
8674 else if (UTF && UTF8_IS_START(*s) && isALNUM_utf8((U8*)s)) {
8675 char *t = s + UTF8SKIP(s);
8676 while (UTF8_IS_CONTINUED(*t) && is_utf8_mark((U8*)t))
8678 if (d + (t - s) > e)
8679 Perl_croak(aTHX_ ident_too_long);
8680 Copy(s, d, t - s, char);
8691 if (PL_lex_state != LEX_NORMAL)
8692 PL_lex_state = LEX_INTERPENDMAYBE;
8695 if (*s == '$' && s[1] &&
8696 (isALNUM_lazy_if(s+1,UTF) || s[1] == '$' || s[1] == '{' || strnEQ(s+1,"::",2)) )
8709 if (*d == '^' && *s && isCONTROLVAR(*s)) {
8714 if (isSPACE(s[-1])) {
8716 const char ch = *s++;
8717 if (!SPACE_OR_TAB(ch)) {
8723 if (isIDFIRST_lazy_if(d,UTF)) {
8727 while ((end < send && isALNUM_lazy_if(end,UTF)) || *end == ':') {
8728 end += UTF8SKIP(end);
8729 while (end < send && UTF8_IS_CONTINUED(*end) && is_utf8_mark((U8*)end))
8730 end += UTF8SKIP(end);
8732 Copy(s, d, end - s, char);
8737 while ((isALNUM(*s) || *s == ':') && d < e)
8740 Perl_croak(aTHX_ ident_too_long);
8743 while (s < send && SPACE_OR_TAB(*s))
8745 if ((*s == '[' || (*s == '{' && strNE(dest, "sub")))) {
8746 if (ckWARN(WARN_AMBIGUOUS) && keyword(dest, d - dest, 0)) {
8747 const char * const brack =
8749 ((*s == '[') ? "[...]" : "{...}");
8750 /* diag_listed_as: Ambiguous use of %c{%s[...]} resolved to %c%s[...] */
8751 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8752 "Ambiguous use of %c{%s%s} resolved to %c%s%s",
8753 funny, dest, brack, funny, dest, brack);
8756 PL_lex_brackstack[PL_lex_brackets++] = (char)(XOPERATOR | XFAKEBRACK);
8757 PL_lex_allbrackets++;
8761 /* Handle extended ${^Foo} variables
8762 * 1999-02-27 mjd-perl-patch@plover.com */
8763 else if (!isALNUM(*d) && !isPRINT(*d) /* isCTRL(d) */
8767 while (isALNUM(*s) && d < e) {
8771 Perl_croak(aTHX_ ident_too_long);
8776 if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets) {
8777 PL_lex_state = LEX_INTERPEND;
8780 if (PL_lex_state == LEX_NORMAL) {
8781 if (ckWARN(WARN_AMBIGUOUS) &&
8782 (keyword(dest, d - dest, 0)
8783 || get_cvn_flags(dest, d - dest, 0)))
8787 Perl_warner(aTHX_ packWARN(WARN_AMBIGUOUS),
8788 "Ambiguous use of %c{%s} resolved to %c%s",
8789 funny, dest, funny, dest);
8794 s = bracket; /* let the parser handle it */
8798 else if (PL_lex_state == LEX_INTERPNORMAL && !PL_lex_brackets && !intuit_more(s))
8799 PL_lex_state = LEX_INTERPEND;
8804 S_pmflag(pTHX_ const char* const valid_flags, U32 * pmfl, char** s, char* charset) {
8806 /* Adds, subtracts to/from 'pmfl' based on regex modifier flags found in
8807 * the parse starting at 's', based on the subset that are valid in this
8808 * context input to this routine in 'valid_flags'. Advances s. Returns
8809 * TRUE if the input was a valid flag, so the next char may be as well;
8810 * otherwise FALSE. 'charset' should point to a NUL upon first call on the
8811 * current regex. This routine will set it to any charset modifier found.
8812 * The caller shouldn't change it. This way, another charset modifier
8813 * encountered in the parse can be detected as an error, as we have decided
8818 if (! strchr(valid_flags, c)) {
8827 CASE_STD_PMMOD_FLAGS_PARSE_SET(pmfl);
8828 case GLOBAL_PAT_MOD: *pmfl |= PMf_GLOBAL; break;
8829 case CONTINUE_PAT_MOD: *pmfl |= PMf_CONTINUE; break;
8830 case ONCE_PAT_MOD: *pmfl |= PMf_KEEP; break;
8831 case KEEPCOPY_PAT_MOD: *pmfl |= RXf_PMf_KEEPCOPY; break;
8832 case NONDESTRUCT_PAT_MOD: *pmfl |= PMf_NONDESTRUCT; break;
8833 case LOCALE_PAT_MOD:
8835 /* In 5.14, qr//lt is legal but deprecated; the 't' means they
8836 * can't be regex modifiers.
8837 * In 5.14, s///le is legal and ambiguous. Try to disambiguate as
8838 * much as easily done. s///lei, for example, has to mean regex
8839 * modifiers if it's not an error (as does any word character
8840 * following the 'e'). Otherwise, we resolve to the backwards-
8841 * compatible, but less likely 's/// le ...', i.e. as meaning
8842 * less-than-or-equal. The reason it's not likely is that s//
8843 * returns a number for code in the field (/r returns a string, but
8844 * that wasn't added until the 5.13 series), and so '<=' should be
8845 * used for comparing, not 'le'. */
8846 if (*((*s) + 1) == 't') {
8849 else if (*((*s) + 1) == 'e' && ! isALNUM(*((*s) + 2))) {
8851 /* 'e' is valid only for substitutes, s///e. If it is not
8852 * valid in the current context, then 'm//le' must mean the
8853 * comparison operator, so use the regular deprecation message.
8855 if (! strchr(valid_flags, 'e')) {
8858 Perl_ck_warner_d(aTHX_ packWARN(WARN_AMBIGUOUS),
8859 "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");
8863 goto multiple_charsets;
8865 set_regex_charset(pmfl, REGEX_LOCALE_CHARSET);
8868 case UNICODE_PAT_MOD:
8869 /* In 5.14, qr//unless and qr//until are legal but deprecated; the
8870 * 'n' means they can't be regex modifiers */
8871 if (*((*s) + 1) == 'n') {
8875 goto multiple_charsets;
8877 set_regex_charset(pmfl, REGEX_UNICODE_CHARSET);
8880 case ASCII_RESTRICT_PAT_MOD:
8881 /* In 5.14, qr//and is legal but deprecated; the 'n' means they
8882 * can't be regex modifiers */
8883 if (*((*s) + 1) == 'n') {
8888 set_regex_charset(pmfl, REGEX_ASCII_RESTRICTED_CHARSET);
8892 /* Error if previous modifier wasn't an 'a', but if it was, see
8893 * if, and accept, a second occurrence (only) */
8895 || get_regex_charset(*pmfl)
8896 != REGEX_ASCII_RESTRICTED_CHARSET)
8898 goto multiple_charsets;
8900 set_regex_charset(pmfl, REGEX_ASCII_MORE_RESTRICTED_CHARSET);
8904 case DEPENDS_PAT_MOD:
8906 goto multiple_charsets;
8908 set_regex_charset(pmfl, REGEX_DEPENDS_CHARSET);
8917 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX),
8918 "Having no space between pattern and following word is deprecated");
8922 if (*charset != c) {
8923 yyerror(Perl_form(aTHX_ "Regexp modifiers \"/%c\" and \"/%c\" are mutually exclusive", *charset, c));
8925 else if (c == 'a') {
8926 yyerror("Regexp modifier \"/a\" may appear a maximum of twice");
8929 yyerror(Perl_form(aTHX_ "Regexp modifier \"/%c\" may not appear twice", c));
8932 /* Pretend that it worked, so will continue processing before dieing */
8938 S_scan_pat(pTHX_ char *start, I32 type)
8942 char *s = scan_str(start,!!PL_madskills,FALSE);
8943 const char * const valid_flags =
8944 (const char *)((type == OP_QR) ? QR_PAT_MODS : M_PAT_MODS);
8945 char charset = '\0'; /* character set modifier */
8950 PERL_ARGS_ASSERT_SCAN_PAT;
8953 const char * const delimiter = skipspace(start);
8957 ? "Search pattern not terminated or ternary operator parsed as search pattern"
8958 : "Search pattern not terminated" ));
8961 pm = (PMOP*)newPMOP(type, 0);
8962 if (PL_multi_open == '?') {
8963 /* This is the only point in the code that sets PMf_ONCE: */
8964 pm->op_pmflags |= PMf_ONCE;
8966 /* Hence it's safe to do this bit of PMOP book-keeping here, which
8967 allows us to restrict the list needed by reset to just the ??
8969 assert(type != OP_TRANS);
8971 MAGIC *mg = mg_find((const SV *)PL_curstash, PERL_MAGIC_symtab);
8974 mg = sv_magicext(MUTABLE_SV(PL_curstash), 0, PERL_MAGIC_symtab, 0, 0,
8977 elements = mg->mg_len / sizeof(PMOP**);
8978 Renewc(mg->mg_ptr, elements + 1, PMOP*, char);
8979 ((PMOP**)mg->mg_ptr) [elements++] = pm;
8980 mg->mg_len = elements * sizeof(PMOP**);
8981 PmopSTASH_set(pm,PL_curstash);
8987 while (*s && S_pmflag(aTHX_ valid_flags, &(pm->op_pmflags), &s, &charset)) {};
8989 if (PL_madskills && modstart != s) {
8990 SV* tmptoken = newSVpvn(modstart, s - modstart);
8991 append_madprops(newMADPROP('m', MAD_SV, tmptoken, 0), (OP*)pm, 0);
8994 /* issue a warning if /c is specified,but /g is not */
8995 if ((pm->op_pmflags & PMf_CONTINUE) && !(pm->op_pmflags & PMf_GLOBAL))
8997 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP),
8998 "Use of /c modifier is meaningless without /g" );
9001 PL_lex_op = (OP*)pm;
9002 pl_yylval.ival = OP_MATCH;
9007 S_scan_subst(pTHX_ char *start)
9014 char charset = '\0'; /* character set modifier */
9019 PERL_ARGS_ASSERT_SCAN_SUBST;
9021 pl_yylval.ival = OP_NULL;
9023 s = scan_str(start,!!PL_madskills,FALSE);
9026 Perl_croak(aTHX_ "Substitution pattern not terminated");
9028 if (s[-1] == PL_multi_open)
9032 CURMAD('q', PL_thisopen);
9033 CURMAD('_', PL_thiswhite);
9034 CURMAD('E', PL_thisstuff);
9035 CURMAD('Q', PL_thisclose);
9036 PL_realtokenstart = s - SvPVX(PL_linestr);
9040 first_start = PL_multi_start;
9041 s = scan_str(s,!!PL_madskills,FALSE);
9044 SvREFCNT_dec(PL_lex_stuff);
9045 PL_lex_stuff = NULL;
9047 Perl_croak(aTHX_ "Substitution replacement not terminated");
9049 PL_multi_start = first_start; /* so whole substitution is taken together */
9051 pm = (PMOP*)newPMOP(OP_SUBST, 0);
9055 CURMAD('z', PL_thisopen);
9056 CURMAD('R', PL_thisstuff);
9057 CURMAD('Z', PL_thisclose);
9063 if (*s == EXEC_PAT_MOD) {
9067 else if (! S_pmflag(aTHX_ S_PAT_MODS, &(pm->op_pmflags), &s, &charset))
9076 curmad('m', newSVpvn(modstart, s - modstart));
9077 append_madprops(PL_thismad, (OP*)pm, 0);
9081 if ((pm->op_pmflags & PMf_CONTINUE)) {
9082 Perl_ck_warner(aTHX_ packWARN(WARN_REGEXP), "Use of /c modifier is meaningless in s///" );
9086 SV * const repl = newSVpvs("");
9088 PL_sublex_info.super_bufptr = s;
9089 PL_sublex_info.super_bufend = PL_bufend;
9091 pm->op_pmflags |= PMf_EVAL;
9094 sv_catpvs(repl, "eval ");
9096 sv_catpvs(repl, "do ");
9098 sv_catpvs(repl, "{");
9099 sv_catsv(repl, PL_lex_repl);
9100 if (strchr(SvPVX(PL_lex_repl), '#'))
9101 sv_catpvs(repl, "\n");
9102 sv_catpvs(repl, "}");
9104 SvREFCNT_dec(PL_lex_repl);
9108 PL_lex_op = (OP*)pm;
9109 pl_yylval.ival = OP_SUBST;
9114 S_scan_trans(pTHX_ char *start)
9123 bool nondestruct = 0;
9128 PERL_ARGS_ASSERT_SCAN_TRANS;
9130 pl_yylval.ival = OP_NULL;
9132 s = scan_str(start,!!PL_madskills,FALSE);
9134 Perl_croak(aTHX_ "Transliteration pattern not terminated");
9136 if (s[-1] == PL_multi_open)
9140 CURMAD('q', PL_thisopen);
9141 CURMAD('_', PL_thiswhite);
9142 CURMAD('E', PL_thisstuff);
9143 CURMAD('Q', PL_thisclose);
9144 PL_realtokenstart = s - SvPVX(PL_linestr);
9148 s = scan_str(s,!!PL_madskills,FALSE);
9151 SvREFCNT_dec(PL_lex_stuff);
9152 PL_lex_stuff = NULL;
9154 Perl_croak(aTHX_ "Transliteration replacement not terminated");
9157 CURMAD('z', PL_thisopen);
9158 CURMAD('R', PL_thisstuff);
9159 CURMAD('Z', PL_thisclose);
9162 complement = del = squash = 0;
9169 complement = OPpTRANS_COMPLEMENT;
9172 del = OPpTRANS_DELETE;
9175 squash = OPpTRANS_SQUASH;
9187 tbl = (short *)PerlMemShared_calloc(complement&&!del?258:256, sizeof(short));
9188 o = newPVOP(nondestruct ? OP_TRANSR : OP_TRANS, 0, (char*)tbl);
9189 o->op_private &= ~OPpTRANS_ALL;
9190 o->op_private |= del|squash|complement|
9191 (DO_UTF8(PL_lex_stuff)? OPpTRANS_FROM_UTF : 0)|
9192 (DO_UTF8(PL_lex_repl) ? OPpTRANS_TO_UTF : 0);
9195 pl_yylval.ival = nondestruct ? OP_TRANSR : OP_TRANS;
9200 curmad('m', newSVpvn(modstart, s - modstart));
9201 append_madprops(PL_thismad, o, 0);
9210 S_scan_heredoc(pTHX_ register char *s)
9214 I32 op_type = OP_SCALAR;
9218 const char *found_newline;
9222 const int outer = (PL_rsfp && !(PL_lex_inwhat == OP_SCALAR));
9224 I32 stuffstart = s - SvPVX(PL_linestr);
9227 PL_realtokenstart = -1;
9230 PERL_ARGS_ASSERT_SCAN_HEREDOC;
9234 e = PL_tokenbuf + sizeof PL_tokenbuf - 1;
9238 while (SPACE_OR_TAB(*peek))
9240 if (*peek == '`' || *peek == '\'' || *peek =='"') {
9243 s = delimcpy(d, e, s, PL_bufend, term, &len);
9253 if (!isALNUM_lazy_if(s,UTF))
9254 deprecate("bare << to mean <<\"\"");
9255 for (; isALNUM_lazy_if(s,UTF); s++) {
9260 if (d >= PL_tokenbuf + sizeof PL_tokenbuf - 1)
9261 Perl_croak(aTHX_ "Delimiter for here document is too long");
9264 len = d - PL_tokenbuf;
9268 tstart = PL_tokenbuf + !outer;
9269 PL_thisclose = newSVpvn(tstart, len - !outer);
9270 tstart = SvPVX(PL_linestr) + stuffstart;
9271 PL_thisopen = newSVpvn(tstart, s - tstart);
9272 stuffstart = s - SvPVX(PL_linestr);
9275 #ifndef PERL_STRICT_CR
9276 d = strchr(s, '\r');
9278 char * const olds = s;
9280 while (s < PL_bufend) {
9286 else if (*s == '\n' && s[1] == '\r') { /* \015\013 on a mac? */
9295 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9302 if ( outer || !(found_newline = (char*)memchr((void*)s, '\n', PL_bufend - s)) ) {
9303 herewas = newSVpvn(s,PL_bufend-s);
9307 herewas = newSVpvn(s-1,found_newline-s+1);
9310 herewas = newSVpvn(s,found_newline-s);
9315 tstart = SvPVX(PL_linestr) + stuffstart;
9317 sv_catpvn(PL_thisstuff, tstart, s - tstart);
9319 PL_thisstuff = newSVpvn(tstart, s - tstart);
9322 s += SvCUR(herewas);
9325 stuffstart = s - SvPVX(PL_linestr);
9331 tmpstr = newSV_type(SVt_PVIV);
9335 SvIV_set(tmpstr, -1);
9337 else if (term == '`') {
9338 op_type = OP_BACKTICK;
9339 SvIV_set(tmpstr, '\\');
9343 PL_multi_start = CopLINE(PL_curcop);
9344 PL_multi_open = PL_multi_close = '<';
9345 term = *PL_tokenbuf;
9346 if (PL_lex_inwhat == OP_SUBST && PL_in_eval && !PL_rsfp) {
9347 char * const bufptr = PL_sublex_info.super_bufptr;
9348 char * const bufend = PL_sublex_info.super_bufend;
9349 char * const olds = s - SvCUR(herewas);
9350 s = strchr(bufptr, '\n');
9354 while (s < bufend &&
9355 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9357 CopLINE_inc(PL_curcop);
9360 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9361 missingterm(PL_tokenbuf);
9363 sv_setpvn(herewas,bufptr,d-bufptr+1);
9364 sv_setpvn(tmpstr,d+1,s-d);
9366 sv_catpvn(herewas,s,bufend-s);
9367 Copy(SvPVX_const(herewas),bufptr,SvCUR(herewas) + 1,char);
9374 while (s < PL_bufend &&
9375 (*s != term || memNE(s,PL_tokenbuf,len)) ) {
9377 CopLINE_inc(PL_curcop);
9379 if (s >= PL_bufend) {
9380 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9381 missingterm(PL_tokenbuf);
9383 sv_setpvn(tmpstr,d+1,s-d);
9387 sv_catpvn(PL_thisstuff, d + 1, s - d);
9389 PL_thisstuff = newSVpvn(d + 1, s - d);
9390 stuffstart = s - SvPVX(PL_linestr);
9394 CopLINE_inc(PL_curcop); /* the preceding stmt passes a newline */
9396 sv_catpvn(herewas,s,PL_bufend-s);
9397 sv_setsv(PL_linestr,herewas);
9398 PL_oldoldbufptr = PL_oldbufptr = PL_bufptr = s = PL_linestart = SvPVX(PL_linestr);
9399 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9400 PL_last_lop = PL_last_uni = NULL;
9403 sv_setpvs(tmpstr,""); /* avoid "uninitialized" warning */
9404 while (s >= PL_bufend) { /* multiple line string? */
9407 tstart = SvPVX(PL_linestr) + stuffstart;
9409 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
9411 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
9415 CopLINE_inc(PL_curcop);
9416 if (!outer || !lex_next_chunk(0)) {
9417 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9418 missingterm(PL_tokenbuf);
9420 CopLINE_dec(PL_curcop);
9423 stuffstart = s - SvPVX(PL_linestr);
9425 CopLINE_inc(PL_curcop);
9426 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9427 PL_last_lop = PL_last_uni = NULL;
9428 #ifndef PERL_STRICT_CR
9429 if (PL_bufend - PL_linestart >= 2) {
9430 if ((PL_bufend[-2] == '\r' && PL_bufend[-1] == '\n') ||
9431 (PL_bufend[-2] == '\n' && PL_bufend[-1] == '\r'))
9433 PL_bufend[-2] = '\n';
9435 SvCUR_set(PL_linestr, PL_bufend - SvPVX_const(PL_linestr));
9437 else if (PL_bufend[-1] == '\r')
9438 PL_bufend[-1] = '\n';
9440 else if (PL_bufend - PL_linestart == 1 && PL_bufend[-1] == '\r')
9441 PL_bufend[-1] = '\n';
9443 if (*s == term && memEQ(s,PL_tokenbuf,len)) {
9444 STRLEN off = PL_bufend - 1 - SvPVX_const(PL_linestr);
9445 *(SvPVX(PL_linestr) + off ) = ' ';
9446 lex_grow_linestr(SvCUR(PL_linestr) + SvCUR(herewas) + 1);
9447 sv_catsv(PL_linestr,herewas);
9448 PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
9449 s = SvPVX(PL_linestr) + off; /* In case PV of PL_linestr moved. */
9453 sv_catsv(tmpstr,PL_linestr);
9458 PL_multi_end = CopLINE(PL_curcop);
9459 if (SvCUR(tmpstr) + 5 < SvLEN(tmpstr)) {
9460 SvPV_shrink_to_cur(tmpstr);
9462 SvREFCNT_dec(herewas);
9464 if (UTF && is_utf8_string((U8*)SvPVX_const(tmpstr), SvCUR(tmpstr)))
9466 else if (PL_encoding)
9467 sv_recode_to_utf8(tmpstr, PL_encoding);
9469 PL_lex_stuff = tmpstr;
9470 pl_yylval.ival = op_type;
9475 takes: current position in input buffer
9476 returns: new position in input buffer
9477 side-effects: pl_yylval and lex_op are set.
9482 <FH> read from filehandle
9483 <pkg::FH> read from package qualified filehandle
9484 <pkg'FH> read from package qualified filehandle
9485 <$fh> read from filehandle in $fh
9491 S_scan_inputsymbol(pTHX_ char *start)
9494 register char *s = start; /* current position in buffer */
9497 char *d = PL_tokenbuf; /* start of temp holding space */
9498 const char * const e = PL_tokenbuf + sizeof PL_tokenbuf; /* end of temp holding space */
9500 PERL_ARGS_ASSERT_SCAN_INPUTSYMBOL;
9502 end = strchr(s, '\n');
9505 s = delimcpy(d, e, s + 1, end, '>', &len); /* extract until > */
9507 /* die if we didn't have space for the contents of the <>,
9508 or if it didn't end, or if we see a newline
9511 if (len >= (I32)sizeof PL_tokenbuf)
9512 Perl_croak(aTHX_ "Excessively long <> operator");
9514 Perl_croak(aTHX_ "Unterminated <> operator");
9519 Remember, only scalar variables are interpreted as filehandles by
9520 this code. Anything more complex (e.g., <$fh{$num}>) will be
9521 treated as a glob() call.
9522 This code makes use of the fact that except for the $ at the front,
9523 a scalar variable and a filehandle look the same.
9525 if (*d == '$' && d[1]) d++;
9527 /* allow <Pkg'VALUE> or <Pkg::VALUE> */
9528 while (*d && (isALNUM_lazy_if(d,UTF) || *d == '\'' || *d == ':'))
9531 /* If we've tried to read what we allow filehandles to look like, and
9532 there's still text left, then it must be a glob() and not a getline.
9533 Use scan_str to pull out the stuff between the <> and treat it
9534 as nothing more than a string.
9537 if (d - PL_tokenbuf != len) {
9538 pl_yylval.ival = OP_GLOB;
9539 s = scan_str(start,!!PL_madskills,FALSE);
9541 Perl_croak(aTHX_ "Glob not terminated");
9545 bool readline_overriden = FALSE;
9548 /* we're in a filehandle read situation */
9551 /* turn <> into <ARGV> */
9553 Copy("ARGV",d,5,char);
9555 /* Check whether readline() is overriden */
9556 gv_readline = gv_fetchpvs("readline", GV_NOTQUAL, SVt_PVCV);
9558 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline))
9560 ((gvp = (GV**)hv_fetchs(PL_globalstash, "readline", FALSE))
9561 && (gv_readline = *gvp) && isGV_with_GP(gv_readline)
9562 && GvCVu(gv_readline) && GvIMPORTED_CV(gv_readline)))
9563 readline_overriden = TRUE;
9565 /* if <$fh>, create the ops to turn the variable into a
9569 /* try to find it in the pad for this block, otherwise find
9570 add symbol table ops
9572 const PADOFFSET tmp = pad_findmy(d, len, 0);
9573 if (tmp != NOT_IN_PAD) {
9574 if (PAD_COMPNAME_FLAGS_isOUR(tmp)) {
9575 HV * const stash = PAD_COMPNAME_OURSTASH(tmp);
9576 HEK * const stashname = HvNAME_HEK(stash);
9577 SV * const sym = sv_2mortal(newSVhek(stashname));
9578 sv_catpvs(sym, "::");
9584 OP * const o = newOP(OP_PADSV, 0);
9586 PL_lex_op = readline_overriden
9587 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9588 op_append_elem(OP_LIST, o,
9589 newCVREF(0, newGVOP(OP_GV,0,gv_readline))))
9590 : (OP*)newUNOP(OP_READLINE, 0, o);
9599 ? (GV_ADDMULTI | GV_ADDINEVAL)
9602 PL_lex_op = readline_overriden
9603 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9604 op_append_elem(OP_LIST,
9605 newUNOP(OP_RV2SV, 0, newGVOP(OP_GV, 0, gv)),
9606 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9607 : (OP*)newUNOP(OP_READLINE, 0,
9608 newUNOP(OP_RV2SV, 0,
9609 newGVOP(OP_GV, 0, gv)));
9611 if (!readline_overriden)
9612 PL_lex_op->op_flags |= OPf_SPECIAL;
9613 /* we created the ops in PL_lex_op, so make pl_yylval.ival a null op */
9614 pl_yylval.ival = OP_NULL;
9617 /* If it's none of the above, it must be a literal filehandle
9618 (<Foo::BAR> or <FOO>) so build a simple readline OP */
9620 GV * const gv = gv_fetchpv(d, GV_ADD, SVt_PVIO);
9621 PL_lex_op = readline_overriden
9622 ? (OP*)newUNOP(OP_ENTERSUB, OPf_STACKED,
9623 op_append_elem(OP_LIST,
9624 newGVOP(OP_GV, 0, gv),
9625 newCVREF(0, newGVOP(OP_GV, 0, gv_readline))))
9626 : (OP*)newUNOP(OP_READLINE, 0, newGVOP(OP_GV, 0, gv));
9627 pl_yylval.ival = OP_NULL;
9636 takes: start position in buffer
9637 keep_quoted preserve \ on the embedded delimiter(s)
9638 keep_delims preserve the delimiters around the string
9639 returns: position to continue reading from buffer
9640 side-effects: multi_start, multi_close, lex_repl or lex_stuff, and
9641 updates the read buffer.
9643 This subroutine pulls a string out of the input. It is called for:
9644 q single quotes q(literal text)
9645 ' single quotes 'literal text'
9646 qq double quotes qq(interpolate $here please)
9647 " double quotes "interpolate $here please"
9648 qx backticks qx(/bin/ls -l)
9649 ` backticks `/bin/ls -l`
9650 qw quote words @EXPORT_OK = qw( func() $spam )
9651 m// regexp match m/this/
9652 s/// regexp substitute s/this/that/
9653 tr/// string transliterate tr/this/that/
9654 y/// string transliterate y/this/that/
9655 ($*@) sub prototypes sub foo ($)
9656 (stuff) sub attr parameters sub foo : attr(stuff)
9657 <> readline or globs <FOO>, <>, <$fh>, or <*.c>
9659 In most of these cases (all but <>, patterns and transliterate)
9660 yylex() calls scan_str(). m// makes yylex() call scan_pat() which
9661 calls scan_str(). s/// makes yylex() call scan_subst() which calls
9662 scan_str(). tr/// and y/// make yylex() call scan_trans() which
9665 It skips whitespace before the string starts, and treats the first
9666 character as the delimiter. If the delimiter is one of ([{< then
9667 the corresponding "close" character )]}> is used as the closing
9668 delimiter. It allows quoting of delimiters, and if the string has
9669 balanced delimiters ([{<>}]) it allows nesting.
9671 On success, the SV with the resulting string is put into lex_stuff or,
9672 if that is already non-NULL, into lex_repl. The second case occurs only
9673 when parsing the RHS of the special constructs s/// and tr/// (y///).
9674 For convenience, the terminating delimiter character is stuffed into
9679 S_scan_str(pTHX_ char *start, int keep_quoted, int keep_delims)
9682 SV *sv; /* scalar value: string */
9683 const char *tmps; /* temp string, used for delimiter matching */
9684 register char *s = start; /* current position in the buffer */
9685 register char term; /* terminating character */
9686 register char *to; /* current position in the sv's data */
9687 I32 brackets = 1; /* bracket nesting level */
9688 bool has_utf8 = FALSE; /* is there any utf8 content? */
9689 I32 termcode; /* terminating char. code */
9690 U8 termstr[UTF8_MAXBYTES]; /* terminating string */
9691 STRLEN termlen; /* length of terminating string */
9692 int last_off = 0; /* last position for nesting bracket */
9698 PERL_ARGS_ASSERT_SCAN_STR;
9700 /* skip space before the delimiter */
9706 if (PL_realtokenstart >= 0) {
9707 stuffstart = PL_realtokenstart;
9708 PL_realtokenstart = -1;
9711 stuffstart = start - SvPVX(PL_linestr);
9713 /* mark where we are, in case we need to report errors */
9716 /* after skipping whitespace, the next character is the terminator */
9719 termcode = termstr[0] = term;
9723 termcode = utf8_to_uvchr((U8*)s, &termlen);
9724 Copy(s, termstr, termlen, U8);
9725 if (!UTF8_IS_INVARIANT(term))
9729 /* mark where we are */
9730 PL_multi_start = CopLINE(PL_curcop);
9731 PL_multi_open = term;
9733 /* find corresponding closing delimiter */
9734 if (term && (tmps = strchr("([{< )]}> )]}>",term)))
9735 termcode = termstr[0] = term = tmps[5];
9737 PL_multi_close = term;
9739 /* create a new SV to hold the contents. 79 is the SV's initial length.
9740 What a random number. */
9741 sv = newSV_type(SVt_PVIV);
9743 SvIV_set(sv, termcode);
9744 (void)SvPOK_only(sv); /* validate pointer */
9746 /* move past delimiter and try to read a complete string */
9748 sv_catpvn(sv, s, termlen);
9751 tstart = SvPVX(PL_linestr) + stuffstart;
9752 if (!PL_thisopen && !keep_delims) {
9753 PL_thisopen = newSVpvn(tstart, s - tstart);
9754 stuffstart = s - SvPVX(PL_linestr);
9758 if (PL_encoding && !UTF) {
9762 int offset = s - SvPVX_const(PL_linestr);
9763 const bool found = sv_cat_decode(sv, PL_encoding, PL_linestr,
9764 &offset, (char*)termstr, termlen);
9765 const char * const ns = SvPVX_const(PL_linestr) + offset;
9766 char * const svlast = SvEND(sv) - 1;
9768 for (; s < ns; s++) {
9769 if (*s == '\n' && !PL_rsfp)
9770 CopLINE_inc(PL_curcop);
9773 goto read_more_line;
9775 /* handle quoted delimiters */
9776 if (SvCUR(sv) > 1 && *(svlast-1) == '\\') {
9778 for (t = svlast-2; t >= SvPVX_const(sv) && *t == '\\';)
9780 if ((svlast-1 - t) % 2) {
9784 SvCUR_set(sv, SvCUR(sv) - 1);
9789 if (PL_multi_open == PL_multi_close) {
9795 for (t = w = SvPVX(sv)+last_off; t < svlast; w++, t++) {
9796 /* At here, all closes are "was quoted" one,
9797 so we don't check PL_multi_close. */
9799 if (!keep_quoted && *(t+1) == PL_multi_open)
9804 else if (*t == PL_multi_open)
9812 SvCUR_set(sv, w - SvPVX_const(sv));
9814 last_off = w - SvPVX(sv);
9815 if (--brackets <= 0)
9821 SvCUR_set(sv, SvCUR(sv) - 1);
9827 /* extend sv if need be */
9828 SvGROW(sv, SvCUR(sv) + (PL_bufend - s) + 1);
9829 /* set 'to' to the next character in the sv's string */
9830 to = SvPVX(sv)+SvCUR(sv);
9832 /* if open delimiter is the close delimiter read unbridle */
9833 if (PL_multi_open == PL_multi_close) {
9834 for (; s < PL_bufend; s++,to++) {
9835 /* embedded newlines increment the current line number */
9836 if (*s == '\n' && !PL_rsfp)
9837 CopLINE_inc(PL_curcop);
9838 /* handle quoted delimiters */
9839 if (*s == '\\' && s+1 < PL_bufend && term != '\\') {
9840 if (!keep_quoted && s[1] == term)
9842 /* any other quotes are simply copied straight through */
9846 /* terminate when run out of buffer (the for() condition), or
9847 have found the terminator */
9848 else if (*s == term) {
9851 if (s+termlen <= PL_bufend && memEQ(s, (char*)termstr, termlen))
9854 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9860 /* if the terminator isn't the same as the start character (e.g.,
9861 matched brackets), we have to allow more in the quoting, and
9862 be prepared for nested brackets.
9865 /* read until we run out of string, or we find the terminator */
9866 for (; s < PL_bufend; s++,to++) {
9867 /* embedded newlines increment the line count */
9868 if (*s == '\n' && !PL_rsfp)
9869 CopLINE_inc(PL_curcop);
9870 /* backslashes can escape the open or closing characters */
9871 if (*s == '\\' && s+1 < PL_bufend) {
9873 ((s[1] == PL_multi_open) || (s[1] == PL_multi_close)))
9878 /* allow nested opens and closes */
9879 else if (*s == PL_multi_close && --brackets <= 0)
9881 else if (*s == PL_multi_open)
9883 else if (!has_utf8 && !UTF8_IS_INVARIANT((U8)*s) && UTF)
9888 /* terminate the copied string and update the sv's end-of-string */
9890 SvCUR_set(sv, to - SvPVX_const(sv));
9893 * this next chunk reads more into the buffer if we're not done yet
9897 break; /* handle case where we are done yet :-) */
9899 #ifndef PERL_STRICT_CR
9900 if (to - SvPVX_const(sv) >= 2) {
9901 if ((to[-2] == '\r' && to[-1] == '\n') ||
9902 (to[-2] == '\n' && to[-1] == '\r'))
9906 SvCUR_set(sv, to - SvPVX_const(sv));
9908 else if (to[-1] == '\r')
9911 else if (to - SvPVX_const(sv) == 1 && to[-1] == '\r')
9916 /* if we're out of file, or a read fails, bail and reset the current
9917 line marker so we can report where the unterminated string began
9921 char * const tstart = SvPVX(PL_linestr) + stuffstart;
9923 sv_catpvn(PL_thisstuff, tstart, PL_bufend - tstart);
9925 PL_thisstuff = newSVpvn(tstart, PL_bufend - tstart);
9928 CopLINE_inc(PL_curcop);
9929 PL_bufptr = PL_bufend;
9930 if (!lex_next_chunk(0)) {
9932 CopLINE_set(PL_curcop, (line_t)PL_multi_start);
9941 /* at this point, we have successfully read the delimited string */
9943 if (!PL_encoding || UTF) {
9946 char * const tstart = SvPVX(PL_linestr) + stuffstart;
9947 const int len = s - tstart;
9949 sv_catpvn(PL_thisstuff, tstart, len);
9951 PL_thisstuff = newSVpvn(tstart, len);
9952 if (!PL_thisclose && !keep_delims)
9953 PL_thisclose = newSVpvn(s,termlen);
9958 sv_catpvn(sv, s, termlen);
9964 char * const tstart = SvPVX(PL_linestr) + stuffstart;
9965 const int len = s - tstart - termlen;
9967 sv_catpvn(PL_thisstuff, tstart, len);
9969 PL_thisstuff = newSVpvn(tstart, len);
9970 if (!PL_thisclose && !keep_delims)
9971 PL_thisclose = newSVpvn(s - termlen,termlen);
9975 if (has_utf8 || PL_encoding)
9978 PL_multi_end = CopLINE(PL_curcop);
9980 /* if we allocated too much space, give some back */
9981 if (SvCUR(sv) + 5 < SvLEN(sv)) {
9982 SvLEN_set(sv, SvCUR(sv) + 1);
9983 SvPV_renew(sv, SvLEN(sv));
9986 /* decide whether this is the first or second quoted string we've read
9999 takes: pointer to position in buffer
10000 returns: pointer to new position in buffer
10001 side-effects: builds ops for the constant in pl_yylval.op
10003 Read a number in any of the formats that Perl accepts:
10005 \d(_?\d)*(\.(\d(_?\d)*)?)?[Ee][\+\-]?(\d(_?\d)*) 12 12.34 12.
10006 \.\d(_?\d)*[Ee][\+\-]?(\d(_?\d)*) .34
10009 0x[0-9A-Fa-f](_?[0-9A-Fa-f])*
10011 Like most scan_ routines, it uses the PL_tokenbuf buffer to hold the
10014 If it reads a number without a decimal point or an exponent, it will
10015 try converting the number to an integer and see if it can do so
10016 without loss of precision.
10020 Perl_scan_num(pTHX_ const char *start, YYSTYPE* lvalp)
10023 register const char *s = start; /* current position in buffer */
10024 register char *d; /* destination in temp buffer */
10025 register char *e; /* end of temp buffer */
10026 NV nv; /* number read, as a double */
10027 SV *sv = NULL; /* place to put the converted number */
10028 bool floatit; /* boolean: int or float? */
10029 const char *lastub = NULL; /* position of last underbar */
10030 static char const number_too_long[] = "Number too long";
10032 PERL_ARGS_ASSERT_SCAN_NUM;
10034 /* We use the first character to decide what type of number this is */
10038 Perl_croak(aTHX_ "panic: scan_num");
10040 /* if it starts with a 0, it could be an octal number, a decimal in
10041 0.13 disguise, or a hexadecimal number, or a binary number. */
10045 u holds the "number so far"
10046 shift the power of 2 of the base
10047 (hex == 4, octal == 3, binary == 1)
10048 overflowed was the number more than we can hold?
10050 Shift is used when we add a digit. It also serves as an "are
10051 we in octal/hex/binary?" indicator to disallow hex characters
10052 when in octal mode.
10057 bool overflowed = FALSE;
10058 bool just_zero = TRUE; /* just plain 0 or binary number? */
10059 static const NV nvshift[5] = { 1.0, 2.0, 4.0, 8.0, 16.0 };
10060 static const char* const bases[5] =
10061 { "", "binary", "", "octal", "hexadecimal" };
10062 static const char* const Bases[5] =
10063 { "", "Binary", "", "Octal", "Hexadecimal" };
10064 static const char* const maxima[5] =
10066 "0b11111111111111111111111111111111",
10070 const char *base, *Base, *max;
10072 /* check for hex */
10073 if (s[1] == 'x' || s[1] == 'X') {
10077 } else if (s[1] == 'b' || s[1] == 'B') {
10082 /* check for a decimal in disguise */
10083 else if (s[1] == '.' || s[1] == 'e' || s[1] == 'E')
10085 /* so it must be octal */
10092 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10093 "Misplaced _ in number");
10097 base = bases[shift];
10098 Base = Bases[shift];
10099 max = maxima[shift];
10101 /* read the rest of the number */
10103 /* x is used in the overflow test,
10104 b is the digit we're adding on. */
10109 /* if we don't mention it, we're done */
10113 /* _ are ignored -- but warned about if consecutive */
10115 if (lastub && s == lastub + 1)
10116 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10117 "Misplaced _ in number");
10121 /* 8 and 9 are not octal */
10122 case '8': case '9':
10124 yyerror(Perl_form(aTHX_ "Illegal octal digit '%c'", *s));
10128 case '2': case '3': case '4':
10129 case '5': case '6': case '7':
10131 yyerror(Perl_form(aTHX_ "Illegal binary digit '%c'", *s));
10134 case '0': case '1':
10135 b = *s++ & 15; /* ASCII digit -> value of digit */
10139 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
10140 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
10141 /* make sure they said 0x */
10144 b = (*s++ & 7) + 9;
10146 /* Prepare to put the digit we have onto the end
10147 of the number so far. We check for overflows.
10153 x = u << shift; /* make room for the digit */
10155 if ((x >> shift) != u
10156 && !(PL_hints & HINT_NEW_BINARY)) {
10159 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10160 "Integer overflow in %s number",
10163 u = x | b; /* add the digit to the end */
10166 n *= nvshift[shift];
10167 /* If an NV has not enough bits in its
10168 * mantissa to represent an UV this summing of
10169 * small low-order numbers is a waste of time
10170 * (because the NV cannot preserve the
10171 * low-order bits anyway): we could just
10172 * remember when did we overflow and in the
10173 * end just multiply n by the right
10181 /* if we get here, we had success: make a scalar value from
10186 /* final misplaced underbar check */
10187 if (s[-1] == '_') {
10188 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10192 if (n > 4294967295.0)
10193 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10194 "%s number > %s non-portable",
10200 if (u > 0xffffffff)
10201 Perl_ck_warner(aTHX_ packWARN(WARN_PORTABLE),
10202 "%s number > %s non-portable",
10207 if (just_zero && (PL_hints & HINT_NEW_INTEGER))
10208 sv = new_constant(start, s - start, "integer",
10209 sv, NULL, NULL, 0);
10210 else if (PL_hints & HINT_NEW_BINARY)
10211 sv = new_constant(start, s - start, "binary", sv, NULL, NULL, 0);
10216 handle decimal numbers.
10217 we're also sent here when we read a 0 as the first digit
10219 case '1': case '2': case '3': case '4': case '5':
10220 case '6': case '7': case '8': case '9': case '.':
10223 e = PL_tokenbuf + sizeof PL_tokenbuf - 6; /* room for various punctuation */
10226 /* read next group of digits and _ and copy into d */
10227 while (isDIGIT(*s) || *s == '_') {
10228 /* skip underscores, checking for misplaced ones
10232 if (lastub && s == lastub + 1)
10233 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10234 "Misplaced _ in number");
10238 /* check for end of fixed-length buffer */
10240 Perl_croak(aTHX_ number_too_long);
10241 /* if we're ok, copy the character */
10246 /* final misplaced underbar check */
10247 if (lastub && s == lastub + 1) {
10248 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX), "Misplaced _ in number");
10251 /* read a decimal portion if there is one. avoid
10252 3..5 being interpreted as the number 3. followed
10255 if (*s == '.' && s[1] != '.') {
10260 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10261 "Misplaced _ in number");
10265 /* copy, ignoring underbars, until we run out of digits.
10267 for (; isDIGIT(*s) || *s == '_'; s++) {
10268 /* fixed length buffer check */
10270 Perl_croak(aTHX_ number_too_long);
10272 if (lastub && s == lastub + 1)
10273 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10274 "Misplaced _ in number");
10280 /* fractional part ending in underbar? */
10281 if (s[-1] == '_') {
10282 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10283 "Misplaced _ in number");
10285 if (*s == '.' && isDIGIT(s[1])) {
10286 /* oops, it's really a v-string, but without the "v" */
10292 /* read exponent part, if present */
10293 if ((*s == 'e' || *s == 'E') && strchr("+-0123456789_", s[1])) {
10297 /* regardless of whether user said 3E5 or 3e5, use lower 'e' */
10298 *d++ = 'e'; /* At least some Mach atof()s don't grok 'E' */
10300 /* stray preinitial _ */
10302 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10303 "Misplaced _ in number");
10307 /* allow positive or negative exponent */
10308 if (*s == '+' || *s == '-')
10311 /* stray initial _ */
10313 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10314 "Misplaced _ in number");
10318 /* read digits of exponent */
10319 while (isDIGIT(*s) || *s == '_') {
10322 Perl_croak(aTHX_ number_too_long);
10326 if (((lastub && s == lastub + 1) ||
10327 (!isDIGIT(s[1]) && s[1] != '_')))
10328 Perl_ck_warner(aTHX_ packWARN(WARN_SYNTAX),
10329 "Misplaced _ in number");
10337 We try to do an integer conversion first if no characters
10338 indicating "float" have been found.
10343 const int flags = grok_number (PL_tokenbuf, d - PL_tokenbuf, &uv);
10345 if (flags == IS_NUMBER_IN_UV) {
10347 sv = newSViv(uv); /* Prefer IVs over UVs. */
10350 } else if (flags == (IS_NUMBER_IN_UV | IS_NUMBER_NEG)) {
10351 if (uv <= (UV) IV_MIN)
10352 sv = newSViv(-(IV)uv);
10359 /* terminate the string */
10361 nv = Atof(PL_tokenbuf);
10366 ? (PL_hints & HINT_NEW_FLOAT) : (PL_hints & HINT_NEW_INTEGER) ) {
10367 const char *const key = floatit ? "float" : "integer";
10368 const STRLEN keylen = floatit ? 5 : 7;
10369 sv = S_new_constant(aTHX_ PL_tokenbuf, d - PL_tokenbuf,
10370 key, keylen, sv, NULL, NULL, 0);
10374 /* if it starts with a v, it could be a v-string */
10377 sv = newSV(5); /* preallocate storage space */
10378 s = scan_vstring(s, PL_bufend, sv);
10382 /* make the op for the constant and return */
10385 lvalp->opval = newSVOP(OP_CONST, 0, sv);
10387 lvalp->opval = NULL;
10393 S_scan_formline(pTHX_ register char *s)
10396 register char *eol;
10398 SV * const stuff = newSVpvs("");
10399 bool needargs = FALSE;
10400 bool eofmt = FALSE;
10402 char *tokenstart = s;
10403 SV* savewhite = NULL;
10405 if (PL_madskills) {
10406 savewhite = PL_thiswhite;
10411 PERL_ARGS_ASSERT_SCAN_FORMLINE;
10413 while (!needargs) {
10416 #ifdef PERL_STRICT_CR
10417 while (SPACE_OR_TAB(*t))
10420 while (SPACE_OR_TAB(*t) || *t == '\r')
10423 if (*t == '\n' || t == PL_bufend) {
10428 if (PL_in_eval && !PL_rsfp) {
10429 eol = (char *) memchr(s,'\n',PL_bufend-s);
10434 eol = PL_bufend = SvPVX(PL_linestr) + SvCUR(PL_linestr);
10436 for (t = s; t < eol; t++) {
10437 if (*t == '~' && t[1] == '~' && SvCUR(stuff)) {
10439 goto enough; /* ~~ must be first line in formline */
10441 if (*t == '@' || *t == '^')
10445 sv_catpvn(stuff, s, eol-s);
10446 #ifndef PERL_STRICT_CR
10447 if (eol-s > 1 && eol[-2] == '\r' && eol[-1] == '\n') {
10448 char *end = SvPVX(stuff) + SvCUR(stuff);
10451 SvCUR_set(stuff, SvCUR(stuff) - 1);
10462 if (PL_madskills) {
10464 sv_catpvn(PL_thistoken, tokenstart, PL_bufend - tokenstart);
10466 PL_thistoken = newSVpvn(tokenstart, PL_bufend - tokenstart);
10469 PL_bufptr = PL_bufend;
10470 CopLINE_inc(PL_curcop);
10471 got_some = lex_next_chunk(0);
10472 CopLINE_dec(PL_curcop);
10475 tokenstart = PL_bufptr;
10483 if (SvCUR(stuff)) {
10486 PL_lex_state = LEX_NORMAL;
10487 start_force(PL_curforce);
10488 NEXTVAL_NEXTTOKE.ival = 0;
10492 PL_lex_state = LEX_FORMLINE;
10494 if (UTF && is_utf8_string((U8*)SvPVX_const(stuff), SvCUR(stuff)))
10496 else if (PL_encoding)
10497 sv_recode_to_utf8(stuff, PL_encoding);
10499 start_force(PL_curforce);
10500 NEXTVAL_NEXTTOKE.opval = (OP*)newSVOP(OP_CONST, 0, stuff);
10502 start_force(PL_curforce);
10503 NEXTVAL_NEXTTOKE.ival = OP_FORMLINE;
10507 SvREFCNT_dec(stuff);
10509 PL_lex_formbrack = 0;
10513 if (PL_madskills) {
10515 sv_catpvn(PL_thistoken, tokenstart, s - tokenstart);
10517 PL_thistoken = newSVpvn(tokenstart, s - tokenstart);
10518 PL_thiswhite = savewhite;
10525 Perl_start_subparse(pTHX_ I32 is_format, U32 flags)
10528 const I32 oldsavestack_ix = PL_savestack_ix;
10529 CV* const outsidecv = PL_compcv;
10532 assert(SvTYPE(PL_compcv) == SVt_PVCV);
10534 SAVEI32(PL_subline);
10535 save_item(PL_subname);
10536 SAVESPTR(PL_compcv);
10538 PL_compcv = MUTABLE_CV(newSV_type(is_format ? SVt_PVFM : SVt_PVCV));
10539 CvFLAGS(PL_compcv) |= flags;
10541 PL_subline = CopLINE(PL_curcop);
10542 CvPADLIST(PL_compcv) = pad_new(padnew_SAVE|padnew_SAVESUB);
10543 CvOUTSIDE(PL_compcv) = MUTABLE_CV(SvREFCNT_inc_simple(outsidecv));
10544 CvOUTSIDE_SEQ(PL_compcv) = PL_cop_seqmax;
10546 return oldsavestack_ix;
10550 #pragma segment Perl_yylex
10553 S_yywarn(pTHX_ const char *const s)
10557 PERL_ARGS_ASSERT_YYWARN;
10559 PL_in_eval |= EVAL_WARNONLY;
10561 PL_in_eval &= ~EVAL_WARNONLY;
10566 Perl_yyerror(pTHX_ const char *const s)
10569 const char *where = NULL;
10570 const char *context = NULL;
10573 int yychar = PL_parser->yychar;
10575 PERL_ARGS_ASSERT_YYERROR;
10577 if (!yychar || (yychar == ';' && !PL_rsfp))
10579 else if (PL_oldoldbufptr && PL_bufptr > PL_oldoldbufptr &&
10580 PL_bufptr - PL_oldoldbufptr < 200 && PL_oldoldbufptr != PL_oldbufptr &&
10581 PL_oldbufptr != PL_bufptr) {
10584 The code below is removed for NetWare because it abends/crashes on NetWare
10585 when the script has error such as not having the closing quotes like:
10586 if ($var eq "value)
10587 Checking of white spaces is anyway done in NetWare code.
10590 while (isSPACE(*PL_oldoldbufptr))
10593 context = PL_oldoldbufptr;
10594 contlen = PL_bufptr - PL_oldoldbufptr;
10596 else if (PL_oldbufptr && PL_bufptr > PL_oldbufptr &&
10597 PL_bufptr - PL_oldbufptr < 200 && PL_oldbufptr != PL_bufptr) {
10600 The code below is removed for NetWare because it abends/crashes on NetWare
10601 when the script has error such as not having the closing quotes like:
10602 if ($var eq "value)
10603 Checking of white spaces is anyway done in NetWare code.
10606 while (isSPACE(*PL_oldbufptr))
10609 context = PL_oldbufptr;
10610 contlen = PL_bufptr - PL_oldbufptr;
10612 else if (yychar > 255)
10613 where = "next token ???";
10614 else if (yychar == -2) { /* YYEMPTY */
10615 if (PL_lex_state == LEX_NORMAL ||
10616 (PL_lex_state == LEX_KNOWNEXT && PL_lex_defer == LEX_NORMAL))
10617 where = "at end of line";
10618 else if (PL_lex_inpat)
10619 where = "within pattern";
10621 where = "within string";
10624 SV * const where_sv = newSVpvs_flags("next char ", SVs_TEMP);
10626 Perl_sv_catpvf(aTHX_ where_sv, "^%c", toCTRL(yychar));
10627 else if (isPRINT_LC(yychar)) {
10628 const char string = yychar;
10629 sv_catpvn(where_sv, &string, 1);
10632 Perl_sv_catpvf(aTHX_ where_sv, "\\%03o", yychar & 255);
10633 where = SvPVX_const(where_sv);
10635 msg = sv_2mortal(newSVpv(s, 0));
10636 Perl_sv_catpvf(aTHX_ msg, " at %s line %"IVdf", ",
10637 OutCopFILE(PL_curcop), (IV)CopLINE(PL_curcop));
10639 Perl_sv_catpvf(aTHX_ msg, "near \"%.*s\"\n", contlen, context);
10641 Perl_sv_catpvf(aTHX_ msg, "%s\n", where);
10642 if (PL_multi_start < PL_multi_end && (U32)(CopLINE(PL_curcop) - PL_multi_end) <= 1) {
10643 Perl_sv_catpvf(aTHX_ msg,
10644 " (Might be a runaway multi-line %c%c string starting on line %"IVdf")\n",
10645 (int)PL_multi_open,(int)PL_multi_close,(IV)PL_multi_start);
10648 if (PL_in_eval & EVAL_WARNONLY) {
10649 Perl_ck_warner_d(aTHX_ packWARN(WARN_SYNTAX), "%"SVf, SVfARG(msg));
10653 if (PL_error_count >= 10) {
10654 if (PL_in_eval && SvCUR(ERRSV))
10655 Perl_croak(aTHX_ "%"SVf"%s has too many errors.\n",
10656 SVfARG(ERRSV), OutCopFILE(PL_curcop));
10658 Perl_croak(aTHX_ "%s has too many errors.\n",
10659 OutCopFILE(PL_curcop));
10662 PL_in_my_stash = NULL;
10666 #pragma segment Main
10670 S_swallow_bom(pTHX_ U8 *s)
10673 const STRLEN slen = SvCUR(PL_linestr);
10675 PERL_ARGS_ASSERT_SWALLOW_BOM;
10679 if (s[1] == 0xFE) {
10680 /* UTF-16 little-endian? (or UTF-32LE?) */
10681 if (s[2] == 0 && s[3] == 0) /* UTF-32 little-endian */
10682 Perl_croak(aTHX_ "Unsupported script encoding UTF-32LE");
10683 #ifndef PERL_NO_UTF16_FILTER
10684 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (BOM)\n");
10686 if (PL_bufend > (char*)s) {
10687 s = add_utf16_textfilter(s, TRUE);
10690 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10695 if (s[1] == 0xFF) { /* UTF-16 big-endian? */
10696 #ifndef PERL_NO_UTF16_FILTER
10697 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (BOM)\n");
10699 if (PL_bufend > (char *)s) {
10700 s = add_utf16_textfilter(s, FALSE);
10703 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10708 if (slen > 2 && s[1] == 0xBB && s[2] == 0xBF) {
10709 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10710 s += 3; /* UTF-8 */
10716 if (s[2] == 0xFE && s[3] == 0xFF) {
10717 /* UTF-32 big-endian */
10718 Perl_croak(aTHX_ "Unsupported script encoding UTF-32BE");
10721 else if (s[2] == 0 && s[3] != 0) {
10724 * are a good indicator of UTF-16BE. */
10725 #ifndef PERL_NO_UTF16_FILTER
10726 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16BE script encoding (no BOM)\n");
10727 s = add_utf16_textfilter(s, FALSE);
10729 Perl_croak(aTHX_ "Unsupported script encoding UTF-16BE");
10735 if (slen > 3 && s[1] == 0x73 && s[2] == 0x66 && s[3] == 0x73) {
10736 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-8 script encoding (BOM)\n");
10737 s += 4; /* UTF-8 */
10743 if (slen > 3 && s[1] == 0 && s[2] != 0 && s[3] == 0) {
10746 * are a good indicator of UTF-16LE. */
10747 #ifndef PERL_NO_UTF16_FILTER
10748 if (DEBUG_p_TEST || DEBUG_T_TEST) PerlIO_printf(Perl_debug_log, "UTF-16LE script encoding (no BOM)\n");
10749 s = add_utf16_textfilter(s, TRUE);
10751 Perl_croak(aTHX_ "Unsupported script encoding UTF-16LE");
10759 #ifndef PERL_NO_UTF16_FILTER
10761 S_utf16_textfilter(pTHX_ int idx, SV *sv, int maxlen)
10764 SV *const filter = FILTER_DATA(idx);
10765 /* We re-use this each time round, throwing the contents away before we
10767 SV *const utf16_buffer = MUTABLE_SV(IoTOP_GV(filter));
10768 SV *const utf8_buffer = filter;
10769 IV status = IoPAGE(filter);
10770 const bool reverse = cBOOL(IoLINES(filter));
10773 PERL_ARGS_ASSERT_UTF16_TEXTFILTER;
10775 /* As we're automatically added, at the lowest level, and hence only called
10776 from this file, we can be sure that we're not called in block mode. Hence
10777 don't bother writing code to deal with block mode. */
10779 Perl_croak(aTHX_ "panic: utf16_textfilter called in block mode (for %d characters)", maxlen);
10782 Perl_croak(aTHX_ "panic: utf16_textfilter called after error (status=%"IVdf")", status);
10784 DEBUG_P(PerlIO_printf(Perl_debug_log,
10785 "utf16_textfilter(%p,%ce): idx=%d maxlen=%d status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10786 FPTR2DPTR(void *, S_utf16_textfilter),
10787 reverse ? 'l' : 'b', idx, maxlen, status,
10788 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10795 /* First, look in our buffer of existing UTF-8 data: */
10796 char *nl = (char *)memchr(SvPVX(utf8_buffer), '\n', SvCUR(utf8_buffer));
10800 } else if (status == 0) {
10802 IoPAGE(filter) = 0;
10803 nl = SvEND(utf8_buffer);
10806 STRLEN got = nl - SvPVX(utf8_buffer);
10807 /* Did we have anything to append? */
10809 sv_catpvn(sv, SvPVX(utf8_buffer), got);
10810 /* Everything else in this code works just fine if SVp_POK isn't
10811 set. This, however, needs it, and we need it to work, else
10812 we loop infinitely because the buffer is never consumed. */
10813 sv_chop(utf8_buffer, nl);
10817 /* OK, not a complete line there, so need to read some more UTF-16.
10818 Read an extra octect if the buffer currently has an odd number. */
10822 if (SvCUR(utf16_buffer) >= 2) {
10823 /* Location of the high octet of the last complete code point.
10824 Gosh, UTF-16 is a pain. All the benefits of variable length,
10825 *coupled* with all the benefits of partial reads and
10827 const U8 *const last_hi = (U8*)SvPVX(utf16_buffer)
10828 + ((SvCUR(utf16_buffer) & ~1) - (reverse ? 1 : 2));
10830 if (*last_hi < 0xd8 || *last_hi > 0xdb) {
10834 /* We have the first half of a surrogate. Read more. */
10835 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter partial surrogate detected at %p\n", last_hi));
10838 status = FILTER_READ(idx + 1, utf16_buffer,
10839 160 + (SvCUR(utf16_buffer) & 1));
10840 DEBUG_P(PerlIO_printf(Perl_debug_log, "utf16_textfilter status=%"IVdf" SvCUR(sv)=%"UVuf"\n", status, (UV)SvCUR(utf16_buffer)));
10841 DEBUG_P({ sv_dump(utf16_buffer); sv_dump(utf8_buffer);});
10844 IoPAGE(filter) = status;
10849 chars = SvCUR(utf16_buffer) >> 1;
10850 have = SvCUR(utf8_buffer);
10851 SvGROW(utf8_buffer, have + chars * 3 + 1);
10854 end = utf16_to_utf8_reversed((U8*)SvPVX(utf16_buffer),
10855 (U8*)SvPVX_const(utf8_buffer) + have,
10856 chars * 2, &newlen);
10858 end = utf16_to_utf8((U8*)SvPVX(utf16_buffer),
10859 (U8*)SvPVX_const(utf8_buffer) + have,
10860 chars * 2, &newlen);
10862 SvCUR_set(utf8_buffer, have + newlen);
10865 /* No need to keep this SV "well-formed" with a '\0' after the end, as
10866 it's private to us, and utf16_to_utf8{,reversed} take a
10867 (pointer,length) pair, rather than a NUL-terminated string. */
10868 if(SvCUR(utf16_buffer) & 1) {
10869 *SvPVX(utf16_buffer) = SvEND(utf16_buffer)[-1];
10870 SvCUR_set(utf16_buffer, 1);
10872 SvCUR_set(utf16_buffer, 0);
10875 DEBUG_P(PerlIO_printf(Perl_debug_log,
10876 "utf16_textfilter: returns, status=%"IVdf" utf16=%"UVuf" utf8=%"UVuf"\n",
10878 (UV)SvCUR(utf16_buffer), (UV)SvCUR(utf8_buffer)));
10879 DEBUG_P({ sv_dump(utf8_buffer); sv_dump(sv);});
10884 S_add_utf16_textfilter(pTHX_ U8 *const s, bool reversed)
10886 SV *filter = filter_add(S_utf16_textfilter, NULL);
10888 PERL_ARGS_ASSERT_ADD_UTF16_TEXTFILTER;
10890 IoTOP_GV(filter) = MUTABLE_GV(newSVpvn((char *)s, PL_bufend - (char*)s));
10891 sv_setpvs(filter, "");
10892 IoLINES(filter) = reversed;
10893 IoPAGE(filter) = 1; /* Not EOF */
10895 /* Sadly, we have to return a valid pointer, come what may, so we have to
10896 ignore any error return from this. */
10897 SvCUR_set(PL_linestr, 0);
10898 if (FILTER_READ(0, PL_linestr, 0)) {
10899 SvUTF8_on(PL_linestr);
10901 SvUTF8_on(PL_linestr);
10903 PL_bufend = SvEND(PL_linestr);
10904 return (U8*)SvPVX(PL_linestr);
10909 Returns a pointer to the next character after the parsed
10910 vstring, as well as updating the passed in sv.
10912 Function must be called like
10915 s = scan_vstring(s,e,sv);
10917 where s and e are the start and end of the string.
10918 The sv should already be large enough to store the vstring
10919 passed in, for performance reasons.
10924 Perl_scan_vstring(pTHX_ const char *s, const char *const e, SV *sv)
10927 const char *pos = s;
10928 const char *start = s;
10930 PERL_ARGS_ASSERT_SCAN_VSTRING;
10932 if (*pos == 'v') pos++; /* get past 'v' */
10933 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
10935 if ( *pos != '.') {
10936 /* this may not be a v-string if followed by => */
10937 const char *next = pos;
10938 while (next < e && isSPACE(*next))
10940 if ((e - next) >= 2 && *next == '=' && next[1] == '>' ) {
10941 /* return string not v-string */
10942 sv_setpvn(sv,(char *)s,pos-s);
10943 return (char *)pos;
10947 if (!isALPHA(*pos)) {
10948 U8 tmpbuf[UTF8_MAXBYTES+1];
10951 s++; /* get past 'v' */
10956 /* this is atoi() that tolerates underscores */
10959 const char *end = pos;
10961 while (--end >= s) {
10963 const UV orev = rev;
10964 rev += (*end - '0') * mult;
10967 Perl_ck_warner_d(aTHX_ packWARN(WARN_OVERFLOW),
10968 "Integer overflow in decimal number");
10972 if (rev > 0x7FFFFFFF)
10973 Perl_croak(aTHX_ "In EBCDIC the v-string components cannot exceed 2147483647");
10975 /* Append native character for the rev point */
10976 tmpend = uvchr_to_utf8(tmpbuf, rev);
10977 sv_catpvn(sv, (const char*)tmpbuf, tmpend - tmpbuf);
10978 if (!UNI_IS_INVARIANT(NATIVE_TO_UNI(rev)))
10980 if (pos + 1 < e && *pos == '.' && isDIGIT(pos[1]))
10986 while (pos < e && (isDIGIT(*pos) || *pos == '_'))
10990 sv_magic(sv,NULL,PERL_MAGIC_vstring,(const char*)start, pos-start);
10997 Perl_keyword_plugin_standard(pTHX_
10998 char *keyword_ptr, STRLEN keyword_len, OP **op_ptr)
11000 PERL_ARGS_ASSERT_KEYWORD_PLUGIN_STANDARD;
11001 PERL_UNUSED_CONTEXT;
11002 PERL_UNUSED_ARG(keyword_ptr);
11003 PERL_UNUSED_ARG(keyword_len);
11004 PERL_UNUSED_ARG(op_ptr);
11005 return KEYWORD_PLUGIN_DECLINE;
11008 #define parse_recdescent(g,p) S_parse_recdescent(aTHX_ g,p)
11010 S_parse_recdescent(pTHX_ int gramtype, I32 fakeeof)
11012 SAVEI32(PL_lex_brackets);
11013 if (PL_lex_brackets > 100)
11014 Renew(PL_lex_brackstack, PL_lex_brackets + 10, char);
11015 PL_lex_brackstack[PL_lex_brackets++] = XFAKEEOF;
11016 SAVEI32(PL_lex_allbrackets);
11017 PL_lex_allbrackets = 0;
11018 SAVEI8(PL_lex_fakeeof);
11019 PL_lex_fakeeof = (U8)fakeeof;
11020 if(yyparse(gramtype) && !PL_parser->error_count)
11021 qerror(Perl_mess(aTHX_ "Parse error"));
11024 #define parse_recdescent_for_op(g,p) S_parse_recdescent_for_op(aTHX_ g,p)
11026 S_parse_recdescent_for_op(pTHX_ int gramtype, I32 fakeeof)
11030 SAVEVPTR(PL_eval_root);
11031 PL_eval_root = NULL;
11032 parse_recdescent(gramtype, fakeeof);
11038 #define parse_expr(p,f) S_parse_expr(aTHX_ p,f)
11040 S_parse_expr(pTHX_ I32 fakeeof, U32 flags)
11043 if (flags & ~PARSE_OPTIONAL)
11044 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_expr");
11045 exprop = parse_recdescent_for_op(GRAMEXPR, fakeeof);
11046 if (!exprop && !(flags & PARSE_OPTIONAL)) {
11047 if (!PL_parser->error_count)
11048 qerror(Perl_mess(aTHX_ "Parse error"));
11049 exprop = newOP(OP_NULL, 0);
11055 =for apidoc Amx|OP *|parse_arithexpr|U32 flags
11057 Parse a Perl arithmetic expression. This may contain operators of precedence
11058 down to the bit shift operators. The expression must be followed (and thus
11059 terminated) either by a comparison or lower-precedence operator or by
11060 something that would normally terminate an expression such as semicolon.
11061 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11062 otherwise it is mandatory. It is up to the caller to ensure that the
11063 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11064 the source of the code to be parsed and the lexical context for the
11067 The op tree representing the expression is returned. If an optional
11068 expression is absent, a null pointer is returned, otherwise the pointer
11071 If an error occurs in parsing or compilation, in most cases a valid op
11072 tree is returned anyway. The error is reflected in the parser state,
11073 normally resulting in a single exception at the top level of parsing
11074 which covers all the compilation errors that occurred. Some compilation
11075 errors, however, will throw an exception immediately.
11081 Perl_parse_arithexpr(pTHX_ U32 flags)
11083 return parse_expr(LEX_FAKEEOF_COMPARE, flags);
11087 =for apidoc Amx|OP *|parse_termexpr|U32 flags
11089 Parse a Perl term expression. This may contain operators of precedence
11090 down to the assignment operators. The expression must be followed (and thus
11091 terminated) either by a comma or lower-precedence operator or by
11092 something that would normally terminate an expression such as semicolon.
11093 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11094 otherwise it is mandatory. It is up to the caller to ensure that the
11095 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11096 the source of the code to be parsed and the lexical context for the
11099 The op tree representing the expression is returned. If an optional
11100 expression is absent, a null pointer is returned, otherwise the pointer
11103 If an error occurs in parsing or compilation, in most cases a valid op
11104 tree is returned anyway. The error is reflected in the parser state,
11105 normally resulting in a single exception at the top level of parsing
11106 which covers all the compilation errors that occurred. Some compilation
11107 errors, however, will throw an exception immediately.
11113 Perl_parse_termexpr(pTHX_ U32 flags)
11115 return parse_expr(LEX_FAKEEOF_COMMA, flags);
11119 =for apidoc Amx|OP *|parse_listexpr|U32 flags
11121 Parse a Perl list expression. This may contain operators of precedence
11122 down to the comma operator. The expression must be followed (and thus
11123 terminated) either by a low-precedence logic operator such as C<or> or by
11124 something that would normally terminate an expression such as semicolon.
11125 If I<flags> includes C<PARSE_OPTIONAL> then the expression is optional,
11126 otherwise it is mandatory. It is up to the caller to ensure that the
11127 dynamic parser state (L</PL_parser> et al) is correctly set to reflect
11128 the source of the code to be parsed and the lexical context for the
11131 The op tree representing the expression is returned. If an optional
11132 expression is absent, a null pointer is returned, otherwise the pointer
11135 If an error occurs in parsing or compilation, in most cases a valid op
11136 tree is returned anyway. The error is reflected in the parser state,
11137 normally resulting in a single exception at the top level of parsing
11138 which covers all the compilation errors that occurred. Some compilation
11139 errors, however, will throw an exception immediately.
11145 Perl_parse_listexpr(pTHX_ U32 flags)
11147 return parse_expr(LEX_FAKEEOF_LOWLOGIC, flags);
11151 =for apidoc Amx|OP *|parse_fullexpr|U32 flags
11153 Parse a single complete Perl expression. This allows the full
11154 expression grammar, including the lowest-precedence operators such
11155 as C<or>. The expression must be followed (and thus terminated) by a
11156 token that an expression would normally be terminated by: end-of-file,
11157 closing bracketing punctuation, semicolon, or one of the keywords that
11158 signals a postfix expression-statement modifier. If I<flags> includes
11159 C<PARSE_OPTIONAL> then the expression is optional, otherwise it is
11160 mandatory. It is up to the caller to ensure that the dynamic parser
11161 state (L</PL_parser> et al) is correctly set to reflect the source of
11162 the code to be parsed and the lexical context for the expression.
11164 The op tree representing the expression is returned. If an optional
11165 expression is absent, a null pointer is returned, otherwise the pointer
11168 If an error occurs in parsing or compilation, in most cases a valid op
11169 tree is returned anyway. The error is reflected in the parser state,
11170 normally resulting in a single exception at the top level of parsing
11171 which covers all the compilation errors that occurred. Some compilation
11172 errors, however, will throw an exception immediately.
11178 Perl_parse_fullexpr(pTHX_ U32 flags)
11180 return parse_expr(LEX_FAKEEOF_NONEXPR, flags);
11184 =for apidoc Amx|OP *|parse_block|U32 flags
11186 Parse a single complete Perl code block. This consists of an opening
11187 brace, a sequence of statements, and a closing brace. The block
11188 constitutes a lexical scope, so C<my> variables and various compile-time
11189 effects can be contained within it. It is up to the caller to ensure
11190 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11191 reflect the source of the code to be parsed and the lexical context for
11194 The op tree representing the code block is returned. This is always a
11195 real op, never a null pointer. It will normally be a C<lineseq> list,
11196 including C<nextstate> or equivalent ops. No ops to construct any kind
11197 of runtime scope are included by virtue of it being a block.
11199 If an error occurs in parsing or compilation, in most cases a valid op
11200 tree (most likely null) is returned anyway. The error is reflected in
11201 the parser state, normally resulting in a single exception at the top
11202 level of parsing which covers all the compilation errors that occurred.
11203 Some compilation errors, however, will throw an exception immediately.
11205 The I<flags> parameter is reserved for future use, and must always
11212 Perl_parse_block(pTHX_ U32 flags)
11215 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_block");
11216 return parse_recdescent_for_op(GRAMBLOCK, LEX_FAKEEOF_NEVER);
11220 =for apidoc Amx|OP *|parse_barestmt|U32 flags
11222 Parse a single unadorned Perl statement. This may be a normal imperative
11223 statement or a declaration that has compile-time effect. It does not
11224 include any label or other affixture. It is up to the caller to ensure
11225 that the dynamic parser state (L</PL_parser> et al) is correctly set to
11226 reflect the source of the code to be parsed and the lexical context for
11229 The op tree representing the statement is returned. This may be a
11230 null pointer if the statement is null, for example if it was actually
11231 a subroutine definition (which has compile-time side effects). If not
11232 null, it will be ops directly implementing the statement, suitable to
11233 pass to L</newSTATEOP>. It will not normally include a C<nextstate> or
11234 equivalent op (except for those embedded in a scope contained entirely
11235 within the statement).
11237 If an error occurs in parsing or compilation, in most cases a valid op
11238 tree (most likely null) is returned anyway. The error is reflected in
11239 the parser state, normally resulting in a single exception at the top
11240 level of parsing which covers all the compilation errors that occurred.
11241 Some compilation errors, however, will throw an exception immediately.
11243 The I<flags> parameter is reserved for future use, and must always
11250 Perl_parse_barestmt(pTHX_ U32 flags)
11253 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_barestmt");
11254 return parse_recdescent_for_op(GRAMBARESTMT, LEX_FAKEEOF_NEVER);
11258 =for apidoc Amx|SV *|parse_label|U32 flags
11260 Parse a single label, possibly optional, of the type that may prefix a
11261 Perl statement. It is up to the caller to ensure that the dynamic parser
11262 state (L</PL_parser> et al) is correctly set to reflect the source of
11263 the code to be parsed. If I<flags> includes C<PARSE_OPTIONAL> then the
11264 label is optional, otherwise it is mandatory.
11266 The name of the label is returned in the form of a fresh scalar. If an
11267 optional label is absent, a null pointer is returned.
11269 If an error occurs in parsing, which can only occur if the label is
11270 mandatory, a valid label is returned anyway. The error is reflected in
11271 the parser state, normally resulting in a single exception at the top
11272 level of parsing which covers all the compilation errors that occurred.
11278 Perl_parse_label(pTHX_ U32 flags)
11280 if (flags & ~PARSE_OPTIONAL)
11281 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_label");
11282 if (PL_lex_state == LEX_KNOWNEXT) {
11283 PL_parser->yychar = yylex();
11284 if (PL_parser->yychar == LABEL) {
11285 char *lpv = pl_yylval.pval;
11286 STRLEN llen = strlen(lpv);
11288 PL_parser->yychar = YYEMPTY;
11289 lsv = newSV_type(SVt_PV);
11290 SvPV_set(lsv, lpv);
11291 SvCUR_set(lsv, llen);
11292 SvLEN_set(lsv, llen+1);
11302 STRLEN wlen, bufptr_pos;
11306 if (!isIDFIRST_A(c))
11310 } while(isWORDCHAR_A(c));
11312 if (word_takes_any_delimeter(s, wlen))
11314 bufptr_pos = s - SvPVX(PL_linestr);
11316 lex_read_space(LEX_KEEP_PREVIOUS);
11318 s = SvPVX(PL_linestr) + bufptr_pos;
11319 if (t[0] == ':' && t[1] != ':') {
11320 PL_oldoldbufptr = PL_oldbufptr;
11323 return newSVpvn(s, wlen);
11327 if (flags & PARSE_OPTIONAL) {
11330 qerror(Perl_mess(aTHX_ "Parse error"));
11331 return newSVpvs("x");
11338 =for apidoc Amx|OP *|parse_fullstmt|U32 flags
11340 Parse a single complete Perl statement. This may be a normal imperative
11341 statement or a declaration that has compile-time effect, and may include
11342 optional labels. It is up to the caller to ensure that the dynamic
11343 parser state (L</PL_parser> et al) is correctly set to reflect the source
11344 of the code to be parsed and the lexical context for the statement.
11346 The op tree representing the statement is returned. This may be a
11347 null pointer if the statement is null, for example if it was actually
11348 a subroutine definition (which has compile-time side effects). If not
11349 null, it will be the result of a L</newSTATEOP> call, normally including
11350 a C<nextstate> or equivalent op.
11352 If an error occurs in parsing or compilation, in most cases a valid op
11353 tree (most likely null) is returned anyway. The error is reflected in
11354 the parser state, normally resulting in a single exception at the top
11355 level of parsing which covers all the compilation errors that occurred.
11356 Some compilation errors, however, will throw an exception immediately.
11358 The I<flags> parameter is reserved for future use, and must always
11365 Perl_parse_fullstmt(pTHX_ U32 flags)
11368 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_fullstmt");
11369 return parse_recdescent_for_op(GRAMFULLSTMT, LEX_FAKEEOF_NEVER);
11373 =for apidoc Amx|OP *|parse_stmtseq|U32 flags
11375 Parse a sequence of zero or more Perl statements. These may be normal
11376 imperative statements, including optional labels, or declarations
11377 that have compile-time effect, or any mixture thereof. The statement
11378 sequence ends when a closing brace or end-of-file is encountered in a
11379 place where a new statement could have validly started. It is up to
11380 the caller to ensure that the dynamic parser state (L</PL_parser> et al)
11381 is correctly set to reflect the source of the code to be parsed and the
11382 lexical context for the statements.
11384 The op tree representing the statement sequence is returned. This may
11385 be a null pointer if the statements were all null, for example if there
11386 were no statements or if there were only subroutine definitions (which
11387 have compile-time side effects). If not null, it will be a C<lineseq>
11388 list, normally including C<nextstate> or equivalent ops.
11390 If an error occurs in parsing or compilation, in most cases a valid op
11391 tree is returned anyway. The error is reflected in the parser state,
11392 normally resulting in a single exception at the top level of parsing
11393 which covers all the compilation errors that occurred. Some compilation
11394 errors, however, will throw an exception immediately.
11396 The I<flags> parameter is reserved for future use, and must always
11403 Perl_parse_stmtseq(pTHX_ U32 flags)
11408 Perl_croak(aTHX_ "Parsing code internal error (%s)", "parse_stmtseq");
11409 stmtseqop = parse_recdescent_for_op(GRAMSTMTSEQ, LEX_FAKEEOF_CLOSING);
11410 c = lex_peek_unichar(0);
11411 if (c != -1 && c != /*{*/'}')
11412 qerror(Perl_mess(aTHX_ "Parse error"));
11417 Perl_munge_qwlist_to_paren_list(pTHX_ OP *qwlist)
11419 PERL_ARGS_ASSERT_MUNGE_QWLIST_TO_PAREN_LIST;
11420 deprecate("qw(...) as parentheses");
11421 force_next((4<<24)|')');
11422 if (qwlist->op_type == OP_STUB) {
11426 start_force(PL_curforce);
11427 NEXTVAL_NEXTTOKE.opval = qwlist;
11430 force_next((2<<24)|'(');
11435 * c-indentation-style: bsd
11436 * c-basic-offset: 4
11437 * indent-tabs-mode: t
11440 * ex: set ts=8 sts=4 sw=4 noet: